Advances in Cartography and GIScience. Volume 2

Lecture Notes in Geoinformation and Cartography SubSeries: Publications of the International Cartographic Association (ICA) Series Editors: William Cartwright, Georg Gartner, Liqiu Meng, Michael P. Peterson

Anne Ruas Editor

Advances in Cartography and GIScience. Volume 2 Selection from ICC 2011, Paris

Editor Anne Ruas Laboratoire COGIT- IGN 73 Avenue de Paris 94160 Saint Mandé France [email protected]

ISSN 1863-2246 e-ISSN 1863-2351 ISBN 978-3-642-19213-5 e-ISBN 978-3-642-19214-2 DOI 10.1007/978-3-642-19214-2 Springer Heidelberg Dordrecht London New York Library of Congress Control Number: 2011926986

© Springer-Verlag Berlin Heidelberg 2011 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer. Violations are liable to prosecution under the German Copyright Law. The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Cover design: SPi Publisher Services Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

Foreword The International Cartographic Association (ICA) has existed since 1959. Initially centered on Cartography, the themes of the Association have changed from years to years integrating new research and technical domains such as web services, location based services or the digitalization and analysis of historical maps. The ICA has 28 commissions and organizes an international conference on Cartography and GIS every two years. This international conference lasts 5 days and gathers between 1000 and 2000 attendants. The ten previous conferences were in Bournemouth (1991), Köln (1993), Barcelona (1995), Stockholm (1997), Ottawa (1999), Beijing (2001), Durban (2003), A Coruña (2005), Moscow (2007) and Santiago de Chile (2009). In 2011 the ICA conference takes place in Paris and for the first time the best ICC papers are published in the new Springer subseries, Publications of the ICA. For the ICC2011, a large international and French scientific committee has been appointed to select the papers and the poster presentations. More than 900 papers were submitted, among which 245 papers were entered for possible ICA Journal or Springer book publication. Of these 245 long papers, only 33% were accepted by means of a double blind review process to ensure a top quality publication. These proceedings thus present the very best ICC2011 papers. The large number of submissions illustrates the importance of the GIS and mapping community all over the world, and we believe that this volume of proceedings is a valuable mirror of our community. The book is composed of 62 papers, organized in 12 sections shared in two volumes. The digital version of this proceeding is available on Springer web site. The first papers deal with map design and analysis. They present methods to map acoustic information, to analyze semantic information via cloud visualization techniques or even to analyze society through the legend of their topographic maps. The second section deals with the use and user issues, focusing not only on the analysis of user needs to identify map content but also on the usability evaluation of geoservices. Section three highlights collaboration tools and processes to either integrate different data sets, or alternatively to propose collaborative and interactive tabletops. Section four focuses on solutions of how to find the appropriate data or services once a need has been defined. Solutions mainly lay on the constitution of ontologies of geographical concepts, names or services. Section five is devoted to generalization. This large section (9 papers) illustrates the dynamism of the research community in this field, encompassing algorithmic solutions in order to select points, to generalize river and road networks, isobathymetric lines or

v

vi

Foreword

polygons. A further paper describes a comprehensive method to combine different generalization processes. The second volume begins with papers related to Map GIS and Education including, for example, the use of Chernhoff face (a schematic human face) to represent variables, and the use of a web server to teach GIS. Section seven presents 7 papers on historical data. There is a real challenge to explore, digitize and analyse historical data. Another paper presents, for example, a GIS for archaeology to support excavation site research. Others provide solutions for the referencing of historical maps so as to facilitate the access and analysis of such data. Section eight presents 2 papers on map projections, one using an empirical process to discover the best projection for existing maps. Section nine presents current work on planet and space cartography, encompassing the conception of appropriate symbology, the integration of multisource data or the determination of nomenclature for extraterrestrial landscapes. The last three sections focus on specific analysis. Section ten describes methods to create information from image processing or to map the resulting information. Section eleven is centered on DTM and terrain analysis, with a paper concentrating on the analysis of glacial areas. Last but not least, section twelve offers models and methods to study specific applications such as urban growth, traffic, epidemiology or language distribution. Some simulation methods are discussed here. Urban growth models including road network expansion and land use development are presented. The extraordinary diversity of papers presented in these proceedings illustrates the dynamic nature and creativity of cartography and GIS today. We hope this volume will be the first of a long and valuable series.

Table of Content

VOLUME 2

Map, GIS and Education

1

Updating Research on Chernoff Faces for School Cartography José Jesús Reyes Nuñez, Anita Rohonczi, Cristina E. Juliarena de Moretti, Ana María Garra, Carmen Alicia Rey, María V. Alves de Castro, Anabella S. Dibiase, Teresa A. Saint Pierre, Mariana A. Campos

3

The Production of Maps for Students in the Context of School Culture Rosangela Doin de Almeida

21

Geoinformation: a social Issue Angelica Carvalho Di Maio, Cilene Gomes, Maria de Lourdes Neves de Oliveira Kurkdjian

35

Passing GIS Knowhow from University Students to secondary School Students: pedagogical Approach in developing youngster’s Capabilities and Understanding in GIS Amal Iaaly , Rola Jadayel, Oussama Jadayel

49

The Methodological Advantages of using Web Server in Teaching GIS Andrea Pődör

73

Historical data: exploration, digitalization, access an analysis

83

Geovisualization and Archaeology: supporting Excavation Site Research Spyridon Tsipidis, Alexandra Koussoulakou, Kostas Kotsakis

85

Virtual Recreation of the Monroy Roman Villa (Extremadura - Spain) Alan D.J. Atkinson ; Jose Juan de Sanjose Blasco ; Jorge Cilleros Recuero; Alberto García Martín and Fernando Berenguer Sempere

109

SEREDONA: a web platform to integrate historical geographic data into current georeferenced frameworks. Eric Grosso

117

vii

viii

Table of Content

Digital Processing and 3D Modelling of an 18th Century Scenographic Map of Bologna Gabriele Bitelli, Giorgia Gatta

129

Web Services and Historical Cadastral Maps: the first Step in the Implementation of the Web C.A.R.T.E. System Brovelli M. A.; Minghini M.; Valentini L

147

A Method for the Visual Representation of Historic Multivariate Point Data Alwyn Davidson, Colin Arrowsmith, and Deb Verhoeven

163

The Atlas and the Globe of Russian Geographical Explorations and Discoveries of the Earth: Concepts and Contents N.N. Komedchikov, V.M. Kotlyakov, A.G. Khropov, A.A. Medvedev, L.N. Zinchuk

179

Crossing Borders: Cartographic and Military Operations and the International Borders in the Libyan Desert before WW II Zsolt Győző Tőrök

187

Map Projection

207

Optimising the Distortions of sinusoidal- elliptical composite Projections Mátyás Gede

209

Using Empirical Map Projections for Modeling Early Nautical Charts Joaquim Alves Gaspar

227

Planet and Space Cartography

249

Requirements for Planetary Symbology in Geographic Information Systems A. Nass, S. van Gasselt, R. Jaumann, H. Asche

251

Venus Mapping at small Scale: Source Data Processing and cartographic Interpretation E Lazarev, J Rodionova

267

Planetary Nomenclature: a Representation of human Culture and alien Landscapes Henrik I. Hargitai, Kira Shingareva

275

A New Version of the Multilingual Glossary of Planetary Cartography Kira Shingareva, Bianna Krasnopevtseva

289

Table of Content

ix

Cartographic Support of a New Landing Site for “Phobos-Grunt” Mission Kira Shingareva, Bianna Krasnopevtseva, Anatoliy Konopikhin, Konstantin Zeljkov, Sergey Dubov

297

Making GeoInformation from Image Analysis

303

Evaluation of the spatial Dynamics of Great Oran (Algeria) using spatial Imagery and GIS Fouzia Bendraoua, Ali Bedidi, Bernard Cervelle

305

Topographical Mapping at 1:50,000 Scale from Satellite Imagery using CARTOSAT-1 Takka El-hadi

321

3D Topographic Mapping using TerraSAR-X Elevation Frank Teufel, Ernest Fahrland, Henning Schrader

333

Terrain Mapping and Analysis

355

An Influence of Spatial Range of Input Data set on Terrain Relief form Classification Homogeneity for Glacial Area Małgorzata Wieczorek

357

Accuracy Assessment of ASTER GDEM in North Shaanxi Xin Yang, Guoan Tang, Wei Zhang, Shijie Zhu

371

DEM based Terrain Factor of Soil Erosion at regional Scale and Soil Erosion Mapping Fayuan Li, Guoan Tang

383

Analysis and Simulation for Application Fields : Urban Growth, Traffic, Epidemiology and Language Locations

397

Urban Growth Modeling with Road Network Expansion and Land Use Development Yikang Rui, Yifang Ban

399

Conception of a GIS-Platform to simulate urban densification based on the analysis of topographic data.

413

x

Foreword

Anne Ruas; Julien Perret; Florence Curie; Annabelle Mas; Anne Puissant; Gregorz Skupinski, Dominique Badariotti; Christiane Weber; Pierre Gancarski; Nicolas Lachiche; Julien Lesbegueries; Agnès Braud Realistic Road Modelling for Driving Simulators using GIS Data Guillaume Despine, Caroline Baillard

431

Modeling and Mapping Traffic-Congested Corridors for Statewide Decision Support Jeong C. Seong, Habtewold Kassa, David Choi

449

How to Map out the Routes of Walkers in a Forestry Environment Considered to be of Risk? The Case of Human Exposure to Lyme Borreliosis in the Forest of Sénart (Île-de-France, France) Vincent Godard, Christelle Meha, Olivier Thomas

457

Estimation of the Locations of the Language-Versions of Wikipedia - a Case Study on Geographic Data Mining Tobias Dahinden

471

List of Authors Abadie, Nathalie Aburizaiza, Ahmad O. Alencar de Mendonça, André Luiz Almeida, Rosangela Doin de Alves de Castro, María V. Anderson-Tarve,r C. Arrowsmith, Colin Asche, Hartmut Atkinson, Alan D.J. Badariotti, Dominique Baillard, Caroline Bailly, G. Ban, Yifang Bedidi, Ali Bendraoua, Fouzia Bitelli, Gabriele Bláha, Jan D. Braud, Agnès Brovelli, M. A. Burghardt, Dirk Buttenfield, B.P. Campos, Mariana A. Canton, Fabio Cartwright, William Cervelle, Bernard Cheng, Tao Choi, David Cilleros Recuero, Jorge Coors, Volker Dahinden, Tobias Davidson, Alwyn de Sanjose Blasco, Jose Juan Despine, Guillaume Di Maio, Angelica Carvalho Dibiase, Anabella S. Dominguès, Catherine

Dubov, Sergey Duchêne, Cécile Dykes, J. El-hadi, Takka Elsley, Monique Engemaier, Rita Fahrland, Ernest Favetta, Frank Flink, Hanna-Marika Florence, Curie Friedmanová, Lucie Fritsch, E Gancarski, Pierre García Martín, Alberto Garra, Ana María Gaspar, Joaquim Alves Gatta, Giorgia Gede, Mátyás Godard, Vincent Gomes, Cilene Grosso, Eric Guilbert, Eric Hahmann, Stefan Haklay, Muki Hammill, William C. Hargitai, Henrik I. Hećimović, Željko Hoarau, Charlotte Iaaly, Amal Jacobson, Daniel R. Jadayel Rola Jadayel, Oussama Jakir, Željka Jaumann, R. Juliarena de Moretti, Cristina E. Karam, Roula Kassa, Habtewold xi

xii

Kent, Alexander J Khropov, A.G. Kilany, Rima Komedchikov, N.N. Konopikhin, Anatoliy Koont,z J.M. Kornfeld, A.-L. Kotlyakov, V.M. Kotsakis, Kostas Koussoulakou, Alexandra Krasnopevtseva, Bianna Kurkdjian, Maria de Lourdes Neves de Oliveira Lachiche, Nicolas Langiù, Giovanni Laurini, Robert Lazarev, E Lecolinet, É. Lesbegueries, Julien Li, Fayuan Li, Zhilin Mas, Annabelle Mechouche, Ammar Medvedev, A.A. Meha, Christelle Minghini, M. Mustière, Sébastien Nass, A. Oksanen, Juha Pauschert, Christian Perkins, Chris Perret, Julien Peters, Stefan Pődör, Andrea Prouteau, Emeric Puissant, Anne Pyysalo, Ulla Reineri, Marco Rey, Carmen Alicia Reyes Nuñez, José Jesús Rice, Matthew T. Riplinger, Emanuel

List of Authors

Rodionova, J Rohonczi, Anita Rönneberg, Mikko Ruas, Anne Rui, Yikang Rumor, Massimo Saint Pierre, Teresa A. Samsonov, Timofey Sarjakoski, L. Tiina Savino, Sandro Schiewe, J. Schrader, Henning Schwarz, Sara Sempere, Fernando Berenguer Seong, Jeong C. Shingareva, Kira Skarlatidou, Artemis Skopeliti, Andriani Skupinski, Gregorz Stamato Delazari, Luciene Stanĕk, Karel Stanislawski, L.V. Štefan, Zvonko Szegö, Janos Tang, Guoan Teufel, Frank Thomas, Olivier Tiede, Carola Tőrök, Zsolt Győző Touya, Guillaume Tsipidis, Spyridon Valentini, L van Gasselt, S. Verhoeven, Deb Viard, A. Wardlaw, Jessica Weber, Christiane Wieczorek, Małgorzata Yang, Xin Zanon, Matteo Zeljkov, Konstantin Zhang, Wei

List of Authors

Zhang, Xunruo Zhou, Qi

xiii

Zhu, Shijie Zinchuk, L.N.

International Scientific Committee Suchith ANAND Gennady ANDRIENKO Regina ARAUJO DE ALMEIDA Yuriy ARTEMYEV Hartmut ASCHE Thierry BADARD Temenoujka BANDROVA Miguel Angel BARNABÉ Manfred BUCHROITHNER Sébastien CAQUARD Patricio CARRASCO William CARTWRIGHT Nicholas CHRISMAN Derek CLARKE Peter COLLIER Josep-Lluis COLOMER Lex COMBER Antony COOPER Philippe DE MAYER Imre Josef DEMHARDT Jason DYKES Sara FABRIKANT David FAIRBAIRN David FORREST

David FRASER Ana Maria GARRA Georg GARTNER Pablo GRAN Eric GUILBERT Lorenz HURNI Dan JACOBSON Bin JIANG Peter JORDAN Milan KONECNY John KOSTELNICK Menno-Jan KRAAK Karel KRIZ Miljenko LAPAINE Jonathan LI Zhilin LI Elri LIEBENBERG Evangelos LIVIERATOS Christina LJUNGBERG Lucia LOVISON-GOLOB William A. MACKANESS Paulo MENEZES Liqiu MENG Takashi MORITA Sébastien MUSTIERE

Itzhak OMER Ferjan ORMELING Karel PAVELKA Chris PERKINS Michael P. PETERSON Barbara PIATTI Alexander PUCHER Ross PURVES José Jesús REYES Stéphane ROCHE Anne RUAS Tiina SARJAKOSKI Monika SESTER Kira B. SHINGAREVA Daan STREBE Vladimir TIKUNOV Tim TRAINOR Maria TSAKIRI-STRATI Lynn USERY Corné VAN ELZAKKER Robert WEIBEL Alexander WOLODTSCHENKO Xiaojun YANG James ZIMBELMAN Wieslawa ZYSZKOWSKA

French Scientific Committee Jean-Paul BORD Alain BOUJU Bénédicte BUCHER Elisabet CHESNEAU Hervé CHEVILLOTTE Sidonie CHRISTOPHE Christophe CLARAMUNT Paule-Annick DAVOINE Thomas DEVOGÈLE Cécile DUCHÊNE Mauro GAIO

Jean-Christophe GAY Jérôme GENSEL Elisabeth HABERT Catherine HOFMANN Thierry JOLIVEAU Didier JOSSELIN Robert LAURINI François LECORDIX Thérèse LIBOUREL Olivier LOISEAUX Hervé MARTIN

Annabelle MAS Hélène MATHIAN Gilles PALSKY Nicolas PAPARODITIS Olivier PARVILLERS Cyril RAY Didier RICHARD Hélène RICHARD Sylvie SERVIGNE Marlene VILLANOVA Christiane WEBER

xv

xvi

Acknowledgements

Acknowledgements This volume is the result of the extraordinary review process done by the international and French scientific committees of the ICC2011. 245 papers have been reviewed in 4 weeks by two reviewers, organised into 36 different themes. I wish to thank Samia Belbachir to help me recieving the papers and reviews on time and Helène Richard who help me for the selection process. I wish to thank François Lecordix, Madame Lecomte from the Comité Français de Cartographie and Charlotte Hoarau from the COGIT Laboratory who helped me for the editing process. Last but not least, I also wish to warmly thank the executive committee of the ICA- William Cartwright, Derek Clarke, David Fairbairn, Georg Gardner, Pablo Gran, Milan Konecny, Menno-Jan Kraak, Zhilin Li and Timothy Trainor- for their constant and friendly help and support.

Map, GIS and Education

Updating Research on Chernoff Faces for School Cartography José Jesús Reyes Nuñez1, Anita Rohonczi1, Cristina E. Juliarena de Moretti2, Ana María Garra2, Carmen Alicia Rey2, María V. Alves de Castro2, Anabella S. Dibiase2, Teresa A. Saint Pierre2, Mariana A. Campos2 1

Department of Cartography and Geoinformatics, Eötvös Loránd University, Budapest, Hungary. [email protected] 2

Centro Argentino de Cartografía, Buenos Aires, Argentina

Abstract Present work includes a short introduction about the Chernoff faces, emphasizing the importance of works related to its’ use in cartography. The possibilities of this method of representation in thematic maps, more specifically in school cartography (adapting the principle followed by Chernoff on pictograms) were studied within an innovative theoretical research. Some of the practical experiences acquired during the theoretical and practical teaching of this method for MSc students on Cartography at Eotvos Lorand University (Budapest, Hungary) are also included. The theoretical research was also tested in an international project counting with the participation of Argentine and Hungarian specialists, presenting some of the more characteristic results briefly. Its conclusions are a starting point to follow the research within an international project with specialists of the Vienna University of Technology, trying to find answers to the questions that remained open.

1- Short introduction to the Chernoff faces This method for data representation was created by Hermann Chernoff (at present Professor Emeritus of Applied Mathematics, Department of Statistics at Harvard University) in 1973 (Figure 1). The essence of his method is the use of the features of a human face to represent different variables, A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_1, © Springer-Verlag Berlin Heidelberg 2011

3

4

J.J.R. Nuñez et al.

changing the parameters that determine a feature according to the values of these variables. His original idea was to create a multivariate symbol easy to be recognized by readers mainly interested on statistical analysis. In the article written to introduce the method (Chernoff 1973) he affirmed that up to 18 themes or variables can be represented at same time.

Figure 1: Herman Chernoff’s portrait, example of the first faces (1973) and a fragment of the first thematic map made by E. Turner (1977)

Only four years after the publication of this article, the method of representation began to be introduced also to cartography abroad, using the human faces to represent data on a map according to the traditional methods of thematic representation. The first and more famous (today considered a classic) example is the map entitled “Life in Los Angeles, 1970”, designed by Eugene Turner and drafted by Richard Doss from the Geography Department at the California State University in 1977 (Figure 1). Turner wrote about this map: “It is probably one of the most interesting maps I've created because the expressions evoke an emotional association with the data. Some people don't like that.” (Turner 2004). Other specialists from different scientific fields began to research the possibilities of the method beginning from the 90’s. In this period the three more important names related to the correct use of Chernoff faces on maps are: • Danny Dorling (University of Newcastle upon Tyne), who obtained his PhD degree on the theme of visualization of spatial structure, combining his cartograms with Chernoff faces to represent the results of data analysis about elections in Great Britain (Dorling 1991). • Elizabeth S. Nelson, who beginning from the second half of the 90’s was having detailed research on specific aspects as feature salience and

Updating Research on Chernoff Faces for School Cartography

5

natural correspondence on Chernoff faces, and the exam of search process using Chernoff faces (Nelson 1997–2007). • Sarah I. Fabrikant (University of Zurich), who have developed numerous research on themes about data visualization, and in 2004 made a map entitled “Chernoff revisited: facing the presidential election” using morphed faces to represent the results of the elections in USA (Fabrikant 2004).

2- Theoretical analysis of the method and practical examples of proposed solutions for its use on maps Originally, this method was not created for the data representation on maps. Herman Chernoff is not a cartographer or a graphic specialist. He proposed a method for the graphic representation of data using a human face, but the characteristics of this method need to be adapted to the cartographic requirements before using it on a map. This is one of the reasons because the cartographic use of this method had not a real, more wide success during the past near 40 years (excepting some high quality maps created by cartographers and geographers, but they are a minority): some of the most important statistical software included the “map representation” of data visualized using Chernoff faces, but it was limited to draw very schematic, sometimes “caricaturistic” (nearly “antihuman”) faces not on a map, but on a sketched representation of a territory (most times countries) delimited only by a very generalized (or a very roughed) borderline. The schematic maps mentioned above reaffirm that the use of this method on maps requires a cartographer’s experience to adapt it for the cartographic conditions in interest of making a readable map with good graphical quality. In 1998 I met Chernoff faces for first time, in an international workshop run by Prof. Henry Castner (Greensboro, USA) during an international ICA Symposium organized in Wroclaw (Poland). The exercises presented by Prof. Castner raised my interest in this topic, following the study of the method during the next years. The preliminary study of research and practical works related to the published Chernoff maps, and the detailed theoretical aspects of the present research were presented in the 24th International Cartographic Conference by the author (Reyes 2009). One main conclusion of this research was the limitation of the number of variables that can be represented by using an easy Chernoff face to a max. of six, applying some principles used successfully in cartography and

6

J.J.R. Nuñez et al.

described by Jacques Bertin in his Graphic Semiology (use of fill and change of size) in 1969 (Figure 2).

Figure 2: The six graphic parameters proposed to change in a Chernoff face for thematic mapping (Reyes 2009)

The reason of this decision can be explained shortly: our main interest was to study the possibilities of the method in school cartography, so we had to make the face easy to be read by pupils from Elementary and Secondary Schools. In other hand, considering the human abilities to recognize graphic differences and to make graphic comparisons, the reading and analysis of numerous variables represented on a face (the maximal number was originally fixed on 18) is a very difficult task that demands too much time and attention. Studying some of the “maps” made with this method (and some variants of the method making it more detailed and growing the number of variables that can be represented), we can see that the time required to read and compare the information became as long as the read of the original statistical database. In those cases the graphic representation of the information does not fill its original objective of facilitating a faster and easier reading of the data. In 2005 this topic was introduced as one of the themes to teach in the final semester of the subject entitled “Thematic Cartography” developed for MSc students of Cartography at Eötvös Loránd University in Budapest, Hungary. During the first years, students worked on creating their own thematic maps using Chernoff faces to represent four variables belonging to a main theme, trying to improve the faces with the use of graphic solutions developed by the traditional thematic cartography and combining the faces with other traditional methods of representation (e.g. choroplets) or using them as proportional symbols too.

Updating Research on Chernoff Faces for School Cartography

7

During the practices numerous positive and negative experiences were acquired and discussed by the author with his students. In these discussions was proposed a possible new direction for a future research, which can be resumed with a short question: Why should only human faces to be used? Chernoff created more than a method of representation using a face: he determined a principle to divide a graphic symbol into its more relevant features or components, using each of these components to represent a different variable. Why not to apply this principle on map symbols, improving their traditional use in cartography? For centuries symbols were mainly used to represent only one theme in the traditional cartography: more often the size (less frequently the shape, the fill or the outline) was changed to represent a theme. Using the principle followed by Chernoff to create the faces, new parameters can be changed within a symbol (Figure 3), and the number of themes to be represented would grow significantly. This theory began to be experimented during the 2007/2008 school year, giving to the students a combined task: first of all to represent thematic data of selected Hungarian counties using Chernoff faces, and later to represent the same data on the same base map applying the Chernoff principle on cartographic symbols, more specifically on pictograms created by them.

Figure 3: Example for the use of the Chernoff principle on a cartographic symbol

Pictograms were selected because its components can be differentiated well, and depending on the theme and the design of the pictogram, it can be easier understandable for map readers than a geometric symbol. In other words:

8

J.J.R. Nuñez et al.

• if a cartographer can choice a pictogram closely related to the main theme, • and the features or components are clearly changed according the variables belonging to this theme. • then the map reading should be easier and faster. During the last two school years very interesting and original results were obtained with a high graphic level of realization. In the Figure 4 we can see fragments of different maps made by the students, working out themes as education, religion, ethnics, etc. In all the maps a general theme was represented using choroplets, constituting the background colour for the Chernoff faces and a compound pictogram was designed to represent other variables. Note that students used to “decorate” the faces with distinctive elements, in some cases using elements related to the topic, or to create a pictogram that “illustrate” the represented theme, trying to help and make easier the map reading.

Figure 4: Fragments of maps made by Hungarian MSc students on Cartography

Updating Research on Chernoff Faces for School Cartography

9

3- International project about the use of the method in schools Our proposals related to the Chernoff method were tested during a representative survey in some selected Hungarian schools. Between 2004 and 2005 was organized a project entitled „Map reading by children in school age: Cartographic education and practice in Hungary and Argentina”, developed under the scope of the bilateral agreement signed by both governments for the support of scientific research. The research on the possible uses of the Chernoff faces in thematic cartography with special attention to school cartography can be considered the continuation of the previously mentioned project, developed in two years (2008 and 2009).

3.1 Design of the questionnaire Three main factors were considered during the previous organizative works of the survey: • The specific characteristics of each educational system • The real possibilities of the participant colleagues to organize the survey (Argentine specialists did not have any kind of financial support to execute the survey in their country, because the bilateral agreement finances only the exchange of specialists) • The design of a questionnaire for pupils with some experience using maps and school atlases Our final decision was to execute the survey for pupils of grades 7 and 8 in Hungarian Elementary Schools, and for pupils of 1st grade in Argentine Secondary Schools. Because of financial limitations the test was printed in a black and white A5 format. The detailed presentation of the survey and the questionnaire was made during the 3rd International Conference on Cartography and GIS (Nessebar, Bulgaria) in June of 2010 (Reyes et al. 2010). After several consultations, the test was formed by four questions to examine four aspects of the use of Chernoff faces: • Use of “traditional” Chernoff faces: the original method created by Chernoff has two main characteristics: only the shape of a face can be changed to represent data, and all the faces should be kept unfilled or

10

J.J.R. Nuñez et al.

filled with the same colour. The decision was taken to determine how difficult can be the reading of the data if only the shape was changed while the size of the faces remained the same. • Use of Chernoff faces applying cartographic principles: the size of the faces was changed to represent a variable and in the Hungarian questionnaire was also changed the fill of the faces to represent a second one. These two parameters were not used by Chernoff in his original method, but they give us the opportunity to examine the grade of interrelation of the use of these parameters with the map reading (e.g. if the fill has or has no influence when the user reads other variables represented in the face). In the Hungarian questionnaire (Figure 5), the selected theme was the comparison of different agricultural products in the Hungary and other countries of the region (Austria, Czech Republic, Slovakia and Poland).

Figure 5: Question presenting a more cartographic version of Chernoff faces in the Hungarian questionnaire (Reyes 2010)

• Applying the Chernoff principle on pictograms: This was one of the more interesting questions of the questionnaire, because the combination of the traditional cartographic pictograms and the principle used by Chernoff to create his faces was tested together (Figure 6). In Argentina

Updating Research on Chernoff Faces for School Cartography

11

and Hungary were selected two different themes (range of parks and squares in some districts of Buenos Aires, and the production of citrus in the southern provinces of Spain), but both themes were presented using the same pictogram (a tree) and using similar elements to represent the different themes. In the Argentine questionnaire (left side of Figure 6), the leafage was used to represent the total area of parks, the trunk represented the total area of larger squares, and the number of fruits represented the total area of smaller squares. Pupils were asked to identify districts with larger green areas, fewer parks and larger area for squares. In the Hungarian one (right side of Figure 6), the total production of citrus was represented with the leafage, and the production of oranges, mandarins and limes with the number of fruits, the number of boughs and the trunk respectively. • - Drawing of thematic data on an outline map using Chernoff faces: Pupils represented the data stored in a table using the preconceived legend.

Figure 6: Question applying the Chernoff principle on pictograms in the Argentine and Hungarian questionnaire

3.2 Survey in both countries The questionnaire was applied between March and June of 2009. Argentine colleagues succeeded in collecting answers from 8 schools placed in the province of Buenos Aires. In Hungary a total of 12 schools participated from three provinces, arriving the majority of the answers from Budapest. In Argentina a total of 818 pupils participated in the survey and the age group widely represented were the 13 years old pupils with 543 participants (Juliarena et al. 2009). A total of 1038 pupils answered the questions

12

J.J.R. Nuñez et al.

of the test in Hungary, and the majority of participants was constituted by 14 years old (437) and 13 years old (350) pupils (Figure 7).

Figure 7: Distribution of participants by age

3.3 General results and analysis of some answers The general results of the survey are presented by questions and countries in table 1 and Figure 8.

Figure 8: Diagram comparing the percentage of correct and incorrect answers by questions in the survey

Together with the general results presented in the table 1, specialists also calculated the partial results by each question, which are presented by different diagrams on the website of the project (Reyes et al. 2009). Based on these diagrams we can have more genuine background information about the results of the survey.

Updating Research on Chernoff Faces for School Cartography

QUESTIO11AIRE Questions “Traditional” Chernoff faces Chernoff faces applying cartographic principles Applying the Chernoff principle on pictograms Drawing thematic data on an outline map with Chernoff faces

13

MAI1 RESULTS OF THE SURVEY ARGE1TI1A HU1GARY Answers Answers Right with one 1o Right with one 1o answers or more answer answers or more answer errors errors 493

313

12

828

207

3

285

527

6

665

367

6

294

520

4

908

123

7

540

257

21

798

211

29

Table 1: General results of the survey

After consultations with the Argentine specialists, we should remark that the Argentine results reflect the need of the use of School Atlases during the teaching of Geography for pupils in Elementary and Secondary Schools, because this constitutes one of the reasons of the lower results of reading data represented in maps. At present, there is a lack of School Atlases in the Argentine Educational System, and teachers and pupils use atlases not designed especially for the national curriculum on Geography. At the same time, the Geography textbooks used in Argentina contain more maps than the similar textbooks in other countries, trying to fill the absence of School Atlases. But as we could experience during our first research on reading thematic maps (Reyes et al. 2005), the textbooks cannot substitute the role to be played by School Atlases during teaching Geography. The more contradictory result was obtained when the pupils gave answer to the question applying the Chernoff principle on a pictogram. As we can see in the table 1, the Argentine result was the second worse general result of the survey (correct answers were only a 36% of the total), while it was the best general result in the Hungarian one (87% of pupils answered correctly this question). In this case, the general result reflects a more negative situation than the analysis of partial results. After the analysis of the

14

J.J.R. Nuñez et al.

partial Argentine results that can be found on the Web (Reyes et al. 2009), it can be seen that the percentage of pupils with only one error in their answers (34%, 281 pupils) is very near to the 36% of pupils who answered correctly the whole question, and both categories together constitute nearly 70% of the participant pupils. Only a 7% of Argentine participants (52 pupils) did not give any correct answer to this question. Other unforeseen result can be appreciated in the question of the Hungarian survey designing Chernoff faces applying also cartographic principles. In this case the size of the faces was changed to represent a data set (production of apples), and the fill was also changed depending on the production of wheat (Figure 5). First we focused to compare the size between faces: which country has a larger production of apples: Czech Republic or Austria?, and the correct answer is both countries, because the size of the faces is the same. At same time the face representing Czech Republic was also filled with a darker grey tone (to represent a larger production of wheat) and it was also smiling (to symbolize a larger volume of agricultural production too). Our interest was to check if these attributes could have or not influence in the pupils’ decision when they had to compare only the size of the faces. A total of 367 answers were wrong, and in 343 of them only one country was indicated (Czech Republic or Austria). In Figure 9 can be seen that only 112 pupils selected the Czech Republic over Austria, so we can affirm that neither the fill nor the smile represented an obstacle during their analysis to answer this question.

Updating Research on Chernoff Faces for School Cartography

15

Figure 9: Diagram presenting the percentage of answers to the question about the use Chernoff faces applying cartographic principles in Hungary

One of the more interesting results obtained in the Hungarian survey were the opinions given by 507 pupils (49% of the participants) about the method. Our first step for its analysis was to categorize the answers as positive or negative, and the result is presented in Figure 10.

Figure 10: Proportion of positive and negative opinions about the survey

16

J.J.R. Nuñez et al.

Reading the answers, we can conclude that a considerable number of the negative opinions about the questionnaires came from pupils who resolved correctly all the questions. The relation of positive and negative answers by schools was also examined and the results are represented in the diagram of the Figure 11 (using only a number for the identification of the schools to keep their anonymity).

Figure 11: Proportion of positive and negative opinions by schools

Four schools had results that emerged from the average: the larger number of negative opinions arrived from two schools (number 1 and 7) with a recognized high level of teaching in the country (but as we can see in the diagram, in the school 1 the number of positive opinions was higher than the negatives ones), while school number 5 can be considered of average level and school number 3 of below average. Mainly in the two first cases the common characteristics of their answers are the use of a polished vocabulary and the writing of the longer and more detailed arguments to express their opinions, sometimes combining negative and positive elements. Some examples: • Not a good idea: it is understandable and logical, but I do not see how help our thinking (14 years old girl) • You need to pay higher attention, but more data can be plotted in less space. But I would not use it every day (14 years old girl)

Updating Research on Chernoff Faces for School Cartography

17

• I like better the traditional symbols (14 years old boy) • You need a longer time to read the legend, but it can help to endear the subject in earlier grades (14 years old boy)

Figure 12: Diagram showing the frequency of the numbered doublets in the pupils’ opinions: (1) interesting – bored, (2) good – wrong, (3) like – do not like, (4) easy – hard, (5) funny – infantile (childlike), (6) understandable, suggestive, unequivocal – incomprehensible, inexplicable

We can affirm that a notable percentage of negative opinions are from pupils better prepared than the majority. This conclusion is also reaffirmed in the Figure 12, which represents some doublets with more frequent occurrence in the opinions. A total of six doublets were selected, expressing contradictory opinions like “interesting – bored”, “easy – hard”, etc. The summarized frequency of these words was compared as seen in the diagram. The doublet number 6 is the only one case, when the number of negative adjectives (incomprehensible, inexplicable) exceed the number of the positive ones (understandable, suggestive, unequivocal) and the selected vocabulary together with the expressed points of view let us infer that the 14 years pupils with a higher level of knowledge and using more often maps and school atlases in the classroom prefer to follow the use of the traditional methods of thematic representation (e.g. choroplet, diagram) for the visualization of data on maps. At same time we cannot omit that they constitute a minority within the participant pupils: only a 34% of the total of opinions and barely a 17% of the total of Hungarian participants. Between the positive opinions (77% of the

18

J.J.R. Nuñez et al.

total) we can find some that directly or indirectly confirm the objectives set by us during the organization of the research and the survey: • I think our age group is accustomed to the traditional symbols, but this is a good idea for the smaller children (13 years old girl) • I like this kind of symbolization, because it is more interesting than „coloring” and we can learn more of it (13 years old boy) • Interesting, how many data can be represented with a face (13 years old boy) • Interesting because many information is drawn in only one image (15 years old girl) • I really liked the nature of these exercises, I would do it more times e.g. if more complex „shapes” are drawn to substitute the faces… (14 years old boy) • Very good exercises, I like that the „image” is divided into several parts. Easy to understand, they could be used in more tests! (15 years old boy) Colleagues interested in this theme can find free access to all the databases, documents, etc related to this project visiting the following website: http://lazarus.elte.hu/hun/dolgozo/jesus/ma0809/proyect2.htm. All the documents are in two languages of the participant countries (Spanish and Hungarian),

4- Present and future plans Some contradictory experiences can be noted when the Argentine and Hungarian results are compared (applying the Chernoff principle on pictograms, reading of data represented by changing only the shape of a face did not provoke more significant difficulties than if the representation is made changing the size and the fill). This survey was made for older (1314 years old) pupils, but one of the conclusions of the previous theoretical research was that use of the faces can be more successfully for children in early grades of Elementary Schools. This idea was reaffirmed by some of the opinions given by the Hungarian pupils as can be read in the previous chapter.

Updating Research on Chernoff Faces for School Cartography

19

Figure 13: Website presenting the new project (fragment)

At same time that the periodical contacts with Argentine colleagues have been kept, expecting to begin new research in this and other themes, the Hungarian team also decided to follow the research in a new international project with the participation of the Institute of Cartography and Geoinformatics of the Vienna University of Technology. According to our common decision, in 2010 both institutions began a project entitled “Further research and survey related to the theoretical and practical results of previous international projects about the possible cartographic uses of the Chernoff faces”, financed by the bilateral agreement for research between both countries (Figure 13). Our main aim is to find answers to the themes with contradictory results in the Argentine-Hungarian survey and to develop new research to complete these results. Based on all the results, the specialists involved on this project plan to work out specific outlines about the possibilities of the use of the Chernoff method in school cartography, which can be useful for their future use in school atlases and other materials related to the geographical education.

Acknowledgments The present research was developed in the framework of activities of the MTA-ELTE Research Group on Cartography and GIS. The theoretical research is supported by the European Union and co-financed by the European Social Fund (grant agreement no. TAMOP 4.2.1./B09/1/KMR-2010-0003). The international surveys were supported by the project ARG-3/2007 of

20

J.J.R. Nuñez et al.

the National Office for Research and Technology of Hungary, and the project 68302 of the Hungarian Scientific Research Fund (OTKA).

References Bertin J (1983) Semiology of graphics. University of Wisconsin Press Chernoff H (1973) The use of faces to represent points in k-dimensional space graphically. Journal of the American Statistical Association, 68:361–367. Dorling D (1991) The Visualization of Spatial Structure. PhD dissertation. Department of Geography, University of Newcastle upon Tyne. http://www.sasi.group.shef.ac.uk/thesis/chapter8.html. Accessed at 9 January 2011 Fabrikant SI (2004) Blue and Red America http://www.geog.ucsb.edu/~sara/html/mapping/election/election04/election.html Accessed at 9 June 2010 Juliarena CE, Garra AM, Rey CA et al. (2009) Posible uso de las fases de Chernoff para la visualización de datos en la cartografía escolar. Boletin CAC 53(45) 2009/1:42–51 Nelson ES (2000) The Impact of Bivariate Symbol Design on Task Perfomance in a Map Setting. Cartographica, 37(4):61–78 Nelson ES (2007) The Face Symbol: Research Issues and Cartographic Potential. Cartographica 42(1):53–64 Nelson ES et al. (1997) Visual Search Processes and the Multivariate Point Symbol. Cartographica 34(4):19–33 Reyes JJ (2009) Ideas for the use of Chernoff faces in school cartography. The World's geospatial solutions. CD Proceedings of ICA 24th ICC. Santiago de Chile Reyes JJ, Juliarena CE, Garra AM et al. (2005) Reading thematic maps in Argentine and Hungarian schools. Mapping Approaches into a Changing World. CD Proceedings of ICA 22nd ICC. A Coruna, Spain Reyes JJ, Juliarena CE, Garra AM et al. (2009) Posibles usos de las fases de Chernoff para la visualización de datos en la cartografía escolar (2do año), http://lazarus.elte.hu/hun/dolgozo/ jesus/ma0809/2/ekutatas.htm. Accessed at 30 December 2010 Reyes JJ, Juliarena CE, Garra AM et al. (2010) Chernoff survey in Argentine and Hungarian schools. CD Proceedings of the Third International Conference on Cartography and GIS. Nessebar, Bulgaria Turner E (2004) Gene’s Map Gallery http://www.csun.edu/~hfgeg005/eturner/gallery/gallery.htm Accessed at 9 January 2011

The Production of Maps for Students in the Context of School Culture Rosangela Doin de Almeida Universidade Estadual Paulista Julio de Mesquita Filho – UNESP-CRC Programa de pós-graduação em Geografia Av. 24 A, 1515 – CEP 13506 900 Rio Claro – SP – Brazil [email protected]; [email protected]

Abstract The objective of this article is to discuss a methodology for the production of maps developed for children at junior school (7 to 10 years), from the research we have been undertaking since 1997 at the Laboratory of Research for the Teaching of Geography and Cartography for the production of maps and local atlases. Firstly, we are going to present the foundation and the process involved to produce these materials. Following that, we will discuss some findings that resulted from these experiments in terms of the adequacy of these cartographical materials for school use and finally, we will present our conclusions about the production of the key for pictorial maps designed for schools.

1- Background and objectives In Brazil the first research projects regarding maps for schools began in the 90's, when the school curriculum defined the study of place (school area and surroundings) as the main theme for junior students (students from 7 to 10 years of age). Even though, the study of the local neighborhood was already present in the previous school curriculum, other factors created a demand for more in-depth studies of this theme. One of these was the sharp increase in migration to medium-sized and large cities, leading to a large contingent of migrants whose children began attending schools in places foreign to the culture of their origin. Besides this, the knowledge of A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_2, © Springer-Verlag Berlin Heidelberg 2011

21

22

R.D. de Almeida

place became more important as a consequence of globalization, due to the need to create a feeling of belonging in the new living spaces. These two key points are related to other crucial education issues, particularly those related to the function of their cultural transmission. We cannot deny the strong relationship between education and culture. According to (Forquin 1993: 12), culture is an asset of knowledge and competencies, of institutions, of values and of symbols, established throughout the generations and characteristic of a particular human community, defined in a more or less broad manner and a more or less exclusive manner. We do not want to say that education transmits the culture or a particular kind of culture, rather a part of a culture, whose selection is included in the school curriculum documents, which according to Yves (Chevallard 1991) relates to the knowledge to be taught. This knowledge concerns something which already existed and is considered as something that should be preserved by means of cultural transmission that takes place at schools. However, the content that is taught and school practices in general deviate from following the prescribed curriculum, and what actually happens is that it results in a conflict of relative factors, for example, the selective memory of the teachers (both to recall and forget certain content, practices and values). (Perrenoud 1993), upon taking into consideration the modus operandi of teachers, favors the idea of bricolage, not with relation to the materials produced, but with relation to the way knowledge is produced and practical tasks are created: making use of the means available, re-using materials, modifying situations, i.e., making use of elements that have originated from different systems. The task attributed to the school of transmitting something of culture not only lies in selecting the knowledge and transmittable cultural materials, but making them “teachable”, that is, turning them into something teachable, memorized via practical map activities and subject to “evaluation in a class, of one year level, to their timetable, to a system of communication and work” (Perrenoud 1993:25), constituting a specific “school culture”. This culture is marked by rituals, routines making the present content in the manuals ineffective. Nowadays, this method is frequently used in the production of maps and cartographic material in schools and consists of a rich variety of material for reflection with respect to the teaching of maps. The social transformations that we are referring to above, lead to disagreements in terms of the school's function of transmitting culture, since the school brings people together of diverse backgrounds, with their own customs in terms of their manner of speaking, eating habits, day-to-day customs, religious beliefs, ethic values and aesthetics. The cultural diver-

The Production of Maps for Students in the Context of School Culture

23

sity of the new subjects involved in the educational process (teachers and students) challenges the methods of schools to deal with the function of cultural transmission. Moreover, the official curriculum guidelines recommend a large amount of flexibility in selecting content and the method of teaching it. A further challenge is to reconcile the demands of the different groups that come in contact with each other at school, which form the current “school culture”. We have received requests from teachers and education authorities to prepare cartographic materials that are more dynamic and better adapted to this new context. This caused us to look for more adequate ways to study local area and represent it cartographically. The most traditional way to produce materials and assist teachers with the teaching of local area quickly became inadequate, because it began with the idea that knowledge which had already been established about the city, its origin, and its geography should be transmitted by means of text and maps. The latter used to be produced according to conventional cartographic knowledge. We realized that it did not work this way with new students (and new teachers), whose needs were more complex, different from those that we assumed to be important at this level of education. In one survey (Viero 2002) carried out for the production of a school atlas for the municipality of Santa Maria (located in Rio Grande do Sul, 270,000 inhabitants), we discovered that the teachers possess an elementary knowledge of the maps. They weren't aware of what would be desirable on maps adequate for students at junior year levels, because the maps displayed a lot of information resulting in confusing, imprecise atlases with excessively detailed maps. We concluded, therefore, that the teachers should be educated as to how the maps and atlases are produced. We also observed that this behavior is supported by erroneous suppositions about the function of teaching today. One of the consequences of globalization was the proliferation of information spread via different means of communication (the media), which added to other social and cultural factors, produces a sense of “loss of identity” at the school and with everything that happens in it. According to some theorists, one effect of globalization is the weakening of previous forms of identification with a national culture and the strengthening of other cultural ties. Despite certain aspects of national identities continuing strongly (such as legal rights), identity with local place has gained greater importance.

24

R.D. de Almeida

Stuart Hall adds that one of the characteristics of globalization is the “space-time compression”, perceived as the reduction of distances and as the impact of the events that occur in one place on other places and people. He believes that time and space are the co-ordinates of all the systems of representation, which are responsible for the production of identities (2005: 71). One of the current functions of the school is, therefore, to enable the formation of identities. The representations created in the school context permit this, among them the maps are responsible for creating spacial identities. However, place and space is no longer the same thing, we have put down roots in places, but space can be quickly traversed by modern means of transport and communication. These considerations take us to the issue of identity and of school culture as the guiding references when taking into account “school cartography”, thinking of it as the whole practice and complete knowledge related to the spacial representations in the school context.

2- Approach and methods School cartography forms part of the didactics of geography. Didactics deals with the relationship between student/knowledge/teacher, which can be represented by a triangle, on which knowledge is at the top. We created a pyramid (figure 1) to demonstrate the relationship between these three items and “school culture” in such a way that: • Knowledge is seen as the social construction that occurs when there is interaction between the subjects; • Teaching has an open character, subject to influences and factors from outside the school environment, which, in part, are introduced by these agents – the teachers and the students; • Learning is a continual transformation of the student's knowledge and a constant search for new teaching practices on the part of the teachers. • The school is considered to be the setting where the students relive their personal experiences, which allows them to reconstruct and co-construct knowledge. Language is the instrument, which allows this construction

The Production of Maps for Students in the Context of School Culture

25

and reconstruction of knowledge, being that the construction of sociospacial knowledge is influenced by cartographic language.

Figure 1: The didactic pyramid

In view of what is presented above, we based our research on the school culture, taking into consideration school knowledge and practices such as social construction. Therefore, we don't see it as knowledge originating from the prescribed curriculum, rather as knowledge replete with meaning and values provided by the groups that we find at educational institutions. We carried out surveys about the production of maps and school atlases for different municipalities in the state of São Paulo. The first of these was conducted between 1998 and 2000 for the production of an atlas for three municipalities (Almeida 2003). Between 2002 and 2004 we undertook some research to discover how teachers use these school atlases. A third experiment was carried out between 2006 and 2008 for the preparation of an atlas for the municipality of Sumaré (SP), which forms part of metropolitan region of Campinas (SP). From these experiments we tried to draw some conclusions which would increase our knowledge of School Cartography. One aspect which stands out is the production of the key used on the maps in these atlases. Before going on to discuss this point, we believe it is necessary to outline the approach and methods that we used in this research.

26

R.D. de Almeida

We outline a collaborative research program that involved our work group and teachers from the public school systems. The issue in question was not “how to teach with an atlas”, but how different subjects relate to knowledge of the place within a school context. This knowledge is present in the daily routine and memory of each person, allowing for incursions into time and space underpinned by each subject’s experiences, which is quite different from elaborating knowledge about something one does not know based on experience. So the organization of the research had to be very permeable to the classroom routine, ensuring the presence of the teachers’ practices and the manifestation of their conceptions, questionings and dilemmas. Thus, the methodological design of the research was abundant in records and included sufficient time for discussion and reflection, allowing for the interlocution of subjects from different social places: university professors and students, school teachers and students, administrators of the Education Bureau, and even people from the community. The overall purpose was to create an atlas that promotes identity and a feeling of belonging through greater knowledge of the region. The projects’ objectives were: • To define themes for the Atlas starting with the issues presented by the schools; • To conduct research for a sociocultural diagnosis of the school population; • To define a thematic axis based on the results of the research and on the Curricular Parameters and adapted to school practices. The methodology begins by considering that local school atlases are cartographic products that are very different in nature to other types of atlases, because they should be appropriate to the school curriculum. This kind of atlas should be organized according to a sequence of concepts and abilities that will be acquired by the students over several years; moreover, they contain topics relating to the daily experiences of the users of the locale. The similarity between the information contained in the atlas and the knowledge of the local users concerning the locale makes the atlas more accessible, enabling them to easily identify errors it may contain. This also makes it easier for the users to make criticisms about the atlas. Therefore, the production of local school atlases requires extensive research to avoid such problems.

The Production of Maps for Students in the Context of School Culture

27

Thus, school cartography research within the context of teacher education and the production of curricula can be conducted in two ways: the first, by adapting theoretical and technical knowledge about cartography to primary and secondary school contexts, thereby establishing a clear difference between university and school knowledge. The second way is examining teaching practices to assess the extent to which they expose learners to new knowledge, rather than simply reproducing knowledge originating from a university curriculum. This implies the need to consider different social backgrounds as a reference when carrying out atlas activities in schools. The social background plays an important role adapted to students’ needs. Therefore, school learning is a social construction, and not an isolated process. The school’s role is to impart certain values and knowledge to students. Collaborative action-research between the university and the local department of education was the methodology we applied here (Clandinin and Connely 2000). This concept underpinned the three aforementioned projects, whose results were used to propose a methodology for the production of local atlases for children. To this end, we came up with the fundamental points of our methodology, which are as follows: 1. Conduct a thorough research of the students’ background and the culture and values associated with their community in order to identify the elements that make up the local culture. 2. Choose a thematic axis around which the contents of the atlas should be organized. The central axis discusses identity and belonging. 3. Develop atlases that are appropriate for each teaching level.

3- Results Among the diverse range of results that emerged from this body of research we are going to discuss some of the findings with respect to the production of the key. We chose to talk about the key because it is through this that children represent and recognize spacial representations. It is important to mention that the activities involving the keys were preceded by activities where the students looked at model maps and practiced drawing the home-school route, which enabled them to work with different elements of a map before actually using the maps.

28

R.D. de Almeida

The atlases mainly consisted of large-scale maps (1: 10 000 or 1: 8 000), where the urban fabric is visible enough to locate the geographical points that identify the route represented, giving them (the maps) an identity. Considering that it is an issue of cultural background, we sought out elements that would allow us to identify these landmarks from the student's perspective (and not just from the teacher's), according to the vision of “kinder culture” (Steinberg and Kincheloe 2001). We are able to cite examples. Children experience a city according to their interests and the perceptive ability they have, to frequent and enjoy this space, therefore, children that travel by car to school capture images of the city at a more rapid pace and in less detail than those that cover this trajectory on foot or by bicycle. A square takes on a certain significance from the relationships that develop among the people that frequent it and if this is peaceful, the meaning of “square” could then be “a place of leisure”, however if there are frequent conflicts among the residents and violent elements, the “square” will take on the meaning of “a dangerous place”. As we can see further on, children create symbols for the key where the predominant meaning (or value) they attribute to these different places, in the first example, could be representing a square as a scene where there are toys, flowers and children, or as in the second example, a place where there are armed bandits. There are many studies regarding the methods children use to learn with maps (Wiegand 2006), which we aren't going to comment about here, however we will make it clear that we adhered to the socio-cultural approach of Lev S. Vygotsky and his co-authors. According to him, cognitive development and human learning are fundamentally social, as knowledge is shared and the senses are established culturally, such that interaction is a way to produce knowledge and transform knowledge. Collaboration and cooperation are also fundamental for learning to take place. Hence, in the experiments that we carried out, the teachers were instructed to organize the students into groups in order to be able to do the activities and conduct a group discussion at the time the work was being carried out. In one of these classroom situations the proposed activity was to make a map of the neighborhood and then present it at a municipal meeting of the Council Budget Committee. The teacher's objective was to produce a pictorial map with the students, which represents the demands of the residents neighboring the school. The students carried out an interview with their parents in order to find out what improvements they would recommend for their neighborhood. Each student drew a key and located the points on the map. Following this, they formed groups so that they could compare their keys and agree on a single key. For this to happen the students were

The Production of Maps for Students in the Context of School Culture

29

required to choose the main improvements recommended and then each group presented their results to the class. The teacher observed that many maps mentioned the same improvements, however with different symbols, which some of the students also noticed. The next step came about automatically – create a single key so that the map could be presented at the meeting. This took a certain amount of time and an ongoing discussion regarding the main improvements for the neighborhood and, mainly, as to how they should be “drawn”. The “symbols” created by the students presented a striking resemblance to the real situation. The majority of them consisted of a setting and not of an object. This caused a certain amount of difficulty to reproduce the key and did not comply with that which we wished to introduce (the idea of a symbol or icon, whose characteristic would evoke a universal idea). For example, one of the improvements was to have all the doctors at a health clinic, which was represented by a picture of a doctor attending a patient on a stretcher (figure 2).

Figure 2: Key for “the need of more doctors”

Presenting a map with a summary of the resident's demands, established another social context for this activity – the negotiations would be done in the community outside the school environment, which required, therefore, a type of cartographic production where the group of students took their ideas to a group of residents of that neighborhood. The intention was to hold a public exhibition of the students' ideas regarding the most important improvements. Thus, the creation of the key required that they establish generalizations, looking for a single meaning, but at the same time retaining the identity of the group that produced it, which was decided on by mutual consensus between the students. To exemplify this better, we have included a passage, which was written by the teacher when they were discussing the symbols for the map:

30

R.D. de Almeida

(...) The students still hadn't realized that the symbols for the same demand were different, so I asked them: -What symbol did you come up with for the school? -small house -backpack -books... -Is there any way of putting it differently? -No, we have to talk about it and choose one -That's right, we have to decide. Is there any symbol on the atlas? -A pencil -but there are different colors, why? (teacher) -Because it is a public, municipal and private school (student) -That's it. The solution in the end was to draw all the symbols for “school” on the board and decide which one was best, by putting it to a vote. This solution was applied to all of the symbols chosen for the map key. The choice of symbol for “security” generated a lot of discussion about the risk of keeping a weapon at home, as one of the students had suggested using a picture of a weapon as the symbol. One student suggested using the symbol which is in the Atlas - a police siren, which was accepted by the group, which meant that the conventional use of the key won out for the first time. Figure 3 shows the symbols chosen for the collective key. Some retained the pictorial character of the first map, others became something a little more iconic, stylistic, generalized (for example, the symbols for “more doctors” for “security”). The personal touch, which was found in the keys of the first maps ceded to symbols that were more universal.

The Production of Maps for Students in the Context of School Culture

31

[Key translation: Leisure center, Day CareCenter, Public School, Street Lighting, Cleaning the Streets, Cleaning empty blocks of land, Doctors, Paving, First Aid, Road Resurfacing, Security] Figure 3: Map key “Demands for the Council Budget Committee”

32

R.D. de Almeida

4- Conclusion and future plans For school cartography the concept of the background map (or basic map) is introduced via cartographic initiation, when the students come in contact with questions related to location, to their point of view and proportion. Now the thematic maps contain a summary or generalization of knowledge that circulated regarding a theme or a problem. Therefore, a certain level of standardization is required for the symbols used in the key, as well as a certain categorization of knowledge, which, in turn, gives rise to a negotiating process in terms of accepting what should remain on the map as something legitimate. Now, the legitimization of a given piece of knowledge, in the context of school situations, results from the relationships established between the subjects (teacher and students) and from the tension between the knowledge to be taught (prescribed by the curriculum) and the knowledge brought by these same subjects. We believe that this argument should be taken into account when choosing the thematic maps presented in the school books and atlases destined for the use of the students. The issue of the key, therefore, is strongly linked to the establishment of meaning, so that the summaries can be presented on the maps. Thus, there is the need to “negotiate” which symbolic forms it will contain. In the early years of junior high school, this plays down the importance of the search for a neutral or correct solution, according to which there can only be a single relationship between the symbols and their meaning. The choice of the symbols draws attention to the meanings, which are attributed to them culturally and socially. A sign takes on different meanings according to the cultural context of its readers, and also possesses a political character, because it transmits inherent conceptions to the symbols themselves. In the example given, using a weapon as a symbol for “security” brings about a strong sense of violence, whereas the use of a police siren can denote “help” or “protection”. The meaning of these educational activities became more explicit upon becoming discoveries made by the students, from knowledge arising from their daily experiences in the place that they live. Thus, the geographical concepts and their representation emerged in a contextualized mode, and the production of knowledge became more significant.

The Production of Maps for Students in the Context of School Culture

33

References Almeida R, Doin de (2003) The Development of School Atlases based on Action Research with Elementary Schools. International Research in Geographical and Environmental Education. V. 12, n. 4, p. 364-369. Chevallard Y (1991) La transposition didactique: Du savoir savant au savoir enseigné. Paris : La Pensée Sauvage. 240pp. Clandinin DJ and Connely EM (2000) Narrative Inquiry. Experience and story in qualitative research. San Francisco: Jossey-Bass Publishers. Forquin JC (1993) Escola e Cultura. As bases sociais e epistemológicas do conhecimento escolar. Tradução de Guacira Lopes Louro – Porto Alegre: Artes Médicas Sul. 208pp. Hall S (2005) A identidade cultural na pós-modernidade. Tradução de Tomaz Tadeu da Silva, Guacira Lopes Louro – 10.ed. Rio de Janeiro: DP&A. 102pp. Perrenoud P (1993) Práticas pedagógicas, profissão docente e formação: perspectivas sociológicas. 2ª ed. Lisboa, Dom Quixote - Instituto de Inovação Educacional. 207pp. Steinberg SR and Kincheloe JL (2001) Introdução. In : Steinberg SR and Kincheloe JL (orgs.). Cultura infantil: a construção corporativa da infância. Rio de Janeiro: Civilização Brasileira. 415pp. Viero LMD (2002) A elaboração de um atlas escolar municipal como uma contribuição para o ensino de geografia – Santa Maria (RS). Dissertação de mestrado. Universidade Estadual Paulista. Instituto de Geociências e Ciências Exatas. Rio Claro. 85pp. Vygotsky LS (1988) A formação social da mente. Trad. José Cipolla Neto, Luis Silveira Menna Barreto e Solange Castro Afeche. São Paulo, Martins Fontes. 168pp. (Coleção Psicologia e Pedagogia - Nova Série). Wiegand P (2006) Learning and teaching with maps. Londres: Routledge. 153pp.

Geoinformation: a social Issue Angelica Carvalho Di Maio1, Cilene Gomes2, Maria de Lourdes Neves de Oliveira Kurkdjian3 1

Universidade Federal Fluminense – Instituto de Geociências, Niterói, RJ [email protected] 2

Universidade Federal do Rio de Janeiro – IPUR, Rio de Janeiro, RJ [email protected] 3

Instituto Nacional de Pesquisas Espaciais – DSR, São José dos Campos, SP [email protected]

Abstract The article proposes to raise a theoretical discussion and to point out some meaningful applications on the relations among geoinformation, citizenship and social participation concerning educative processes and regional and urban planning. Considering that the socialization of information increases the knowledge of the world, which also increases the possibility to interact and transform it, some projects that address the issue of geoinformation as an input for social action, some already completed and others under development, are subject of a brief presentation, as well as future projects that may also contribute to a social cartography.

1- Background and objectives The theme of “geoinformation as a social issue” has been a field research due to academic knowledge on Cartography, Remote Sensing and Geographic Information System (GIS) as well as Geography, Architecture and Urbanism, and more specifically in Regional and Urban Planning. The analysis carried out individually or in cooperation, on the application of geotechnology for the organization and generation of new information has been established as an argument in favor of knowledge of the place where A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_3, © Springer-Verlag Berlin Heidelberg 2011

35

36

A.C. Di Maio et al.

one lives and, thus, supports the development of activities in various fields of social life. In this sense, the organization of databases and the use of geoprocessing (techniques of digital Cartography, image processing, spatial analysis, WEBGIS and so on) have served as instrumental resources in applications, basically directed to the processes of knowledge production on the human space, education in these themes, and regional and urban planning. These works have been accomplished in different Brazilian universities, and to a large extent, have been at the root of experiences and learnings obtained from the National Institute for Space Research (INPE), which was one of the pioneering institutions in the development of geotechnology and geoinformation in Brazil. In the last decades, along with other institutions in the country, INPE has been dedicated to environment studies and monitoring, to education and regional and urban planning through the use of geotechnologies, among others. Considering the above background and the opportunity for a dialogue that begins with a joint effort, the general aim of this work was to create a theoretical discussion about maps, GIS and society in order to highlight relations between geoinformation and social issue that here implies the educational process of citizenship reconstruction - a process that is also associated with the essential preparation for the political participation of different actors in the path of society development and the remodeling of their common living space. In reference to the Brazilian social and spatial reality individualized in the study of some regions and cities, and taking for granted, in the same context, the enhanced diffusion of the access and use of geoinformation, which took place in the last two decades, mainly through the network resources, free and open source software, and multimedia systems, another goal of this paper was to point out some recur-ring applications and/or of special meaning for a critical and exemplified approach of the general relations among technology and education, socialization of information, citizenship and social participation. Furthermore, some developing projects were also objects of a brief presentation and reassessment aiming at a future methodological guidance.

Geoinformation: a social Issue

37

2- Approach & methods From the theoretical point of view, the focus of the geoinformation issue arises from the basic understanding that space is information and is likely to be represented by maps and geographic information systems, which are greatly useful for spatial analysis and the base for education and regional and urban planning actions. We live in a time when the advance of scientific and technological knowledge in the information and communication fields has become remarkable and of extreme social and economic value. In this con-text, the geographic space differentiation can also be observed by the scientific, technological and informational density that various regions and cities in a country have incorporated in their territories, because of the widespread dissemination of innovations in various fields of economic and social activities. Modern Cartography and geoprocessing technologies as well as those of organization, storage and delivery of information in net-work systems are at the core of this diffusion process which, in Brazil, has increased, especially in the 1990s and 2000s. But in this process a major difficulty still stands: large portions of Brazilian society have no access, training or condition for a targeted and critical use of these technological resources especially of the huge amounts of information that can be arranged, generated or transmitted. Thus, the issue of digital inclusion does not dissociate from the possibility of a conscious appropriation of technology. Hence we understand that we are still giving the first steps on the path of the establishment of a society of information and know ledge, not only in the sense of a universal access, but mainly in the perspective of an awareness of the world and/or of the place where one lives more and more favorable to the production and appropriation of a local knowledge about the living place, to the social inter-actions that can lead to increasing levels of social organization and political participation, and at last, to the exercise of citizenship rights. Concerning the socialization of information, it is important to pay attention to the fact that modern man lives in an information society which refuses him the right to be informed. This leads to a large segment of society to live in ignorance of the world they live and of the decisions that affect them. In general, these decisions are made on the basis of the information

38

A.C. Di Maio et al.

we lack, which does not contribute to the formation of a full citizenship (Santos 1987). In fact, a way to measure the extent of societies’ development would be in its ability to produce, integrate and create dynamic information circuits on society and on the place people live, which can help to raise the level of sensitivity and involvement of citizens in public or collective issues. In this case, geography has an important role in the context of the current and future societies; new geography was tasked with the mission of conceiving the GeoCitizen (Julião 1999). That is, the apprehension and the production of geographical space knowledge is a matter of education and citizenship. It is therefore an essential tool to awakening responsibilities towards our way of living. Thus, it is possible to perform an educational task, aimed at clarifying the individuals about their citizenship, especially when they feel owner and understand their world, country, city and place of living (Damiani 1999). Hence, the perspective of GIS introduction in education, in the teaching of spatial science and, particularly, of Geography, matches the goal of reinforcing the school-environment relationship, since it is known that the access to education is a fundamental condition for participation in contemporary world (Tristão 2008). The socialization of information increases the knowledge of the world, which increases the possibility of interacting and transforming it. From the earliest stages of schooling, Geography education can and should aim at showing the students that citizenship is also the feeling of belonging to a reality in which society and nature interrelations form an integrated whole - constantly changing – of which he is part and therefore needs to know and feel like a participating member, emotionally connected, responsible and historically committed (Ministry of Education -MEC 1999). Adapted to the educational Geography curricula, GIS tends to facilitate a dynamic interaction through the exploitation of geographic in-formation, which comes from various sources and is organized in an integrated manner and in different scalar cuttings. Spatial perception and cartographic language are resources of great interest to support Geography education, as they establish a more direct relationship with the reality in which one lives. Thus, they are crucial in both the evolution of the cognitive structures and young people’s intellectual growth. The representation of spatial phenomena helps the formation of ideas and of a base for the comparison of ideas

Geoinformation: a social Issue

39

and the interactive practice enhances actions that result in knowledge (Silva 2008). Hence, the development of Cartography applications and a new generation of GIS for electronic media and Internet-oriented architectures, mainly in terms of interactivity, are also pointed as the most democratic way of disseminating information and of creating motivational education environments. These, despite of constituting a bridge between what is taught, apprehended and learnt and the real world, do not replace the direct learning relationship with the concreteness of the real world. In the context of social interactions mediated by new cartographic and GIS technologies, it must be considered that the attributes of cartographic products determine how geographical space is perceived and mentally understood. Therefore, there is a need for an understandable and easier transfer of concepts to different potential users, in order to encourage the participation (by integrating the users’ previously acquired knowledge), the decision making, the formulation of studies and plans, and so on. This meets the Social Cartography. In this crucial moment to the future, and of understanding the relationships among mappings, geographic information systems and society, it must be mentioned, in fact, the new research line named Social Cartography (Acselrad 2008; 2010). It opposes, among other issues, the spatial concept and its representation - including geo-graphic information and cartographic technological questions - but also the citizens’ political participation in the social organization and in the local production of knowledge based on experiences or daily involvement in different territories. Therefore, the Social Cartography, including GIS, would serve as a support to the social interaction processes and participative-action of most distinguished social agents on their way to a gradual reversal of the social alienation or lack of information processes, particularly to processes of political inequality and social and spatial segregation. From the methodological point of view, the study of significant applications is firstly and very briefly based on the trajectory of INPE, from the 1970s up to nowadays. This study should also provide additional references raised by the authoress’ experiences in the framework of INPE’s own cooperation projects or in the scope of researches made in different Brazilian universities including projects currently being developed.

40

A.C. Di Maio et al.

In the general framework of this brief account aiming at significant applications, it is important to highlight some technological innovations for the distribution of data, products and presentations. The use of free technologies and open source software represents, for example, a great advantage in developing a WEBGIS application, since they enable the reduction of the final application cost, greater spatial information dissemination and interactivity. The application of multimedia techniques in mapping provides tools for cartographic visualization through multiple types of information (images, texts, etc.). In this universe, another significant innovation is the animation, or virtual reality, as it also offers new tools and possibilities for the presentation of cartographic products. All these improvements interfere with the process of information transmission and acquisition, encouraging concepts dissemination, supporting the exploration of new ideas and the integration of new knowledge (community organizations, for example) and thus, contributing in the process of social interaction, positioning and decision for action. In the design of GIS tailored to geographical and environmental education, the objectives of Geography curricula should be considered. Here the primary concern would be to encompass the relation-ship between natural and social processes, the different existence modes and production processes and the spatial organization by means of landscape analysis and social life related phenomena (MEC 1999). In this sense, the choice of basic functions for a GIS to be used in basic education should be guided by the possibilities of setting up activities and analysis of interest which also leads to the understanding of the best way of using GIS in the classroom. Maps and GIS can be used by ordinary people to improve their lives. This conviction is better stated when we say that the decisions made by someone better informed and better-trained in the use of GIS will be, in a long term, better than the others’ (Goodchild 1998). Indeed the world understanding and human consciousness will have been intensified by the integrated geographic information offered by GIS technology. In the Geography teaching processes and also in those of architecture, urban planning and environmental sciences - and next to the corresponding theoretical predictions - these informational contents have become indispensable in our time. In the participative processes, these contents may constitute the matter and the energy of social and territorial transformation practices. They may be the measure of choices and values for personal and social life reorganization.

Geoinformation: a social Issue

41

3- Results

3.1 The trajectory of INPE Considering the institutional aspect, INPE is a representative mark of development and diffusion of geotechnologies and their various applications which determines a clear tendency towards a new social cooperation and public service practices. Nowadays, geotechnology applications are very important for scientific and technological progress, as well as for the Brazilian territory management, contributing to the citizens’ extensive access to these technologies. In fact, as specified below, a basic distinction in the fields of geoinformation application in INPE includes not only the development of science and technology projects within the Space Science and educational programs and research, but also the initiatives and projects in order to strengthen interactions between institution and society (in a broader scope), which can be observed by the use of dissemination resources on the Internet and in real time and by the new cooperation oriented to public policies and services that are focused on the environmental and territorial issue. In this context, INPE has been contributing on satellites for earth observation, on generation and distribution of images via the Internet, including free ones, on researches in the field of remote sensing and its applications and on educational programs and development of free GIS softwares as SPRING (Georeferenced Information Processing System), SPRINGWEB and TERRAVIEW. Today, one of INPE’s distinguished areas is the environment management which has been developed on the Rain Forest and Amazonia. The highlighted activities are the real time detection of burnings and deforestation evaluation, the growth of soya bean plantation and wood exploration occurrence, and the climate changes data. As to this latter activity, it is pointed out the Center for Earth System Science (CCST) of INPE with emphasis placed on extreme events and natural disaster studies. It is worthwhile highlighting the recent creation of the Space and Society Program with the objective of better learning about Brazilian social and territorial realities and which intends to map the Brazilian territory in order

42

A.C. Di Maio et al.

to provide subsidies to the elaboration of proposals for public and territorial management policies. The Northeast Regional Center created the Interactive Atlas, the Atlas of Social, Economic and Environmental Northeast and the Municipality Maps and thus, inaugurated a new program on the local institutional action, which is available at Internet and can be taken as an example of the institutional proposal. This simplified view of INPE’s trajectory serves as backdrop to some research experiences in INPE as well as in different universities in Brazil, in the field of urban and regional planning and basic education. Experiences and applications in urban and regional planning One of these experiences was the elaboration of MAVALE (Vale do Paraíba and the Northern Coast of São Paulo Region Macro-zoning) project (Kurkdjian et al. 1992), which represents a pioneer institutional experience in the field of regional planning and interdisciplinarity through the use of remote sensing techniques applied to a de-tailed characterization of the region on environmental, social and spatial aspects and prospective mappings. Other experiences refer to remote sensing applications which have been developed by means of partnerships between city halls and planning agencies for land use mapping on municipal scale. Several supervised studies have been developed within graduate and undergraduate levels, based on geotechnologies, with the objective of having a more refined understanding of the spatial organization of different municipalities of Vale do Paraíba region. The highlighted activities are: studies concerning problems related to the territorial organization such as the accelerated urban expansion and the environmental degradation, the spatial segregation, the formation of peripheries by clandestine settlements, the urban poverty, and other urban problems as the land use and occupation in conflict with the existing legislation for airports’ neighborhood. Other experiences resulted from the participation in projects con-ducted by the Northeast Regional Center (INPE) and Rio Grande do Norte Federal University, both located in the city of Natal. These projects were especially intended for GIS and mapping elaboration for Rio Grande do Norte State. For Natal city, the objective was also to organize a GIS for the historic district named Ribeira. The database consisted of photographies and cadastral

Geoinformation: a social Issue

43

data for buildings of interest to heritage architecture and generation of several maps through the SPRING software’s queries tool. The results, together with other database, subsidized the elaboration of the district revitalization plan (Tinoco et al. 2008). In order to make a comparative summary, such experiences and applications converge to the knowledge of the territory connected to the processes of scientific research and education and, on the other hand, to the processes of planned action. In this sense, the creation of GIS maps meets primarily the academic community and public agencies, in varying levels of complexity, depending on different geographical scales and objectives of applications and methods and technologies employed, basically causing the progressive advance, detailed knowledge of territories and regions and cities as essential support for public actions in different time perspectives.

3.2 Geotechnology in Education Today, despite a large amount of data and free programs such as satellite imagery and geographic information systems that has been available on the Internet, there are major challenges to overcome for widespread use of geotechnology combined with informatics in the public system of education in Brazil. They are: the very implementation and operation of computer labs in schools, training teachers in using new technologies and the development of appropriate material specifically for educational purposes in elementary and secondary education. The development of an educational digital program based on local data, the elaboration of material with the objective to disseminate the use of geotechnologies for the primary and secondary schools and the teachers’ training represent the activities which have been defined and executed by the GEODEN (Digital Geotechnologies in Education) and GEOIDEA (Geotechnologies as tool for Digital In-clusion and Environmental Education) projects. These projects also aim to contribute on changing the situation of digital inclusion in the public schools system in Brazil, therefore seek to integrate the use and appropriation of technological approach in the educational environment by means of geotechnology. These projects contribute with applications in the socialization of geoinformation for teaching purposes and the promotion of citizenship. Their scientific, social and educational

44

A.C. Di Maio et al.

importance is based on socializing new knowledge; creating a teaching technique; presenting studies over dynamic processes which are updated on different scales; using innovative material to be applied to students and to the population in general - an introduction to Cartography through GIS. GEODEN project (Di Maio 2004; 2007) may be accessed by Inter-net at http://www.uff.br/geoden. The educative site was structured in modules which contain themes related to Geography (spatial perception), Cartography and Remote Sensing. The exercises are performed from the public domain GIS EDUSPRING, the customized SPRING (Câmara et al. 1996) software for educational purposes. GEOIDEA project (Di Maio et al. 2009) is a free interactive multimedia CD-ROM which contains topics related to Cartography, space technology and environmental issues with focus upon Brazilian biomes, watershed and nature conservation units. The proposed activities use EDUSPRING software. Within the proposal of developing a Geographic Information System to suit specific application requirements for primary and secondary schools, a specialized version of SPRING 5.0/INPE, named EDUSPRING (SPRING for Education), was created under the GEOIDEA Project. The original program had been suited and reduced to approximately 70% of its size, losing some advanced functions which had not direct application for those schools levels. The applicative was successfully used in schools where the projects have already been implemented. However, some drawbacks were found, for example, public schools have not provided the best logistical conditions for the use of geotechnologies, but the reality in the country is not to equip public schools with one computer for each student. Far from it, but the important thing is to work with reality and create a way to adapt the use of new technologies to the conditions of each institution. The problems faced in the systematization of the use of computer labs in the schools were also a positive learning aspect due to the importance of the suitability of new practices in schools in real situation. After all, just what is actually used can be improved.

3.3 The perspective of social cartography The experiences in the field of Social Cartography found in the literature reviewed lead to the observation of new possibilities and contradictions with regard to relations among geoinformation, citizenship and social participation.

Geoinformation: a social Issue

45

Based on recent researches (Acselrad 2008; 2010), it is understood that, in the essence of territorial data social production exists different political subjects which are more or less in accordance with the proposal of a consciously and promising built cooperation for the whole society. That is to say, on the other hand, the production of participative mapping which involves subordinated or not assisted communities under political and social viewpoint may comprise conflicting relations and new forms of rights and territorial identity dispossession. If in the participative processes of a local knowledge production, an active voice is recognized or conceded to different local communities, the more sensible point of this situation continues to be the ethic and political character of the social use of produced information. That is, the right to think about the world and interpret social and territorial local problems may be, by the participatory method and the use of maps and GIS, a force for social and political statement and also an additional ideological mechanism to control hegemonic agents not engaged in the genuine processes of citizens’ socialization.

4- Conclusion and future plans For educational purposes and regional and urban planning activities, with regard to citizenship, there is no doubt that geoinformation use has acquired a differentiated and highlighted meaning today. In the process of diffusion of geotechnology in recent decades, without taking into account the limitations on the innovative mindset, the training on its use and its still limited access, there is some consensus about its strategic potential for reality comprehension and elaboration of public policies for environmental monitoring, control and management with social participation. But if until the decade of 1990, these resources were used as subsidies for a few decision officers, nowadays they have become potentially strong tools for resizing social and political spheres and the consequent redistribution of posts and assignments in the perspective of building a more egalitarian and human society. But if geotechnologies are not in themselves the solution to social problems in general, the information they can mobilize encourages new ways of knowledge and action. Just as the positive impacts of geotechnologies inclusion in the school teaching practices are found, it will not be different in distinguished social spaces in which they flourish. Contact with the

46

A.C. Di Maio et al.

social and spatial reality of the living place, through social interactive environments shaped by information, makes its understanding less abstract, more dynamic and inclusive (Di Maio 2004), and can naturally lead to positive impact on organizational processes and social awareness. In cases of spontaneous or conducted participation, spatial information is or may be the matter and the energy of social and territorial practices transformation. They can be the tools for the choices and values for social and personal life reorganization. In any space where social life happens, geoinformation can consolidate interactive and self-awareness practices and can generate political clout and thus establish itself as a key ingredient to the reversal of social differences, yet so large in our society. As a perspective for future action to consolidate the social use of geoinformation, the RIsO (Supportive Information Network for Rio de Janeiro) Project, under development into the Fluminense Federal University, aims to create an Internet portal with georeferenced in-formations related to available educational, cultural and sports free activities with focus on citizenship in Rio de Janeiro metropolitan area. These activities should compose a map of social action for youth, from the use of WEBGIS, GIS and Social Cartography concepts, and tools from Google Earth. The system will be disclosed in public schools for students and students’ parents and society as a whole, and they may also add information and participate in the construction of the Project. Thus, it takes into account that Cartography may be a tool to approach the places and the world (MEC 1999), it brings closer to citizens, especially the disadvantaged ones, the appropriation of the feeling of belonging to public space and with access to opportunities. Other actions involve the dissemination of educational projects with activities supported by practical applications in geotechnologies and free software in Portuguese speaking countries as Guinea Bissau, in Africa, and Brazil, especially for public school teachers and students (Nosoline in press). A distance course in Environmental Education to elementary public school teachers, also under development with support from the Ministry of Education, provides the inclusion of GIS in activities and in the methodology to be adopted throughout the course. As part of the Laboratory of Studies and Strategies for Participatory Planning of the Master Course on Urban and Regional Planning, at UNIVAP (Vale do Paraíba University), another project under development is called Vale do Paraíba Observatory Center (SP). The project central aim is the

Geoinformation: a social Issue

47

creation of an internet portal where information and social interaction are the elements for articulating collective processes of knowledge production, discussions about the region and the mobilization agents, favorable to the enrichment of civic participation and to the remodeling of principles and actions for social change and planning In short, for the current stage of democratization of social and political life of Brazil, which establishes in law the shareholding requirement in the process of public policy formulation and planning, it does make sense to use GIS technology for knowledge of the living place and its broader regional context and the educational process for the full exercise of citizenship.

Abbreviations CCST - Center for Earth System Science GEODEN - Digital Geotechnologies in Education GIS - Geographic Information System GEOIDEA - Geotechnologies as tool for Digital Inclusion and Environmental Education MAVALE – Vale do Paraíba and the Northern Coast of São Paulo Region Macro-zoning MEC- Ministry of Education INPE - National Institute for Space Research RIsO - Supportive Information Network for Rio de Janeiro SPRING - Georeferenced Information Processing System UNIVAP – Vale do Paraíba University

References Acselrad H (ed) (2008) Cartografias Sociais e Território. IPPUR/UFRJ, Rio de Janeiro Acselrad H (ed) (2010) Cartografia social e dinâmicas territoriais: marcos para o debate. IPPUR/UFRJ, Rio de Janeiro Câmara G, Souza R, Freitas U, Garrido, J (1996) SPRING: Integrating remote sensing and GIS by object-oriented data modelling. Computers & Graphics 20: (3) 395-403 Damiani A (1999) A geografia e a construção da cidadania. In: Carlos AF (ed) A Geografia da Sala de Aula, Contexto. São Paulo Di Maio A (2004) Geotecnologias Digitais do Ensino Médio: Avaliação Prática de seu potencial. Tese de Doutorado, Universidade Estadual Paulista Di Maio A (2007) GEODEN: geotecnologias digitais no ensino básico por meio da Internet. In: INPE (ed) Anais do XIII Simpósio Brasileiro de Sensoriamento Remoto, INPE, São José dos Campos, p.1457 -1464. Di Maio A, Francisco C, Levy C, Pinto C, Nunes E, Carvalho M, Dornelas T (2009) GEOIDEA - Geotecnologia como instrumento da inclusão digital e educação ambiental. In: INPE (ed)

48

A.C. Di Maio et al.

Anais do XIV Simpósio Brasileiro de Sensoriamento Remoto, INPE, São José dos Campos, p. 2397-2404. Goodchild M (1998) Rediscovering de World Through GIS - Prospects for a Second Age of Geographical Discovery. In: GIS PlaNET'98 Proceedings, Lisboa Julião R (1999) Geografia, Informação e Sociedade. GeoInova - Revista do Departamento de Geografia e Planejamento Regional nº 0: 95-108 Kurkdjian ML, Valério Filho M, Veneziani P et al (1992) Macrozoneamento da Região do Vale do Paraíba e Litoral Norte do Estado de São Paulo. INPE, São José dos Campos MEC - Ministério da Educação (1999) Parâmetros Curriculares Nacionais. História e Geografia (Ensino Fundamental), v. 5. SEF, Brasília Nosoline I (in press) O uso das geotecnologias como recurso didático nas aulas de geografia. Dissertação, Universidade Federal de Viçosa Santos M (1987) O Espaço Cidadão. Nobel, São Paulo Silva M (2008) Os professores e o desafio comunicacional da cibercultura. In: Freire W (ed) Tecnologia e educação: as mídias na prática docente. Ed. Wak, Rio de Janeiro Tinoco M, Bentes D, Trigueiro E (2008) Plano de reabilitação de áreas urbanas centrais – PRAC/Ribeira. Edufrn, Natal Tristão M (2008) Educação Ambiental na Formação de Professores: redes e saberes. Annablume e Facitec, São Paulo,Vitória

Passing GIS Knowhow from University Students to secondary School Students: pedagogical Approach in developing youngster’s Capabilities and Understanding in GIS Amal Iaaly 1, Rola Jadayel, 2, Oussama Jadayel 3 1

Amal Iaaly, University of Balamand P.O.Box: 100, Tripoli, Lebanon [email protected] 2

Rola Jadayel, University of Balamand P.O.Box: 100, Tripoli, Lebanon [email protected] 3

Oussama Jadayel, University of Balamand P.O.Box: 100, Tripoli, Lebanon [email protected]

Abstract The present work reports an effort done by the GIS Center at the University of Balamand (UOB) to spread spatial awareness and GIS knowhow among youth. It also reports a change from conventional methodology in teaching the GENG 310 GIS course to a modern one that incorporated a community-driven project-based learning. To implement such change and promote the use of GIS within the community and spread spatial awareness, it was decided to incorporate the “UOB Recycles” project within the framework of the GIS course work. “UOB Recycles” is a project launched by the University of Balamand as an individual effort toward the emerging pollution situation in Lebanon. It aims to promote recycling activities and implant the sense of civic and environmental responsibility among school students. This project constitutes an interdisciplinary approach that will enable university students to deploy GIS skills to address this problem, and pass on their GIS knowledge to school students using constructivist pedagogical learning approach.

A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_4, © Springer-Verlag Berlin Heidelberg 2011

49

50

A. Iaaly et al.

Modern educational methodologies are highlighted as complementary approaches in the learning process at University level. Additionally, Community based programs are presented as an integral part in the development of individuals at early ages. The work focuses on Service Learning and Project Based Learning as powerful concepts in promoting education. It shows how students throughout these processes learn to deploy their academic knowledge to meet community needs, develop practical skills and build leadership qualities. It also examines how students sense civic responsibilities, develop team spirit and share knowledge through experimental learning. The experience of the University of Balamand (UOB) in Lebanon is presented in this paper as a supportive example. An innovative education technique is presented under the umbrella of an interdisciplinary project which succeeded in passing technological advances to younger generation through partnership between the University, a number of schools and the community. The collaborative work of several units at UOB is detailed to show a successful learning experience with students from the Faculty of Engineering at UOB. Results are highlighted at two levels (1) community service level and (2) educational level. The projection of the University on its community is illustrated to provide a good model which could be adopted to promote spatial awareness, environmental respect, recycling behavior and team work spirit through Service Learning and Project Based Learning. This experience resulted in an innovative pedagogical educational model which provided a smooth integration of the GIS science at the high school level. The work illustrates an alteration from the classical instructional model versus a student based active learning process: Youth teaching Youth using direct instructions and immediate feedback with minor dependency on the instructor.

Passing GIS Knowhow from University Students to secondary School Students

51

1- Introduction

1.1 Role of the University

1.1.1 Historical Background The classical dominant model of the University emerged during the nineteenth century, in which teaching and research were combined in the search for absolute truth (Robert A 2010). At that time education was designed to serve the social elite (Robert A 2010). Over the years, the structure and aims of universities remained constant.It was until 1963, that the Robbins Report (Robert A 2010), sought to introduce a minor change to the classical university model without radically changing it. The democratization of the University model aimed to make universities accessible to the public thus the emergence of “Poly-technique” as an alternative to university education. Gradually research has increasingly become detached from teaching. Research funding was directed to become more socially and economically relevant. Consequently the concept of academic freedom was developed (Robert A 2010). Finally, as suggested by Robert A (2010), to seek guidance from the past the 'idea of the university' is best seen a set of tensions, permanently present, but resolved differently according to time and place rather than a fixed set of characteristics. Compensation or coming down too heavily on one side of these balances will usually mean that the aims of the university are being simplified and distorted.

1.1.2 Towards Society It is believed that the solution to any problem at national level is a unified effort of four stake holders: the Government, the Private sector and NGO’s, Citizens, and Educational institutions. In particular, Universities, being at the core of every society, are seen to play a primary role in its development. They are to identify, analyze and address problems in the

52

A. Iaaly et al.

society. Consequently they are to suggest pragmatic solutions, set proper ground for their implementation, and direct individuals to use their academic knowledge to serve their community.

Figure 1: The four stake holders in the society.

1.1.3 Towards Students University life is well known to be a main factor in shaping one’s personality. Individuals at this stage are exposed to diverse aspects of life; they are expected to make choices, face challenges and practice decision making. This fact shaped University education along the years to converge towards preparing well rounded individuals with high graduate qualities answering the needs of an increasingly diverse population and a global economy. It became axiomatic that developing new skills beyond academic achievements is vital in order to help succeed in our contemporary highly competitive job market.

Figure 2: Contemporary graduates’ qualities

Passing GIS Knowhow from University Students to secondary School Students

53

From that perspective and during the last two decades, more attention has been given to extracurricular activities and community service programs. Most educational institutions started integrating modern educational methodologies (such as service learning, project based learning, etc...) in parallel with the classical learning methodology, seeing in that the proper approach to preparing individuals capable of assuming responsible roles in service to society. Preparing critical thinkers who can demonstrate leadership capabilities, ability to work efficiently with others, and practical experience applying their academic knowledge efficiently in serving humanity became a major role for Universities.

1.2 Modern Educational Techniques: Supportive Tools for Classical teaching Modern educational methodologies (such as Project Based Learning and Service Learning) were developed to complement classical learning and hence provide students with a monitored exposure to real world problems which have always captured students’ interest and provoked serious thinking (Wikipedia 2011). Project Based Learning (PjBL) is organized around an open-ended Driving Question or Challenge allowing some degree of student “voice and choice”; it incorporates feedback and revision, and results in a publicly presented product or performance. The student therefore develops communication skills, research methods and learns to take ownership of his success (Wikipedia 2011). Another academic approach to integrating real life problems in the educational system is Service Learning which relies on the clear identification and definition of the learning components and outcomes of a student undergoing a community service experience in the same manner community partners’ requirements were defined (Jadayel O and Nahas G 2009). In both methodologies, students and academics are confronted with very specialized and challenging learning opportunities in which practical solutions are to be found for very specific problems (Jadayel O and Nahas G 2009), therefore adding a new component to the traditional teaching/ learning processes.

54

A. Iaaly et al.

2- Background and Objective

2.1 GIS Education in Perspective A Geographical Information System (GIS) is a generic term which refers to the combination of hardware, software, data, people and methods. It is a strong science which is capable of handling, storing, updating, presenting and querying data based on their geographic occurrence. It therefore enables answering crucial questions otherwise impossible. GIS has been listed among the 25 most important developments affecting the life of humanity in the 20th century (Cook et al. 1994). Nowadays, GIS is no more perceived as a technological tool but an educational tool that contributes to creating an inquiry and project based learning environment that will help students develop spatial thinking skills creating pedagogical advantages. Being a highly interactive science of applied nature, it is realized that GIS is most powerful when linked to real life situations and used to solving community related issues. GIS education was incorporated in many university geographic educational programs since 1980. It grew to become an integral part in most university programs and even a degree by itself. GIS was integrated in high school education curricula as well, due to its strong spatial analysis capabilities, and its importance in geography studies, social studies and science education. The main aim of this integration is for students to develop both geographical and spatial thinking capabilities and to link geography to the real world. Despite its significant function in real-world geography, GIS had not yet been exploited efficiently in geography education at school level (Wiegand 2001). The slow incorporation of GIS into schools is faced by many barriers and challenges. The first difficulty is the inadequate training of teachers on the use of the GIS software due to its complex technological and composite aspect. This training must be embraced by a well structured infrastructure and continual professional support due to the fast pace of GIS software package upgrade. The second barrier is the lack of exemplary models for integrating GIS within schools (Bednarz and Audet 1999). The third barrier is the lack of explicit guidance on how to plan pedagogy around it, and the lack of effective pedagogic practices and teaching models.

Passing GIS Knowhow from University Students to secondary School Students

55

2.1.1 Nation’s Perception of GIS in Education GIS education and GIS in education is not perceived as a requirement in the Lebanese educational system. Individual efforts done by the private sector in parallel with universities within the educational sector are held to promote the use and implementation of this technology and to date have had limited success. In 1996, Environmental Research Institute ESRI along with their Lebanese representatives Khatib & Alami attempting to introduce GIS to schools. In this respect, professional training and free software packages were provided to schools in order to promote the use of GIS in education. Nowadays, almost all universities in Lebanon introduced GIS to their students each taking a different approach in teaching. Their main objective is to create a pool of GIS experts to better serve the emerging and evergrowing requirements of the professional job markets.

2.1.2 The University of Balamand Approach to GIS Education The University of Balamand, established its GIS Center (GISC) in 1999 within the faculty of Engineering to serve faculty, students, and staff by coordinating the acquisition, instruction, deployment, and development of Geographic Information Technologies on UOB campus. The particularity of this center is in its belonging to an educational institution; the fact that shaped its mission towards advancing knowledge in GIS technologies through real projects using modern educational techniques. The GISC developed to become a center of excellence in Lebanon and the region and it won national and international awards on many occasions. The center is student based; it aims at engaging youth in Community Based Projects using experimental practices hence promoting learning through doing. Its primary objective is to prepare skilled GIS individuals who can participate in building, using and maintaining GIS applications. The GIS Center introduced in parallel to its current activities, a well structured GIS introductory- level course aiming to teach basic principles of geographic information sciences, cartography and remote sensing within the Faculty of Engineering.

56

A. Iaaly et al.

The introduction to GIS course GENG 310 was structured in a 3 credit traditional format of higher education courses including 1.5 hours per week of lecture course to introduce the theories and fundamentals of GIS, and 1.5 hours per week hands-on lab exercises on the ArcGIST Software. Although the course proved to be very popular among students being an introduction to a new powerful science, it was noted that it only focused on the conceptual elements of GIS and on the software skills and thus limiting student’s ability in spatial problem solving and critical thinking. Therefore it was decided to introduce the concept of project based learning in order to provide an opportunity for students to practice the skills acquired in the classroom and implement them in a real-life project.“UOB recycles” project constituted a good medium to practice project based learning and achieve the set goals.

Figure.3: Students at the GIS center

2.2 The Pollution Problem in Lebanon The Environmental Performance Index (EPI) in 2010 study, ranked Lebanon at the 90th position out of 163 countries indicating an alarming need to address the pollution problem at the governmental, social and personal levels (Yale Center of Environmental Law & Policy et al. 2010). In parallel, the Pew Global Attitudes Project in 2007 showed that Lebanon ranked at the bottom three out of forty six nations in terms of public awareness and environmental concern (Newsweek 2008). In a country like Lebanon, the problem is managerial in as much as it is environmental. It is realized that in the absence of a general plan in waste

Passing GIS Knowhow from University Students to secondary School Students

57

management, efforts should be directed in two dimensions: (1) spreading environmental awareness and (2) setting an action plan for managing the process of waste disposal in general.

2.3 “UOB Recycles”: Project Outline Recycling by definition is the act of processing used or abandoned materials for use in creating new products (American Heritage Dictionary ed. 2009). Essentially, it is an attitude, a behavior that all citizens should develop and adopt as waste management is primarily a personal initiative in the absence of a global policy at the national level. While this awareness behavior in Lebanon is on the rise, more efforts are yet to be expended before it becomes effective. Educational institutions, NGOs and Municipalities may then provide proper framework for such an operation and hence play a major role in spreading this culture. “UOB recycles” is a project set to meet that philosophy. It is managed by the Office of Student Affairs (OSA), a University unit lead by the Dean of Students. OSA is mainly concerned with developing programs and activities that are intended to improve the quality of life at the University and provide proper ground for the development of a well rounded student (OSA, UOB web-site). The project started as a student based awareness campaign initiated by the Nature Club at UOB and developed later to become an interdisciplinary project across the University as a whole. It attracted students and faculty members from different disciplines. A committee was then formed for proper planning and vision. The committee believed that students are the main element in the success of the project. They are potential managers, powerful evaluators, and would be the most influential messengers in the society (Jadayel R et al. 2010). Hence students from various backgrounds were engaged in the project diverse activities as volunteers or under the umbrella of UOB’s community service program: Service Education, Experience through Doing (SEED). SEED is a program based on the philosophy of service learning, seeking to meet the real needs of the nation by building partnerships between the University and the community (SEED, UOB Website). The GIS Center saw in this project a good medium to link with the society and an efficient mechanism for students to practice the skills acquired in the classroom and implement them in a real-life project. University students main task was to pass their GIS know how to school students and to

58

A. Iaaly et al.

supervise the production of digital maps in partnership with a number of local schools.

2.4 Objectives The objective of this work is to highlight an educational model in introducing GIS to school students using pedagogic approach of Youth Teaching Youth directly. This educational model focuses on active learning with GIS through altering the instructional delivery methods. These methods are based on exploiting project and finding solution to community problem rather than focusing on technological software training or in-class teacher tutoring. Active learning aims at engaging students in field-work, data analysis, critical thinking and problem solving and make them part of the decision making process.

3- Approach and Methodology

3.1 Approach A partnership was established between the OSA, the SEED office and the GIS Center giving the project a particular characteristic, merging academia and civic engagement. From that perspective, “UOB recycles” constituted a good medium for the GIS center to achieve its academic goals. Furthermore, it provided proper ground for the SEED office to project on and link with the community. The three units, each having its own objectives, benefited from this partnership and participated jointly in addressing a national problem to which pragmatic solutions were suggested thus fulfilling the mission of the University in serving its community. It was realized that in this unique learning practice, University students will not only experience a tangible practice on the GIS project life-cycle development in a real-life community project, but will also foster caring toward the community and the environment, develop spatial awareness, in

Passing GIS Knowhow from University Students to secondary School Students

59

addition to transferring their GIS skills to younger students who took part in this project as will be seen in the coming sections.

3.2 Obstacles Despite the emerging opportunity provided by the “UOB Recycles” through establishing a partnership between the University of Balamand and the surrounding schools many impediments were faced at the implementation phase. Beyond the apparent barrier related to the rigid and overloaded curriculum adopted in the Lebanese school, and the scarcity in computer literacy among both teachers and students, there were many other obstacles to be overcome in order to spread spatial awareness. These are summarized as follows: • Lack of spatial thinking and understanding among teachers , students and community members • Lack of time allocation for teachers to receive adequate GIS training • Lack of active learning perception and practices among teachers • Lack of ICT infrastructure within school premises • Lack of funds allocation for GIS software purchase • Lack of an effective exemplary GIS educational model

3.3 Overcoming Obstacles: Innovative Student-Based Pedagogic Training It was realized that a complete new approach must be developed in order to overcome the above mentioned obstacles and introduce GIS to both teachers and students in the participating school, knowing that their GIS knowledge was minimal at that time. The aim was to first create a motivated learning community through incorporating pedagogical activities in teaching complex GIS tools, and second to provide access to the hardware and software. The technical and pedagogical training on GIS technology will be provided through both workshops and student-to-student training sessions.

60

A. Iaaly et al.

Thus university students enrolled within the GENG 310 course will diffuse basic GIS skills directly to school students in an informal manner with no dependency on the instructor. The youth teaching youth informal approach in training lead to an attenuation of the complex aspect of GIS technology, and simplify its application. Furthermore it was decided to make use of the UOB GIS Center facilities and make it available for school students in order to overcome the problem of cost and technology. To increase motivation among school students a Map Gallery competition was scheduled with attractive prizes for the three runner ups to stimulate the learning process.

3.4 Methodology

3.4.1 The Awareness Campaign and Related Workshops Upon implementation, schools were not conscious of the exact capabilities and powers of the GIS software in spatial analysis. Therefore, it was decided to initiate the project through a workshop targeting students and teachers from participating schools. The workshop took place at UOB in the objective of introducing the usage and capabilities of the software in managing the recycling activities and in producing thematic maps part of the “UOB recycles” project.

Figure 4: GIS workshop targeting students from participating schools

Passing GIS Knowhow from University Students to secondary School Students

61

3.4.2 GIS Pedagogical Training The methodology adopted at the GIS center was to assign two volunteer students from UOB who are knowledgeable with GIS, to work with selected interested students (12-17 years old) from a partner school. The main responsibilities of UOB students were to tutor and train students from schools on the use of GPS, escort them during their visits to engage community members, and teach students how to produce maps using GIS software in order to display locations, build a database, and create a school recycling community. Students from schools, on the other hand, were responsible for the production of thematic school community recycling maps.

Figure 5: GPS field work data collection

University students adopted learning by doing as a main learning methodology in order to facilitate the understanding and the implementation of GIS work with the younger students. Schools students visited the GIS center on a weekly basis during the after school hours in order to learn how to display collected GPS points on the GIS software. They were also trained on how to create, edit, and display maps of the school recycling community.

Figure 6: Student- to -Student Project Based GIS training

62

A. Iaaly et al.

Furthermore school students were given a detailed training on how to create a layout using GIS that was of major importance for visual presentation. The layout design training was a focal point in order to effectively present the final results. It included an extensive overview on the: substance, the text elements, the layers, the legend, the background selection, the bounding frame, scale, North arrow and other map elements. Layouts in GIS are considered valuable means to share results, analysis and information visually with users with non-GIS proficiency. Afterwards, school students worked solely in the design of school community recycling maps that will be included in a Map Gallery Competition to be detailed thereafter.

Figure 7a: Sample 1: Map prepared by a school students team

Passing GIS Knowhow from University Students to secondary School Students

63

Figure 7b: Sample 2: Map prepared by school students team

3.4.3 Community Work A major part of the project is to promote recycling activity within the community, therefore students from participating schools were highly motivated to engage the largest number of community partners in the recycling project; parents, teachers, and staff were all working hard to present the best they had for the project. UOB students on the other side worked closely with every team offering assistance at any level whenever needed.

Figure 8: School students engaged in community work

64

A. Iaaly et al.

The community work required an ongoing and interrelated cycle of continuous collaboration and communication between the University of Balamand, participating schools, and community partners (households, supermarkets, banks, publishers, etc…) reflecting all recycling activities. The following diagram shows the interaction and cycles of the three stakeholders under the “UOB recycles” community-based GIS project and it illustrates the role of both groups of students and involved members of the community (Jadayel R et al. 2010).

Figure 9: Diagram showing connection between various partners in “UOB Recycles”

The left hand side of the diagram shows the interaction between UOB students and students from participating schools in the context of the project. It reflects a collaborative work engaging both groups of students which aimed at 1) serving the community and 2) improving their academic achievement. The results of this work are analyzed as a positive learning experience reflecting on both groups of students. Firstly, University students got the chance to practice what they have learned in the classroom (GIS course) by actual implementation in a real

Passing GIS Knowhow from University Students to secondary School Students

65

life project to which they could connect; they also developed leadership qualities when passing their knowhow to students from participating schools. Secondly, students from participating schools learned basic GIS and GPS skills, acquired communication and leadership qualities as they were addressing members from their community in an attempt to engage them in the recycling activities. The right hand side of the diagram shows the interaction between the University and the community. This is a phase of the project which is currently being implemented. Students are targeting municipalities and NGOs in a proposed mechanism yet to be finalized.

3.4.4 Map Gallery Competition A major aspect of the project was a Map Gallery competition launched among participating schools for the academic year 2009-2010.The theme of the competition was recycling and it involved production of thematic maps reflecting a school’s recycling activities. Different teams from participating schools presented their work which was subject to voting by the audience who attended the event. The competition created an atmosphere of competiveness and enthusiasm hence promoting the GIS science in an informal and pleasant educational perspective.

Figure 10: Students in the Map Gallery

Figure 11: Award Distribution Ceremony

66

A. Iaaly et al.

4- Results

4.1 Academic Achievements in GIS Within the context of classical teaching, students do not get a real chance to apply their newly acquired GIS skills in real-life situations. The project constituted a good medium for students enrolled in the GIS course to enhance their GIS knowhow and improve their spatial thinking skills and problem solving techniques. It also provided a rich learning experience to those students who joined it and resulted in significantly high achievement of the course learning outcomes as opposed to the ones acquired by students who were enrolled in the course solely. Students working with the project sensed real outputs and could link GIS to their surrounding and learned therefore through doing in a real life project scenario. A comparative study was conducted among students enrolled in the GIS course aiming at evaluating the results of students before and after project based learning implementation. Both sets of data represented results of the students given a hypothetical real life problem scenario which they had to suggest solution. It was observed that the solution suggested by students on the first project given at the beginning of the semester was rational but not realistic as they did not take into consideration the geographic constraints. On the other hand, the results on the second project were very relevant and reflected satisfying results in terms spatial thinking and analysis. It worth noting that the involvements of students in “UOB recycles” community service project enhanced the correlation between theoretical solution to a given problem and practical implementation The following chart represents the detailed results on both projects in which the higher achievement in the second project is noted.

Passing GIS Knowhow from University Students to secondary School Students

67

Figure 12: Chart projecting the students’ performance in GIS projects

4.2 GIS and Youth The impact of that educational experience was quite impressive on both groups of students. Along with the positive educational results noted, students developed leadership qualities, team work spirit, conflict resolution skills, planning and auditing methodologies, communication and presentation skills. They succeeded in spreading environmental awareness within their community and more importantly they developed spatial awareness, and a sense of appreciation to the powers of GIS technologies especially when directed to answer community related problems (Jadayel R et al. 2010). The following diagram shows the impact of such learning experience on the students on three levels: Academic, personal and Social aspects (Jadayel R et al. 2010).

68

A. Iaaly et al.

Figure 13: Diagram assessing the outcomes of the “UOB Recycles” project

4.3 GIS and Schools “Problem solving lies at the heart of GIS use and GIS technology is designed to organize, analyze and foster interpretation of spatial information” (Thomson D and Buttenfield B 1997). Spatial reasoning is focal for educational development, as it is seen to set proper ground for pragmatic skills development among youngsters’ logical spatial reasoning skills develop through discipline observation, description and measurement and are better adopted at early age. Hence the needs for the development of a pedagogically oriented learning approach to provide a smooth integration of GIS into school education. The project provided a smart and pioneering methodology in exposing school students at early age to the GIS technology. The informal approach in GIS training proved to be successful in promoting spatial reasoning, enhancing the GIS knowhow, and presenting the science as an interesting and motivating learning experience.

Passing GIS Knowhow from University Students to secondary School Students

69

4.4 GIS and Society As mentioned earlier, the pollution problem in Lebanon is managerial in as much as it is environmental. The project provided a structured academic approach in two dimensions: (1) spreading environmental awareness and (2) setting an action plan for managing the process of recycling in general. It therefore succeeded in gaining the trust of many NGO’s and municipalities in the region who sensed the powers of the GIS technologies in data visualization and data analysis. In particular the center succeeded in presenting the GIS science to its society as powerful tool of highly flexible and interdisciplinary nature. When first implemented, the project aimed at promoting recycling activities within the university premises only. It developed gradually to incorporate eighteen schools from the surrounding region and overgrew to involve one hundred seventy eight community partners.

Figure 14: UOB recycling Community Partners

70

A. Iaaly et al.

5- Conclusion and Future Plan

5.1 Conclusion In conclusion, the work presented a successful attempt in spreading spatial awareness and promoting GIS knowhow as it is becoming more prominent at all levels of education internationally. The overarching goal in this project is to present GIS as a strong mean for data visualization and data analysis as opposed to a dry technological tool. Practically, GIS can be introduced to youth at an early age in an active learning process through problem solving, project based learning, and service learning. This study strived to find an effective pedagogical approach to integrate GIS within school education. It emphasized the strength of projects of interdisciplinary nature when properly designed showing the power of each discipline and the ways knowledge is complimented. Moreover, it is noted that students connect more to what they are learning when working on community based projects.

5.2 Future Plan The GIS center will continue to cease opportunities in the context of community service. It will constantly aim at promoting GIS knowhow among youth and present the Geographic Information Systems as a powerful science of wide applications. For the coming six academic years, the plan is to engage more schools each year within the same mechanism, so as to reach the full number of schools in the region of North Lebanon. Moreover and for this coming academic year the center will approach interested municipalities to study the possibility of future collaboration and hopefully invite them to become active partners in the project. Within this framework, the GIS center will continue to collaborate with the Office of Student Affairs and the SEED office to spread environmental awareness, spatial awareness and encourage students to play active roles in their societies.

Passing GIS Knowhow from University Students to secondary School Students

71

Acknowledgments The authors would like to acknowledge the support of the: - Office of Students Affairs at UOB - GIS Center at UOB - Nature Club at UOB - Partner schools in the region of North Lebanon - Members of the community who participated in the project - The University of Balamand for providing its facilities for the success of this learning experience.

Abbreviations ArcGIS T: GIS Software Package EPI: Environmental Performance Index ESRI: Environmental Research Institute ICT: Information and Communication Technology NGO: Non Governmental Institutions OSA: Office of Student Affairs PjBL: project Based Learning SEED: Service Education, Experience through Doing: The Community Service program at UOB GENG 310: Introduction to GIS Course GIS: Geographic Information System GPS: Global Positioning System UOB: University of Balamand

References Cook WJ, Collins S, Flynn MK, Guttman M, Cohen W, and Budiansky S, (1994) 25 breakthroughs that are changing the way we live and work. U.S.News and World Report, 2 May. Jadayel R, Iaaly A, Jadayel and Oussama (2010) Service Learning and Scholarship: Experimental Learning and Community Partnership for Common Good. Proceedings of the 39th IGIPSEFI Annual Conference, Ternava, Slovakia, 19-22 Sept., 2010. Jadayel, Oussama, Nahas, Georges (2009) Community service and scholarship: prospects and challenges for Lebanese engineering institutions. Proceedings of the 38th IGIP-SEFI Annual Conference, Graz, Austria, 6-9 Sept., 2009. McPherson K (2005) Service Learning. In New Horizons. [Online] 10. 2005 [cit. 2010 – 2002].http://www.newhorizons.org/strategies/service_learning/front_service.htm. Newsweek Magazine (2008). The New Green Leaders. Newswek Magazine article. Online http://www.newsweek.com/2008/04/26/the-new-green-leaders.html. Accessed 15 January 2011

72

A. Iaaly et al.

Office of Students Affairs, University of Balamand Web-Site. http://www.balamand.edu.lb/english/OSA.asp Robert A (2010) The ‘Idea of a University’ today. History and Policy. http://www.historyandpolicy.org/papers/policy-paper-98.html.Accessed 9 December 2011. The American Heritage® Dictionary of the English Language, Fourth Edition copyright ©2000 by Houghton Mifflin Company. Updated in 2009. Published by Houghton Mifflin Company. http://www.thefreedictionary.com The SEED Program, University of Balamand Web-Site. http://www.balamand.edu.lb/english/Seed.asp Thompson D, Buttenfield B (1997) Learning with GIS, Learning about GIS. UCGIS White paper. http://www.ucgis.org/priorities/education/priorities/learning.htm. Accessed 10 January, 2011. Wiegand P (2001) Forum Geographical Information Systems (GIS) in Education. International Research in Geographical and Environmental Education. 10 (1) pp.1-4. Wikipedia Contributors (ed. 2010) Project Based Learning. Wikipedia, The Free Encyclopedia.http://en.wikipedia.org/wiki/Project-based_learning.Accessed 10 October 2010. Yale Center for Environmental Law 7 Policy, Yale University Web-Site . http://epi.yale.edu/countries

The Methodological Advantages of using Web Server in Teaching GIS Andrea Pődör Faculty of Geoinformatics, University of West Hungary H-8000 Székesfehérvár, Pirosalma utca 1-3. Hungary [email protected]

Abstract The paper presents a project based method in teaching geoinformatics with the integration of web server. In the paper the author discusses some different learning techniques and outlines how geoinformatics course work was designed at the University of West Hungary to integrate some of these new learning methods. The paper shows some practical example of a subject that models the workflow from data acquisition through to data publication. Some practical example demonstrate that the teacher act as a mentor during the classes to help the students in acquire knowledge in a way which best suits them, they share their knowledge, their experiences with each other and they learn to cooperate. Moreover, the quality and the accuracy of their work is improving during the semester, because the success of the “project” is dependent on the collective work of each student.

1- Introduction With the immense development of the implementation of Geographic Information Systems (GIS) in the different fields connected to spatial data management, the publication of spatial data became an essential part of a

A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_5, © Springer-Verlag Berlin Heidelberg 2011

73

74

A. Pődör

well prepared GIS. Recently, GIS has shown progress in two fields: mobile and web applications (Csutorás and Pődör 2009). In order to prepare the students for enhancing their competitiveness in the labour market it is necessary to implement various web mapping solutions in education. Incidentally, GIS web server solutions proved to be a good methodological tool in the hand of teachers in two ways. One is enhancing the collaborative activity of the students; the other is the augmentation of the quality of their work.

2- Background and objectives The Faculty of Geoinformatics at the University of West Hungary offers courses in geoinformatics for Bachelor of Science (BSc) students who will be surveyors and land managers. Receiving their bachelor degree, the students usually get a job at companies where they have to do practical work. It is also essential for them to be able to work as a team member and later as a leader, who conducts and plans the workflow of others. The students study geoinformatics for three semesters. In these semesters they acquire knowledge of geoinformatics from the basic level until they are able to plan a special GIS application in their last semester. During the semesters the same commercial GIS programs are used, but there are some optional subjects which give the students a chance to get to know other GIS software as well. These semesters allow the students to permit their skills in manipulation, storage, processing and analyses of spatial data. Though the students have the opportunity in one semester to learn the basic rules of cartography and visualization, their main interest is usually limited to large-scale mapping. There are special subjects for those students who are interested in GIS and feel the need to deepen their knowledge in this field.

The Methodological Advantages of using Web Server in Teaching GIS

75

3- Approach and Methods It is a very difficult task to explain to the students that accuracy should have a very important role in their common workflow. The departmental GIS server was applied in their GIS course to present this problem. If someone makes an error when working for a company it can cost a huge amount of money and it can even generate false information. With the help of the GIS server it was easy to visualize for the students how their inaccuracy appeared on the map. They were taught the basic usage of the GIS server (such as downloading and uploading data) and the server was also used an interactive tool to measure the students’ activity. The task of the students was to create a statistical analysis for different settlements of Hungary using the data available from the official website of the Hungarian Central Statistical Office. As not all the data were in digital format, they had to build a geodatabase and digitize a part of the given area. Each of them had a different part of the country to process. They had to be aware of the topology and the topological rules. In the end, the accuracy concerning digitizing and keeping the topological rules were measured. The quality of the work of the students was shown by the web server. The legend was created according to topological failures (Figure 1): • • • • •

The items must have cover each other are blue. The items larger than cluster tolerance are yellow. The items which must be covered by each other are pink. The items must not overlap each other are red. The green colour means that there are gaps between items.

At the moment when the students opened the GIS web server containing their work they were able to follow visually how many failures they made and they were able to compare each other’s performance.

76

A. Pődör

Figure 1: Topological failures appearing on the map

Though by the end of the fourth semester the students have a basic knowledge of the capability of GIS server applications, they will not have learnt yet how to publish data without having programming knowledge.

4- Collaborative learning method in the GIS course Different learning methods have been described by scientists. The most best known is the work of Dewey, Piaget and Vygotsky on collaboration and interaction between equals (Wood 1994; Pound 2005). Other theories include behaviourism, learning styles, multiple intelligences, constructivism, constructionism right brain/left brain thinking (Cuthell 2005). Education for many years supported the learning process called behaviourism in the curriculum of surveyors at the University of West Hungary. Behaviourism is a learning process which is built on repetition and confirmation. The teacher transmits the learning material, the student memorizes, repeats and writes it. Behaviourism was the leading learning method for most of the past century. The other traditional method is cognitivism, which means that during the learning process we create a model about the outside world and we make it interior. Teachers should present an appropriate model with their meta communication (Kulcsár 2008).

The Methodological Advantages of using Web Server in Teaching GIS

77

Learning Style Theory presumes that students learn in different ways. David Kolb in his work detected that there are four different learning styles (feeling, watching, thinking, doing) and the best way to improve our knowledge is using the learning style which suits us best (Smith 2001). This theory is supported by Howard Gardner’s study about multiple intelligence: his assumption proposes that different kinds of intelligence exist in human beings (Allen, et al. 2007), so they can acquire knowledge in different ways. Throughout the past years the Faculty of Geoinformatics made efforts to introduce new types of learning methods in education. New types of learning methods, like Bruner’s constructivist (Bruner, 1966) theory, made it necessary that the traditional educational methods completely needed to be reformed. His theory affirms that learning is an active process and that students construct new ideas through their own knowledge. In this process, the students are able to select information, form hypothesis and make decisions. On the other hand, the teacher has a different role. Teachers have to transform lesson resources into a form that the students can understand and encourage them to gain experiences on their own and engage the student in dialogue. Using Bruner’s method the teacher should design the curriculum in a way that it builds on what the students have already known and makes a progress and development with the students’ active work in new fields (Smith 2002). In this manner the Faculty made attempts to reform some of its GIS course. Under the influence of the constructivist theories of Piaget, Vygotsky and Bruner, Papert developed his constructionist theory (Sefton-Green 2004). His idea is that the teacher is not to teach the students; instead of this the teacher becomes a mediator of studying. Students construct understanding and draw their own conclusion through creative experimentation. The learning theory for the digital age is connectivism which is used in computer science and it is based on the hypothesis that knowledge does not exist in the head of an individual person but it is present in the world (Siemens 2005). This new paradigm revived the disciplines like “Activity theory” and “Distributed cognition” consider knowledge to exist within systems which are accessed through people participating in activities.

78

A. Pődör

Most of these theories influence the modern teaching methods and give ideas and opportunity to the teacher to implement them in order to improve the effectiveness of learning. In the Faculty of Geoinformatics the teachers tried to follow the theories above and made it available for the students to acquire knowledge in a way which best suited them. The GIS subject “Mobile and web GIS” can support at least the last two forms of the four theoretical learning methods (behaviourisms, cognitivism, constructivism, connectivisms). This kind of subject can aid those learning principles where the students are in the centre. Some characteristics of the classes are: • • • •

Active learning against passive, The autonomy of the students, Assuring extracurricular learning, The connection among students are equal, the teacher acts rather like a mentor.

This subject can model the workflow from data acquisition, database design, analyses to data publication. Each year the teacher generates a project which is suitable to show the students the whole process. The subject is an elective one for those who are interested in GIS. The duration of the subject is 2 hours per week in one semester, which is altogether 28 hours. While the subject lasts for only a semester, the teacher initiates a project based learning, where the students have the opportunity to experience the various GIS solutions themselves: the teacher gives some technical details to aid the students in acquiring the basics of the software, but the students have the freedom to form the sequence of their study. Namely, if they are more interested in certain particulars, they have a chance to learn it in details. The number of contact hours is very limited to fully work out a complicated project. The essential thing is that first an appropriate project goal has to be found. The project must be simple enough to be able to fulfil the requirements of the teaching plan. The students need to gather some data, process, analyze and publish them. Last

The Methodological Advantages of using Web Server in Teaching GIS

79

year the project was dealing with the selective waste deposits of the town Székesfehérvár. As this was a successful project, it is worth looking at it as an example. In the project, the teacher defined the steps together with the students: what should be made, what the milestones of the projects were, how students should cooperate with each other, what kind of data was needed. Although the project was a simple one, the students had to go through several steps consistently: • • • • • • •

Geodatabase design, Data acquisition, Processing of the collected data, Control of the accuracy of the data, Analysis of the data, Querying of the data, Publication of the data.

In a process like this, the students need to get acquainted with mobile, desktop and web applications. There are some basic tasks of the subject which each student has to complete. They have to go out to the field to gather data with mobile equipment using a mobile GIS application. Before they go out, the students with the help of the teacher are designing the database and load the geodatabase and the base map on the mobile equipment. While the students are gathering the data, they apply the same domains and subtypes to build a uniform data structure. They have to process their data in a computer lab in a desktop application and they publish it on the web. During the semester the students and the teacher are working together on the basics, but the students have to fulfil their task on their own by integrating the data gathered by the other students of the group. There are tasks that the teacher and the students discuss, but the better students can also develop the publishing phase on their own. In the case of the selective waste deposit project, the database structure was very simple. Only fundamental data were stored in the database such as the type of the wastes, the coordinates, addresses of the deposits, a picture about each deposit and emptying time of the bins. Each student had to measure approximately five deposits. They loaded the data into the Spatial Data Engine of the GIS server and with the help of APIs provided

80

A. Pődör

by the producer of the software each student was able to make a web application to publish the data. The APIs were easy to change and gave opportunities for best students to alter the initial codes to improve more their service (Figure 2). In the final class the students had to give a presentation about their experiences and show their web application results.

Figure 2:. Waste deposits in Székesfehérvár as examples of simple and more complex solutions by the students

5- Results Throughout the semester each student is able to acquire a basic knowledge concerning mobile and web GIS applications, but those who are more interested can improve their familiarity with GIS or even use their programming knowledge to build a service on their own.

The Methodological Advantages of using Web Server in Teaching GIS

81

As this a relatively new subject, the author’s experience covers three years only. However, the subject is proved to be successful and effective. The project on selective waste deposits raised the interest of the local administration and waste management company, which expressed its interest in developing the system further. The students can use the new knowledge gained during the project when in preparation for their degree thesis.

6- Conclusions As the GIS shows a rapid progress, it is very difficult to keep up with the improvements from year to year. This project based approach is important, because the teacher can give floor to the students to improve and acquire knowledge by themselves. Further they share their knowledge, their experiences with each other and they learn to cooperate. Moreover, the quality and the accuracy of their work is improving during the semester, because the success of the “project” is dependent on the collective work of each student. The students experience independent learning by doing and collaborative working as well. This project based learning method with GIS web server can give an opportunity to the teacher and the students to establish a relationship with the business community. Furthermore it can give a chance for the best students to create some outstanding publishing service. On the other hand the teacher gains methodological experiences which can be integrated into the teaching plan of the following semesters.

Acknowledgments

I would like to acknowledge the contributions of the project 06524/100/3 TÁMOP-4.2.1/B09/1/KONV-2010-0006 to the development of my paper.

82

A. Pődör

References Allen I E, Seaman J and Garrett J (2007) Blending In: The Extent and Promise of Blended Education in the United States. Needham, MA: The Sloan Consortium, March 2007. Retrieved May 27, 2010 from http://www.sloan-c.org/publications/survey/pdf/Blending_In.pdf. Bruner J S (1966) Toward a Theory of Instruction, Cambridge, Mass.: Belkapp Press. 176 + x pages. Csutorás B, Pődör A (2008): Térinformatikai szerver használata az oktatásban.(GIS web server in education). XVII. Térinformatika az oktatásban szimpózium, Budapest. Cuthell J (2005). Learning theory and e-pedagogy, MirandaNet. Kulcsár Zs (2008): Az integratív e-learning felé, retrieved 20 August 2010 from http://www.crescendo.hu, Pound L (2005). How Children Learn. Leamington Spa, Step Forward Publishing Ltd. Sefton-Green J (2004). Literature Review in Informal Learning with Technology Outside School. Bristol, Futurelab. Siemens G: Connectivism: A Learning Theory for the Digital Age, International Journal of Instructional Technology and Distance Learning, Vol. 2 No. 1, Jan 2005. Retrieved 20 August 2010 from http://www.itdl.org/Journal/Jan_05/article01.htm Smith M K (2001).'David A. Kolb on experiential learning', the encyclopedia of informal education. Retrieved 20. August 2010 from http://www.infed.org/b-explrn.htm. Smith M K (2002). Jerome Bruner and the Process of Education. Retrieved 26 August 2010, from http://www.infed.org/thinkers/bruner.htm. Wood D (1994). How Children Think and Learn. Oxford, Blackwell.

Historical data: exploration, digitalization, access an analysis

Geovisualization and Archaeology: supporting Excavation Site Research Spyridon Tsipidis, Alexandra Koussoulakou, Kostas Kotsakis Aristotle University of Thessaloniki, Greece [email protected], [email protected], [email protected]

Abstract Archaeology is a science where geographical and spatial factors are of capital importance; in this context Geo-visualization and Archaeology provide interesting challenges for each other and they can both benefit from a combined approach of their interests. Archaeological excavations in particular constitute an excellent field for geo-visualization applications, since they generate large amounts of data with complex structures in 3D space and in time. Consequently, visualization methods and tools can provide support to archaeological excavation analysis. This paper presents a visualization environment created for use by archaeologists in the prehistoric excavation site of Paliambela in Northern Greece. It is currently fully operational and is used on a steady basis in the excavation field. The system enables the archaeologist to create his/her own paths in information querying and synthesis and to save any concluded interpretations. This task is undertaken through the design of custom tools, based on principles arising through archaeological methodology and theory, structuring a useful geo-visualization framework for the assistance of archaeological interpretation. It is this need that the environment presented here attempts to fulfill.

1- Background and objectives Despite any individual differences, an archaeological excavation can be seen as an multi-scale and high-detailed observation process, resulting in the formation of a rich structured information archive. This archive is shaped by the systematic recording of different kinds of observations

A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_6, © Springer-Verlag Berlin Heidelberg 2011

85

86

S. Tsipidis et al.

related to the revealed archaeological evidence and the broader archaeological space, like: • • • • •

Catalogued attributes, Free comments, On-site actions or decisions, Objects spatio-temporal relations, Photos illustrating the revealed findings and architectural structures, as well as • Drawings/plans depicting the different stages of the excavation process (Lock 2003, p.85). Any such comment has an evident spatial reference as it describes real archaeological objects with specific spatial attributes and relations. The spatial relationships between objects, represent significant archaeological information with particular interpretative significance. The emphasis of excavation practice in systematic recording and spatial documentation can be explained by the fact that the excavation is actually “an experiment that cannot be repeated” (Kotsakis 1998). In this sense, what remains after the completion of excavation process, is a highly detailed archive of observations and related photographic and topographic records, reporting on the undertaken actions-decisions in the field. The synthesis of these archived information, presents the medium through which the archaeologist is attempting to reconstruct an image about the excavation site, in any post excavation analysis (Kotsakis 1998). The task of primary observations analysis and formulation of interpretive conclusions can be expressed through the concept of archaeological interpretation. Archaeological interpretation, is best described as process not subjected to strict rules, which resists any standardization, predicating the free composition of any type of recorded information with reference to the wider archaeological area. It is obvious that such a complex analytical approach cannot be served well through fragmented and time-consuming searching in the pages of analogue excavation diaries and static plans. In summary: • • • • •

the difficulty to re-excavate or re-visit the site, the diversity of archaeological evidence, the complex relationships between archaeological information, the need to link excavation observations with the excavation space and the complex logic of archaeological analysis and interpretation

Geovisualization and Archaeology: supporting Excavation Site Research

87

necessitate the formulation of a workflow for dynamic investigation and analysis of excavation archive through a space-centered organization and representation of the information recorded in excavation diaries.

2- Approach & methods

2.1 Methodological approach Focusing on dynamic cartographic design and implementing techniques to support visual analysis of complex, large in size and heterogeneous information across space and time, geovisualization appears as the most suitable methodological framework for the current effort. However, a review along the relevant literature, indicates a plethora of interrelated theoretical concepts and alternative techniques concerning various design stages. Moreover, clear is the inability of practical implementation and functional exploitation of all these theoretical issues within a solid broader-acceptable geovisualization design theory. For this reason, a mainly conceptual approach, proposed by Howard and MacEachren (1996), is adopted by the current project. The advantage of this approach lies in its generic character, taking seriously into account the specificities of each case study, avoiding in this way any systemization of the design process, and allowing a case study driven formation of the design principles. Howard and MacEachren propose the organization of the whole design process within three successive levels of analysis: • Conceptual level • Operational level and • Implementation level Within the first level - the conceptual, important questions that need to be answered before taking any specific design effort are stressed. These questions can be formulated as follows: • For whom is the system designed?

88

S. Tsipidis et al.

• What special needs are met by the system? • What should be the results of working with this system? • How is the goal reached? The second, operational level, outlines the functions needed to achieve the defined goals. These operations are directed to the spatial, temporal and attribute aspects of data and cover the several tasks of data processing and visualization. The final level of implementation refers to all the elements that constitute the dynamic environment and help the user interact with the system (see also Koussoulakou and Stylianidis, 1999).

2.2 Conceptual level

2.2.1 For whom is the system designed? The current application aims at assisting archaeologists review-revisit the excavation site and its inclusions, inspect their actions in the field, compare, synthesize and analyze the complex archaeological information. In other words, archaeologists constitute the target user group, placed in the centre of the whole design process, with the proposed cartographic environment shaped according to their requirements and analytical needs.

2.2.2 What needs are met by the system? / What should be the results? Although the co-operation between cartography and archaeology has a history of almost three decades, currently expressed through numerous archaeological GIS applications, we could identify some critical issues that prevent the development or restrict the usefulness of current archaeological intra-site GIS approaches. First of all there is an obvious lack of necessary experience regarding this specific research field. With few exceptions, GIS have not been utilized as single unifying platforms, for the effective management, visualization, targeted investigation and analysis of archaeological information (Merlo

Geovisualization and Archaeology: supporting Excavation Site Research

89

2004), and co-essentially, have not been able to adapt their usability to the specific needs of archaeologists - users. The continuous changes in the formation of excavation area caused by excavation process highlight the necessity for realistic visualization of the collected-removed archaeological spatial entities. However, while archaeologists excavate in three dimensional space (real world), in most cases they analyze and interpret the archaeological information via two dimensional plans (Conolly & Lake 2006, p.38). Conversely, a visual overview of the archaeological site through both horizontal and vertical angles can play a significant role towards the better understanding of the complex relationships between archaeological material. Another significant GIS restricting issue is linked to the inability of GIS to effectively incorporate the time dimension. Despite the recent efforts, the functional management of spatiotemporal information is usually undertaken by prototype GIS approaches, where time is a given or a measurable variable, and the temporal values are expressed through definite numeral intervals. In archaeology, the dating of archaeological evidence is not indicated directly at the time of its discovery. This is, on the contrary, determined through archaeological analysis and interpretation process. A number of temporal analysis techniques are applied to any excavation, structuring a wide range of temporal categories with different characteristics that accompany any distinct archaeological evidence. A cartographic environment that seeks to assist archaeological interpretation, is necessary to incorporate dynamically each information reflecting temporal aspects of excavation data. A review on current archaeological GIS systems reveals their weakness to establish a digital medium that supports archaeological interpretive process. Interpretation reasoning can be seen as a reduction from the part to the total, as the understanding of the overall configuration of the site is reached in the final stages of archaeological analysis. In relation to Bertin’s (1967, 1983) reading levels, the archaeological analysis starts posing questions at the local level (e.g. ceramic analysis of a single layer), proceeds by questioning in the intermediate level (e.g. distinguishing typological characteristics among several trench layers) and concludes by posing questions in the general level (determining “phases” of site history).

90

S. Tsipidis et al.

The questions raised in all three reading levels-development stages of the excavation, fall into four general categories: thematic, spatial, hierarchical and the temporal queries.

2.2.3 How is this goal reached? Most of pre-mentioned actions cannot be fulfilled through the use of the current commercial GIS packages as they presuppose a high level of user’s expertise and familiarity with relevant terminology. GIS cannot then be utilized as an effective mean of visualization, analysis and synthesis of archaeological information, if the archaeologist-user is not able to easily navigate and act within its environment. Following this logic, there is an obvious need for customization and simplification of the GIS environment, aiming at the formulation of a suitable archaeological-oriented interface. Based on the previous discussion, a list of important conditions can be formulated as follows: 1. 3D Visualization of the excavation spatial data, 2. Effective presentation of the temporal characteristics and relationships among archaeological evidence 3. Dynamic presentation, correlation and comparison of thematic, spatial and temporal properties of archaeological data assisting the interpretation process 4. Dynamic searching - filtering among archaeological information 5. Implementation of complex queries assisting interpretation reasoning 6. Design of custom tools enabling targeted archaeological stratigraphic analysis 7. Design of a custom archaeologically centered interface enhancing user interaction with the data

Geovisualization and Archaeology: supporting Excavation Site Research

91

3- Results

3.1 Operational level This section mainly concerns the implementation of 3D modeling techniques and specific representation tasks, structuring the basic excavation spatial entities and facilitating spatial information. Before mentioning the applied techniques a short description of excavation spatial entities is given.

3.1.1 Observation Units The archaeological site can be seen as a complex matrix of “sediment”, formatted in the course of time through successive events of natural effects and anthropogenic activities. In the course of excavation this matrix is decomposed into discrete entities (excavation units, finds, samples, features) recording in detail their spatial, temporal and thematic properties. These units of observation are the basic means of organizing the excavation information archive. Following the conceptual framework of the data modeling process as described by Katsianis (2009), it was possible to define the basic spatial entities that correspond to the basic units of observation in an archaeological excavation and allow the access to database information through the mapping environment. These are presented in Figure 1.

3.1.2 Observation Units 3D modeling The geometrical characteristics recorded in the field, the actual shape of the excavation entities and the archaeological analytical approach upon each observation unit class, constitute the factors that affect the classification of each observation unit class within two broad categories: volumetric and non- volumetric objects -phenomena.

92

S. Tsipidis et al.

Figure1: Excavation units of observation: a) excavation trenches, b) excavation unit within a trench, c) an excavation plan, d) artifacts – small finds, e) an archaeological feature

Trench limits are defined at the beginning of the excavation process on the ground surface and the excavation works are taking place within these limits following a vertical formation. In this sense, trenches is useful to be represented as vector objects, defining the limits of the excavation works. Elements of specific archaeological significance, are systematically depicted through excavation plans. Each digital photograph-plan, is photogrametrically transformed and combined resulting to series of photomosaics (Patias et.al 1999). The final ptotomosaic is then embedded in the GIS environment and each depicted entity of archaeological significance is digitized and stored. The next step is the DTM construction using control

Geovisualization and Archaeology: supporting Excavation Site Research

93

points taken during the excavation on the corresponding surface in the field. Excavation units and archaeological features present the most complex excavated objects, as they refer to three-dimensional volumetric objects and their size and shape varies significantly in 3D. The appropriate 3D data type and methods for the effective modeling of volumetric objects is chosen based on the practical needs of the archaeological excavation. The fundamental need arising through the excavation process is: the easy-effortless and rapid process of the very large number of archaeological entities identified during each excavation day, with the lowest cost in terms of time, money and storage size. Comprising the advantages and disadvantages of 3D processing methodologies applied to excavation data as well as the nature of excavation recording methodology (Katsianis et.al. 2008), the B-rep surface model is chosen as the most effective process, achieving realistic representations and avoiding the integration of complex algorithms (3D triangulation) and time-consuming modeling techniques. High experimentation took place in an effort to choose the most appropriate 3D representation method of excavation units, following as mentioned before the B-rep surface model approach (Tsipidis et.al. 2005). This experimentation resulted in the design of a fully automated methodology that fulfills the pre-mentioned conditions. The methodology is based on a programming routine (using MS Visual Basic and ESRI “ArcObjects” library) that applies Delaunay triangulation, constructs multipatch files, that can be treated within the ESRI ArcGIS ® environment (Fig. 2). The primary information used by the proposed method is actually the upper and lower points recorded in the field and presenting the limits and depth of each excavation unit (Figure 3).

Figure 2: Excavation unit 3D modeling

Figure 3: Excavation unit recorded points

94

S. Tsipidis et al.

An alternative representation of the excavation units as unit sections, was decided in order to facilitate stratigraphic analysis. According to this methodology the user-archaeologist defines a custom section plane, with the modeled unit sections presenting the intersection areas between the 3D excavation units and the vertical to them custom plane.

Figure 4: a) Layers depicted on trench sides, b) 3D modeled excavation units, c & d) extracted unit sections

In this way, archaeologists are able to observe the formation and characteristics of excavation units and compare them with the sequence of successive stratigraphic layers, vertically depicted on the sides of each trench (Figure 4). The procedure is again fully automated, using Visual Basic programming and ESRI ArcObjects library.

3.1.3 Modeling 3D excavation Features Archaeological features present the most complex three-dimensional excavation objects, as their shape does not follow any fixed, general form (as in the case of excavation units). For this reason, the methodology developed for the construction of three dimensional excavation units is not applicable for this category of archaeological objects. In order to uncover this problematic issue, the use of SketchUp 3D CAD software was decided. Within this environment, the external faces of each feature are constructed manually, and exported in ArcGIS environment (Figure 5).

Geovisualization and Archaeology: supporting Excavation Site Research

95

Figure 5: a) An archaeological feature and its b) digital model

3.1.4 Representation of the processed data Each excavation spatial object acts as the recovery framework of the thematic and temporal information captured in the field and stored in the database. In this way, the spatial reference of any recorded thematic or temporal information is achieved. The emphasis is now placed on the identification and implementation of representation/rendering tasks that facilitate the visual exploitation of archaeological object differences. The modeled excavation spatial entities are represented in threedimensional digital space in the form of point, linear, surface and volumetric objects. For each type of geographic phenomenon (point, line, surface or volumetric object), particular visual (or retinal) variables (Bertin 1967, 1983) can be utilized optimizing the final visual output and enhancing the visual discrimination of the map elements along their spatial, temporal and thematic differences. Based on this fact, the adoption of the most suitable rendering practices is undertaken in two ways: • The design of special symbols, referring visually and conceptually to the actual - real excavation object • The use of retinal variables to enhance the identification of clusters, relationships and differences between represented excavation objects along their spatial, temporal and thematic properties

96

S. Tsipidis et al.

For the implementation of the pre-mentioned tasks, ESRI ArcGIS® GIS software was used taking advantage of the capabilities offered by the accompanying three-dimensional platform ArcScene®. Additionally, Google SketchUp® software was used especially for the design of the custom symbols embedded in the 3D digital map. Several examples of the current approach are presented below (Figure 6, Figure 7, Figure 8).

Figure 6: Color discrimination based on soil texture

Figure 7: Archaeological terminology driven textures

Geovisualization and Archaeology: supporting Excavation Site Research

97

Figure 8: Some of the symbols referring to specific categories of excavation finds

3.2 Implementation level

3.2.1 Interface overview The proposed interface is designed with Visual Basic programming based on ESRI ArcObjects Library. The environment has the structure of a typical Windows application (Figure 9), consisting of windows and menus.

Figure 9: Overview of interface environment

98

S. Tsipidis et al.

The specific analytical and managerial needs stated at conceptual level, combined with additional considerations emerged through the active participation and technical support of the Paliambela prehistoric excavation program from 2002 to 2010, set the main design principles of the custom archaeological interface. These needs are translated, taking the form of dynamic functions - structural elements of the custom interface described in detail in the following paragraphs.

3.2.2. Interface functions

3.2.2.1 Dynamic search-filtering Archaeological data can be visualized through dynamic searching among the vast amount of information recorded in the database. The user chooses among the available categories of archaeological-observation units and specifies spatial, temporal and thematic constraints (Figure 10), that the data need to follow.

Figure 10: Defining data attributes

The user can handle the visualized information using the table of contents on the left part of the interface.

3.2.2.2 Facilitating navigation - orientation The user can zoom in, zoom out, pan and rotate within the map frame allowing the inspection of the visualized elements from every possible angle.

Geovisualization and Archaeology: supporting Excavation Site Research

99

An overview window indicates the relative position of the visualized data in relation to the broader site area and a 1m north arrow appears within the 3D frame, assisting the orientation and enabling the user to perceive the actual size of the visualized entities (Figure 11).

Figure 11: North arrow and overview window

3.2.2.3 Informing the user about entities attributes By pointing over an object, the identity of the object is displayed on the map along with its coordinates (x, y, z) (Figure 12). By double clicking on an excavation object, the object is highlighted in three-dimensional and two-dimensional map frames, while its detailed properties are displayed within specific interrelated forms including textual information, images, lists of spatially related objects and inclusions (bone, shell, ceramic counts) (Figure 13).

100

S. Tsipidis et al.

Figure 12: Onscreen object selection and inspection of object characteristics

Α

Figure 13: Excavation Unit attributes

3.2.2.4 3D Realistic visualization All displayed information can be classified and visualized according to their thematic and temporal properties through the visualization tool (Figure 14).

Geovisualization and Archaeology: supporting Excavation Site Research

101

Figure 14: Thematic classification: Small finds according to material type and excavation units according to field interpretation type

Each object class is once again displayed and can be separately handled within the table of contents (Figure 15).

Figure15: Deactivation of lithics and ceramic small finds from table of contents

Dynamic graphs can also be activated enabling the selection of an object category through the chart bars and highlighting the selected list of objects within the 2D and 3D maps (Figure 16).

102

S. Tsipidis et al.

Figure 16: Selecting data through a dynamic graph

Figure 17: Presentation of objects temporal aspects and level of dating confidence

An alternative temporal classification can be undertaken through Temporal Graph. Excavation objects are represented as dynamic buttons within a

Geovisualization and Archaeology: supporting Excavation Site Research

103

chart displaying both archaeological time (Neolithic, Byzantine) and excavation-real time (i.e 20/09/2010). Moreover, the level of dating confidence (low – medium – high confidence) is presented (Figure 17). Finally a slider offers the opportunity to define a certain temporal period and select the corresponding objects (Figure 18).

Figure 18: Selection of objects by defining a certain time period

3.2.2.5 Querying the data The user can formulate complex queries along the thematic, temporal and spatial characteristics of the data, resulting on the selection of the corresponding objects within the map frames. In the following figure (Figure 19), the user wants to select bone finds that are within a radius of 0.5m from a defined excavation unit.

104

S. Tsipidis et al.

Figure 19: Defining spatial query parameters

As a result a single bone fragment is identified and highlighted in the map and its properties are thoroughly displayed on the corresponding forms (Figure 20).

Figure 20: A spatial query result

3.2.2.6 Supporting Stratigraphic analysis Specialised tools are implemented in order to facilitate archaeological stratigraphic analysis. The stratigraphic layers appearing on the sides or within each trench can be visually compared with the position of excavated archaeological objects, enabling generic correlations between the individual archaeological entities.

Geovisualization and Archaeology: supporting Excavation Site Research

105

Figure 21: Correlating stratigraphic layers (depicted on trench sides) with excavation units

In the former figures, the correlation between a number of excavation units (Figure 21, d) and a stratigraphic layer depicted on the side trenches (Figure 21, c) is obvious. Based on this example, these excavation units consist a larger unified stratigraphic layer entity. The grouping of excavation entities and the formation of new larger spatial objects is supported through the Create Layer tool. The user selects the entities forming the new layer objects and defines its properties with the help of dynamic charts (Figure 22).

Figure 22: Defining the new layer components and attributes

106

S. Tsipidis et al.

4- Conclusion and discussion Concluding, geo-visualization techniques can be very useful in cases of complex phenomena with evident spatiotemporal diversions and extents. The adoption of participants’ observations and participatory design process is considered critical when the potential users are specialists or analysts. The adoption of these approaches enhances geo-visualization techniques, focusing in the design of flexible, useful and cost effective environments suited to the user specific needs. By participating actively in the Paliambela excavation works since 2002, it was possible to track the real user needs, formulate a vocabulary that could be adopted by the final interface, outline the ways that the archaeologists are accustomed to view and analyze data, and clarify the specific symbolic standards that they use. Based on this experience, the methodologies followed are fully compatible with archaeological excavation methods and adapted to archaeologists’ needs and suggestions. In this way, the basic methodological characteristics remain constant regardless of the case study, with any future design actions to focus on minor customization of the interface components according to each archaeologist - analyst special needs. Archaeological theory and archaeologists’ needs shaped the several aspects of the proposed design process, such as the effective integration of the third spatial dimension. Depth in archaeology (as well as geology), is a concept extremely important in analysis, as it reflects significant temporal aspects and implies temporal relations among spatial data. Within this frame, low cost field techniques and rapid modeling procedures were defined and applied in real time conditions since 2005 at the Paliambela prehistoric excavation, enabling the functional integration of the third spatial dimension both from visual, analytical and practical perspective. Time in archaeology is not defined only according to the relative depth – relative temporal position among the data. Every object is characterized by a set of temporal properties, referring to different “times” with distinct temporal aspects (range, order, duration, uncertainty), with the effective incorporation of all this temporal information to be considered crucial for the analysis. When dealing with such variant temporal information, the design of a single interactive frame (such as the proposed temporal graph) depicting all the possible “times” and temporal variability of the data, is

Geovisualization and Archaeology: supporting Excavation Site Research

107

really helpful for the potential system users. Linking temporal graphs with the actual spatial elements of reference in 3D space is further enhancing the understanding of temporal diversity across space. The overall interface functionality has been thoroughly tested in the last three years, since it is used as the main visualization and analysis tool of Paliambela excavation. According to the archaeologists the interface is acting “as the mediating aspect to knowledge” (Kotsakis 1999), enabling them to proceed in sophisticated correlations and uncover hidden aspects of archaeological material, establish temporal relations of archaeological data, re-evaluate past excavation actions and assist in the formulation of interpretative conclusions regarding site history.

References Bertin J (1967) Sémiologie graphique. Mouton, Paris Bertin J (1983) Semiology of graphics. The university of Wisconsin Press, Madison Conolly J, Lake M (2006) Geographical information systems in archaeology. Cambridge University Press, Cambridge Howard D, MacEachren A (1996) Interface design for geographic visualization: Tools for representing reliability. Cartogr and GIS 23:59-77 Katsianis M (2009) Excavation methodology and information system development for the management of archaeological data. PhD thesis, School of History and Archaeology, Aristotle University of Thessaloniki Katsianis M, Tsipidis S, Kotsakis K, Koussoulakou A (2008) A 3D Digital Workflow for Archaeological Intra-Site Research Using GIS. J Archaeol Sci 35(3): 655-667 Kotsakis K (1998) Excavation. In: Educational Greek Encyclopedia, Ekdotiki Athinon, Athens Kotsakis K (1999) Objects and Narratives. Material culture interpretation in modern Archaeology. Eptakyklos, Athens Koussoulakou A, Stylianidis E (1999) The use of GIS for the visual exploration of archaeological spatio-temporal data. Cartogr and GIS 26(2): 152-160 Lock G (2003) Using Computers in Archaeology: Towards Virtual Pasts. Routledge Merlo S (2004) The “Contemporary Mind”. 3D GIS as a Challenge in Excavation Practice. In: Ausserer KF, Börner W, Goriany M, Karlhuber-Vöckl L (eds.) Enter the Past. The E-Way into the Four Dimensions of Cultural Heritage, Computer Applications and Quantitative Methods in Archaeology (CAA 2003). Archaeopress, Oxford Patias P, Stylianidis E, Tsioukas V, Gemenetzis D (1999) Rapid Photogrammetric survey and GIS documentation of pre-historic excavation sites. In: Proc. XVII CIPA Symposium, Mapping and Preservation for the New Millenium. Olinda Tsipidis S (2009) Geovisualization of Spatiotemporal archaeological data. PhD Thesis, School of Rural and Surveying Engineering, Aristotle University of Thessaloniki Tsipidis S, Koussoulakou A, Kotsakis K (2005) 3D GIS Visualization of Archaeological Excavation Data. In: International Cartographic Conference (ICC 2005). La Coruna Witmore, CL (2004) On multiple fields. Between the material world and media: two cases from the Peloponnesus, Greece. Archaeol Dialogues 11(2): 133–64

Virtual Recreation of the Monroy Roman Villa (Extremadura - Spain) Alan D.J. Atkinson; Jose Juan de Sanjose Blasco; Jorge Cilleros Recuero; Alberto García Martín and Fernando Berenguer Sempere Dep. Expresion Grafica, University of Extremadura, Cáceres, Spain [email protected]

Abstract Recreation by means of virtual network environments facilitates the study, analysis and dissemination of information about our cultural and historical heritage. In the Geomatic Engineering and Urban Heritage Research Group (IGPU) of the University of Extremadura, a virtual recreation of the excavated part of the Monroy Roman villa was made by the following process: (1) historical study and preliminary documentation to define the villa; (2) Collection of current information through a topographical survey of the villa grounds; (3) Generation in ArchiCAD of the volumes of the building; (4) Generation of textures in 3DStudio Max and Photoshop to define the buildings, the furniture and the surroundings of the villa; (5) Generation of a virtual tour video using Adobe Premiere and 3DStudio Max; and (6) Generation of a virtual visit controlled by the user, by means of panoramic flash images, using the software Tourweaver. In this way and with a relatively low cost, we have been able to generate an interactive virtual network environment for the study of the conditions of life in a Roman agricultural villa of the fourth century A.D.

1- Background and objectives The first archaeological work at the Monroy Roman villa (in Cáceres, Extremadura), was undertaken between 1970 and 1973 by the University of Salamanca. Numerous saddles and fragments of pottery were found, and various zones of the “pars urbana” of the villa (the landlord’s or owner’s dwelling) were discovered: the entrance, the peristyle, and some of the rooms and their mosaics. During the following ten years, the Roman villa A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_7, © Springer-Verlag Berlin Heidelberg 2011

109

110

A.D.J. Atkinson et al.

suffered a period of total abandonment. This caused a great deterioration of all the archaeological remains and the disappearance of many of them. Between 1981 and 1985 excavations were resumed under the direction of the Chair of Archaeology of the University of Extremadura, Dr Enrique Cerrillo Martín (Cerrillo et al. 1991, AAEA 2006). During these years the site was restored by removing the rubble from the peristyle, the patio and entrance to the gallery, the residential area and bathing complex, the warehouse on the southern hill, and the western section of the workshops. The ruins that were excavated comprised a substantial area of the Roman villa, 500 metres in length and 100 in breadth, between two hillsides and a brook. Broadly speaking, a Roman villa is a centre dedicated to producing and processing agricultural goods and livestock. It is divided into three zones: the ‘urban’ area, or the residence of the proprietor and his family; the ‘rustic’ area, or warehouses and work spaces; and the production area, for the manufacture of agrarian products (mills, ovens, wool processing, etc.). The buildings of the Monroy Roman villa were raised with almost no ditches for foundations. The present height of the walls does not exceed 50 to 60 cm and their average width is 60 cm (see figure 1). The materials employed in its construction are blocks of white quartz and slate joined with mud or lime and sand. The upper part of the walls was aligned with a layer of lime mortar. These fragile structures were capable of supporting the weight of roofs consisting of wooden beams covered with regula in the proprietor’s section and with vegetation in that of the servants and slaves (Herrera et al. 1989).

Figure 1: Archaeological site and floor plan of the “pars urbana”.

The archaeological excavations have provided the information necessary in order to be able to reconstruct virtually the villa and the probable lifestyle of its inhabitants. The main objective is to generate the most faithful

Virtual Recreation of the Monroy Roman Villa (Extremadura - Spain)

111

3D model possible, based on archaeological studies, which can be used in an automated and/or autonomous manner to show both the architectural environment and the mode of life in the villa in the Roman era (Cilleros and García 2010).

2- Approach & methods The methodology for the work undertaken consisted in six phases: 1. Historical study and preliminary documentation to define the villa. In the preliminary study we relied on the collaboration of Dr Enrique Cerrillo Martín (director of the excavations) for the definition of the principal characteristics of the villa: the construction materials, the estimated heights of the buildings, exterior and interior finishes, decoration, furniture, etc. 2. Collection of current information through a topographical survey of the site. To obtain the topographic plan of the site a Topcon 502E total station was employed and 3D points were recorded on the villa grounds. By means of these points a three-dimensional model was obtained of the site in its current state. Based on this model, a floor plan of the building was generated (see figure 1). 3. Generation in ArchiCAD of the volumes of the building. The floor plan of the building was imported to ArchiCAD and the volumes were generated by inserting the estimated heights in this type of construction (figure 2). These heights were based on the archaeological studies.

Figure 2: Generation of volumes (walls and roofs) by means of ArchiCAD.

112

A.D.J. Atkinson et al.

Figure 3: Final results of the “bump mapping" technique as used for enhancing the realism of a simulated stone floor.

1. Generation of textures in 3DStudio Max and PhotoShop to define the buildings, the furniture, and the surroundings of the villa. The 3D model generated with ArchiCAD was imported to 3DStudio Max. Employing this software together with Photoshop, textures were created which were applied to the building and the furniture of the villa. One of the techniques used was “bump mapping" for simulating bumps and wrinkles on the surface of an object (figure 3). 3DStudio Max was also used for drawing the furniture and objects which complete the 3D model, the environment surrounding the villa (trees, grass, sheep, etc.), and the sky. Finally, 360º spherical panoramic views were created. 2. Generation of a virtual tour video using 3DStudio Max and Adobe Premiere. 14 separate videos of the 3D model, following tour routes, were generated with 3DStudio Max, with a resolution of 60 fps (frames per second). Afterwards, these were joined and edited with Adobe Premiere, and the filters, texts, and soundtrack which had been created for the tour were added. 3. Generation of a virtual visit controlled by the user, by means of panoramic flash images, using the software Tourweaver. 360º panoramic views (created with 3DStudio Max) were imported into Tourweaver and a viewer was generated which permits the user to interact with and navigate through the 3D model of the villa: to look at the villa from 18 different points of view, approach and back away from the model, and go in and out of the rooms by means of links to the panoramic views. Figure 4 shows the map with the locations of the panoramic views and six sample visualizations.

Virtual Recreation of the Monroy Roman Villa (Extremadura - Spain)

113

Figure 4: From left to right: locations of the panoramic views and virtual navigation tools, aerial view, principal entrance, peristyle and patio, tablinum and kitchen.

The methodology used in this project was previously used in part for generating 3D realistic models of the historical buildings of the heritage of Extremadura (Atkinson et al. 2010; Durán and Sanjosé 2008).

3- Results The result obtained is a digital model of the Roman villa, reconstructed on the basis of the ruins and remains encountered in the various archaeological excavations. On the basis of this model it was possible to generate the final products which are available to the public. As a final result, then, two distinct products were obtained: • In the first place, a video was generated following various routes both around the villa and through its interior. This ‘virtual walk’ does not require the user’s interaction, being a video which automatically shows the whole milieu of the Roman villa and its principal buildings: the entrance, peristyle, patio, gallery, kitchen, servants’ quarters, etc. • Secondly, an interactive virtual visit was generated, using 18 panoramic views with a spherical camera, both around and within the Roman villa. This product provides substantial interaction with the user because he decides which building to enter, in which direction to look, and the level of zoom, and hence can approach and back away in an easy, quick, and very intuitive manner. Both products are available on the free-access website of the Geomatic Engineering and Urban Heritage Research Group

114

A.D.J. Atkinson et al.

(http://www1.unex.es/eweb/igpu, IGPU, 2011). Our website is now under construction, but soon there will be a new English and Spanish version.

4- Conclusion and future plans From an archaeological point of view, the work carried out by a multidisciplinary team of archaeologists, cartographers, and structural engineers has permitted the generation of an interactive virtual-reality product of great added value both for the research itself and for the dissemination of results. The work carried out in designing and adapting the textures to the materials found in the construction of the villa was very thorough. Hence a very significant part of the time spent in generating the 3D model was used in adapting all the textures simulated by VR to the walls, roofs, floors, and the materials with which various implements were fabricated in the Roman era. These kinds of product (videos and virtual visits) add great value to the discoveries made on archaeological sites, in terms of their dissemination to museums and historical interpretation centres. The good results achieved permit us to generate new virtual environments based on archaeological and historical studies of different monuments in Extremadura. During 2011 the first results will be available on the web from a regional research project working on an interactive multimedia atlas of various monuments of Extremadura: the Aqueduct of San Lázaro (Mérida), Plaza Santa María, Plaza de las Veletas and Plaza Mayor (Cáceres), the Monroy Roman villa, etc.

Acknowledgments We would like to express our gratitude to the Junta de Extremadura for the financing of the project PRI09A025 (Regional Research Project), and to Dr Enrique Cerrillo (Chair of Archaeology, University of Extremadura) for the historical information.

Virtual Recreation of the Monroy Roman Villa (Extremadura - Spain)

115

References AAEA - Anejos del Archivo Español de Arqueología XXXIX (2006) Villas Tardoantiguas en el Mediterráneo Occidental. Atkinson ADJ, Sanjosé Blasco JJ, Matías Bejarano J (2010) Applied 3D Photogrammetric Studies for the Historical Heritage of Extremadura (Spain). ISPRS Commission V, ISSN 1682-1777, Vol XXXVIII, Part 5, Newcastle, U.K. Cerrillo Martín E, Herrera G, Castillo Castillo J, Hernández López M, Alvarado Gonzalo M, Molano Brías J (1991) Excavaciones arqueológicas en la villa romana de Los Términos, Monroy (Cáceres). Actuaciones y propuestas de futuro. Extremadura arqueológica, Nº. 2, Junta de Extremadura, Mérida, pp. 379-386. Cilleros Recuero J, García Martín A (2010) Levantamiento topográfico y modelado tridimensional de la villa romana de Monroy y paseos virtuales y estudios patológicos de las plazas de San Mateo y de las Veletas del casco histórico de Cáceres. Proyecto Fin de Carrera de Arquitectura Técnica, Escuela Politécnica, Universidad de Extremadura. Durán G, Sanjosé Blasco JJ (2008) Técnicas fotogramétricas aplicadas al Patrimonio. Actuaciones en la ciudad monumental de Cáceres. Cuadernos de Restauración (Revista del Ilustre Colegio Oficial de Doctores y Licenciados en Bellas Artes). Sevilla (Spain), num. 7, pp. 47-56. Herrera G, Domínguez Burrieza FJ, Cerrillo Martín E (1989) Técnicas constructivas en la villa romana de Monroy y ejemplos actuales. Alcántara: revista del Seminario de Estudios Cacereños, ISSN 0210-9859, Nº. 18, pp. 91-110 IGPU – Grupo de Investigación en Ingeniería Geomática y Patrimonio Urbano (2011) Website of the Geomatic Engineering and Urban Heritage Research Group: http://www1.unex.es/eweb/igpu/ Accessed 10 February 2011

SEREDONA: a web platform to integrate historical geographic data into current georeferenced frameworks. Eric Grosso Institut Géographique National, Laboratoire COGIT – Université la Rochelle, L3i 73, avenue de Paris F-94165 Saint-Mandé Cedex [email protected]

Abstract In order to benefit as much as possible from geographical information, users have expressed a strong need to integrate their data with other data. The need to integrate historical data is particularly expressed as it often contains invaluable information which is often unmapped or not represented in current maps or data. Unfortunately, easily accessible data integration treatments are not yet available for users. This paper aims to introduce a solution to solve this problem by explaining the building of a data integration web platform, called SEREDONA, a French acronym for “SErvice de REcalage de DONnées Anciennes” which can be translated as follows: “Web service platform for the adjustment of historical data”. The approach to build this platform, its functionalities and its architecture are fully described. In particular the architecture is based on OGC Web Services (Web Map Service, Web Feature Service and Web Processing Service), Open Source components and the principles of a service oriented architecture. Thus, it allows using the developed components in a more global spatial data infrastructure or separately in other contexts.

1- Introduction Over the last few years the creation and diffusion of geographical information have considerably increased. In order to benefit from this information as much as possible, either to carry out better analysis or to study the temporal evolution of georeferenced data, users have expressed a A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_8, © Springer-Verlag Berlin Heidelberg 2011

117

118

E. Grosso

strong need to couple their data with other data provided both by producers or other users. The need to integrate historical data is particularly expressed as it often contains invaluable information which is often unmapped or not represented in current maps or data. Historical data is of particular interest to ecologists (study of forest evolutions, comparison of ground occupation on various dates, study of climate evolution, etc.) (see e.g. Sanderson and Brown 2007), archaeologists, historians, and also to research scientists who work in the field of simulation (research of evolution rules based on historical data). Moreover, to enable a better exploitation, analysis and processing of the historical data and to give a more meaningful result, this data has to be vectorised. To enable users to couple this data, a data integration process is needed (Parent and Spaccapietra 2000, Sheth and Larson 1990). This integration can be done by integrating all user data into a common frame of reference which is usually the most detailed database available to the institutional geographical data producers. Unfortunately, easily accessible data integration treatments are not yet available for users. This paper aims to introduce a solution to solve this problem by explaining the building of a data integration web platform based on an OGC Web services architecture, called SEREDONA, a French acronym for “SErvice de REcalage de DONnées Anciennes” which can be translated as follows: “Web service platform for the adjustment of historical data”.

2- Background and objectives In order to better understand the different issues of the integration of historical geographic data into current georeferenced frameworks through a web platform, we present a quick overview of work relative to these fields.

2.1 Historical geographic data integration Over the last few decades, historical geographic data has been digitised for archiving and preservation purposes. This phenomenon has consequently contributed to a wider diffusion of this data. Although it is possible to use data in its actual state or digitalised, users realise that this data becomes more useful if its content is directly integrated into a recent frame of

SEREDONA: a web platform to integrate historical geographic data

119

reference, thus enabling that data to be manipulated using a Geographical Information System (GIS). Indeed, the process of integration of historical data into a recent frame of reference increases the possibility of analysis with current data. An integration process is thus essential. Currently, the main method used to integrate this data consists of georeferencing the digitalised image onto the ancient document. Nevertheless, some users such as archaeologists, historians, ecologists or more generally research scientists, begin to vectorise this data in order to obtain better analysis and to improve the study of geographical phenomenon. Currently, several projects use this solution, e.g. the ALPAGE project (Noizet 2009) or a study of the French forest evolution (Dupouey et al. 2007). Moreover, some new solutions about the georeferencing of vector data begin to be proposed in the International Cartographic Association commission on “Digital Technologies in Cartographic Heritage” (Cajthaml 2010, Grosso 2010). These solutions of historical data integration should allow data enrichment (Hampe and Sester 2002), enable the detection and correction of inconsistencies (Egenhofer et al. 1994) and also the evaluation of data quality. Unfortunately, the tools to integrate historical geographic vector data are not yet easily available for users.

2.2 Web geo-services In a context of geographic information, interoperability issues need to be taken into account. Several groups publish geographic norms and standards, notably the International Organisation for Standardisation (ISO) through the technical committee TC/211 relative to geographic information, and the Open Geospatial Consortium (OGC). In our context of building a web platform which has to deal with geographic data, the OGC standards are specifically adapted. Indeed, the OGC standards answer to the issues of diffusion and processing of geographical data through Web geo-services. Firstly, geographical data access services have been introduced through Web Feature Services (WFS) for vector data and through Web Coverage Services (WCS) for raster data. Secondly, to enable data visualisation, the Web Map Service (WMS) standard has been designed. The latter is associated to a Style Layer Descriptor (SLD) and now to a Symbology Encoding (SE) to allow users to control the visual portrayal of the geospatial data. The most well known implementations of all these last OGC standards are MapServer and GeoServer. Then, the Web Processing

120

E. Grosso

Service (WPS) standard has been defined, which aims at providing an open framework to any geo-processing service. Implementations of this standard (WPS 0.4 then WPS 1.0) have been carried out by the 52° North project (Foerster and Stoter 2006). More recent work deals with improving the discovery, use, orchestration and chaining of geo-processing services in order to improve efficiency (Lemmens et al. 2006, Schäffer and Foerster 2008). All the projects provide solutions to build web architectures based on complex geo-processing services. However, they do not propose to specifically develop a data integration service.

2.3 Objectives On one hand, the overview above shows that tools to integrate geographic vector data – notably historical geographic vector data – into current georeferenced frameworks are not yet available but that some solutions exist. On the other hand, there are a lot of standards and implementations in terms of services which are efficient and usable. Therefore, this article has the main objective to propose and describe the architecture of a web platform, SEREDONA, which provides tools to integrate easily historical geographic data into current georeferenced frameworks. This platform has to provide users a solution to visualise and co-visualise their data with both data produced by institutions or national mapping agencies, and provided by other users. Thus, this solution has to provide an online geo-data manager. This platform also has to provide an easy access to geo-processes and finally, has to check the integrity of users’ data, notably to detect the errors. We present now our approach and methods to design the SEREDONA web platform.

3- Approach and methods To assure the best possible interoperability, firstly SEREDONA is built around the OGC standards – mainly those described in the paragraph 2.2 –. Indeed, these standards are now fully recognised and adopted theoretically

SEREDONA: a web platform to integrate historical geographic data

121

and technically by both the geomatics community and the institutions, such as in the european INSPIRE directive. Secondly, the SEREDONA platform is made of separate functional modules. This separation allows the reuse and flexibility of these modules in other GI applications, following one of the main principles of Service Oriented Architecture (SOA). The SEREDONA platform is based on four main axis: data management, data visualisation, data integration and data analysis. These axis are described below.

3.1 Data management: add, delete and edit users’ data To display their geographic data online, users have to use web map displayers such as OpenLayers or OpenScales. These displayers require some specific data formats such as GML, KML or require an access through WMS and WFS. In a context of historical geographic data, these users are historians, archeologists or ecologists, and manipulate mainly geographic data in a ESRI Shapefile format. Consequently, being rarely non specialists in GIS, let-alone in geodata conversion or geodata server installation, they encounter difficulties in displaying their data. The same issues are raised if users want to geo-process their data online. Thus, one of our goals is to allow users to manage their data easily through the SEREDONA platform in providing them tools to dynamically add (from at least a Shapefile format), delete and edit their data on the platform. This online geo-data management requires the creation of users accounts. Therefore, users must register to have an access to the platform. Finally, to allow users to share their data and to collaborate each user can give access to their data (or a part of it) on a read-only basis to another user. Such a management provides to users a full control of their data, control which is often asked in a context of web applications.

122

E. Grosso

3.2 Data visualization: visualize and co-visualise data The online visualisation of historical geographic vector data is not straightforward. Indeed, the spatial and semantic information contained in this data is often difficult to identify after vectorisation. To add the image of the old corresponding map in the background to show the original context could be a possible solution. Nevertheless, due to the fact that the image is too voluminous, it is almost impossible to upload and display it in a readable way. Thus, the platform has to provide users the possibilities to define their own legends and to add on the web map some semantic information contained in the data (e.g. placename). The co-visualisation raises the problem of the compatibility between map projections. Therefore, a geoservice of reprojections is necessary.

3.3 Data integration: integrate and adjust data Users have first to select some of their data in order to integrate it into current georeferenced frameworks. Then, two map displayers have to be shown: the first one contains user data and the second one contains institutional data. The institutional data here are the data provided by the national mapping agency through its national geoportal which contains both images of vector database and maps. Then, users have to select some control points and launch the computation of the georeferencing process. Different levels of users have to be considered for this task, from beginners to experts. If users are beginners, the platform computes several possible spatial transformations: affine transformations, Helmert transformations (four or seven parameters), transformations based on a gravitating model (Langlois 1994), triangulation and rubber sheeting, second (or higher) order polynomial transformations, thin-plate spline method, etc. A table with all the information about transformations (errors on each control point, RMS, RMS associated to cross-validation, transformation grid, map of errors, etc.) is displayed to help the beginners to choose the transformations adapted to their needs. Moreover, some control points can be removed. At the opposite end of the scale, experts can decide to do the computation with only the transformation(s) they choose. The new georeferenced data can be saved through the online geo-data management.

SEREDONA: a web platform to integrate historical geographic data

123

3.4 Data analysis: detect possible data errors The data analysis module allows users to check if their geographic data are valid or not, considering the correctness of geometries, the homogeneity of geometric type (point, linestring, etc.), the topology errors, etc. This analysis process appears to us as crucial. Indeed, if a geo-processing service encounters a problem in a context of service chaining, it implies usually a global error for the entire process, such as described in (Grosso et al. 2009). This analysis is mainly required when historical geographic vector data are produced by non specialists of GI. In the next section we present the results in terms of technical choices and of architecture in response to the approach presented above.

4- Results: the SEREDONA architecture Since many years, all our software developments are realised in the objectoriented language Java. In order to take benefits from all our existing developments, the choice has been made to keep and use this language. Consequently, our architecture has to be built on Java technology. That is why the Apache Struts 2 framework, which is an extensible framework for creating enterprise-ready Java web applications [Struts 2, 2010], is used. The latter allows us to use a Model-View-Controller (MVC) architecture which enables the reuse of the different modules of our architecture in other contexts. At the View level, in order to make accessible the platform to foreigners, the platform is internationalised (French and English languages are provided) based on the “locale” parameter or on the country code of the browser. Based on this framework, we can now present the technical choices of the SEREDONA platform.

4.1 Data management: add, delete and edit users’ data GeoServer is used (http://geoserver.org) to allow users to dynamically load their data on our web platform. Indeed, it is the only server to provide an interface to dynamically load spatial data, thanks to the RESTful

124

E. Grosso

GeoServer extension. To be able to use this extension, many different technical choices are possible but, as recommended by the Geoserver community, the cURL library is the best adapted to this purpose (http://curl.haxx.se). Consequently, we have developed an ad hoc cURL based Java binding adapted to our architecture in order to be able to use the GeoServer RESTful extension. It is thus possible to add a new “workspace” or to delete one, to upload or remove a “data store” (which can be a Shapefile or a PostGIS database), to change a “feature type style”, etc. Consequently, thanks to this RESTful extension, the online users data management is made possible. It is also possible to use and manage data only through their WMS and WFS urls provided by GeoServer – in this way, they become distributed data –. Firstly, we have used the extension with spatial data in an ESRI Shapefile format. This was successful but in order to develop the platform genericity in terms of data format, we also have developed an implementation which is able to use the extension with spatial data loaded in PostGIS databases. In terms of performance, considering the WMS benchmarks of geodata diffusion servers realised every year in the FOSS4G conference (http://wiki.osgeo.org/wiki/FOSS4G_Benchmark), GeoServer provides data at the same speed both from a PostGIS database and from an ESRI Shapefile format. Once this platform functionality has been developed, we created a model to take into account users and their data. Each user has to register to the platform to be able to use it. Once registered, the user can load one or several datasets which contain one or several data – a data represents for example a shapefile –. Data can also be added to an existing dataset. Each data or dataset can be removed, edited (e.g. a name modification) or shared with another user. Each data can be accessible for users through WMS(-C) or WFS(-T) URLs (thus through the mechanisms provided by GeoServer) and not directly on servers in an ESRI Shapefile format or in PostgreSQL/PostGIS databases. The target here is to follow the recommendation of data access provided by (Friis-Christensen et al. 2009), both for users and for geo-processing services. All information about users, datasets and data are stored in a PostgreSQL database and access to this database is based on Hibernate, an object/relational mapping framework for Java, which is supported by the Apache Struts 2 framework. To improve performance, a Data Access Object pattern has been followed and two levels of memory cache have been implemented, using Jboss Cache and MemCached as memory object caching systems.

SEREDONA: a web platform to integrate historical geographic data

125

4.2 Data visualisation: visualise and co-visualise data Users can visualise their data through the OpenLayers, a web map displayer fully compatible with GeoServer. Users can choose one or several data to display. The platform checks if the map projection is the same for all data. If it is not the case, users are invited to select one of the projections which allows to display the corresponding data. Moreover, our platform is also based on the API JavaScript of the French GeoPortal, which is based on Open Layers. This API is used considering the French context of this work. This second displayer enables, still in respect to map projections, to co-visualise institutional and users data. In order to display use of WMS is necessary. Indeed, a huge amount of data can be difficult to display on such web map displayers using WFS or WFS-T. Nevertheless, to improve performance, data are provided through WMS-C, not only through WMS. Finally, to help users in the display of their data, they have the choice to create their own legend through a web interface. This legend is stored in a SLD compatible format and then sent to GeoServer using the same cURL application than the one described in the paragraph above.

4.3 Data integration: integrate and adjust data Users can integrate their data using an interface based on two web map displayers: OpenLayers is used to display users data and the API JavaScript of the French GeoPortal to display institutional data. A JavaScript implementation enables to add control points and to launch the georeferencing process. It is possible to adjust vector data with an affine, a polynomial or a Thin Plate Spline (TPS) transformation. This process is implemented using WPS in the 52°North platform. The latter is a Java project. Consequently, it is chosen in order to reuse our existing Java developments, even if other WPS platforms exist, e.g. PyWPS. Finally, by using a WPS platform, we allow to provide data algorithms through the internet, which implies that users can have access to WPS geoprocessing treatments through a web browser, not only though a desktop GIS, e.g. using uDig (http://udig.refractions.net/). To improve performance, we follow once again the recommendations of data access provided by (Friis-Christensen et al. 2009), by sending to WPS

126

E. Grosso

treatments the WFS URLs of data and not directly data in a GML format. The data integration module also proposes to help users during this process. Many web pages are dynamically created to provide users with all information they need to make the choice of the transformation adapted to their problem. For each transformation, results from the WPS process are shown, e.g. a table of residual errors or the global RMS. Charts of residual errors are also dynamically created via WPS in order to have a general point of view about errors, using JFreeChart (http://www.jfree.org/jfreechart). Finally, another WPS process creates different maps associated with the transformations: punctual residual errors, grid of transformation, interpolated errors map, etc. These maps are created thanks to our own algorithms or the well-known R software (http://www.r-project.org). For the latter, a specific implementation has been done to bind Java results and R through a WPS.

4.4 Data analysis: detect possible data errors The data analysis module is also implemented through WPS. It is based on the JTS – Java Topology Suite – library (http://www.vividsolutions.com/jts). JTS enables to check if a set of geographic data is valid or not at geometric and topologic levels, based on the fact that JTS conforms to the Simple Features Specification for SQL (OGC standard). The data analysis module creates dynamically web pages with the results of this validation. These results are presented thanks to statistics and maps of validation errors. The architecture of the SEREDONA platform is based only on Open Source components in order to be able to reuse a part of this architecture in other contexts.

5- Conclusion and future plans This paper introduced a solution to the problem of historical geographic data integration by explaining the building of the SEREDONA web platform based on an OGC Web services architecture. The functionalities have been presented, showing the different modules of the platform: an

SEREDONA: a web platform to integrate historical geographic data

127

online geographic data management module, a spatial data visualisation and co-visualisation module, a data integration module, and a data analysis module. The technical building of these modules raised many problems. These problems are solved thanks to the architecture and the technical choices made. The architecture is based on OGC Web Services, Open Source components and the principles of a service oriented architecture. Thus, it allows using the developed components in a more global spatial data infrastructure or separately in other contexts. At some points, it allows also for example that the SEREDONA platform could be used in the architecture of services designed in INSPIRE, as it is fully interoperable with the latter. Finally, the SEREDONA web platform is still under development and should be complete at the end of February 2011, and we intend to add more functions in the near future.

References Apache Struts 2, Website of the project: http://struts.apache.org/2.x/, Accessed 14 February 2011. Cajthaml J (2010) Digital technologies in analysis of the Müller's map of Bohemia, 5th ICA International Workshop on Digital Approaches in Cartographic Heritage, 22–24 February, Vienna, Austria. Dupouey J-L, Bachacou J, Cosserat R, Aberdam S, Vallauri D, Chappart G, Corvisier De Villèle M-A (2007) Vers la réalisation d'une carte géoréférencée des forêts anciennes de France. Revue du Comité Français de Cartographie (CFC), Vol. 191, 85-98. Egenhofer M, Clementini E, Di Felice P (1994) Evaluating inconsistencies among multiple representations. In: The Sixth International Symposium on Spatial Data Handling (SDH), Edinburgh, Scotland, pp.901–920. Foerster T, Stoter J (2006) Establishing an OGC Web Processing Service for generalization processes. In: ICA, Workshop on Generalization and Multiple Representation, Portland, USA. Friis-Christensen A, Lucchi R, Lutz M, Ostländer N (2009) Service chaining architectures for applications implementing distributed geographic information processing. International Journal of Geographical Information Science 23(5): 561-580. Grosso E (2010) Integration of historical geographic data into current georeferenced frameworks: A user-centred approach, 5th ICA International Workshop on Digital Approaches in Cartographic Heritage, 22–24 February, Vienna, Austria. Grosso E, Bouju A, Mustière S (2009) Data Integration GeoService: A First Proposed Approach Using Historical Geographic Data. In Proceedings of the 9th international Symposium on Web and Wireless Geographical information Systems (Maynooth, December 07 - 08, 2009). J. D. Carswell, A. S. Fotheringham, and G. Mcardle (eds), Lecture Notes in Computer Science (LNCS), Springer-Verlag, Berlin, Heidelberg, pp.103-119. Hampe M, Sester M (2002) Real-time integration and generalization of spatial data for mobile applications. In: Geowissenschaftliche Mitteilungen, Maps and the Internet, Wien, pp.167– 175. Langlois P (1994) Une transformation élastique du plan basée sur un modèle d'interaction spatiale, applications en géomatique. Journées de la Recherche sur les SIG, GDR 1041

128

E. Grosso

Cassini, INSA Lyon. pp.241–250. Lemmens R, Granell C, Wytzisk A, De By R, Gould M, Van Oosterom P (2006) Semantic and syntactic service descriptions at work in geo-service chaining. In: Proceedings of the 9th AGILE International Conference on Geographic Information Science, Visegrad, Hungary. Noizet H (2009) Les plans d'îlots Vasserot, support d'un système de l'information géographique historique de Paris. In: EAV, La revue de l'école nationale supérieure d'architecture de Versailles, 14, pp.86-95. Parent C, Spaccapietra S (2000) Database Integration: The Key to Data Interoperability. In: Papazoglou M., Spaccapietra S. and Tari Z. (eds), the MIT Press., Advances in ObjectOriented Data Modeling, MIT Press, pp.221-253. Sanderson E W, Brown M (2007) Mannahatta: An ecological first look at the Manhattan landscape prior to Henry Hudson. Northeastern Naturalist, vol. 14, 4, pp.545-570. Schäffer B, Foerster T (2008) A client for distributed geo-processing and workflow design. In: Journal of location based services, Vol. 2 (3), pp.194–210. Sheth A, Larson J (1990) Federated database systems for managing distributed, heterogeneous and autonomous databases. In: ACM Computing Surveys (22:3), pp.183–236.

Digital Processing and 3D Modelling of an 18th Century Scenographic Map of Bologna Gabriele Bitelli, Giorgia Gatta DICAM Dept. - University of Bologna viale Risorgimento 2, 40136 Bologna Italy [email protected], [email protected]

Abstract Metric recovery and digital processing of historical cartography not only allow preserving mapping heritage but also give new possibilities for the use of this information, unachievable using the analogical support. The often poor metric quality of an ancient map is balanced by the value and the richness of its contents and also by the potentialities that the same map can offer in a digital form. In this work some tests of digital elaboration, performed on a perspective map of Bologna (Italy), dating 1702, are reported: georeferencing, analysis of map deformation induced by the adopted transformation technique, three-dimensional modelling and texturization of historical buildings with present images, fusion of data coming from different sources and insertion in Earth-viewer environment.

1- Introduction Digital processing of historical maps together with the possibility to preserve cartographic heritage, can offer new potentialities for using and studying historical information; for instance, historical cartography in digital form can be used for interesting comparison and integration with current data, for urban development studies and change detection procedures, or it can be used in a digital environment to achieve a higher expressivity level of information (Balletti 2006, Bitelli & Gatta 2008, Boutoura & Livieratos 2006, Livieratos 2006). In digital management of historical cartography, the first stage is the conversion of the ancient map in a digital form, employing specific processes A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_9, © Springer-Verlag Berlin Heidelberg 2011

129

130

G. Bitelli and G. Gatta

in order to reduce deformations of the support as much as possible and to guarantee high precision preserving the information content of the map. Usually high resolution scanners (“direct acquisition”) or in some cases 3D surveying methods as photogrammetry or laser scanning (“indirect acquisition”) are adopted. As a rule, the second stage is map georeferencing, to give back to the map its native metric content; selecting ground control points on modern maps is a very delicate operation, because selected points must have not been subject to significant variations; this fact is very important in order to maintain or improve the often limited metric quality of the ancient maps. By means of georeferencing, it is furthermore possible to understand the characteristics of the projection of the historical map, evaluating and representing the degree of deformation induced by that type of cartographic transformation; as a matter of fact, the projection of the ancient map is usually unknown or lacking, and the unavoidable deformation is unpredictable. The map, now converted into a digital form, can be elaborated by modern software tools. The georeferenced map can be superimposed on current base maps, digital terrain models or satellite images, inserted into a GIS environment as a specific layer or processed to obtain a new representation form; in the shown example, a new graphic view is given to a historical perspective map, resembling a 3D city model. Moreover, in this environment, fusion of data coming from different sources with the historical information recorded in the map is possible. The paper reports some experiences of digital processing performed on a 18th century perspective map of Bologna (Italy). Although the main purpose was to derive from the map a 3D model for the historical town and explore some data fusion possibilities, the other mentioned stages were performed and are here presented.

2- The subject of the study The subject of the research is a map of Bologna, dating 1702 (Figure 1); its author is Filippo de’ Gnudi. The map represents the city of Bologna, which in the 18th century was surrounded by walls with twelve gates; the outside of the city is described by the cartographer on the same map.

Digital Processing and 3D Modelling of an 18th Century scenographic Map of Bologna

131

The map is named “Ichnoscenografia” by its author, to point out the scenographic representation as well as its ichnographic value. As a matter of fact, the map offers a scenographic view of Bologna (Casamorata 1947, Comelli 1914, Ricci 1985), providing a “bird’s eye” vision on the city: fronts and buildings façades (churches, town halls or common houses) and other details (e.g. gardens), when visible in perspective view, are very carefully drawn.

Figure 1: A modern remake (L. Confortini) of the Ichnoscenographia of Bologna by Filippo de’ Gnudi (1702). On the right some details of the scenographic view: the church of S. Petronio and the two Towers. The North direction is reported

Moreover, the map has an ichnographic value: it is equipped with a geometric scale in Bolognese perches (1 Bolognese perch = 3.80 m), with a compass-card and various declarations on the realization method. In fact, on one hand the author does not want to break with tradition (until 17th century all maps of the city were scenographic representations), on the other hand he wants to keep up to date (the first ichnographic map of Bologna dates 1692, only 10 years before). The analyzed map represents a transition between the two approaches: it is a scenographic representation aiming to be quite accurate in geometry. It is to be pointed out that the scenographic representation was adopted here for the last time in a map of

132

G. Bitelli and G. Gatta

Bologna; a few years later, in the 1712, the first planimetric map without elevation, derived from the first general topographic survey of the city, with the aim to calculate the overall surface of roads, would have been drawn up, and afterwards technical maps for cadastral applications would have been born. The map orientation is towards the hills (South); the aerial apparent point of view is located North-West of the city, between Porta Galliera and Porta Lame. The drawing of the map appears adapted to the frame. Here the forma urbis of Bologna, the first evidence of which is given us by a Vatican drawn of 1575, is clearly outlined by the walls; it would have remained unchanged until the 20th century. As highlighted by Ricci (1985), for the first time in the cartographic history of Bologna a map clearly was showing the morphological difference between the two areas at the opposite ends of the Decumano: to the East the “Longobard addition” (with its circular shape, because of military reasons), to the West the linear urban development along the radial roads (there is a large number of monasteries along the walls). The difference between East and West was showed also by toponymy: in the eastern side only, place-names have a Longobard origin (e.g. via Alemagna = arimannia). Moreover, there had been differences in the local speech between the two sides until 14th century. The inhabitants of Bologna at the beginning of 18th century were 6500070000. Their number had recently increased, as we can deduce from the map by de’ Gnudi: here many tower and tower-houses are represented, belonging to consolidate households, along with some projections and arches to connect houses belonging to a same owner (this expedient was frequently adopted, although it was forbidden by the municipal corporation). All these particulars can be found because of the map amplitude and its drawing details. Moreover, in the map a great number of religious buildings is visible: in fact at the beginning of 18th century about 1/6 of the landed properties was occupied by churches and monasteries, even though religious population was only 6% of the whole (Ricci 1985). The map by de’ Gnudi shows an high level of technical skilfulness, and the explicit representation of the building fronts by axonometry makes the map very attractive, even for the non-specialists. Nevertheless, despite its geometric scale and its ichnographic value, the representation is not always true: some qualitative tests (see Paragraph 3.3) show in certain cases some enlargement of the streets (for a better visibility of façades?), and maybe a vertical exaggeration too.

Digital Processing and 3D Modelling of an 18th Century scenographic Map of Bologna

133

The map by de’ Gnudi, a sample of which is now preserved at the Archiginnasio Library of Bologna, is a copper etching of about 1 m2 (dimensions: 0,990 x 1,045 m). It derives from the assemblage on canvas of 9 prints from single matrixes; as a matter of fact, seams are clearly visible, and the drawing not always matches from a print to the adjacent one. The examined original map shows some preservation problems of the canvas support: both folds and wear and tear traces (Figure 2).

Figure 2: Details of material and preservation problems of the original map: a) map folding; b) seams due to assemblage of the prints; c) traces of map wear and tear. (Concession by Archiginnasio Library, Bologna)

134

G. Bitelli and G. Gatta

Because of assemblage seams (that would have hindered our work in the 3D reconstruction phase) as well as preservation problems of the original map, for the purpose of the present work a remake of the original map has been elaborated (Figure 1). This map was made by Loreno Confortini (with historical advice by Mario Fanti), and edited by Sabiem in 1985. The copy was realized with a very high accuracy in respect to the original map, nevertheless it is characterized by some differences in comparison to the original: seams are not visible because the drawing is continuous, and a frame is added, in order to localize places by means of letters and numbers, with reference to a list in the right side. A copy by Confortini of the original map was converted in digital form through high resolution scanning.

3- Methods and results

3.1 Geocoding and georeferencing The first stage of map elaboration consisted in geocoding and georeferencing. These well known procedures (Balletti et al. 2000) were conducted in order to assign the scanned map a geodetic datum and a cartographic projection, even though in this case probably no information concerning the use of a cartographic rule in the map drawing was available. The performed steps were: 1. choice of an appropriate geometric transformation (Helmert and robust Helmert, affine, projective, polynomial, and finite elements transformations have been tested) from the image reference system to a cartographic reference system; 2. calculation of the transformation parameters from a set of Ground Control Points (GCPs) of known coordinates in the two reference systems; 3. accuracy evaluation (on the basis of the residuals on the GCPs, as RMS) and eventual re-calculation of the parameters (phase 2) after re-arrangement of type, number and location of GCPs; 4. choice of a resampling method (Nearest Neighbour, Bilinear Interpolation, Cubic Convolution) and generation of the new image (i.e. application of calculated parameters to every pixel of the image); 5. storage of the new image with geocoding and georeferencing metadata.

Digital Processing and 3D Modelling of an 18th Century scenographic Map of Bologna

135

In general, it is possible to derive coordinates of GCPs from a dedicated topographic survey or from another cartographic base of the same area, of good quality and well known features. In historical map management, this stage is commonly very delicate, owing to various reasons: i) the frequent lack of geodetic and cartographic information, or the adoption of unknown geographic reference systems in ancient maps; ii) recognition of unchanged GCPs (on the ground or on a recent map), which is usually a very difficult task because of changes occurred with time. In the peculiar case of this ancient map of Bologna, there are two more reasons: iii) difficulties in identification of homologous points, due to the subjective drawing style of the historical map; iv) the presence of points that cannot be seen because of the perspective view. The only problem that is possible to face, although not solve, is the second one. In this case the GCPs were selected after a careful evaluation - side by side - of data coming from historical research about the town of Bologna and the elements (usually buildings) visible on the ancient map. The analysis was based on descriptive documentation related to single buildings (Bocchi et al. 1998) and maps produced by the historians related to this period. From a first inspection concerning about 300 points, only 130 were thought reliable with a sufficient level of certainty: these points belong to buildings that do not appear to have been subject to significant variations (Figure 3).

Figure 3: An extract of the list (a) of the GCPs selected on the basis of descriptive (b) and graphic (c) documentation (Bocchi et al. 1998) in the ancient map (d), their coordinates deriving from the current numerical municipal map (e)

136

G. Bitelli and G. Gatta

The cartographic coordinates for the GCPs were derived from the current 1:2000 numerical cartographic base of the Municipality (UTM-ED50 system, fuse 32); ortophotos and satellite images have also been considered. After an examination of the transformation results, 50 points were deleted; this way, the final GCPs list is composed by 80 points. Different mathematical transformations were tested: Helmert and robustHelmert, affine, polynomial, finite elements, etc. Some statistical values related to a second order polynomial transformation, performed with a specific software tool, are reported in Table 1. The adopted resampling method was bilinear interpolation. Parameters

Type

Number of GCPs Residuals [m]

Values 80

range, x

0 ÷ 16

range, y

0 ÷ 16

average, x

6

average, y

6

RMS error [m]

range

1 ÷ 19

RMS error [m]

total

10

Table 1: Statistical analysis of the residuals from a 2nd order polynomial transformation

3.2 Study of map deformations A map is a flat representation (in plane coordinates) of a surface that cannot be developed on a plane, so that every map has in itself an intrinsic and inevitable deformation. Frequently, the cartographic transformation adopted to draw up the ancient map is unknown (in some cases a transformation rule does not even exist), thus it is a difficult task to understand if the transformation induces deformations that are acceptable for our modern standards. To investigate the cartographic transformation of the historical map, it is only possible to display the degree of map deformation induced by the type of cartographic realization. The deformations recorded in the map and highlighted by the process not only derive from the type of cartographic transformation, but also from different reasons (i.e. deformation of the analogical support, ancient surveying and drawing techniques, errors related to the used topographical instruments). Therefore, the different sources of deformation are not immediately distinguishable among themselves (Gatta 2010).

Digital Processing and 3D Modelling of an 18th Century scenographic Map of Bologna

137

Regarding the studied map of Bologna, an analysis of cartographic properties was performed by identification and representation of the ancient map deformation, to evaluate the metric quality of the map before the 3D modelling stage. Visualization of map deformations (Jenny et al. 2007) derives from the calculation of the transformation parameters (from a Helmert with Hampel estimator transformation, in this case) between the ancient map and the reference modern map, based on the same GCPs used for the georeferencing stage. Map deformations were visualized by the following tools, as showed in figure 4: • distortion grid (UTM-ED50 cartographic system, mesh size 100 m); • displacement related to each GCP; • isolines showing local scale variation (interval equal to 1:12.5); the mean scale resulted 1:2660, whereas the scale calculated basing on the graphic scale bar was 1:2500. The mean values of scale and rotation angle in respect with the cartographic North are reported in table 2. It must be underlined that in this particular case, in which the analysed map is a remaking of the original one, and it is characterized by some little differences in comparison to the original (see Chapter 2), the process could be not completely right. In fact, the deformations on map which result from the aforementioned study could be not only connected to the type of original ancient cartographic realization, but also connected to the work of modern copying and remaking; and indeed, deformations could have been reduced by the copy itself. Notwithstanding this, the analysis appears interesting, in order to understand if in the map there are areas more critical than others from a metric point of view.

138

G. Bitelli and G. Gatta

Figure 4: Study of deformations: the distortion grid in blue, the used GCPs in yellow (with green displacements vectors magnified x5), scale isolines in red, some scale values in black

Parameters

Type

Number of GCPs Scale factor

Counterclockwise rotation angle [°]

Value 80

range

2750 ÷ 2530

average

2660

calculated basing on the graphic scale

2500 164.4

Table 2: Mean values of scale and rotation angle in respect with the cartographic North

Digital Processing and 3D Modelling of an 18th Century scenographic Map of Bologna

139

3.3 Comparison and integration with current data In order to check in a direct way the georeferencing process, the resampled historical map was inserted into a GIS environment and compared with the current 1:2000 numerical municipal map, from which the GCPs coordinates used in the georeferencing stage were derived: in figure 5a the superimposition of the current map on the ancient map is shown. The resulting discrepancies between the two maps are compatible with the GCPs residuals obtained in the georeferencing stage. Another interesting comparison was performed through superimposition of the georeferenced map on recent high resolution satellite images (Figure 5b), by means of specific software tools. Some further measuring tests were carried out, basing on both the numerical cartographic base of the Municipality used in the georeferencing stage and descriptive documentation related to single buildings. By these tests it was found that the width of many buildings is quite accurate (e.g. the church of S. Petronio), whereas there is a widespread enlargement of streets, a vertical exaggeration of some buildings (e.g. towers) and a shortening of others. Probably, reasons of not always true representation are sometimes a better visibility of façades (enlargement of streets, shortening of buildings) and sometimes exaltation of symbolic buildings of Bologna (e.g. the two Towers). Comparison of historical cartography, in digital form, with current data is not only necessary to check georeferencing, but also interesting for researchers. Georeferenced data inserted into a GIS give us the possibility to ask detailed queries, for instance on single buildings. By the same way, historical data provided by digitized and georeferenced ancient maps, inserted into a HGIS (Historical GIS, a growing application in historical Cultural Heritage environment), allow us to ask queries about elements represented in maps and to achieve information related to that specific historical period.

140

G. Bitelli and G. Gatta

Figure 5: Comparison of the historical georeferenced map with current data: a) the current municipal map (in the box a detail); b) high resolution satellite image

Digital Processing and 3D Modelling of an 18th Century scenographic Map of Bologna

141

3.4 3D model derivation The main purpose of the study was to test the possibility of performing a three-dimensional reconstruction on the digitized map, taking advantage of the particular kind of city representation in this ancient map. The planimetric information (building perimeter) of the historical map and the altimetric one (building height, readable from the scenographic representation of the map) were exploited, in order to derive a simple 3D model of the central part of the city for the epoch. By means of 3D modelling software (developed for scenes and architectural design, but which can be adapted for environmental and territorial applications), the 3D model generation was performed in four steps (Figure 6): • demarcation of the base perimeter of each interesting building (nontrivial operation because of perspective view); • building extrusion, according to the ancient map: each building can be represented by a 3D solid element, starting from its planimetric and altimetric information; • façade rendering: texture mapping on the 3D solid element, with images derived from fronts depicted in the digital map, after their cutting and correcting from distortion due to the native representation; • generation of the complete model (Figure 7).

Figure 6: The steps of 3D model generation (in the example, the church of S. Petronio): a) demarcation of the building base perimeter, b) extrusion, c) texturization, d) generation of the solid

142

G. Bitelli and G. Gatta

Figure 7: Texturized 3D model of a part of the city centre (the region is hooped in orange in the map)

The created model is not really a 3D City Model, but a product maintaining all the descriptive aspects and the information of the original map, associated to a 3D visualization. Notwithstanding this, the test demonstrated that many advantages from the new 3D visualization are added to the historical map: an unusual and more suggestive view, easiness of reading and understanding of the urban environment (even by non-specialists), a great flexibility of use in many applications (virtual flies, re-creation of specific points of view, historical studies, etc.). A digital processing can give new and interesting possibilities to a historical map, allowing integration of map historical information with data coming from different sources, not only historical but also more recent; alphanumerical or descriptive data (e.g. documents related to that period), can be preserved in external archives and linked with the single building into the model. Integration give us additional information, allowing to achieve a higher expressivity level in a new digital multimedia environment. Regarding only the representation aspect, for instance, it is possible to exploit the “modern paradigm” of ancient cartography with a bird’s eye view, using recent oblique aerial images of the same portion of the city taken from different points of view (e.g. the Pictometry® technology,

Digital Processing and 3D Modelling of an 18th Century scenographic Map of Bologna

143

implemented in Bing Maps 3DTM environment, by Microsoft). The result of capturing and mapping these aerial images onto the historical 3D model is an unusual and suggestive three-dimensional product, historical and modern at the same time. The example of figure 8 refers to the texturization of the church of S. Petronio with this kind of images and its insertion in the previous model. It shows one of the advantages of a similar process: to provide to the historical 3D model the information which is lacking in the ancient map, because of the perspective view (in every building two sides are hidden). Of course the technique is suitable only for buildings where historical research has proven the lack of significant variations through time.

Figure 8: Texture mapping of the 3D solid element of the church of S. Petronio from recent oblique aerial images (a), and its insertion in the 3D historical model (b)

3.5 Interfacing with Web-based systems The final aim of our research was the fusion of historical data with modern ones, as well as the fusion of data coming from different sources. For this

144

G. Bitelli and G. Gatta

reason, an interesting use of the historical 3D model was tested: the superimposition of the historical model on the present day situation, by means of Earth Viewers, as Google EarthTM (Figure 9)). A suggestive view of the 3D model was created by georeferencing (in the UTM-WGS84 system) and superimposing it on recent satellite imagery available in Google EarthTM (the database made by high and medium resolution satellite imagery, warped on a SRTM - radar derived - global DTM). Implementation in the Earth Viewer of the DTM allows vision of the 3D model in a threedimensional context. A similar data integration is also possible in WebGIS platforms. Earth Viewers are not at all WebGIS platforms, but, as Web based systems, they adopt the advantage of communication via Web. As a matter of fact, an additional opportunity they offer is to share the obtained model with other users. This is a great advantage, showing that today virtual environment is essential in communication and widespread of cartographic information to specialists as well as non-specialists.

Figure 9: Superimposition of the historical 3D model on recent satellite imagery

Digital Processing and 3D Modelling of an 18th Century scenographic Map of Bologna

145

4- Conclusion Cartographic Heritage is an essential support for the study of city evolution. The modern digital instruments allow new modalities to approach the contents of Cartographic Heritage: recovery, georeferencing and digital management of historical cartography. Moreover, virtual environments can play a basic role in scientific communication or divulgation of cartographic issues to non-specialists. In this work some experiences related to digital processing of a remaking of a 18th century map of Bologna have been described, in order to derive some quantitative considerations after map georeferencing step and to provide a new aspect for the historical information. In particular, the scenographic representation of the ancient map has been exploited to obtain a 3D textured model: the resulting product, more expressive and attractive, simulates the 18th century view of the city, and can be interactively explored. The historical model has been improved on texturization by using modern oblique aerial imagery from different points of view. In addition, the superimposition of the historical 3D model on recent satellite imagery shows by which way digital historical maps can be used in shared image-based GIS systems or in Web-based systems supporting 3D datasets, in order to increase the usefulness and the fruition of these data.

Acknowledgments We would like to thank the Archiginnasio Library of Bologna, for the consultation of the map by Filippo de’ Gnudi and the staff of Gabinetto Disegni e Stampe for their useful information. We also are grateful to G. Mazza, for the 3D modelling. Numerical cartography was realized by the Municipality of Bologna, U.O. SIT.

References Balletti C (2006) Georeference in the analysis of the geometric content of early maps. EPerimetron, 1(1):32-42 Balletti C, Guerra F, Monti C (2000) Analytical methods and new technologies for geometrical analysis and georeferenced visualisation of historical maps. ISPRS XXXII, 6W8/1, Ljubljana

146

G. Bitelli and G. Gatta

Bitelli G, Gatta G (2008) Cartografia storica: valorizzazione e fruizione in ambiente digitale. Geomedia, 12(4):42-44 Bocchi F, De Angelis C, Dondarini R, Greco G, Morigi Govi C, Ortalli J, Preti A, Passatelli G, Tarozzi F (1998) Atlante storico delle città italiane, Emilia-Romagna, Bologna. Grafis, Bologna Boutoura C, Livieratos E (2006) Some fundamentals for the study of the geometry of early maps by comparative methods. E-Perimetron, 1(1):60-70 Casamorata C (1947) Quattro mappe di Bologna del XVII e XVIII secolo esistenti nella cartoteca dell’Istituto Geografico Militare. L’universo, 2:203-213 Comelli GB (1914) Piante e vedute della città di Bologna. Berti, Bologna Gatta G (2010) Valorizzazione di cartografia storica attraverso moderne tecniche geomatiche: recupero metrico, elaborazione e consultazione in ambiente digitale. PhD thesis, University of Bologna Jenny B, Weber A, Hurni L (2007) Visualizing the Planimetric Accuracy of Historical Maps with MapAnalyst. Cartographica, 42(1):89-94 Livieratos E (2006) On the Study of the Geometric Properties of Historical Cartographic Representations. Cartographica, 41(2):165-175 Ricci G (1985) Le città nella storia d’Italia: Bologna. Laterza, Bari

Web Services and Historical Cadastral Maps: the first Step in the Implementation of the Web C.A.R.T.E. System Brovelli M. A.; Minghini M.; Valentini L. Politecnico di Milano, DIIAR, Polo Regionale di Como via Valleggio 11, 22100, Como, Italy [email protected], marco.minghini, luana.valentini)@mail.polimi.it

Abstract

The State Archive of Como in Northern Italy has begun a process of digitization of its map heritage, with the support of interested municipalities. In its archive approximately 15000 historical cadastral maps of Como province are preserved: the Theresian cadastral maps (1718-1722), the “Lombardo-Veneto” ones (1854-1858), the “updates” (1898) and for some locations the maps of 1905. Besides their artistic value, those maps describe with great accuracy the status of the territory and therefore they are a valuable tool for scholars and professionals working on this area. After the digitization step, historical maps must be reprojected in the actual Italian reference system and warped to make them overlap as much as possible the recent maps (used as references). In the paper some methods to fulfil this task are discussed, describing all the related problems and the solution proposed. Maps must also be accompanied by metadata, conformed with the Italian standard; in the project metadata are managed by a Web Catalogue Service and therefore they are searchable on the Internet. Simultaneously in recent years Internet GIS tools, for instance Web Mapping Services, have been developed to serve distributed maps and geographic information. In the prototype of the system, named Web C.A.R.T.E. (Catalogo e Archivio delle Rappresentazioni del Territorio e delle sue Evoluzioni) and presented in the paper, such a kind of solution is proposed: the publication interface uses a map server and an ad hoc client that allows users to quickly and easily interact with the maps.

A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 147 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_10, © Springer-Verlag Berlin Heidelberg 2011

148

M.A. Brovelli et al.

1- Introduction The State Archive of Como owns a considerable amount of historical maps of 246 municipalities belonging both to Como and Lecco districts (Italy). In agreement with the interested municipalities, the Archive has started a digitization process of those maps. Reproductions of maps will be made available to the public in the usual manner prescribed by the legislation in force at the State Archive, office of the Ministry of Heritage and Culture. The available maps belong to different cadastral series: • • • •

Theresian cadastral maps (1718 – 1722); Lombardo-Veneto cadastre (1854 – 1858); updates (1898); 1905 maps (only for some places).

In 1718 Emperor Carl VI appointed the Census Council and from 1721 to 1723 maps have been surveyed. This cadastre came into force from 1760, under Maria Teresa. Maps belonging to this cadastral set are not suitable for building descriptions because they were drawn in order to describe and measure only lands. In other words, this cadastre is not geometric but descriptive with respect to building parcels. In fact, in these maps buildings are represented without details, such as the subdivision in parcels. Only in the walled city of Como we can find the representation of inner porches and other details. The map scale is 1:2000. The following cadastral survey of Como district, the so called “LombardoVeneto”, started in 1854 and, differently from the Theresian register, it can be considered a true geometric cadastre also with respect to buildings at the scale of 1:2000 (with some attachments at 1:1000 or 1:500 for details). Further updates have been done afterwards (1898) for the Buildings Cadastre of Italian Kingdom. Some maps of the 1905 Lands Cadastre are also available. Besides their artistic value, these maps describe with significant accuracy the territory that they represent, turning out to be precious instruments for scholars and professionals working on this area (e.g. for urban planning or restoration plans). To enhance this important cartographic heritage the project ‘Web C.A.R.T.E.’ (Web Catalogo e Archivio delle Rappresentazioni del Territorio e delle sue Evoluzioni – Web Catalogue and Archive of the Territory

Web Services and Historical Cadastral Maps

149

and its Evolution Representations) has been started, with the support of the ‘Fondazione Provinciale della Comunità Comasca Onlus’ (District Foundation of Como Community Onlus). With the dawn of the Geoweb2.0 in the cartographic field, meant as the set of whole the web applications that allow users to query and interact with maps, it becomes possible to integrate and merge different data, such as ortophotos and maps provided by companies, public administrations or services as Yahoo or Google. One of the most significant examples in the current context of geocatalogues is FAO GeoNetwork, that provides Internet access to interactive maps, satellite imagery and related spatial databases maintained by FAO and its partners. All the maps are supplied with the related metadata to identify the source, the date of realization, the reference system, the use constraints and all the other important information associated to the maps. Concerning historical maps, in recent years many historical archives, libraries or cartographic collections have started to put their imagery available on the Internet. Among them, some have just created a catalogue with the available digitized maps, such as the ‘Institut Cartogràfic de Catalunya’, focused mainly on Spanish maps, and the ‘Archivio Storico Capitolino’ (Rome Historical Archive). Some others have built up a really web geo-catalogue service, with the maps projected in an actual reference system, like for instance the ‘David Rumsey Map Collection’, with over 22000 maps and images online focused on rare 18th and 19th centuries maps and other cartographic materials, or the ‘National Library of Scotland’, that provides the Ordnance Survey large scale Scottish town plans (1847-1895). In this paper all the processing phases to build up a web geo-catalogue will be described. The first preliminary step is the digitization of the maps. Nowadays this operation is raised by the onward equipment of proper hardware for papery data digitization by archives and historical collections and techniques are already well known (Fleet 2009). The following preprocessing step consists in the georeferencing and warping of digital images in order to obtain a map that overlaps with the recent cartography, projected in the Italian reference system. Those procedures will be shown in the following section: ‘Preprocessing: georeferencing and warping’.

150

M.A. Brovelli et al.

Besides the georeferencing step, another preprocessing phase is needed. All the maps, in fact, have to be supplied with all the related documents in terms of metadata, that can identify and describe the content of displayed geographic data. This is useful for the on-line identification, searching and management of data themselves. In the section: ‘Preprocessing: web geocatalogue’ the software used and the implementation done will be described. The last step consists in the publication of the maps using a proper webGIS, that needs both a cartographic server and a client. This phase will be described in the section: ‘Web technologies’. This project is still at an early stage: a summary of the first results and future development will be presented in section: ‘Conclusions’.

2- Preprocessing: georeferencing and warping As mentioned before, maps need a pre-elaboration step in order to place them in the actual reference system. It is necessary, in fact, to pass from image coordinates to actual ground coordinates. In practice, the known coordinates in the reference system that we want to adopt have to be associated to points that are well recognizable on the old map. This step can be performed using tools already implemented in several GIS software, such as ArcGIS (Georeferencing Tool), OpenJump, GRASS and GDAL (gdal_translate and gdalwarp), or specific software, like for example PCI Geomatica OrthoEngine, an image processing tool designed to handle images from standard aerial, digital and satellite sensors. Before starting with the georeferencing operation, it is useful to take into account that most of the cadastral maps are available divided in sheets, pieces of the same cartographic map that have been surveyed and produced independently one from the others. As a consequence, besides differences in terms of preservation status, sheets are not perfectly matching and generally streets, municipalities boundaries and buildings shapes have a bad coincidence in adjacent maps. Moreover there are sheets representing areas in which the urban configuration has totally changed over time, so it would be very difficult to identify points of actual known coordinates in these regions. Therefore a choice has been done: instead of georeferencing each

Web Services and Historical Cadastral Maps

151

sheet separately, a complete map for each series has been obtained by unifying them in a proper way; the georeferencing procedure has then been applied to the entire map. To produce a single map starting from sheets, it is sufficient to apply to each of them a roto-translation and, when necessary, a scale variation: details in the deformed sheets are then more likely to coincide each other. These operations can be done using any digital image processing software that implements them, like Adobe Photoshop. Once the complete digital cadastral maps have been created, it is possible to proceed to the real georeferencing procedure. This operation has been tested both in a common GIS package like ArcGIS and in PCI OrthoEngine. The algorithms used by ArcGIS Georeferencing Tool are the polynomial (from 1st to 3rd order), the splines and the adjustment (a combination of a polynomial transformation and a TIN interpolation). To georeference a raster dataset the user must have an existing spatial data (target data), such as a vector feature class, in the desired map coordinate system. The process involves identifying a series of ground control points (GCPs) with known image and ground coordinates that link locations on the raster dataset with locations in the spatially referenced data (target data). Control points are locations that can be accurately identified on the raster map and in real-world coordinates. There are many different types of features that can be used as identifiable locations, such as road or stream intersections, the corner of an established field, street corners, or the intersection of two hedgerows. Control points are used to build the transformation that will convert the raster dataset from its existing location to the spatially correct location. As a general rule, GCPs should be well spread out over the entire raster dataset and not concentrated in one area. In this case it is not possible to insert check points (CPs), points of known coordinates that are not included in the computation of the model but used to check the result of the transformation (to appreciate the difference between known coordinates and the ones calculated with the interpolation model). At the end the user can check the quality of the process looking at the final table that reports GCPs calculated coordinates and residuals. In Geomatica PCI OrthoEngine the interpolation methods implemented are: polynomial (of user-specified order, from 1st to 5th), rational functions and thin plate spline. In this case it is not compulsory to have a georeferenced vector file, but just points of known coordinates. Loaded the input image, the user can identify with great precision GCPs or CPs location simply by clicking on them or writing their pixel coordinates in the apposite

152

M.A. Brovelli et al.

forms. True coordinates can instead be derived with the same collimation approach seen in ArcGIS from a vector file, a raster dataset or an already georeferenced image, but also from a text file that has been previously prepared (for example if the coordinates come from a GPS survey). Also in this case the user can check the residuals in a table. Since it has been seen, by examining the RMSs of residuals, that ArcGIS does not improve the quality of the output and furthermore PCI OrthoEngine allows also the cross-validation of the adopted model, for the Web C.A.R.T.E. project the use of PCI OrthoEngine has been preferred and about a hundred well-spread points has been identified in each map. These points have been derived from Como digital cartography at the scale 1:2000 identifying most of them as building corners quite unchanged over time. According to the normal practice, about the 80% of the points has been treated as GCPs, while the remaining 20% has been used as CPs. Using the same subdivision between GCPs and CPs for each map, the performances of polynomial models having different-order (till the maximum allowed, the 5th) and thin plate spline have been evaluated and compared. The accuracy of polynomial models, that make use of the selected GCPs to build a least-squares transformation, proved to be generally increasing at the increment of polynomial order. Furthermore the behaviour of polynomial models on the CPs, expressed by the RMS (Root Mean Square) index evaluated on the transformation residuals, revealed to be always better than the thin plate spline. In fact, this is an exact interpolation method, i.e. the interpolating surface must pass through all the points on which the model is built (GCPs). The effect is clearly visible in the following figure, referred to the map for which the biggest differences between the two kinds of model have been detected: thin plate spline cause inadmissible local effects in regions with no GCPs, while polynomial models tend to filter and therefore to generate a smooth resulting surface.

Web Services and Historical Cadastral Maps

153

Figure 1: Details of a topographic map of the walled city of Como (end of XVII century) georeferenced using thin plate spline (on the left) and a 5th order polynomial (on the right)

For each map a statistical Fisher test has then been performed to compare the residuals on CPs obtained with different order polynomials, in order to evaluate if each polynomial of increased order is able to explain the data in a way that is significantly better than the previous order one. The results show that, using the typical significance level of 5%, it is always sufficient a 1st or 2nd order polynomial model to explain data in a statistically significant way. In other words, polynomials of higher order, even if they generally give smaller residuals on CPs, don’t decisively improve the global interpolation given by the model, but are responsible only for local behaviours. Finally also some maps of residuals have been generated using Matlab: as in the example below, GCPs and CPs are indicated directly on the cadastral map; from each CP starts an arrow that shows the direction of the residual (difference between real and estimated positions of that point) and has a length proportional to its module. This representation allows to easily understand where the most inaccurate areas of the maps are located, and makes it possible to predict if inaccuracies are due to low surveying quality of the maps themselves, heavy urban changes occurred over time or a scarce correspondence between adjacent sheets.

154

M.A. Brovelli et al.

Figure 2: Residuals in the walled city of Como Theresian cadastral map

3- Preprocessing: web geo-catalogue The usage of geographic data in decision-making processes related to territory governance has been significantly increased in recent years, and many are the Institutions and Public Agencies involved in their production and management. This raised the need to develop instruments able to ensure data research, accessibility, exchange and to know about their use restrictions. In other words, different systems handling geographic data must be interoperable with each other: this goal cannot be reached without data

Web Services and Historical Cadastral Maps

155

documentation in terms of metadata (literally the “data over data”), which are a collection of information to identify and describe a set of objects. Metadata allow to unequivocally detect the needed document and get information about its spatial, temporal and qualitative content and its availability. The diffusion of metadata usage in cartographic field has received a great impulse by Spatial Data Infrastructures (SDI), that are structures consisting of institutional agreements, technologies, data and people able to promote the sharing and the efficient use of geographic information among all their components. To implement a good SDI, metadata must be organized in a well-defined scheme that conforms to national and international standards. The draft of meta-information is currently regulated by the ISO TC211 (International Organization for Standardization - Technical Committee 211) with the standard ISO 19115 Geographic Information – Metadata. This work aims to establish a structured set of specifications in whatever regards methods, tools and services for geographic data management (including their definition and description) and the way to acquire, process, analyze, access, present and transfer such data between different users, systems and locations. The standard ISO 19115 has been accepted at the European level by the CEN TC87 (European Committee for Standardization – Technical Committee 87) and at the Italian level by the norm UNI EN ISO 19115:2005. The CNIPA (Centro Nazionale per l’Informatica nella Pubblica Amministrazione – National Centre for Informatics in Public Administration), named DigitPA (Digital Public Administration) after 2009, referred to the standard ISO to create its own standard, published in 2006 and consisting in the guidelines to create the RNDT (Repertorio Nazionale dei Dati Territoriali – National Inventory of Territorial Data). This instrument has been established by Art. 59 of Digital Administration Code and aims to define a new metadata standard to promote sharing, exchange and accessibility of geographic information at a national level. According to the specifications contained in the European Directive INSPIRE (INfrastructures for SPatial InfoRmation in Europe), the RNDT implements the standard ISO 19115:2003 Geographic Information – Metadata by defining a minimum set of metadata, the so-called ‘Core Metadata’, that has to be used for all kinds of geographic data regarding Italian Public Administrations. The historical cadastral maps used in the Web C.A.R.T.E. project have therefore been catalogued, in terms of metadata, according to the CNIPA guidelines (basically remained the same after DigitPA update of April 2009). The RNDT standard conceives a hierarchical structure to define

156

M.A. Brovelli et al.

every set of data: in fact the metadata can be applied at the different levels of datasets, series (aggregations of datasets that share some properties) and sections (subsets of the same dataset that again have certain characteristics in common). The metadata related to datasets that depend on the same series are partially shared and inherited from the series itself; similarly the common information of more sections derives from their same dataset. Actually an exact definition of dataset (and, as a consequence, of series and section) does not exist: it depends on the type of data that has to be described, on the institutional environment in which the data is produced or on the way it is provided and managed. Nevertheless it’s possible to generalize, saying that series are complete products, datasets are distinct entities which constitute a series of data, while sections represent the smallest units in which a product is provided. This definition can thus be adapted to document the cartographic production of Como and Lecco studied in the project: particularly the series have been identified as the single historical cadastral sets (like the Theresian one, the Lombardo-Veneto one and so on), the datasets of a certain series as the administrative regions represented in the maps (like the walled city of Como or other municipalities) and the sections of the same dataset as the cadastral sheets in which the corresponding map is divided. According to the CNIPA guidelines, the ‘Core Metadata’ created for the available maps are grouped in the following fields: • information about metadata: it includes the univocal identification of the current metadata file and that of the higher rank one (if present), the language and date of metadata, the name and version of the reference standard and the accessibility and usage constrains; • information about data identification: it embraces all the characteristics of the treated geographic data such as the title, date, language, characters set and type of data, their identification and description, the keywords and the Thesaurus they belong to, the person or agency responsible for the data and the information for contacting him/it, and finally the geographic localization of data (minimum and maximum latitude, longitude and height); • information about data constrains: it expresses if there are, and what exactly they are, possible limitations on the usage of data or constrains on their accessibility and fruition; • information about data quality: it gives the positional accuracy of the map; • information about data reference system: it consists in the indication of the map spatial reference system;

Web Services and Historical Cadastral Maps

157

• information about data origin and production process: in this field it’s possible to explain where data come from (the State Archive of Como) and which are the georeferencing processes previously executed; • information about data distribution: it contains the distribution format (name and version) of data, the indications (name and contact) of the data distributor and, if present, the on-line resource in which data are published. The final operation in this step of adding metadata to georeferenced cadastral maps is to make them (thought now as complete geographic data correlated with metadata) searchable, accessible and manageable by users. In other words there’s the need to implement a software, named a web geocatalogue, which allows scholars, professionals and generic users to quickly detect on the Internet information regarding the maps, their spatial extent, their quality, where they are located, their availability and the possible usage restrictions, who is the responsible, etc. A geo-catalogue is therefore a computer instrument that implements functions of storage, management and consultation of geographic data. Different software can satisfy these standardized cataloguing operations: among them, the choice has been done to use GeoNetwork Open Source, originally developed by FAO (2001, first version in 2003) to exploit the potential of the Internet for sharing maps, improving access to cartographic data, archiving them, knowing their origin, owner and manager and, after all, for taking better decisions in the environmental field. The interoperability reached by the software has been recently increased thanks to the system of harvesting, that allows users to get information from databases located in different parts of the world returning, as the result of a request, a wide selection of the discovered resources in a single web page. The web geo-catalogue for the Web C.A.R.T.E. project has been built through a cartographic web service realized by the Geomatics Laboratory of Politecnico di Milano, Polo Regionale di Como. The system aims to manage and distribute geographic information and the corresponding metadata: it adapts the original standard of GeoNetwork, which follows different specifications from the Italian ones, to the reference rules contained in the RNDT of CNIPA, in such a way that only the ‘Core Metadata’ can be inserted and searched. An example of the used GeoNetwork interface for metadata editing is given by the following figure.

158

M.A. Brovelli et al.

Figure 3: GeoNetwork interface for the Lombardo-Veneto cadastral map of the walled city of Como

In the initial phase of the project the web geo-catalogue has been developed only in a prototype form; next phase will be its evaluation and judgement by scholars and experts in the historical and cadastral mapping field.

4- Web Technologies The last step consists in the implementation of the webGIS, so that the final user can visualize historical data. Digitized maps, georeferenced and provided with their metadata, can be distributed on line using a cartographic server and a client. At this purpose FOSS (Free and Open Source Software) solutions have been chosen, so that it is possible to have access to the source code and adapt the software to user needs. For the server side

Web Services and Historical Cadastral Maps

159

GeoServer and MapServer have been evaluated, both supported by an active community of users and developers. For our purposes both those servers can handle also large raster files thanks to the map tiling procedure and their performances are comparable (Aime 2010). Also for the client side many FOSS solutions are available, in particular the research team has experience with p.mapper (that depends on MapServer), OpenLayers (independent from the server side) with or without MapFish libraries. For this initial testing phase, the system works with MapServer for the server side and p.mapper for the client one (as shown in the following figure), but more suitable solutions to manage large raster maps are not excluded to be adopted in the future.

Figure4: Detail of a Theresian map superimposed to Como orthophoto visualized in the webGIS

The final user of such a system can be an ordinary person, a professional working on this area (e.g. for restoration plans) but also a historical scholar who wants to browse old maps and compare them with the current situation. Therefore, the proposed solution has to be tested by a possible user to understand if it satisfies his/her needs and if it is a user-friendly environment also for ‘non GIS-experts’. In the current p.mapper client it is possible to navigate the maps, to overlap maps belonging to different epochs, changing their transparency in order to compare them, and to visualize as background a recent orthophoto of the area, so that the user can understand how the city evolved in time.

160

M.A. Brovelli et al.

Nowadays also non-experts are familiar with 2D and 3D web viewers, such as Google Earth or Street View, therefore a future improvement will be the possibility to visualize in one side of the screen old and current maps and on the other side a 3D GIS visualizer, in order to improve user’s perception of the interested area. This solution will be developed using OpenLayers on the client side, with the support of GeoExt libraries.

5- Conclusions The main characteristics of the system named Web C.A.R.T.E. (Catalogo e Archivio delle Rappresentazioni del Territorio e delle sue Evoluzioni) have been presented in the paper. The project, financed by the ‘Fondazione Provinciale della Comunità Comasca Onlus’, aims to develop an on-line cadastral archive to increase and exploit the historical and cultural value of Como cartographic heritage. The maps, provided in digital form by the State Archive of Como, have first to be geometrically corrected and georeferenced to evolve into products that are comparable with the current cartography. The creation of a geo-catalogue makes their metadata searchable and available on the web, giving an important benefit to any user interested in data content and usage. Finally an ad hoc webGIS allows users to easily interact with the maps and derive useful information by visualizing them together with nowadays cartographic products. The system is still at an early stage and it will need to be deeply tested and evaluated by the users for which it has been thought (common non-experts GIS users but also professionals and scholars in the cadastral and historical field), in such a way that all the needed procedures and technologies can be chosen to obtain the best possible performances.

Acknowledgments This research was partially supported by grants of the ‘Fondazione Provinciale della Comunità Comasca Onlus’ in the frame of the project Web C.A.R.T.E. (Catalogo e Archivio delle Rappresentazioni del Territorio e delle sue Evoluzioni).

Web Services and Historical Cadastral Maps

161

References Aime A (2010) WMS Benchmarking. Presentation adapted and updated for FOSS4G Barcelona CNIPA - Segreteria Tecnica del Comitato tecnico nazionale per il coordinamento informatico dei dati territoriali (2006) Repertorio Nazionale dei Dati Territoriali - Linee guida per l’applicazione dello Standard ISO 19115 Geographic Information – Metadata. http://www.cnipa.gov.it/site/_files/Repertorio_LineeGuida_ISO19115_v03.pdf. Accessed 18 January 2011 Fleet C (2009) The ABC of map digitization. Map Library, National Library of Scotland Gianoncelli M, Della Torre S (1984) Microanalisi di una città: proprietà e uso delle case della Città murata di Como dal Cinquecento all’Ottocento. New Press, Como Henrie D (2009) Ordnance Survey historic town plans of Scotland (1847-1895): Geo-referencing and web delivery with ArcIMS and OpenLayers. e-Perimetron, Vol. 4, No. 1, 73-85 Rumsey D, Williams M (2002) Historical maps in GIS. In: Knowles AK (Ed) Past time, past place: GIS for history, pp.1 – 18. ESRI Press

A Method for the Visual Representation of Historic Multivariate Point Data Alwyn Davidson¹, Colin Arrowsmith¹, and Deb Verhoeven² ¹School of Mathematical and Geospatial Sciences ²School of Applied Communication RMIT University, Melbourne, Australia [email protected]

Abstract The visual representation of multivariate spatial and temporal data is important for interpreting and analysing historical geographic patterns that change over time. The introduction of geospatial technologies in historical scholarship has challenged the suitability of current visual representations due to the need for greater temporal emphasis and the tracking of historical events over time. This research presents a holistic multivariate approach to historical visual representation for point based historical data. The method has been developed through extending the spatial presence in information graphics and through meaningful spatial classification. This paper demonstrates the benefits gained from integrating historical, geographic, temporal, and attribute data through the development of a case study on the history of Melbourne’s cinema venues between 1946 and 1986.

1- Introduction The ability to visually convey historical information is a current issue in historical research. When this historical information is given a geographical context, the complexity of the analysis, interpretation, and subsequent visualisation increases dramatically. Through use of a Historical Geographic Information System (Historical GIS) it has been possible to integrate the disciplines of geography and history. Historical GIS incorporates methods of analysis, data mining, and visual representation of historical spatial data for the aid of historical enquiry, the support of historical

A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 163 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_11, © Springer-Verlag Berlin Heidelberg 2011

164

A. Davidson et al.

scholarship, and the analysis of the geographical significance of historical events. The use of temporal point data within Historical GIS has received increasing interest in many research projects focused on historical events. This interest is evident in the variety of projects that Historical GIS has attracted such as documenting the distribution of witch trials in England in the late 1600s (Ray 2002); challenging the causes of the Great Plains Dust Bowl of the 1930s (Cunfer 2008); and the creation of a European Atlas of Literature (Piatti et al. 2008). The uniqueness of spatial historical data, through its often unconventional spatial data sources and temporal significance, requires an approach to visual representation that is equally unique in its treatment of the data characteristics. This study recognises that there is a need to develop an approach that considers in more depth the treatment of time and space, multivariate analysis and interpretation, and the importance of visual communication. A review of the visualisation techniques adopted for Historical GIS projects highlights the lack of multivariate representation for point feature data. Whilst there are a number of Historical GIS projects that include multiple variables in analysis (Cunfer 2008; Gong and Tiller 2009), there is little attempt to create a multivariate holistic representation for historic investigation. This research aims to do this through extending the spatial dimension of time-series graphs to enhance the explanatory power of graphic displays (Tufte 2001), and through developing a multivariate temporal visualisation through a cartographic technique rather than technology. This paper aims to demonstrate the value of visual representations for historical research in the context of exploring the history of cinema venues in Melbourne, Australia between 1946 and 1986.

2- Background Combining data that is historical, geographic, and thematically changing requires an approach to visual representation that can show the relationships and patterns of all elements concurrently. The relatively new field of Historical GIS has tentatively begun to address the challenges that have arisen from current attempts to visualise the type of data found in historical scholarship. A review of approaches to visual representation in Historical GIS and the challenges that they pose are discussed below.

A Method for the Visual Representation of Historic Multivariate Point Data

165

2.1 Historical GIS The ability to ask geographical questions, to gain new insights from historical investigation, and to present these insights and findings in a way that stresses the geographical context of the research are the aspects that drives the field of Historical GIS. Applying GIS to different historical projects has been the focus of most research in this area, where historical researchers using GIS often refer to it as following a GIS approach (Gregory and Ell 2007). However, the emergence of Historical GIS has given GIS a historical context not only in its application but in its scholarly practice, and supports the claim that it is more than just a collection of methods; it is now being recognised as a subfield within historical studies (Knowles 2008). Historical GIS challenges the traditional mode of communication for historians – narrative communication. As a result, the visual methods used in Historical GIS need to be effective and persuasive in presenting results, supporting evidence, and geographical context so much more so than if applied to disciplines that are familiar with visual communication. A review of the visualisation techniques used in Historical GIS reveals a number of approaches, the main types being: differentiating and ordered visual variables (Cunfer 2008; Donahue 2008); isarithmic (DeBats and Lethbridge 2005), and choropleth statistical surfaces (Healey and Delve 2007); charts and graphs (Gregory 2000); and three-dimensional representations (Shimizu and Fuse 2006). The treatment of time utilised by almost all of the projects reviewed fell into either a snapshot approach, or timeline/time-series approach. Whilst both are very effective in the spatial sciences, the challenges associated with interpreting change through multiple frames (the snapshot approach), and the lack of geographical context for spatial understanding (a timeline/time-series approach) has hampered their suitability for the representation and analysis of historical events.

3- Approach The nature of historical events is challenging to describe, especially in the form of a data structure. Based on a review of the important aspects of historical events at point specific locations, a set of criteria was established for the development of the visualisation, being:

166

1. 2. 3. 4.

A. Davidson et al.

temporal accuracies were to be maintained aggregation of location and attributes were to be avoided the ability to display multiple attributes concurrently and, to represent the temporal dimension in one image

This criteria is hard to satisfy through conventional techniques found in Historical GIS. As a result, visualising historical events requires an approach beyond the univariate and bivariate representations currently provided. Despite advances in multivariate representations, it remains difficult to present a holistic view of multivariate spatial patterns (Guo et al. 2005). This research proposes a new method of visualising multivariate historical data that addresses the challenges and criteria identified above by creating a visualisation that centres on two main concepts; (1) extending the spatial presence in information graphics; and (2) creating a holistic treatment of time. Both of which will be discussed in the subsections below.

3.1 Extending the Spatial Presence in Data Graphics Data Graphics make use of a combination of visual variables such as points, lines, symbols, shading, and colour to visually display measured quantities (Tufte 2001). The works of William Playfair (1759-1823) led the way in the principles for fundamental graphical designs, developing time-series, scatterplots, and multivariate displays. Tufte (2001) argues that the use of time-series displays coupled with the spatial dimension, so that the data are moving over space as well as over time, is especially effective for enhancing the explanatory power of time-series data. Timeseries graphs have been used extensively in Historical GIS to display information, yet most do not convey the data in its geographical context. Instead, their treatment of location or of the spatial dimension is more abstract. By taking an abstracted view of space, this research has extended the timeseries graphic technique to create a more position oriented representation. A simple graph of a chosen variable against a regular division of time was chosen as the platform of the visualisation. The geographical space was then divided into eight cardinal direction grids, of 45 degree segments, originating from a central location of influence (in this case the CBD). Therefore, the classification of direction from the centre was given to a venue based on where their coordinates fell within the direction grid. This produced a framework of eight time-series graphs (Figure 1).

A Method for the Visual Representation of Historic Multivariate Point Data

167

Figure 1: The graphic framework of the combined time-series graphs

3.2 Creating a Holistic Treatment of Time The time-series approach has created a view of time that encompasses the entire temporal dimension. The ability to recognise change over time was one of the main issues for the data, so for the ease of visual interpretation the temporal dimension was dealt with within the one holistic display. Due to the need of maintaining the exact temporal records collected, it was impossible to present the data using the most common method of time treatment in Historical GIS; snapshots, without resorting to hundreds of sequential images to cover the temporal granularity needed. Time-series enabled the integrity of the temporal records to be maintained.

168

A. Davidson et al.

4- Development of the Visualisation The history of cinema venues provides an excellent case study for developing a method of visualising change over time for point feature geographic entities. This is because venues, whilst geographically static, have associated variables that change over time such as seating capacity, change in ownership, and change in screen numbers. In the context of this research, the term cinema venue has been adopted from the work of cinema historian Robert Allen, defining venue as the coming together of physical location, agency, and event (Allen 2008). The use of humanities data within Historical GIS is an area of increasing interest largely due to the recent spatial turn in humanities studies (Jessop 2008). Several authors have attempted to explain this spatial turn in cinema through the term Cinematic Cartography which is characterised loosely by the ways in which cinema and cartography have converged (Roberts 2010), forming a hybrid form of cartography (Caquard and Fraser Taylor 2009). Roberts (2010) provides a typology of the varying thematic areas within cinematic cartography, one of which is Mapping Film Consumption and Production . Historical GIS is particularly prevalent in this field and has been used both as a way of mapping the geographies of film production and consumption (Caquard 2009b), and in the analysis of geographical patterns of cinema operation and influence (Klenotic 2007; Verhoeven et al. 2010). The visual techniques used to communicate the information found in these projects include highlighting cinema venues on historical maps (The University Library 2008), tree graphs of Markov Chain analysis (Verhoeven et al. 2010), differentiating and ordering visual variables (Hallam 2009), choropleth mapping (Verhoeven et al. 2009), and graphs (Caquard 2009a). These methods present a number of challenges for visualising historical data such as: the lack of multivariate representations, the static treatment of time, and the lack of visual representations for interpretation and analysis. The following subsections will attempt to address these challenges in relation to the development of a new visualisation method.

4.1 Case Study: Melbourne’s Cinema Venues The approach discussed above was applied to historic cinema venue data from the Cinema and Audiences in Australia Research Project (CAARP) database for the city of Melbourne, Australia. CAARP holds information

A Method for the Visual Representation of Historic Multivariate Point Data

169

on every known venue screening films (whether it be a public hall or technical college) incorporating opening and closing dates, name, capacity, screen numbers, primary purpose, ownership, management, and spatial location (latitude and longitude). Each of these records is time stamped so changes are easily distinguished. The data was extracted from the online database and organised simply into a time stamped relational database for use in ArcGIS. In this case study, the developed approach was used to visually represent chosen variables concerning cinema venue operation in order to explore, investigate, and interpret multivariate spatial patterns over time. The temporal scale chosen was between 1946 and 1986 and a total of 289 cinema venues operated during this period in Melbourne. There has been very little Australian research into distribution and exhibition practises, especially during the post-war period (Verhoeven 2006), despite it being a time of much cultural change, especially because of the introduction of television in 1956. This research attempts to assist in addressing this deficit.

4.2 Treatment of Space and Time The spatial organisation of each individual venue is based on the variables of distance and direction. This classification method reduces visual complexity in a meaningful way by taking the spatial coordinate values, street addresses and suburbs and transforming them into classes of direction and distance from Melbourne CBD. Therefore, classification is given to a venue based on where their coordinates fall within the distance/direction grid. Distance has been used in classifying historical spatial data in the area of relational graphics (Fyfe and Holdsworth 2009; Gregory 2008). The method has been adopted for forming part of the classification for Melbourne’s cinema venues due to the central location of the city and the influence of distance from the CBD on cinema distribution and exhibition practises. The classes adopted for the visualisation divide the geography of Melbourne into six radial intervals from the CBD and are shown in Table 1, and graphically in Figure 2a. Just as distance is an influential geographical variable for cinema distribution, so to is the directional location of cinemas from the CBD. The classification of eight cardinal directions have been outlined above, and have been combined with the distance classification to create two venue variables

170

A. Davidson et al.

based around the spatial dimension (see Fig. 2). By not strictly mapping individual venue coordinates, the treatment of space is more flexible and can be utilised in much the same way as any other variable, such as seating capacity or ownership. The spatial classification has the advantages of reflecting the geographical nature of cinema operation whilst still maintaining a position oriented view. The time-series approach to time treatment was chosen to produce a holistic representation of time and to accurately track the development of a venue over time. Through this method, it is possible to track every individual cinema venue separately throughout the expanse of the timeline and identify the time of change and the type of change that took place. Distance Class Number

1

2

3

4

5

6

Distance Interval From CBD (Kms)

0–1

1–5

5 – 10

10 – 20 20 – 50 50+

Table 1: Distances adopted for distance classification

Figure 2a : Distance classification

Figure 2b : Direction classification

4.3 From Point to Line After the coordinates are used for the spatial classification of each individual cinema venue, the exact location is no longer required. Instead of representing the venues as points in space, these positions are transformed into lines that rest within the spatial framework. Because of its continual nature, a line can be tracked through a time-series far more effectively than

A Method for the Visual Representation of Historic Multivariate Point Data

171

a point. Another advantage of using lines instead of points for representing the venues is the capacity for greater variability. Just as a point can vary the primary visual variables of size, shape, and colour, so too can a line utilising line width, line style, and colour. But when these graphic elements are transferred onto the graphic space, a linear feature can vary in its shape (through changes in direction and curves) and length where a point cannot. This means greater flexibility in use and also greater capacity for representing multivariate data. By combining these visual variables, it is possible to represent multiple variables in the one display and depict the time in which each of these variables changed. For example, Fig. 3 depicts the history of a cinema venue between 1946 and 1970. The length of the line indicates the lifetime of the cinema, and we can see that it closes in 1967. The central axis shows capacity, and therefore any change in the direction of the curve indicates a change in capacity (a change occurred in 1952, showing a small increase in capacity and then again in 1964, showing a decrease to approximately 750). The colour, line style and line width indicate additional variables; distance, primary purpose, and number of screens respectively. By considering a cinema venue not as a location but as a historical lifetime, it is possible to embrace a more abstracted view of space and achieve a more holistic and multivariate representation for comparison.

Figure 3: The history of a single cinema venue

172

A. Davidson et al.

4.4 Geographic and Temporal Analysis The main functions of analysis and comparison can be achieved in a single display due to combining multiple variables, the spatial dimension, and the entire temporal span. It is possible to analyse the changes and characteristics of individual venues and to compare individual venues to other venues in the same area. Visual comparisons and pattern detection can be made for geographic, attribute, and temporal characteristics. Additional analyses can be achieved by manipulation of the main display through selecting those venues based on attribute and spatial classification. This allows certain aspects to be investigated more thoroughly and relationships between variables to be viewed clearly and simply. Fig. 4 shows a set of data for both the main display (Fig. 4a) and then once a selection has taken place (Fig. 4b). In this case, the user has specified to view only those cinemas with a capacity of less than 1000 in order to simplify the results and make the interpretation of patterns clearer.

Figure 4a: All capacity records for Northern venues Figure 4b: Northern venues with a capacity less than 1000

5- Case Study Results This research project involved developing a visual representation in order to answer a number of geographical questions for the data, exploring the geographical distribution, patterns, and differences for all questions. Such questions included: what is the longevity of cinemas operating in 1946, and how does change affect the longevity of large cinema companies? An initial exploratory analysis was performed to identify those variables that

A Method for the Visual Representation of Historic Multivariate Point Data

173

were important to display for each question and the relevant data was extracted for visualisation. To show the effectiveness of the derived visualisation only one geographical question will be explored in the subsections below; how does change affect the longevity of large cinema companies?

5.1 Development Large cinema companies in Melbourne operated throughout 1946-86. Exploring the effect change has on the longevity of large cinema companies required assigning visual variables to the chosen attributes, these were: 1. 2. 3. 4.

Distance from the CBD – the central axis against the timeline Name of the cinema company – line colour Number of changes between 1946-86 – line width Role of the cinema company – line style

The spatial classification of direction and the temporal dimension is fixed for all questions; time is divided by radial intervals from the centre, and direction is depicted by the cardinal position on the grid. Fig. 5 shows the resulting visual representation of the geographical question. Each line refers to an individual cinema venue and its operation can be traced through time. Many of the venues existed before the study period of 1946 and therefore show their opening date as 1946. Each of the curved lines begins on a grid line to indicate the distance from the CBD (the closer to the central line the closer to the city), and ends on its closing date at the central line. This emphasises the end of the operation of the cinema and also allows for greater visual interpretation when exploring closing dates. The integration of all visual variables presents a single display of all elements concurrently, encouraging multivariate and temporal visual spatial analysis.

5.2 Case Study Findings and Results The resulting visual representation of the changing nature of large cinema companies demonstrates the capacity for thorough visual analysis of multivariate data over space and time. The total number of lines indicates the number of large cinema venues operating during the time period, yet it is easy to see that this number is not uniform over space or time. The more densely populated North and East of the city have a strong presence of

174

A. Davidson et al.

large companies, the majority of which are found within 20km of the city centre. The large number of converging lines between 1956 and 1966 refers to a large number of closures during this period and is found throughout the extent of Melbourne. The sheer number of closures is also evident if we look at the longevity of those cinemas that were operating in 1946; for only a handful of these cinemas lasted past 1986. We can see a cluster of new cinemas opening between 1965 and 1975, occurring after a short break from the peak of closures, distributed mainly in the East and North of the city. The lack of new cinemas opening in the West outside of the CBD is clearly shown. Indeed, there are no large cinema companies operating in this area past 1975. Further associations and findings can be found when we interpret the data in relation to the other variables of change, name, and company role. The dominance of the blue curves is indicative of the dominance of Hoyts cinema venues. The large numbers of Hoyts cinemas are found throughout Melbourne and at varying distances from the CBD. However, the majority of these cinemas do not survive past 1970, and of those that do they experience some form of change indicated by the width of the line. This suggests that a Hoyts cinema operating in 1946 would most likely not last past 1970 unless they made adjustments to their cinema such as adding another screen or decreasing seating capacity. Another interesting association is the comparison between Hoyts and the Greater Union Theatres. Greater Union, represented in red, shows that all venues that existed in 1946 adjusted to the times and had a longer lifespan than the average Hoyts; making a possible association between change and longevity.

A Method for the Visual Representation of Historic Multivariate Point Data

175

Figure 5: The resultant visual representation of the geographical question, how does change affect the longevity of large cinema companies?

If this venue layer is combined with information about the cinema industry or societal events, some direction of explanation may be provided. For example, the introduction of television in Melbourne did not occur until around 1956 and this may go to some lengths to explain the large amount of closures shortly after this, although further data on the uptake of TV consumption would be required to definitively secure this observation. By

176

A. Davidson et al.

creating visual access to the data, historians can think of the information in a geographical context, and analyse the data in a way that is not possible without the introduction of time and space.

6- Summary and Conclusions This research has developed a method for the treatment of historical point data in visual representations. It has sought to extend the use of multivariate representations in Historical GIS through an approach that incorporates a holistic treatment of time and a spatial extension of simple data graphics. The method consists of several components that are combined to produce a final visual representation which includes a spatial extension of time-series graphics and the abstraction of point feature data. These components have been developed to successfully communicate multivariate and temporal spatial information and produce insights into patterns of historical point data. By creating a holistic representation, visual interpretations of spatial patterns for multiple attributes and the entire temporal dimension can be analysed without resorting to comparisons of multiple displays of temporal and attribute snapshots. The history of cinema venues has provided the developed method with a case study to validate the success of the final visual representation. It has created visual access to historical cinema records, allowed relationships between attributes to be explored, and provided a visual representation for the analysis of multivariate and temporal dimensions. It is possible that the success of this application can be applied to almost any historical events for which location and time are recorded, provided they have a central point of influence and a temporal lifespan. Future developments will focus on the automation of the visualisation process for screen display to provide a method that is interactive and question specific. This will allow the analytical capabilities of the method to be fully realised. However, one of the main benefits of creating a holistic visual representation is the ability to work with hardcopy paper representations. Hard copies allow the entirety of the data to be viewed whilst particular areas can be examined. It gives the option of highlighting interesting patterns and differences for later inspection. Usability testing will be required to further the development of the method, focusing particularly on the use for historical investigation. Future research will also

A Method for the Visual Representation of Historic Multivariate Point Data

177

assess the method of its ability to be used in other historical applications, and will attempt to provide a geographical and graphic context to a variety of historical events.

References Allen RC (2008) Going to the Show, The University of North Carolina. http://docsouth.unc.edu/gtts/about-venue.html. Accessed: 21 July 2010 Caquard S (2009a) Foreshadowing Contemporary Digital Cartography: A Historical Reivew of Cinematic Maps in Films. The Cartographic Journal - Cinematic Cartography Special Issue, February 2009 46:46-55 Caquard S (2009b) Reframing the Digital Cartographic Frame: Examples from the Cybercartographic Atlas of Canadian Cinema. Paper presented to: International Cartographic Conference, Santiago, Chile Caquard S, Fraser Taylor DR (2009) Editorial - What is Cinematic Cartography? The Cartographic Journal - Cinematic Cartography Special Issue, February 2009 46:5-8 Cunfer G (2008) Scaling the Dust Bowl. In: Knowles AK (ed) Placing History: How Maps, Spatial Data, and GIS Are Changing Historical Scholarship, ESRI Press, Redlands, California DeBats DA, Lethbridge M (2005) GIS and the City: Nineteenth-Century Residential Patterns. Historical Geography 33:78-98 Donahue B (2008) Mapping Husbandry in Concord: GIS as a Tool for Environmental History. In: Knowles AK (ed) Placing History: How Maps, Spatial Data, and GIS Are Changing Historical Scholarship, ESRI Press, Redlands, California Fyfe DA, Holdsworth DW (2009) Signatures of Commerce in Small-Town Hotel Guest Registers. Social Science History 33:17-45 Gong G, Tiller J (2009) Exploring Vegetation Patterns Along an Undefined Boundary: Eastern Harrison County, Texas, Late Spring, 1838. Social Science Computer Review 27:363-379 Gregory IN (2000) Longitudinal Analysis of Age- and Gender-Specific Migration Patterns in England and Wales: A GIS-Based Approach. Social Science History 24:471-503 Gregory IN (2008) Different Places, Different Stories: Infant Mortality Decline in England and Wales, 1851-1911. Annals of the Association of American Geographers 98:773-794 Gregory IN, Ell P (2007) Historical GIS: Technologies, Methodologies and Scholarship. Cambridge Studies in Historical Geography, Cambridge University Press, New York Guo D, Gahegan M, MacEachren AM, Zhou B (2005) Multivariate Analysis and Geovisualization with an Integrated Geographic Knowledge Discovery Approach. Cartography and Geographic Information Science 32:113-132 Hallam J, and Roberts, L. (2009) Projecting place: Mapping the city in film. Paper presented to: E-science Workshops, 2009 5th IEEE International Conference on, , Oxford Healey RG, Delve J (2007) Integrating GIS and data warehousing in a Web environment: A case study of the US 1880 Census. International Journal of Geographical Information Science 21:603-624 Jessop M (2008) The Inhibition of Geographical Information in Digital Humanities Scholarship. Literary and Linguistic Computing 23:39-50 Klenotic J (2007), “One moving picture show is enough... for a town like Colebrook”: a Comparative Social Geographic Analysis of Movie Exhibition in New Hampshire, 18961920. Paper presented to: “The Glow in their Eyes”. Global Perspective on Film Cultures, Film Exhibition and Cinemagoing, Ghent

178

A. Davidson et al.

Knowles AK (2008) GIS and History. In: Knowles AK (ed) Placing History: How Maps, Spatial Data, and GIS are Changing Historical Scholarship, ESRI Press, Redlands, California Piatti B, Bar H, Reuschel A, Hurni L, Cartwright W (2008), Mapping Literature: Towards a Geography of Fiction. Paper presented to: Cartography and Art - Art and Cartography Conference, Vienna Ray BC (2002) Teaching the Salem Witch Trials. In: Knowles AK (ed) Past Time, Past Place: GIS for History, ESRI Press, Redlands, California Roberts L (2010), Cinematic Cartography: Towards a Spatial Anthropology of the Moving Image. Paper presented to: Mapping, Memory, and the City, Liverpool Shimizu E, Fuse T (2006) A Method for Visualizing the Landscapes of Old-Time Cities Using GIS. In: Okabe A (ed) GIS-based Studies in the Humanities and Social Sciences, CRC Press, NW The University Library (2008) Going to the Show, The University of North Carolina. http://docsouth.unc.edu/gtts/index.html. Accessed: 20 September 2010 Tufte ER (2001) The Visual Display of Quantitative Information. 2nd edn. Graphis Press LLC, Cheshire, Connecticut Verhoeven D (2006) Film and Video. In: Cunningham S, Turner G (eds) Media and Communictions in Australia, Allen & Unwin, Sydney Verhoeven D, Arrowsmith C, Davidson A (2010) The Role of GIS and Spatial Analyses for Understanding Historical Change in the Post-War Film Industry: an Interdisciplinary Case Study Approach. Paper presented to: Mapping the City in Film: A Geo-historical Analysis An Interdisciplinary Conference, University of Liverpool Verhoeven D, Bowles K, Arrowsmith C (2009) Mapping the Movies: reflections on the use of geographical technologies for historical cinema audience research. In: Ross M, Grauer M, Freileben B (eds) Digital Tools in Film Studies, Transcript Verlag, Bielefeld

The Atlas and the Globe of Russian Geographical Explorations and Discoveries of the Earth: Concepts and Contents N.N. Komedchikov, V.M. Kotlyakov, A.G. Khropov, A.A. Medvedev, L.N. Zinchuk Institute of Geography, Russian Academy of Sciences Moscow, Russia [email protected]

Abstract The Institute of Geography of the Russian Academy of Sciences with the assistance of the Russian Geographical Society begun with the compilation of the great three-volume Atlas of Russian geographical explorations and discoveries of the Earth. The first volume is devoted to the Russian geographical explorations and discoveries before 1845, the year of the establishment of the Russian Geographical Society. The expeditions of the Imperial Academy of Sciences, that have contributed significantly to the exploration, description, and mapping of Russia, as well as the circumnavigations of Russian sailors and their discoveries in the World ocean are emphasized in the first volume. The compilation of the Atlas is carried out simultaneously with the creation of a virtual globe representing animated routes of Russian explorers and navigators. The Atlas is provided with many old maps and drawings of artists who participated in expeditions

Concepts and Content of the Atlas and Globe This atlas is being developed as an integrated illustrated cartographical work devoted to the history of geographical discoveries made by Russian explorers during the whole period of Russian nationhood, beginning with A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 179 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_12, © Springer-Verlag Berlin Heidelberg 2011

180

N.N. Komedchikov et al.

the first geographical information which became known in the Ancient Rus’, and ending with the results of systematic task-oriented geographical explorations of dry land and oceans up to the present day. This atlas reflects not only the history of geographical exploration of Russia’ territory on different stages of its existence, but also the contribution of Russian explorers and discoverers to the global process of accumulation of knowledge about all, sometimes even very remote areas of the Earth. It gives the detailed picture of expedition activities of the Russian Academy of Sciences, the Russian Geographical Society, and other Russian institutions and authorities engaged in exploration of Russia and the whole world. The thematic contents of the atlas are given in a chronological order, and inside each chronological group the material is systematized on the basis of either the subject-chronological or the spatial-chronological principle. The subject-chronological principle implies that the material is presented according to the definite themes – military campaigns, embassies, seavoyages, etc. – and then, within each of them, in a chronological order. According to the spatial-chronological principle the material is grouped due to the areas – the European North, Siberia, the Arctic Regions, etc. – and then, within each group, in a chronological order. Such arrangement of the materials inside the atlas allows to present the historical outline of Russian geographical explorations and discoveries of the Earth in the most compact, most complete, and the most logical way. Basing on these principles, and taking into consideration the large volume (more than 1,000 pages) ‘The Atlas of Russian Geographical Explorations and Discoveries of the Earth’ is being developed in the form of three separate issues with the following chronological frames: 1) up to the middle of the 19th century (up to 1845 – the year of the establishment of the Russian Geographical Society); 2) from the middle of the 19th century to the early 20th century (1845– 1917); 3) from the early 20th century to the early 21st century (from 1917 up to the present). In accordance with the atlas title, two basic key and interrelated concepts are its contents highlights: ‘geographical exploration’ and ‘geographical discovery’. The geographical exploration is considered as a process of special study of geographical objects aimed at the discovering of new

The Atlas and the Globe of Russian Geographical Explorations and Discoveries

181

knowledge about them, as a scientific process of obtaining new geographical knowledge of the Earth and the processes and phenomena taking place on it. The concept ‘geographical discovery’ combines the territorial discovery, i.e. the discovery of new geographical objects which are shown on maps or given in special geographical descriptions (inventories) as well as clarification of specific regularities in the development of geographical environment (for example, the discovery of latitudinal zonation law) which were not known before. During different historical stages the character of geographical discoveries changed – from mainly territorial discoveries up to the middle of the 19th century to territorial and theoretical geographical discoveries in the 19th and 20th centuries. The period of territorial discoveries finished in the second half of the 20th century, when all ‘white spots’ discovered from the world map; the period of mainly theoretical geographical discoveries followed. Not every geographical exploration results in a geographical discovery. In the history of geography up to the end of the 17th century there were a lot of accidental geographical discoveries made by pure chance, and only later they began to occur as a result of aimed geographical explorations of new lands by special scientific expeditions. This atlas describes the circumstances and specific details of individual discoveries. Besides the territorial discoveries and explorations, the atlas also describes oceanographical discoveries and explorations made by Russian navigators in the waters of the World Ocean. Special emphasis is made on the Russian priority of some or others discoveries having the world or national importance, and the priority of geographical discovery is attributed to a person who was the first to “find out the given object and give the first substantial information about it, which allow to mark this object on the map of the Earth” (Fradkin 1972, p. 46). The first documentary recording of a geographical discovery on the map or its first verbal description is the main argument in the priority reasoning of some or other discovery. Thereby the demonstration of such maps or of fragments of descriptions made by discoverers are considered to be very important and essential. Just written information, sketches, and maps made by Russian pathfinders and navigators and characterizing the areas of the Northern Eurasia, the Far East, the Central Asia, and NorthWest America provided the basis for the most West-European maps and

182

N.N. Komedchikov et al.

became a constituent part of the world geographical and cartographical knowledge. The Atlas presents the process of the main Russian explorations and discoveries connected with removing of ‘white spots’ from the world maps in the historically consequent, geographically accurate, and cartographically clear way; describes their characteristics and specific historical context of the time they took place, their significance, as well as the role in this process of discoverers and explorers themselves. The route maps themselves as well as encyclopedically short but informative text notes to them accompanied by various illustrations – the pictures of explorers and discoverers, photos of objects (artifacts) connected with them, gravures, pictures, drawings, and photos devoted to these events, old maps, fragments of manuscripts recording some or other explorations and discoveries, etc. – contribute to the most full and comprehensive presentation of the process of explorations and discoveries. All this allows to describe the journeys of Russian discoveries more excitingly, to characterize their achievements and various motives: an interest in new lands, financial well-being, military eagerness, civil duty, political necessity, scientific interests, and simply curiosity. By that the atlas of a new type is created for the first time in the Russian cartography practice, it presents the contribution of Russian explorers into geographical understanding of the Earth in a comprehensive and historically documentary way. The Russian literature on the history of geography and cartography has accumulated a bulk of knowledge concerning Russian explorations and discoveries of the Earth. The works of A.I. Alekseyev, A.I. Andreyev, M.I. Belov, L.S. Berg, M.S. Bodnarskiy, V.Yu. Vize, V.F. Gnucheva, L.A. Goldenberg, V.I. Grekov, V.A. Yesakov, A.V. Yefimov, N.N. Zubov, D.M. Lebedev, I.P. and V.I. Magidovich, M.G. Novlyanskaya, B.P. Polevoy, A.V. Postnikov, N.G. Fradkin, Yu.M. Shokal’skiy, and many others are devoted to comprehensive and detailed investigation of both individual expeditions and the total contribution of the Russians into geographical understanding of the world. The maps of travelers and expeditions routes are presented in many Russian geographical atlases. One of the first cartographical summaries of Russian explorations and discoveries was the album of famous Russian travelers and geographers compiled by the Institute of Geography of the USSR Academy of Sciences with the participation of the USSR Geographical Society which was issued in 1948 (Berg et al. 1948). The album includes route maps, portraits and short biographical information about the travelers. In 1959, a special ‘Atlas

The Atlas and the Globe of Russian Geographical Explorations and Discoveries

183

of the History of Geographical Discoveries and Explorations’ (Salishchev et al 1959) as a study guide for students in their process of geography and cartography historical studies was issued. Another fundamental ‘Atlas of Geographical Discoveries in Siberia and North-West America in the 17th – 18th centuries’ edited by A.V. Yefimov was issued in 1964 (Yefimov 1964); it contains discoverers’ sketches and maps made during their journeys and summarizing maps with archeographic notes and lists of geographical names presented on the maps. The maps of the main sea voyages and expeditions are presented in the second volume of the ‘Sea Atlas’ (Demin 1953) and in four volumes of the ‘Atlas of the Oceans’ (Gorshkov 1974-2005) . The first (2004) and in the fourth (2008) volumes of the ‘National Atlas of Russia’ also contain maps of Russian geographical explorations and discoveries. All these works will be used during the compilation of the new ‘Atlas of Russian geographical explorations and discoveries of the Earth’. Besides that, a lot of documents and materials from the Archive of the Russian Geographical Society, the Russian State Archive of Ancient Acts, the Archive of the Russian Academy of Sciences, the Military History Archive, the Central Naval Archive, as well as from the collections of the State Historical Museum, of manuscripts and maps departments of the Library of the Russian Academy of Sciences, the Russian National Library, and the Russian State Library are used during the compilation of the atlas maps, writing the texts, and the selection of illustrations. In our opinion, foreign cartographical publications present and evaluate Russian geographical discoveries insufficiently. The only exception is the famous English ‘Atlas of exploration’ (1997) having some maps of the expedition routes of Yermak, Poyarkov, Dezhnev, Khabarov, Przheval’skiy, the Great Northern Expedition, and Bellingshausen. The methods of map depiction of geographical discoveries and explorations in the new atlas are mainly the same: color background shows either the political division to end of the period shown on the map or topography (in subdued tones). The use of some or another color background is determined by map contents – embassies, scientific expeditions, sea voyages, etc. The routes themselves are shown by means of symbols of moving (arrows, lines) in different colors. If necessary, the areas covered by detailed studies are presented with the use of different hatchings and solid colors. Geographical names on the maps are given in the form which was common to the end of the period shown on the map. The names given by

184

N.N. Komedchikov et al.

the discoverers and explorers are shown in a special way. For the more convenient use of the atlas it is provided with the index of geographical names including their modern forms. The atlas is prepared with the use of computerized publishing programs which enable to have two products in the end: an electronic (interactive) and a traditional paper (printed) versions of the atlas. The globe of Russian geographical explorations and discoveries is being developed only in the electronic environment in the form of a virtual interactive product – a virtual globe. It is based on a mathematical model of a sphere with a 10° grid and the parameters of the Krasovskiy ellipsoid. The model of the sphere and the system of its management has been developed with the help of the language ActionScript 3. The routes of travelers and expeditions are presented on the globe by means of animation images (2D, 2.5D and 3D animation); their frames combined in a virtual globe with the help of the program Flash. The first issue named ‘Russian Geographical Explorations and Discoveries of the Earth up to 1845’ contains the following main subject themes: the most important trade routes, military campaigns, pilgrimages, embassies, expeditions of pathfinders and sailors, the first atlases and new maps of Russia, the first instrumental land and sea surveys, the first sea descriptions, expeditions organized by the Senate and the Academy of Sciences; circumnavigatory expeditions. Thereby the first issue of the atlas will simultaneously show both the exploration of Eurasia territories by the peoples of Russia and heroic pursuits of Russian pathfinders and sailors trying to reach the unknown regions of the world and showing the incredible courage and stamina. The second issue named ‘Russian Geographical Explorations and Discoveries of the Earth in 1845-1917’ will be devoted to the scientific expeditions of the Russian Geographical Society to different corners of the world as well as to the expeditions and surveys of Russia carried out by other Russian departments, institutions, and authorities. The results of oceanographic explorations of the World Ocean will also be considered and presented. The third issue named ‘Russian Geographic Explorations and Discoveries of the Earth from 1917 up to the present’ will be devoted to discoveries and explorations of the Soviet and post-Soviet periods which are characterized by heroic explorations of the Arctic regions, Antarctica, oceanographic explorations of the World Ocean, activities aimed at the

The Atlas and the Globe of Russian Geographical Explorations and Discoveries

185

studies of productive forces and natural resources of the country by the Academy of Sciences, the completion of the 1:1,000,000 mapping of Russia and the whole world, as well as the creation of the topographic map covering the whole country on a scale of 1:25,000, the beginning of the Earth exploration from the outer space, and the compilation of fundamental world atlases which record all territorial discoveries and the level of modern geographical knowledge of the Earth. The atlas is planned to be issued in full to the 170th anniversary of the Russian Geographical Society. As for the coverage of events, contents, the depth of scientific consideration, and the form of presentation, the ‘Atlas of Russian Geographical Explorations and Discoveries of the Earth’ unlikely has any analogue in the world. It will certainly be popular not only among the specialists in the spheres of geography and history, but also by wide public. It will be especially important for purposes of education.

References Atlas of Exploration (1997) Foreword by J. Hemming. Oxford University Press, New York Berg LS, Grigor’yev AA, & Baranskiy NN (eds.) (1948) Russkiye geografy i puteshestvenniki (Russian Geographers and travelers). Issue 1. USSR Academy of Sciences. Institute of Geography. USSR Geographical Society. Iskusstvo, Moscow & Leningrad (in Russian) Demin LA (ed) (1953) Morskoy atlas (Sea Atlas). Volume II. Physical geography. Leningrad (in Russian) Fradkin NG (1972) Geograficheskiye otkrytiya i nauchnoye poznaniye Zemli (Geographical discoveries and scientific understanding of the Earth). Mysl’, Moscow (in Russian) Gorshkov SG (ed) (1974-2005) Atlas okeanov (Atlas of the Oceans). GUNiO, Leningrad (St Petersburg). The Pacific Ocean (1974); The Atlantic and the Indian Oceans (1977); The Arctic Ocean (1980); The Antarctic Regions (2005) (in Russian) Salishchev KA, Yefimov AV, et al. (eds.) (1959) Atlas istorii geograficheskikh otkrytiy i issledovaniy (Atlas of the History of Geographical Discoveries and Explorations). GUGK, Moscow (in Russian) Yefimov AV (ed) (1964) Atlas geograficheskikh otkrytiy v Sibiri i v Severo-Zapadnoy Amerike XVII–XVIII vv. (Atlas of Geographical Discoveries in Siberia and North-West America in the 17th – 18th centuries). USSR Academy of Sciences. The N.N. Miklukho-Maklay Institute of Ethnography. Nauka, Moscow (in Russian)

Crossing Borders: Cartographic and Military Operations and the International Borders in the Libyan Desert before WW II Zsolt Győző Tőrök Department of Cartography and Geoinformatics, Eötvös Loránd University, Budapest [email protected]

Abstract In the early 20th century the Libyan Desert in the Eastern Sahara was one of the largest blank spots on the modern world map. The harsh geographic conditions and the featureless terrain made traditional European surveys and cartographic representation difficult. Expedition maps remained important sources for cartographic and military intelligence after WW I, when the international borders were demarcated by colonial powers. This article examines the Italian mapping of the remote Kufra region in the 1930s in connection with contemporary desert expeditions. The unknown region was divided by the invisible international border. In search of the legendary Zerzura the real ‘English patient’ and companion explorers crossed that border. The topographic maps compiled by the Italian military and colonial authorities effectively supported imperialistic territorial claims and the Libyan-Sudanese border was changed in 1934. The topographic survey mission to the Uweinat Mountain remained an exception however, and the generally unreliable and inaccurate topographic maps of the region were not revised until WW II.

1- Exploring and mapping the Libyan Desert The interior of the Libyan Desert remained unknown for so long because of the extreme climate and the rigid terrain. The major problem was drinking water which could be rarely found in the vast area. Therefore, for the traditional camel caravans continuous water supply was extremely difficult. Consequently, the regions beyond the range of camel remained A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 187 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_13, © Springer-Verlag Berlin Heidelberg 2011

188

Z.G. Tőrök

inaccessible until the introduction of the new expedition technology in the 20th century. In 1873-74 Gerhard Rohlf’s large and well equipped German expedition visited and surveyed the westernmost Egyptian oasis, Dakhla. A novel method, terrestrial photogrammetry, was used here for first time by a scientific expedition (Jordan 1875). Rohlfs planned to traverse the unknown central part of the Libyan Desert to reach the more or less mysteriousLibyanoasis,Kufra (Fig. 1).

Figure 1: Detail of the map showing Rohlfs expedition’s route with the turning point ‘Regenfeld’ (Rohlfs 1876)

At some point on the way towards the target in southwest he suddenly headed northwest, and marched to Siwa in Egypt. His large caravan was the first to cross the unexplored Great Sand Sea, but the project was not accomplished until 1879, when Rohlfs reached the Libyan oasis from Tripoli.

Crossing Borders: Cartographic and Military Operations and the International Borders

189

The failure was not due to the terrain or climate, however. Social and political conditions were also extremely hostile in the region. The Muslim religious movement, generally mentioned as Senussia, controlled the oases and the caravan routes in the Libyan Desert. The followers of the fundamentalist religious teaching resisted to Western civilization and fought against the British colonial army in the Sudan before their retreat to Kufra. In the 19th century this oasis was practically inaccessible to Europeans and the followers of the Great Sheikh did everything they could to hinder the progress of expeditions in their territory. Rohlfs and the other German scholars could feel the everyday suspicion and hostility of the local inhabitants and these circumstances may explain the reorientation of the expedition, which may saved the life of its members. In the early 20th century, the romantic English explorer, Harding King, with the support of the Royal Geographical Society in London decided to start searching for the legendary oasis, Zerzura. In Dakhla, the westernmost oasis in Egypt, he observed the migrating birds coming from the southwest with freshly eaten olives in their stomachs. Based on his experiments, he calculated the distance of his ‘olive’ oasis, and made three attempts to locate it. In 1911 he reached a point approximately 250 kilometres south-west from the oasis, but his camel caravan had to return because his native guide tampered with his water supplies. Harding King published his reports in journal of the Royal Geographical Society and, in 1925, wrote a fascinating, even mysterious, book (Harding-King 1925). The map he constructed and enclosed to that was based on native information, an amalgam of geographical information of highly different reliability as it was distilled mainly from lies. In modern terms, it may be considered a mental map, constructed by a talented and experienced European explorer (Fig. 2).

190

Z.G. Tőrök

Figure 2: Harding-King’s ‘mental map’ of the inner part of the Libyan Desert (Harding-King 1925)

In 1911 Senussi militants regularly invaded the former Libyan and Egyptian territories of the Ottoman Empire. After World War I new Italian and British colonial powers, took over these possessions along the Mediterranean coast. The British military patrols, chasing the guerilla parties, have experimented first with a new means of transport, the

Crossing Borders: Cartographic and Military Operations and the International Borders

191

automobile. In 1917, the English director of the Egyptian Geological Survey, Dr. John Ball’s motorized expedition discovered an ancient water depot west of Dakhla. In 1923, the Egyptian Ahmed Hassanein’s caravan reached Kufra and he could even continue the caravan route further south towards Sudan. Hassanein, an Oxford graduate, was not European and kept good relations with the Senussia, so he could manage to re-discover and determine the position of the Arkenu and the Uweinat mountains (Hassanein 1925). The greatest figure of Egyptian desert exploration was Prince Kemal by Din, who applied modern expedition technology and used caterpillar Citroën motorcars in the desert. From 1924, accompanied by Dr. Ball, he made several long-distance exploratory journeys in the interior of the desert. From the Egyptian side he managed to reach Uweinat Mountain; and in 1925, to north-east from here, he discovered the huge sandstone plateau Gilf Kebir, the Great Wall. The report of his expeditions, including route maps, was published in a French geographical journal and became available for the international public.

2- Demarcating the international border in the desert Ottoman Egypt was occupied by the British colonial army in 1882, and then the country remained under protectorate until 1914, when the country became independent. In World War I the coastal strip of Libya, belonging to the Ottoman Empire, was occupied by Italy. However, in the new situation the international border between the new countries became a political and cartographical problem. The vagueness of the borders between the two provinces was traditional. In 1841 the order of the Sultan on the borders of Egypt referred to a map hanging in the office in the Grand Vizier. This map was, however, not published and was found in Constantinople only when the borders were defined by international treaties. The treaty between Egypt and Italy was signed on December 6, 1925, and was revised in the following year. Starting from Sollum at the coast the border was demarcated in two successive sessions: in May-June, 1926 and in February-April, 1927. It must be emphasized here that the actual demarcation extended only to the northernmost, about 350 kilometer-long section of the border. From the southernmost border point the international border was simply drawn on

192

Z.G. Tőrök

paper along the 29th meridian East of Greenwich until the 22º Northern latitude. The point of the intersection with the Sudanese border in the south fell in the Uweinat area.

Figure 3: Detail of the map enclosed to the report of the border demarcation commissions with the signatures of the surveyors. Author’s photo.

The Egyptian border demarcation commission was lead by Patrick Clayton, an experienced surveyor and cartographer of the Desert Survey. The commander of the Italian party was Captain Campo the military topographer was Lt. Gallino(Fig. 3). Both commissions used cars: Clayton the Desert Survey’s new Ford, which was guarded by the motorized desert patrol; the Italian party’s four Fiat lorries were escorted by fifteen Italian ‘mecharisti’ on camelback. The two commissions got on well with each other during the missions and, although they worked separately, cooperated as well. To calculate accurate geographical position the English Clayton used a radio to receive time signals from Cairo. The Italian report mentioned this method and did not forget to add with some proud that they used Marconi’s invention for astronomical position determination. Along with the one dozen astronomical points, closer to the seacoast they could already use the fixed points of the Egyptian triangulation chain, which was developed there by Clayton in the previous year. At about at the same time, the southern border between Egypt and Sudan was surveyed by Lewellyn Beadnell. At Williams Pass in the Great Sand Sea a 600 meter-long base line was measured, and a graphic triangulation network was developed to the 29º

Crossing Borders: Cartographic and Military Operations and the International Borders

193

15’ north latitude. The temporary marks were placed along the border line according to the terrain and visibility requirements. Actually, the border markers were slightly displaced, although their position difference did not exceed 150 meters. The demarcation commissions prepared a detailed report, which was reproduced in limited numbers by the Survey of Egypt. Each report was numbered and signed by members of the commissions. The border markers, altogether 178, were represented on a 1: 250 000 maps with lettering in French. The northern, coastal area was depicted on more detailed, 1: 10 000 scale topographic maps. The maps were authenticated by the handwritten signatures of the surveyors.

3- The ‘English Patient’ in Italian Libya The character of ‘Count Almasy’ in the film ‘The English Patient’ was based on a real person. László Ede Almásy (1895-1951) (Fig. 4) came from a noble but untitled Hungarian family. Almásy led expeditions that were part test-drives and part safaris in Egypt and Sudan. In 1929 with two cars he journeyed across east Africa, the Sudan, and Egypt. In the course of this 12,000-kilometer trip Almásy, accompanied by Prince Ferdinand von Liechtenstein and the Austrian film maker Rudi Mayer, rediscovered an old caravan path - the Darb el Arbain, or Road of Forty - the ancient trade route connecting Egypt and inner Africa (Almásy 1935).

Figure 4: László Ede Almásy in front of the Grand Hotel in Khartoum, Sudan in 1929. Courtesy of Kurt Mayer Film, Vienna.

194

Z.G. Tőrök

During this expedition Almásy fell in love with the immense desert. The story about the lost oasis of Zerzura, mentioned in a medieval Arabic manuscript written for treasure hunters (Johnson 1930), particularly lured him. Sir John Gardner Wilkinson in 1818 also heard about Zerzura from the inhabitants of the Dakhla Oasis in Egypt and recorded it in his writings (Wilkinson 1835). In the period of the scientific cartography of the Enlightenment, the critical method of the eminent French geographer D'Anville resulted in huge blank spaces on his 1747 map of Africa. These empty spaces were to be filled by route surveys of new expeditions which were practically the only source of information. The Napoleonic topographical survey of Egypt by the French colonial army did not extend to desert areas, and vast areas remained unmapped during the ensuing century (Godlewska 1988). By the time Almásy entered the world of Zerzura seekers in the early 1930s, only the innermost section of the Libyan Desert had remained unmapped. Automobiles made it possible to explore those unknown territories. In those years, using specially equipped Fords, the English officer Major Ralph Bagnold, and his companion explored enormous tracts of the Libyan Desert. Despite special efforts the British could not find the lost oasis. The map showing the routes of Bagnold’s 1929 and 1930 expedition routes was the best expedition map of the region. ‘Owing to unforeseen delays in receipt of material’ the large sheet was published in The Geographical Journal as an appendix, but half a year later than the paper it illustrated (Bagnold 1931b: 525) (Fig. 5). The mapping of the ‘blank spot’ required unusual cartographic solutions to the problem of presenting nothing. Bagnold’s map was filled with detail, but these minor remarks and notes would certainly be ignored by any compiler of a geographical map on a scale of 1: 1 million. Cartographers would not depict topographic features such as a 'dead camel’' on a normal map. In lack of conventional signs, the map includes textual notes as the graphic language of European geography apparently became insufficient to represent the 'flat featureless sand plain', in other words the 'nothing'. The map makers, in order to avoid the blank spaces syndrome, created a scientific illusion. They represented blank spaces but their scientific technical context already suggested the whole territory had been surveyed. Map spaces were left blank as results of the process of survey and map construction, and now they became evidences of the scientific method. What to represent when nothing was found, was indeed a serious cartographic problem.

Crossing Borders: Cartographic and Military Operations and the International Borders

195

Figure 5: The Uweinat region of the map of the Libyan Desert showing expedition routes, including Bagnold’s journeys in (Bagnold 1931b)

4- The Zerzura Quest Almásy, who spoke six languages, including Arabic, was welcome in the Egyptian court where Prince Kemal el Din acted as his patron. After consulting scientific reports, maps, and historical documents, and conducting interviews with native Bedouins, Almásy concluded that Zerzura should be somewhere in the unexplored Gilf Kebir region, on the route from the Dachla Oasis to Kufra. In 1931 he attempted to use his light aircraft to do reconnaissance, but unfortunately crashed in Syria on the way.

196

Z.G. Tőrök

In the following year a young English baron, Sir Robert Clayton EastClayton, joined Almásy's quest. His plane ‘Rupert’ (type Havilland Gipsy Moth I) figured prominently in the expedition. Wing-Commander Penderel of the Royal Air Force and Patrick Clayton of the Desert Survey, were the other British members. Their presence indicated growing interest of British intelligence in the region. Almasy, who was always practically penniless, had to cooperate with Egyptian governmental and military organizations, who returned his service with granting permission and providing the expedition equipment and vehicles. In 1931 the Italian colonial army occupied Kufra and any journey into the Uweinat region became highly important source for British intelligence. In 1932 the Almásy-Clayton expedition reached the western escarpment of the Gilf Kebir when they run out of supplies. On April 27 Almásy undertook a dangerous trip across unknown territory to fetch water and petrol from the nearest oasis, Kufra. His arrival surprised the Italian officers of the colony. Almásy's unexpected visit proved that the desert was not impassable and called military attention to border conflicts. However, he was warmly welcomed and guided round the colony by the Italian commander, Major Ottavio Rolle. Almásy returned to the expedition camp on the following day and learnt that his companions’ reconnaissance flight was successful. On May 1 they also located another wadi. However, despite Almásy’s efforts, they could not find the entrance and the expedition eventually ran out of petrol and water and had to return to Cairo, where they announced the discovery of the ‘lost oasis’. Unfortunately, Almásy lost his supporters, both Prince Kemal el Din and Sir Clayton, in late 1932. Early 1933 Patrick Clayton, while surveying the Great Sand Sea could make a detour, and explored both valleys. He then returned to Cairo, where he met Lady Clayton, Sir Clayton's young widow, and together they visited the valleys again. Meanwhile, Almásy was having difficulty raising money. His international expedition did not set out until March 1933, along with Wing-Commander Penderel (RAF), Arnold Hoellriegel (penname, actually Richard Bermann, an Austrian journalist), Hans Casparius (a German photographer), and László Kádár (a Hungarian geographer). They mapped the southern and eastern sides of the Gilf Kebir, and discovered the Aqaba Pass, the Gap, notched between two sides of the plateau.

Crossing Borders: Cartographic and Military Operations and the International Borders

197

On April 17, 1933 Almásy’s expedition arrived at Kufra again, just missing the British expedition which had left on the same morning. Despite all the problems and this disappointing news, Almásy did not give it up. He managed to get information from a native Tebu guide that convinced him there was a third, undiscovered wadi in the Gilf Kebir. From Kufra, Almásy led his expedition to the western side of the Gilf, where he discovered the third valley. After this success, they drove to Ain Dua, a well in the Uweinat Mountains. They could find an Italian military patrol camping there and the scientific expedition of Lodovico Caporiaccio. In Italian territory discovered Almásy the prehistoric rock paintings of antelopes, giraffes, and even swimmers, the historic evidences that the Sahara had not always been a desert.

5- Cartography and intelligence The 1933 Almásy-expedition was closely escorted in Kufra by the Italian military intelligence officer. A few days after their departure the officer sent a secret report to Rome, with minute details of their visit. The paragraph mentioning expedition route maps is worth attention as it explicitly refers to some expedition map that Almásy had promised and eventually gave to the Italians. The report says the map showed the exact location of the valleys discovered in the Gilf Kebir, however, its cartographic value should not be overestimated (Fig. 6). Actually, already in November 1932 Major Rolle’s military expedition followed Almásy’s path to the Gilf Kebir and the Italians also visited the valleys on Egyptian territory. In general, however, the expeditions’ route maps were important sources of geographical information and this fact explains why colonial and military authorities showed special interest in them. The expedition maps and survey documents Almásy and contemporary explorers produced in the field became substantial documents for mapping projects of very different nature.

198

Z.G. Tőrök

Figure 6: Italian military sketch representing explorations in the western Gilf Kebir. (Institute and Museum for Military History, Budapest, author’s photo.)

The Istituto Geografico Militare in Florence and Colonel Enrico Agostini and his colonial cartographers at the Uffizio Studi in Benghazi closely cooperated in the production of the topographic base map of Italian Libya. The study of the sheets of the ‘Carta dimostrativa della Libia’ series reveals that the representation of the Kufra region on these topographical map sheets was based on the contemporary expedition route maps like the one Almásy gave to the Italians in 1933 in Kufra.

Crossing Borders: Cartographic and Military Operations and the International Borders

199

This cartographic document could be very similar to the Italian military sketch, found in the collection of the Military Historical Institute and Museum in Budapest. The 1: 1 million scale sheet is a compilation, and was based on British material. The geographical name ‘Kufra’ on the map, instead of the Italian form ‘Cufra’ is a clear evidence for this. However, the material used is nothing to do with intelligence and romantic spy stories. The source is the report on the 1932 Almásy-Clayton expedition, which was published in The Geographical Journal with a map (Clayton 1933).

Figure 7: Detail of the 1: 400 000 Italian topographical map sheet A’rchenu with the routes of the 1933 Almásy- Penderel expedition (1934)

To include the strategically important oasis, not represented on Clayton’s sketch, the map was extended, a little more than half a degree, to the west. In the additional strip another important map element could be plotted,

200

Z.G. Tőrök

namely Hassanein’s 1923 track, leading from Kufra to Uweinat. The third element the Italian map compiler added to his source is the representation of the international border between the colony Cirenaica (from 1932 Libia) and Egypt. In contrast, it must be noted that the British map in the Geographical Journal does not show any border in the region. Apparently, the British still considered the desert open space. Its content is reflected on the 1: 400 000 Italian topographic map sheet A’rchenu, where the expedition route crossing the international border and the ‘Almasy-Penderel’ note still appear (Fig. 7).

6- The Uweinat border problem (1934) The international border between Egypt and Sudan was delimited by British colonial authorities in 1884. The 22º northern latitude was not a natural border and Anglo-Egyptian Sudan traditionally included the Sarra triangle (around the Sarra well) in the southern part of Libya. To counterbalance the French influence west of the river Niger the Sudanese colonial forces occupied Eastern Sudan in 1898. According to the 1919 Anglo-French Convention the international border could be differently interpreted, especially because the agreement was not represented on an official map. Italian claims also referred to the tradition and this vagueness. In 1932 Bagnold and his British expedition met and dined on a friendly footing with an Italian party at the Sarra well. In April 1933, first Patrick Clayton, and shortly after the Almásy-Penderel expedition, visited Kufra. The second visit to the oasis convinced the Italians that Almásy was an English spy. At the same time, for the same reason, he became even more suspicious to the English military intelligence in Egypt. The Zerzura expeditions foreshadowed the possibility of enemy operations. This explains why special surveying parties were sent by the Italian army to Uweinat and the Gilf in 1933. In that year, when Almásy visited Ain Dua again he found there an Italian military post (Almásy 1935, 1939, 1997). Among the officers was Captain Marchesi of Kufra. His staff of Italian military topographers surveyed and mapped the important border region. The 1: 100.000 topographic map of Uweinat (Fig. 8), one of the most beautiful products of Italian military cartography, was published by the

Crossing Borders: Cartographic and Military Operations and the International Borders

201

Istituto Militare in Florence. This map was based on actual topographic survey and is in sharp contrast with the compiled map sheets of low accuracy and reliability. However, as soon as they were published in the official series, the difference obscured and went unnoticed by the users.

Figure 8: Detail of the map of the Uweinat, surveyed by the Italian topographic mission (1: 100 000, Istituto Geografice Militare, 1934)

By 1934 the political issue had escalated and the international border was no longer a line on paper. In November 1933 Almásy’s archeological expedition returned from the desert to Wadi Halfa and reported a permanent Italian military post at Ain Dua, Uweinat. Air reconnaissance during the year ascertained the Italian occupation at Sarra well, and airfields were marked on both locations. The British reaction was a similar demonstration of power: in January 1934 the RAF and Sudan Defense Force were ordered to occupy oasis Merga and Karkur Murr at Uweinat. Interestingly enough, in March 1934 both Italian and British officers met Almásy, whose archaeological expedition camped on the British side of the Uweinat. On July 20, 1934 the Sarra triangle was eventually ceded to Italy in the Rome Agreement and the problem seemed to have been solved. It is worth

202

Z.G. Tőrök

mentioning that the Italian topographic mission led by Oreste Marchesi marked the new border, the point of confluence a year earlier. The new border between Libya and Sudan was established by the border demarcation committees led by Colonel Agostini and Colonel Wyatt.

7- Enemy operations in WW II The political and military situation of the world was changing quickly. In 1939 Almásy had to leave Egypt, but by 1941 he was in the Libyan Desert again as General Rommel’s desert expert. On the Allied side Bagnold formed the light car patrols of the Long Range Desert Group, including many former English explorers. Almásy's most famous mission was Operation Salaam, in which he took two German spies from Libya through Allied lines to Egypt in 1942. His war diary of ‘Operation Salaam’ is kept in the Imperial War Museum in London (Almásy 1997). The description of the route of the special unit includes many references to desert navigation and maps. Almásy complained about the ‘useless Italian map’ and on his return journey, already in Libya, but still behind Allied lines, he ironically noted the ‘blank Italian map’ again and made a sarcastic remark in connection with the ‘missed triangulation’.

8- Conclusions: new border and bad maps The new Libyan-Sudanese border was demarcated in 1934 by the topographers who previously surveyed and mapped the region. From this perspective the topographic campaign in the Kufra military zone can be interpreted as a possessive act. The earlier and contemporary expeditions apparently played an important role in the cartography of the region. Especially international expeditions were also interpreted as representation of enemy military and political interest in the colonial territory. On the other hand, route maps remained important sources and topographic map sheets were generally not based on actual survey. A remarkable exception is the 1933-34 mapping of the Uweinat region, which was closely related to Italian territorial claims and served the demarcation of the new Sudanese-Libyan border.

Crossing Borders: Cartographic and Military Operations and the International Borders

203

Figure 9: Detail of sheet ‘Uweinat of the 1: 1 million International Map of the World (Survey of Egypt, 1942 edition, author’s collection).

In 1942, during WWII the sheet ‘Uweinat was published by the Survey of Egypt in the 1:1 million International World Map series (Fig. 9). Although the expedition routes in red are still emphasized, they are sharply cut by the strong, black signature of a new Egyptian-Libyan at the international

204

Z.G. Tőrök

border. While explorers crossed the borders, the Italian and AngloEgyptian and Sudanese topographic map series constructed a new political-military space consisting of territories for colonial and military administration. The cartographic works related to the demarcation of international borders in the Libyan Desert demonstrate the geopolitical determinations of the cartographic discourse on power. For a few years explorers’ tracks could cross the international borders in the field and on the map, but those virtual lines soon became more real and more permanent than any other dream of the imperial cartographer.

References Brownlie I, Burns I R (1979): African Boundaries. A legal and diplomatic encyclopedia. C. Hurst and Co., London Almásy L (1935) Az ismeretlen Szahara. Franklin, Budapest Almásy LE (1936) Récentes explorationes dans la Désert Libyque (1932-1936). Publ.de la Soc. R. de Géographie d’Egypte, Schindler, Cairo Almásy L (1938) Levegőben … Homokon … Franklin, Budapest Almásy LE (1939) Unbekannte Sahara: mit Flugzeug und Auto in der Libyschen Wüste. Brockhaus, Leipzig Almásy L (1943) Rommel seregénél Líbiában. Stádium, Budapest Almásy LE (1997) Schwimmer in der Wüste: auf der Suche nach der Oase Zarzura. Haymon, Innsbruck Ball J (1927) Problems of the Lybian Desert. Geogr. J. 70: 21-38; 105-128; 209-224. Bagnold RA (1930) Journeys in the Libyan Desert in 1929 and 1930: A paper read at the Evening Meeting of the Society on 20 April 1931, by Major R.A. Bagnold. Geogr. J. 78:1339. Bagnold RA (1930) Journeys in the Libyan Desert in 1929 and 1930: Note on the Map by Major R.A. Bagnold, Remarks by Mr. W.J. Harding-King, Appendices on Survey Methods and results, Barometric Heights and Meteorology, the Kufara- Egypt caravan Route, ‘Ain Kiyeh, and Botanical Notes. Geogr. J. 79:524-535. Browne WG (1799) Travels in Africa, Egypt and Syria, from the Year 1792 to 1798. Cadwell, London Caporiacco L, Graziosi P (1934) Le pitture rupestre di Ain Doua. Centro di Studi Coloniali e Instit. Geogr. Militare, Firenze Clayton PA (1933) The Western side of the Gilf Kebir. Geogr. J. 81: 254-259. Clayton P (1998) Desert Explorer - A biography of Colonel P.A. Clayton. Zerzura, Cargreen Desio A. et al. (1934-39) Missione Scientifica della Reale Accademia d'Italia a Cufra (1931-IX), Vol. I – III. Reale Accademia d'Italia, Roma Godlewska A (1988) The Napoleonic survey of Egypt. A masterpiece of cartographic compilation and early nineteenth-century fieldwork. Cartographica 25 GR (1934) La spedizione Almásy-Penderel nel Deserto Libico. Bull. R. Soc. Geogr. Ital. 578580. Hassanein Bey AM (1925) The Lost Oases. Century, New York & London Hoellriegel A. (=Bermann R) (1938) Zarzura, die Oase der kleinen Vögel. Die Geschichte e. Expedition in d. Libysche Wüste. Orell Füssli, Zürich

Crossing Borders: Cartographic and Military Operations and the International Borders

205

Johnson Pasha EA (1930) Zerzura. Geog. J. 75:59-61. Jordan W (1875) Die geographischen Resultate der von G. Rohlfs geführten Expedition in die libysche Wüste: öffentl. Vortrag gehalten im Museum zu Carlsruhe, am 16. Dezember 1874 (= Sammlung gemeinverständl. wiss. Vorträge, 218). Habel, Berlin Kádár L (1934) A study of the Sand See in the Libyan Desert. Geogr. J. 83: 470-478. Kemal el Dine (1928) L'exploration du desert de Libye. La Geographie 50: 171-183; 320-336. Negro G (1991) Il Great Sand Sea e la sua esplorazione (Sud-ouest dell’Egitto). Sahara 4:71- 81. Penderel HWG J (1934) The Gilf Kebir. Geogr. J. 83:449-456. Rohlfs FG (1875) Drei Monate in der libyschen Wüste. Mit Beiträgen von P. Ascherson, W. Jordan und K. Zittel, etc. (= Expedition z. Erforschung d. libyschen Wüste, 1). T. Fischer, Cassel Rolle O (1933): Ricognizione zona a cavallo del 25ºmeridiano (adiacenze Gelf el-Chebir) eseguita del Magg. Rolle nel novembre 1932. Boll. Geogr. 16:23-26. Shaw WBK (1945) Long Range Desert Group. The Story of its Work in Libya, 1940-1943. Collins, London The Geogr. J.. (1930). The Zerzura Problem. Geogr. J. 75:48 Traversi C (1964) L’Italia in Africa. Storia della cartografia coloniale Italiana. Poligrafico dello Stato, Roma Török Z (1989) Almásy László szerepe a Kelet-Szahara kutatásában. Földr. Múz. Tan. 8:21-26. Török Z (1997) Az ismeretlen Szahara magyar felfedezője, Almásy László. Földr. Közl. 71:7786. Török Z (1998) Salaam Almásy. Almásy László életregénye.[The biography of László Almásy.] ELTE Eötvös,Budapest Török Z (2002) Almásy László és a Líbiai-sivatag expedíciós térképezése. Stud. Cart. 12:107113. Török Z (2004) Der letzte österreichisch-ungarische Entdecker: László Almásy und die Expeditionskartographie der Libyschen Wüste. Wiener Schr. zur Geogr. u. Kart. 16:131-141. Török ZG (2009) The ‘English’ patient, fools, foxes and rats: exploration, mapping and war in the Libyan Desert. In: Liebenberg E (ed.): Symposium on “Shifting Boundaries: Cartography in the 19th and 20th centuries,” Portsmouth University, Portsmouth, United Kingdom, 10-12 September 2008, ICA Commission on the History of Cartography, 1-14. Wilkinson IG (1835) Topography of Thebes and General View of Egypt. John Murray, London

Map Projection

Optimising the Distortions of sinusoidal- elliptical composite Projections Gede Mátyás Eötvös Loránd University Budapest, Hungary [email protected]

Abstract The sinusoidal and the elliptical projections are two large groups of pseudocylindrical projections. A sinusoidal-elliptical composite projection combines the members of these groups by applying one projection for the polar regions and another for the equatorial area. Although several attempts were made to create equal-area projections of this kind, none of these solutions were perfect. The main issues were the break of meridian lines at the connection latitude and the different scale of the two parts. This paper’s aim is to fill this gap by deducing two different, parametrisable sinusoidal-elliptical combined projections.

1- Introduction The first sinusoidal-elliptical projection was introduced by J. P. Goode in 1925 (Bugayevskiy and Snyder 1995). He attached the Mercator-Sanson (Sanson-Flamsteed or sinusoidal) projection to the Mollweide, attaching them at the ±40.7367° latitude as this parallel has the same length in both projections. The result was far not perfect as the meridians are refracted at the attaching latitude; moreover, the outline of the map is concave here. The Hungarian cartographer, Érdi-Krausz György connected a Wagnertransformed form of the sinusoidal projection to Mollweide at the 60° or the 70° latitude (Érdi-Krausz 1968). This solution is more attractive than the previous, but has a great disadvantage: the scale of the two parts is not the same. That is, although each part is equal-area, the whole projection A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 209 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_14, © Springer-Verlag Berlin Heidelberg 2011

210

G. Mátyás

loses this attribute. The meridians are still refracted, but not so hard, and this error was smoothed by the editors on the maps. Juhász Péter showed two possible solutions for this problem in his MSc degree thesis, and later in his articles (Juhász 2002, 2003). In the first solution the parameters of the Wagner-transformation are numerically optimised to prevent the break at the 60° latitude. The second, more detailed one uses a „mixer function” to make the transition between the two parts smooth. This paper enhances the first solution of Juhász. A relatively easy mathematic method is introduced first which calculates the exact parameters of the Wagner transformation to gain a smooth connection between the Mollweide and the transformed sinusoidal projection using any boundary latitude. The next part shows additional azimuthal transformations to get better distortion values. Finally, a very general, three-parameters sinusoidalelliptical equal-area composite projection is introduced, with a discussion on the relationship between the parameters and the distortions.

2- The Base Projections

2.1 The sinusoidal projection The well-known sinusoidal projection (also called Sanson-Flamsteed or Mercator-Sanson in the continental Europe) is a pseudocylindrical projection with true scale parallels and central meridian. These attributes give the formulas also: y =ϕ x = λ ⋅ cos ϕ

The first equation is for the true scale central meridian and the second for the true scale parallels. The projection is furthermore equal-area (Figure 1).

Optimising the Distortions of sinusoidal- elliptical composite Projections

211

Figure 1: The sinusoidal projection

2.2 The Wagner transformation The Wagner transformation (its original name is Umbeziffern von Kartennetzen which means “re-numbering of cartographic grids”) was introduced by Karlheinz Wagner in the 1930’s. This transformation uses a smaller part of a projection’s grid, magnified to keep the original area of the projection (Wagner 1962). If the original projection is equal-area, the renumbered one keeps this nature. If this transformation is applied on the sinusoidal projection, using a part of it with relative dimensions n (horizontally) and m (vertically), the formulas for the new projection are as follows:

x= y=

1 mn 1 mn

n ⋅ λ ⋅ 1 − m 2 sin 2 ϕ ⋅ arcsin(m sin ϕ )

As the parameters m and n can vary between 0 and 1, these formulas describe a set of projections named Mercator-Sanson series. More expressive parameters can be the length of the central meridian (p) and the pole line (q) as a proportion of the equator’s length. The values of m and n can be easily calculated from these ones:

212

G. Mátyás

m = 1 − q2 n=

arcsin m pπ

Figure 2 shows an example of the Wagner-transformed sinusoidal projection.

Figure 2: The p = 0.55, q = 0.63 member of the Mercator-Sanson series

2.3 The Mollweide (elliptical) projection This equal-area projection depicts the Earth’s surface into an ellipse with 2: 1 axis proportions (Fig. 3). The formulas are:

x=

2 2

π

⋅ arcλ ⋅ cosψ

y = 2 sinψ

where ψ can be calculated using the following formula: sin ϕ =

sin(2ψ ) + 2ψ

π

Optimising the Distortions of sinusoidal- elliptical composite Projections

213

Figure 3: The Mollweide projection

2.4 The Érdi-Krausz projection The Érdi-Krausz projection (Figure 4) connects the Mollweide projection to a member of the Mercator-Sanson series with parameters p=0,4; q=0,6; at the 60° (or rarely at the 70°) latitude. The disadvantages of this projection are that the scales of the two parts are not the same, so although the parts separately are equal-area, the whole projection loses this nature; and the meridians are refracted at the boundary latitude.

Figure 4: The Érdi-Krausz projection

214

G. Mátyás

3- Finding parameters for perfect connection Let us find such m and n parameters for a given φc connecting latitude for the Wagner transformation that the resulting projection (from here “projection A”) can be connected to the Mollweide’s (“projection B”) at φc connecting latitude without break and magnification. The constraints are the following: a) The connecting latitude (φc) should have the same length in the two projections. The length of φc in projection A: lA =

2π

n 1 − m 2 sin 2 ϕc = 2π

mn

(

n 1 − m 2 sin 2 ϕc m

)

In projection B:

lB = 2π ⋅

2 2

π

cosψ c

The first constraint is l A = lB , that is 2 2

π

(

)

n 1 − m 2 sin 2 ϕc = m

cosψ c =

2 n ⎛⎜ ⎛ sin (2ψ c ) + 2ψ c ⎞ ⎞⎟ . 1 − m2 ⎜ ⎟ m ⎜⎝ π ⎝ ⎠ ⎟⎠

b) The meridians should not refract, i. e. the Θ angular distortion of meridians at φc latitude should be equal for the two projections. cot Θ A = −mnλ sin ϕ .

Using the sin ϕ =

sin (2ψ ) + 2ψ

π

Optimising the Distortions of sinusoidal- elliptical composite Projections

215

substitution we get cot ΘA = − mnλ cot ΘB = −

2

π

sin 2ψ + 2ψ

π

λ tanψ

Therefore, the second constraint: cot ΘA = cot ΘB

that is 2 tanψ c = mn(sin 2ψ c + 2ψ c ) .

The two constraints give quadratic equations of m and n with ψc as parameter. The solutions are:

n=

8

π2

⋅

sin 2ψ c 4 tan 2 ψ c , + sin (2ψ c ) + 2ψ c π2

2 tanψ c sin (2ψ c ) + 2ψ c m= n

This means that there exists such m and n for any given φc (ψc) connecting latitude that a member of the Mercator-Sanson series with those parameters can be connected to the Mollweide projection at φc without break. The lack of break implies the continuity of distortion indicators also: The

h=

∂x 1 ⋅ ∂λ cosϕ

linear distortion along the latitude lines is the same for projections A and B at φc, and hk sin Θ = 1 because they are equal-area projections, and it implies the equality of the k linear distortion along the meridians for both projections at φc.

216

G. Mátyás

We still need the offset of the Mollweide parts for the combined projection, which is the difference of the ordinates of the φc latitude in the two projections:

Δ = 2 sinψ c −

arcsin(m sin ϕ c ) mn

The absolute values of Mollweide ordinates should be decreased by this amount.

Figure 5: The central meridian / equator ratio (ρ) as a function of φc

4- Transforming the projection If you draw the projection grid based on the previous formulas with different φc values, you can see that the central meridian / equator ratio (ρ) is decreasing by the increase of φc (see Figure 5). (This ratio is approx. 0.46 for the original Érdi-Krausz projection). Let us apply an affine transformation to make this ratio independent of φc. A new parameter, K, will be needed, and the formulas need to be modified as follows: projection “A”:

Optimising the Distortions of sinusoidal- elliptical composite Projections

xA =

yA =

1 K

217

n λ 1 − m 2 sin 2 ϕ m

K mn

cot Θ A = −

arcsin(msinϕ ) mnλ sin ϕ K2

projection “B”:

xB =

2 2 λcosψ Kπ

y B = K 2 sinψ cot ΘB = −

2λ tanψ K 2π

As the x formulas are multiplied and the y’s are divided by K, the projections remain equal-area. K also can be expressed as a function of ρ:

K=

ρ nπ arcsin(m sin ϕ c ) + 2mn (1 − sinψ c )

Figure 6 shows this projection when φc = 60° and ρ = 0.5.

218

G. Mátyás

Figure 6: The sinusoidal-elliptical combined projection with parameters φc=60°, ρ=0.5

5- Examining the distortions The parameters φc and K can be freely chosen for this projection. It is worth examining how the average distortions depend on these values.

5.1 Projection distortions The projection distortions can be indicated by the Airy or the Kavrayskiy angular distortion criteria. As the projection is equal-area there is no need to care about area distortions. At a given point, the Airy criteria is 2

⎛a ⎞ κ A = ⎜ − 1⎟ , while the Kavrayskiy is ⎝b ⎠ a κ K = ln 2 b

The average of distortions is usually calculated on a half hemisphere, between the Equator and the 85° parallel, because the polar regions are always very distorted (Frančula 1971). Thus the formula is: κ=

1 π sin 85°

85° 180°

∫ ∫ κ cosϕ dλdϕ 0

0

Optimising the Distortions of sinusoidal- elliptical composite Projections

219

Figure 7 and 8 show the average angular distortions’ dependence on φc and ρ. The minimum of the Airy distortion criteria is 2.0861 (at φc =45°, ρ=0.8). The Kavrayskiy criteria is minimal at an affine transformed (ρ=0.6) variant of the pure Mollweide projection.

Figure 7: Graph of the average Airy angular distortion dependence on φc and ρ

Figure 8: Graph of the average Kavrayskiy angular distortion

220

G. Mátyás

6- Generalizing the projection Let us substitute the Mollweide part of the projection by an arbitrary elliptical equal-area projection where the only constraints are that the latitude lines are straight parallel equidistant lines and the meridians are parts of an ellipse. First let us define its outline. It is a part of an ellipse, parametrised by the following way: a 2γ central angle part of an r radius circle (which is touching the pole point) is horizontally scaled by q to gain the outline (Fig. 9). q

ϕ ϕ α γ

c

r

Figure 9: Parameters of the elliptical projection part

Let us use the angle α instead of ϕ as it is shown in the figure. Due to the equivalency of the projection, the area of the gray shape needs to be equal to the surface of the corresponding spherical triangle. This gives the relation between ϕ and α:

q

r2 ⎛ sin 2α ⎞ ⎜α − ⎟ = π (1 − sin ϕ ) 2⎝ 2 ⎠

As α=γ at the boundary parallel φc, r can be expressed by q:

r=

2π (1 − sin ϕ c ) . sin 2γ ⎞ ⎛ q⎜ γ − ⎟ 2 ⎠ ⎝

Thus, the formulas for the elliptical projection part are the following:

Optimising the Distortions of sinusoidal- elliptical composite Projections

x=

qr

π

221

λ sin α

y = yc + r (cos α − cos γ )

where yc is the ordinate of the connecting parallel φc. The angular distortion of meridians will be

cot Θ = −

q

π

λ cot α .

The constraints for attaching the Mercator-Sanson series part are the same as before – the length of the boundary parallel and the angular distortion of the meridians must match at the attaching line:

(

)

n qr 1 − m 2 sin 2 ϕ c = sin γ m π q mn sin ϕ c = cot γ

π

These are quadratic equations of m and n. The solutions:

m=

cot γ sin ϕ c qr 2

π n=

sin 2 γ + cot γ sin ϕ c

q cot γ 1 ⋅ π sin ϕ c m

Therefore the projection can be defined by the φc, γ and q parameters. It is advisable to substitute the parameter q with the more expressive ρ (central meridian / equator ratio). The length of the half Equator: xϕ =0°,λ =180° = The central meridian: yϕ =90° =

arcsin(m sin ϕ c ) mn

n π. m + r (1 − cos γ ) .

222

G. Mátyás

arcsin(m sin ϕ c ) + r (1 − cos γ ) mn Therefore mpe = . n π m

If you wish to express q with ρ, you will get a non-linear equation, so it is advisable to calculate it numerically using the formulas above.

Figure 10 φc =60°, γ=64°, ρ=0.4625

Using adequate parameters you can gain a projection with an outline similar to the one of the Érdi-Krausz projection, but without any break. Naturally, the latitude lines will be different as this projection is fully equal-area. Fig. 10 shows the projection and parameters best fitting the original Érdi-Krausz projection with 60° boundary parallels. Fig. 11-13 show the κ¯K Kavrayskiy average angular distortion dependence on the parameters.

Optimising the Distortions of sinusoidal- elliptical composite Projections

Figure 11: The average angular distortion at φc=80°

Figure 12: The average angular distortion at γ=88°

223

224

G. Mátyás

Figure 13: The average angular distortion at ρ=0.57

The minimum κ¯K is 0.455 with parameters φc=80°, γ=88°, ρ=0.57. The shape of the projection however, is not really eye-appealing using these values. If the angular distortions are not very important, several different, special-shaped projections can be produced with the adequate parameter settings.

7- Summary The research made it clear that the numeric values of distortions and the aesthetic features can be improved only at each other’s expense. But if it is decided which factor is more important, a wide palette of projections can be produced with the methods described above.

Optimising the Distortions of sinusoidal- elliptical composite Projections

225

Acknowledgments The European Union and the European Social Fund have provided financial support to the project under the grant agreement no. TÁMOP 4.2.1./B-09/1/KMR-2010-0003.

References Bugayevskiy LM, Snyder JP (1995) Map Projections: A Reference Manual. Taylor & Francis, London, Bristol Érdi-Krausz Gy (1968) Combined equal area projection for world maps, Hungarian Cartographic Studies, Földmérési Intézet, Budapest. pp 44-49. Frančula N (1971) Die vorteilhaftesten Abbildungen in der Atlaskartographie. Dissertation. Institut für Kartographie und Topographie der Universität Bonn Györffy J (1990) Anmerkung zur Frage der besten echten Zylinderabbildungen. Kartographische Nachrichten, Nr. 4, Bonn Juhász P (2002) Optimal projections by means of convex linear combination. Studia Cartologica 2002., pp. 43–54, ELTE Eötvös Kiadó Juhász P (2003) Repairing of the Érdi-Krausz projection. Geodézia és Kartográfia, 2003/6. pp. 13–19., Budapest. Snyder JP (1987) Map Projections – A Working Manual – U.S. Government Printing Office, Washington Wagner K (1962) Kartographische Netzentwürfe. Bibliographisches Institut, Mannheim

Using Empirical Map Projections for Modeling Early Nautical Charts Joaquim Alves Gaspar Centro Interuniversitário de História da Ciência e das Tecnologias (CIUHCT) University of Lisbon, Faculty of Sciences [email protected]

Abstract A numerical model using the concept of multidimensional scaling, generalized to distances and directions measured on the surface of the Earth, is presented and tested, with the objective of simulating the main geometric features of early nautical charts. Starting with a sample of points defined by their latitudes and longitudes, the process consists in rearranging their positions in a plane so that the differences between the initial (spherical) and final (planar) distances and directions between them are minimized. The geometry of the Cantino planisphere (1502) is simulated and the output is compared with the geographic grid implicit to the original chart, with satisfactory results. The model proved to be an effective and easy-to-use research tool and may be used, not only for simulating and assessing the various factors affecting the geometry of early nautical charts, but also for educational purposes e.g. illustrating the properties of map projections.

1- Introduction The concept of ‘empirical map projection’, to designate those representations of the Earth’s surface where numerical techniques are used to obtain useful geometric properties, not taking into account the geographic coordinates, was introduced by (Tobler 1977). Some map A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 227 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_15, © Springer-Verlag Berlin Heidelberg 2011

228

J.A. Gaspar

projections can be easily reproduced using this approach. For example, by transferring to a plane the positions of a sample of places defined by their distances and directions measured along great circles from some central point, one obtains an azimuthal equidistant projection. If, instead of great circles, loxodromes are used, a loximuthal projection is reproduced. Knowing that it is formally impossible to conserve, on a plane, all distances and directions measured on the spherical surface of the Earth, an interesting question is whether one can construct a representation, using the same empirical approach, where those quantities are approximately conserved for a given area. A solution to the problem was first suggested by (Tobler 1977). Starting with a sample of spherical distances between a set of regularly spaced latitude and longitude intersections, Tobler proceeded to determine the plane coordinates of the points, so that the differences between the original and the adjusted distances were minimized. In the research presented here, the process suggested by Tobler was generalized to both distances and directions, and applied to two types of spherical lines: great circles (or ‘orthodromes’) and rhumb-lines (or ‘loxodromes’). With the application of this kind of modeling to the study of early nautical charts it is expected to better understand how they were constructed and used, and how the various factors associated with the acquisition and representation of the information affected their geometry.

2- Navigation and charting It is a known historical fact that pre-Mercator nautical charts were drawn by transferring directly to the plane the magnetic courses and distances observed on the curved surface of the Earth, as if it were flat1. Not due to the ignorance of cosmographers and pilots but because of the constraints imposed by the navigational methods of the time. Two methods for determining the position of the ship at sea were used in the Middle Ages

1

The construction process is described in some sources, the oldest known being the ‘Treatise in defense of the navigational chart’ (1537), by the Portuguese mathematician Pedro Nunes (Nunes 2002, 120-84).

Using Empirical Map Projections for Modeling Early Nautical Charts

229

and Renaissance: the method of the ‘point of fantasy’, based on the directions given by the compass and the distances estimated by the pilots; and, from the beginning of the sixteenth century on, the ‘set point’ method, based on astronomically-observed latitudes.

point of departure

N

point of departure

N

d Course

Course

ϕ PF

SP

Figure 1: Point of fantasy (PF: left) and set point (SP: right) (ϕ is the latitude). Due to the error introduced in the course by magnetic declination and the uncertainty in the estimated distances, the point of fantasy and the set point did not usually coincide.

In the method of the point of fantasy (Figure 1, left), the position of the ship was determined as the intersection between a straight segment representing her track, oriented according to the magnetic course steered, and an arc centered on the last known position, representing the distance sailed. In the set point method, the latitude always prevailed over the course and the distance, and the position was usually found at the intersection between the parallel of the observed latitude (represented as a horizontal line) and the ship’s track. In the absence of any errors in the observed latitude, compass direction (e.g. magnetic declination) and estimated distance, the point of fantasy and the set point coincide. The adoption of one or the other method for cartographic purposes leads to different geometries. While with the method of the point of fantasy, the

230

J.A. Gaspar

spatial variation of the magnetic declination is reflected upon a variable orientation of both meridians and parallels, in the case of the set point method the parallels are always represented as straight, east-west oriented and equally-spaced lines. This makes relatively easy to identify the charting methods used in an old nautical chart, once the grid of meridians and parallels implicit to the representation is interpolated2. An important issue concerning all pre-Mercator nautical cartography is the geometric inconsistency associated with the charting methods, which consists in a position of a place being dependent on the particular set of routes that were used for represent it on the chart. For example, the longitudinal position of the island of Terceira (Azores), relative to Lisbon, will be different whether it is determined using a single rhumb-line track connecting the two places or, alternatively, a route composed of two tracks, with a scale in Madeira Island (Figure 2). This is due to impossibility of representing a spherical triangle on a plane without distorting lengths and angles. While the north-south component of a spherical rhumb-line track is always conserved when the track is represented on a plane, the same does not happen with the longitudinal component, due to the convergence of meridians.

2

In pre-Mercator nautical charts meridians and parallels are not explicitly depicted. However the geographic graticule implicit to the representations can be estimated by interpolation, using a sample of control points of known latitudes and longitudes, positively identified in the old and in a modern map. In this research the freeware application MapAnalyst (http://mapanalyst.cartography.ch/), developed by Bernhard Jenny and Adrian Webster (2005-2009), was used for that purpose. For a description of the application and its use in the visualization of the planimetric accuracy of historical maps (see Jenny et al. 2006; 2007).

Using Empirical Map Projections for Modeling Early Nautical Charts

C

A

C’1

A’ 803’ 59

59

4’

4’

4’ 59

59 4’

803’

C’2

231

B

B’

Figure 2: The inconsistency of the charting process. At left, the relative positions and rhumb-line distances between Lisbon (A), Madeira (B) and Terceira (C), as measured on a spherical Earth. At right, the position of point C (Terceira) as determined using two different tracks (AC and ABC), by plotting directly on a plane the angles and distances measured on the surface (adapted from Gaspar 2010, p. 32).

When the represented area is relatively small, the angles and lengths measured on the sphere approximately coincide with the corresponding planar ones and the inconsistencies can be ignored. That is the case of the representation of the Mediterranean and Black Sea in the medieval and Renaissance portolan charts, where the errors associated with the navigational methods of the time mask the inconsistencies of the charting method. Additionally, those inconsistencies tended to be minimized by the use of redundant information originated in a dense mesh of maritime routes covering the area. The same approach can hardly be applied to large oceanic basins, like the Atlantic or the Indian Oceans, due to their much larger size and to the absence of a similar network. In these cases, the geometry of the representations became determined by a particular set of more or less well-defined routes. It is interesting to realize how similar the shape of Africa is in all nautical charts of the fifteenth through the seventeenth centuries, which suggests that the set of routes underlying their construction did not change with time. The geometric inconsistencies affecting nautical cartography during the maritime discovery and expansion periods were not unknown by the pilots, who were aware that only certain courses were, in principle, accurately represented on the charts. Furthermore, and due to the influence of magnetic declination in the set point method, the distances between places

232

J.A. Gaspar

could be strongly distorted3. These limitations were compensated for with a careful record of estimated distances along the usual routes, which the pilots kept in their rutters.

3- The EMP model A suitable numerical approach for simulating the geometry of early charts is given by ‘multidimensional scaling’ (or mds), a multivariate statistical technique often used in Social Sciences for exploring similarities and dissimilarities between objects. Suppose that one wants to estimate the relative positions of a group of places knowing only, with uncertain accuracy, some distances between them. Starting with an arbitrary initial arrangement of the places on the plane, the process consists in re-arranging their positions, using a least squares approach, so that the differences between the initial (given) and the final (adjusted) distances are minimized. The process is equivalent to the methods used in geodetical surveying for adjusting measurements obtained in trilateration. In the model here presented the concept was generalized to distances and directions measured on a spherical model of the Earth and represented on a plane surface.

3

The estimated distances were not usually taken into account in the construction of the charts, where the positions of the places were plotted according to their latitudes and directions relative to other places. This was due to the fact that the course was a much more important piece of navigational information than the distance. With large values of magnetic declinations, or courses close to east or west, the corresponding distortions in the charted distances could be very large. This was first recognized by D. João de Castro in 1537, in his ‘Rutter from Lisbon to Goa (see Cortesão & Albuquerque 1968, Vol. 1, 1-169).

Using Empirical Map Projections for Modeling Early Nautical Charts

233

Figure 3: Interface of the EMP model application

Figure 3 shows the visual interface of a computer application, the EMP model (‘Empirical Map Projection’), which was developed using the empirical approach introduced above. The input of the EMP model is a set of points defined by their geographic coordinates, from which spherical distances and directions are calculated; the output is a cartographic representation where the coastlines and the grid of meridians and parallels, interpolated from the final positions of the points, are shown. The input (‘Input points’, in Figure 3) is defined either as a regular grid of meridians and parallel intersections (‘Use geographic grid’) or as a set of points and tracks, in an external file (‘Use track’). Two types of spherical lines connecting pair of points can be chosen (‘Method’, in Figure 3): ‘Orthodromes’ (arcs of great circles) and ‘Loxodromes’ (or ‘rhumb-lines’). For the loxodrome case, two types of charting methods are considered: the method of the ‘point of fantasy’, in which the positions are determined on the plane according to the rhumb-

234

J.A. Gaspar

line course and distance between points; and the ‘set point’ method, in which the positions are determined according to the latitudes and course (or distance) between them. A ‘mixed model’ is also considered, for simulating those representations where both the point of fantasy and the set point methods were used (which is very common in the nautical cartography of the Renaissance). In this case, it is possible to attach to each individual track a specific charting method. Magnetic declination is given in the form of a matrix of values containing the spatial distribution for a given year (‘Magnetic declination’, in the figure), or as punctual values attached to each point of the input file, when the external track option is chosen. At the bottom of the ‘Method’ box in the application interface a sliding scale is shown, used to adjust the relative weighting (w) attributed to distances and directions in the optimization process, when the method of the point of fantasy is used4. For w = 0, only distances are considered and the output is identical to the results obtained by (Tobler 1977); for w = 1, only directions are considered; for a value of w between 0 and 1, both distances and directions are considered. It is also possible to impose certain restrictions to the pairs of points used as input (‘Constraints’, in Figure 3). For example, one may consider only the tracks with origin in a given position (‘Only from position…’) or to restrict the input to distances less than a certain value (‘Distances less than…’). The ‘Distortion info’ box in the interface allows distortion information to be represented in the output: ‘Tissot ellipses’ and isolines of constant maximum angular distortion (‘Max angular distortion’). The ‘Show lines’ box is used for representing arcs of rhumb-lines or great circles radiating from a chosen position (‘Origin’) as well as the corresponding lines of constant distance (‘Distance lines’). Once the geographic limits of the area have been set and all the other options about the method, magnetic declination, tracks, etc. have been chosen, the ‘Plot’ button is used for starting the optimization process and

4

The optimization process only applies to the cases where the tracks form closed loops. In all other cases an exact solution exists and the weighting has no effect on the final result.

Using Empirical Map Projections for Modeling Early Nautical Charts

235

present the results in the form of a map. The longitudes (λ) and latitudes (ϕ) of the input points, measured in decimal degrees, are used by the application as the initial plane coordinates, x and y. An iteration routine will then start with the purpose of gradually adjusting these coordinates. The iteration process ends when their values cease to show a significant change or a pre-defined maximum number of iterations is reached. The result is then presented as a map, where the meridians, parallels and coastlines are interpolated from the final coordinates of the points. A series of examples is presented next to illustrate the use of the model and the influence of the various parameters on the geometry of the resulting maps.

Example 1: Great-circle distances and directions from a point (Figure 4, left) • • • • •

Method = Orthodromes w = 0.5 (distances and directions) Input points = Use geographic grid Distortion info = Tissot ellipses Constraints = Only from position lat=0; long=0

This is an exact solution corresponding to an azimuthal equidistant projection centered on the Equator.

Example 2: Great-circle distances (Figure 4, right) • • • • •

Method = Orthodromes w = 0 (distances only) Input points = Use geographic grid Distortion info = Tissot ellipses Constraints = No constraints

236

J.A. Gaspar

This is an approximate solution, where the error in the great circles distances between all pairs of points is minimized.

Figure 4: Examples 1 and 2. At left: great circle distances and directions measured from a point (azimuthal equidistant projection); at right: great circle distances (minimum error)

Example 3: Rhumb-line distances and directions from a point (Figure 5, left) • • • • •

Method = Loxodromes, Point of fantasy w = 0.5 (distances and directions) Input points = Use geographic grid Constraints = Only from position lat=60; long=0 Show lines = Loxodromes, Distance circles (origin: lat=60; long=0)

This is an exact solution corresponding to a loximuthal projection centered on 60° N, in which rhumb-line distances and directions measured from the center are conserved.

Using Empirical Map Projections for Modeling Early Nautical Charts

237

Figure 5: Examples 3 and 4. At left: rhumb-line distances and directions measured from a point (loximuthal projection, centred on 60° N; 0°); at right: rhumb-line directions (Mercator projection)

Example 4: Rhumb line directions (Figure 5, right) • • • • •

Method = Loxodromes, Point of fantasy w = 1 (directions only) Input points = Use geographic grid Distortion info = Tissot ellipses Constraints = No constraints

This is an exact solution corresponding to the Mercator projection, in which all rhumb-lines are represented by straight segments and rhumb-line courses are conserved.

Example 5: Portolan chart model – North Atlantic and Mediterranean (Figure 6, left) • • • • • •

Method = Loxodromes, Point of fantasy w = 0.5 (distances and directions) Input points = Use geographic grid Constraints = Distances less than 50 degrees Show lines = Loxodromes (origin: lat=40; long= -10) Magnetic declination = 1500.

238

J.A. Gaspar

This is a hypothetic result of the application of the portolan chart model, based on the point of fantasy, to the representation of the North Atlantic and the Mediterranean, considering the effect of magnetic declination as of 1500 and using only distances less than 50 degrees.

Figure 6: Examples 5 and 6. At left: method of the point of fantasy; at right: set point method. In both cases, magnetic declination as of 1500 and distances less than 50 degrees (parallels and meridians spaced 10 degrees). The lines radiating from the centre are loxodromes

Example 6: Latitude model – North Atlantic and Mediterranean (Figure 6, right) • • • • • •

Method = Loxodromes, Set point method w = 0.5 (distances and directions) Input points = Use geographic grid Constraints = Distances less than 50 degrees Show lines = Loxodromes (origin: lat=40; long= -10) Magnetic declination = 1500.

This is a hypothetic result of the application of the latitude model, based on the set point method, to the representation of the North Atlantic and the Mediterranean, considering the effect of magnetic declination as of 1500 and using only distances less than 50 degrees.

Using Empirical Map Projections for Modeling Early Nautical Charts

239

Example 7: Minumum error world map (Figure 7) • • • •

Method = Loxodromes, Point of fantasy w = 0.5 (distances and directions) Distortion info = Max angular distortion Magnetic declination = Nil

This is a representation of the world where the distortion of rhumb-line directions and distances is minimized. Because magnetic declination is zero everywhere, the solution is identical to considering the set point method. 70 40

80

60 50 20

30

10 5

5

40

6050 30

80

70

60 50

40

20

34 2

20

10

20

30

10 10

2

30

20

50 60

30

20

40 80 70

20

5 43

10

10

40

5

20

40

30

5060

80

50 0 6

70

Figure 7: Example 7. A minimum error world map showing lines of maximum angular distortion (degrees). Method of the point of fantasy, no constraints

Examples 1, 3 and 4 illustrate exact solutions and are presented as test results. In Figure 6 notice how the same spatial distribution of the magnetic declination is reflected differently on the geometry of the two types of representations: curved meridians in the first case (method of the point of fantasy); straight and equally spaced meridians, in the second (set point method). Figure 7 illustrates a hypothetical use of the model to construct a minimum error projection where the distortion of rhumb-line directions and distances is minimized.

240

J.A. Gaspar

4- Modeling early nautical charts In this section two sets of simulations of the geometry of the Cantino planisphere are presented, the first applying to the Mediterranean and Black Sea, and the second, to the Atlantic and Indian Oceans. The Cantino planisphere, made by an anonymous Portuguese cartographer in 1502, is one of the most precious monuments of our cartographic heritage. It shows the world, as it was known after the exploratory missions of the end of the fifteenth and the beginning of the sixteenth century, represented in the way of the navigational charts of the Renaissance. Although no graphical scale of latitudes is explicitly shown, such scale is implicit in the position of the Equator and Arctic Circle5. Contrarily to some portolan charts of the sixteenth century, where latitude scales were added to the old patterns, with no relation to the real latitudes of the places, this scale plays a fundamental role in the new cartographic model of which the Cantino planisphere is one of the earliest known examples6. Two distinct cartographic models were used to draw the chart: the portolan-chart model, based on the method of the point of fantasy; and the latitude model, based on the set point method. The first was used to represent the Mediterranean, Black Sea, northern Europe and the Caribbean Sea; the second, to represent part of the Brazilian coast, the southern Atlantic and the Indian Ocean.

5

And not between the Equator and the tropics, which are misplaced on the chart. For a detailed discussion see (Gaspar 2010, 142-46). 6

The planisphere of Juan de la Cosa, dated 1500, also depicts the Equator and Tropic of Cancer but their positions are approximate.

Using Empirical Map Projections for Modeling Early Nautical Charts

241

Figure 8: Geographic grid implicit to the Cantino planisphere in the area covering the Mediterranean and Black Sea. Notice the counterclockwise tilt of meridians and parallels

Figure 8 shows the geographic grid of meridians and parallels implicit to the Cantino planisphere, for the area of the Mediterranean and Black Sea, interpolated with MapAnalyst. Notice how the axis of the Mediterranean is tilted counterclockwise, as a result of the uncorrected magnetic declination, confirming that no astronomical observations are incorporated in the representation.

Figure 9: Model output for w = 0.8, dmax = 15° and magnetic declination as of 1200. The vectors represent the displacements relative to the grid implicit to the Cantino planisphere. The areas of the circles are proportional to the corresponding displacements

242

J.A. Gaspar

To simulate the grid shown in Figure 8, several outputs of the EMP model were produced, using the method of the point of fantasy and different combinations of the parameters w (relative weighting of distances and directions) and dmax (maximum distance allowed), and different spatial distributions of the magnetic declination, estimated using the geomagnetic model of Korte and Constable (2005). The results were compared with the interpolated grid of the original chart, using MapAnalyst. To evaluate their quality, the following parameters were used: - s : root-mean-square displacement, s =

∑d

2 i

N;

- µ : average displacement, μ = ∑ d i N ;

− α : average rotation angle between the chart’s implicit graticule and the model output, where di is the distance between the positions of control point i in the chart’s implicit graticule and in the model output, and N is the number of points. The best results were obtained for w between 0.7 and 0.8, dmax between 10° and 15° and magnetic declination as of 1200. Figure 9 shows the model output for w = 0.8, dmax = 15° and magnetic declination as of 1200. The values of the quality parameters, for this case, were s = 0.7, μ = 1.0 and α = 0.7°, where the units of s and μ are close to the length of one degree of latitude. It is interesting to notice how the best matches were obtained for a value of w between 0.7 and 0.8, which indicates that, although both courses and distances were used to construct the chart, a larger weight was given to the courses. Such result is expected not only because a larger uncertainty was associated with the estimated distances but also because courses were usually a more important piece of navigational information than distances. As for the maximum distance allowed, dmax, notice that 15° corresponds to about 900 nautical miles, which is a little less than half the longitudinal length of the Mediterranean. Also interesting is the average tilt of the interpolated grid (about 8°), approximately matching the average value of the magnetic declination in the region, during the period 1200-1300. This suggests that the orientation of the Mediterranean was not corrected since

Using Empirical Map Projections for Modeling Early Nautical Charts

243

the time the earliest portolan charts were produced, in the thirteenth century. The results confirm the conclusions previously reached by Gaspar (2008), for the case of Jorge de Aguiar’s chart of 1492. The second set of simulations applies to the area of the chart representing the Atlantic and Indian Oceans, to the east of the Hindustani peninsula (including the Mediterranean and Black Sea). Figure 10 shows the geographic grid implicit to the Cantino planisphere, which was estimated on the basis of about 240 control points. The visual inspection of the grid clearly suggests that neither the Mediterranean and Black Sea, nor the northern Atlantic, were represented using observed latitudes.

Figure 10: Interpolated geographic grid of the Cantino planisphere for a set of control points from which positions known to be wrongly placed on the chart were excluded. The small circles are the control points

Contrarily to the previous simulations, in which the model’s input was a sample of meridian and parallel intersections, a set of tracks supposedly representative of the routes utilized for constructing the chart was used in this case. The method of the point of fantasy, with w = 0.8, was applied to the Mediterranean, Europe and the Caribbean Sea, and the set point

244

J.A. Gaspar

method, to Africa, Brazil and the Indian Ocean. Three simulations were made, for different distributions of the magnetic declination. All values refer to 1500, except in the Mediterranean, where the year of 1300 was chosen: Simulation A: magnetic declination derived from the observations made by D. João de Castro in 1538 (South Atlantic and Indian Oceans) and from the outputs of the geomagnetic model of Korte & Constable (2005) (North Atlantic, Mediterranean and North Sea); Simulation B: magnetic declination as yielded by the geomagnetic model of Korte and Constable (2005); Simulation C: magnetic declination zero everywhere.

A

C

B

Cantino

Figure 11: Output of simulations A through C compared to the Cantino grid. The vectors represent the displacements of the control points relative to the Cantino grid. The areas of the green circles are proportional to the displacements

The relative quality of the simulations was assessed using two parameters introduced before: the root-mean square displacement (s) and average positional displacement (µ). Figure 11 shows the output of the three

Using Empirical Map Projections for Modeling Early Nautical Charts

245

simulations, compared with the original grid. As expected, the results confirm that the output of Simulation A is closer to the geometry of the Cantino planisphere than the other two, which validates the spatial distribution of the magnetic declination derived from historical observations. Figure 12 shows the output of simulation A, together with the tracks used as input. 40W 60N

20W

0º 20E 40E

60W 40N 60E

20N

0º

20S

Figure 12: Output of simulation A showing the tracks used as input. Meridians and parallels are spaced five degrees. Dashed lines indicate the use of the set point method

In general, the main geometric features of the grid, including the convergence of the meridians and the spacing between parallels in the different areas of the chart, are well reproduced by the model. Notice how the strong distortion of the meridians in the area of the Red Sea, caused by magnetic declination, is well simulated. Still no detailed analysis of the output is justifiable due to the various other factors affecting the geometry of the chart. That is the case of the scaling errors in the representation of Northern Europe and the Caribbean Sea, and the latitude errors in some parts of the African coast, originated in the construction of the planisphere.

246

J.A. Gaspar

5- Final remarks A numerical model with the purpose of simulating the geometry of early nautical charts was presented and tested. Two sets of simulations of the Cantino planisphere were produced: one for the area covering the Mediterranean and Black Sea; and the other for the area covering the Atlantic and Indian Oceans. The results were satisfactory, in the sense that the main geometric features of the chart were well replicated by the model. Still the existence of various errors and mistakes, originated in its construction, prevented a more detailed comparison between the original and the simulations. The EMP model proved to be an effective and easy-to-use research tool. It was used, not only for producing the graphical simulations illustrated in the article, but also for quickly assessing the influence of the various factors affecting the geometry of the old charts. However only a limited fraction of its capacities, focused on a relatively narrow cartographic subject, was applied to this research. Other applications are possible, such as the production of ad-hoc maps with specific geometric features or the illustration of map projections’ properties for educational purposes. As an empirically-developed prototype based on geometric considerations, the EMP model is still a relatively inefficient instrument and can be improved in several ways. In particular, a mathematical formalization of the optimization algorithm is desirable, with the purpose of proving the convergence of the iterative process and help improving its computational efficiency.

References Cortesão A, Albuquerque L (1968) Obras completas de D. João de Castro, 4 vols. Academia Internacional de Cultura Portuguesa, Coimbra Gaspar JA (2008) Dead reckoning and magnetic declination: unveiling the mystery of portolan charts’. e- Perimetron, Vol. 3, No. 4: 191-203 Gaspar JA (2010) From the Portolan Chart of the Mediterranean to the Latitude Chart of the Atlantic: Cartometric Analysis and Modeling. Doctoral dissertation, Universidade Nova de Lisboa

Using Empirical Map Projections for Modeling Early Nautical Charts

247

Jenny B (2006) MapAnalyst: a digital tool for the analysis of the planimetric accuracy of historical maps’. e-Perimetron, Vol. 1, No. 3: 239-245 Jenny B, Weber A, Hurni L (2007) Visualizing the Planimetric Accuracy of Historical Maps with MapAnalyst’. Cartographica, Vol. 42, No. 1: 89-94 Korte M, Constable, C (2005) Continuous geomagnetic field models for the past 7 millennia: 2. CALS7K. Geochemistry, Geophysics, Geosystems, Volume 6, Number 1. AGU and the Geochemical Society Nunes Pedro (2002) – Obras, Vol. I: Tratado da Sphera; Astronomici Introductorii de Spaere Epitome. Fundação Calouste Gulbenkian, Lisboa [1st edition: Lisboa, 1537] Tobler W (1977) Numerical Approaches to Map projections. In Kretschmer E. (ed.) Studies in Theoretical Cartography, p. 51-64. Deuticke, Viena

Planet and Space Cartography

Requirements for Planetary Symbology in Geographic Information Systems A. Nass1, 2, S. van Gasselt3, R. Jaumann2, 3, H. Asche2 1

Institute for Planetary Research, Department of Planetary Geology, German Aerospace Center (DLR), Berlin, Germany [email protected] 2

Geoinformation Science Research Group, Department of Geography, University of Potsdam, Potsdam, Germany 3

Planetary Sciences and Remote Sensing, Institute of Geological Sciences, Freie Universitaet Berlin, Berlin, Germany.

Abstract A large number of developments have significantly influenced digital mapping in the last few years which also affected the field of planetary mapping. Like in Earth-based geosciences the mapping nowadays are mainly conducted in Geographic Information Systems (GIS). In order to simplify the geological/geomorphological planetary mapping process on the mapper’s side, and harmonise the variety of mapping results we currently work on embedding a set of standardized mapping symbols within GIS. Such a symbol catalogue enables the user to visualise unit entities that were delineated during analysing and interpreting planetary surface in a homogeneous and cartographically correct way. As the symbols are to be used by different mappers at various locations, our main aim is to develop a portable symbol catalogue. In this context, we here describe a number of different approaches in order to find an efficient storage of symbol sets and to create an appropriate interface between a symbology layer and the (physical) underlying database model. This entire task forms one component in an overarching project that focuses on the generation of an extensible and modular geodatabase model to meet growing scientific and technical needs in planetary mapping.

A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 251 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_16, © Springer-Verlag Berlin Heidelberg 2011

252

A. Nass et al.

1- Introduction and Background The field of digital mapping has seen a large number of new developments over the last few years. Moving beyond analogue mapmaking, key innovations range from [a] mapping with the help of vector-based graphic software, to [b] more complex GIS-based mapping which provides a common spatial domain and context for planetary data, as well as [c] recent endeavours to provide platforms for interactive web-based mapping. Within this context, major efforts have been made to arrange and manage data efficiently in geospatial databases and to support standardization initiatives (e.g. for digital cartographic symbolization). In view of the steadily growing international interest in the exploration of the planets in our Solar System, the planetary mapping community has been engaged in important systematic mapping activities. Surveycoordinated and agency-funded scientific mapping projects have been engaged in producing geological and/or geomorphological maps of planetary surfaces, in particular of the Moon, Mars and Venus. Since the early 1960s, such programs have spawned a wealth of analogue and digital map sheets, map series and catalogues (e.g., LPI 2006, LPI 2009, Astrogeology Science Center 2010). A variety of individual maps, map sheets and catalogues were created with a thematic map focus, including e.g. the series of multilingual (relief) maps of terrestrial planets and their moons (e.g., Shingareva and Krasnopevtseva 2001, Shingareva et al. 2002 Shingareva et al. 2003 Buchroithner 1999). Numerous thematic maps have emerged from scientific research projects. Auxiliary cartographic information and data collections for planetary map production are being archived by the United States Geological Survey (USGS) (US Geology Survey 2010) who are developing and continuously updating their material in publications and documentations like the Gazetteer of Planetary Nomenclature. Another project conducted by ELTE University (Hargitai 2006) deals with the visualization and nomenclature of planetary maps. The Commission on Planetary Cartography (ICA) was established in 1999 (Shingareva et al. 2006) with the overall aim of harmonizing international planetary cartographic activities and developing reference materials to support the global dissemination of planetary cartographic information. It has launched a number of projects, such as the creation of multilingual planetary maps. Today, digital mapping is usually conducted in and supported by GI software environments. In terms of technical improvements such as GIS-based integration, analysis and visualization of data from other planets, planetary mapping has undergone essentially the same developments as the Earth-

Requirements for Planetary Symbology in Geographic Information Systems

253

based geosciences. Yet, making maps of other planets comes with a number of specific requirements concerning planetary cartographic systems and the definition of reference bodies. Some projects focus on automated cartographic data processing to generate topographic and thematic planetary maps (Gehrke et al. 2006), or work on improvements in the fields of data integration, management, processing aspects and analysis within a geodatabase context. One area currently ongoing vast improvements is that of planetary cartographic and map symbolization and its implementation within a GIS framework, based on the Digital Cartographic Standard for Geologic Map Symbolization (FGDC 2006) developed by the USGS on behalf of the Federal Geographic Data Committee (FGDC). This document lays down rules for the cartographic visualization of geological and geomorphological features both on terrestrial and planetary maps, and defines a catalogue of point, line, and areal symbols. This FGDC symbology can serve as a reference standard for GIS-based mapping, too. In order to simplify geological and geomorphological mapping while making efficient use of the standardized symbols within a GI system we created each particular symbol of the planetary geological features (appendix 25). As a response to working-group requirements, a set of additional symbols was created, and the entire collection was implemented in a commercial GIS environment. As the symbols are to be used by different mappers at various locations, we are currently seeking to develop an efficient way of making the GIS symbol catalogue portable. This task is one component in our presently superior project that focuses on the generation of an extensible, scalable, generic and modular database model to meet growing scientific and technical needs in planetary mapping (Nass et al. 2010, van Gasselt and Nass 2010, van Gasselt and Nass 2011). However, all modules will first be generated as stand-alone packages for individual but proprietary usage and will be integrated into the steadily growing database model in several subsequent steps. At the current project phase, we are drawing up different scenarios to find an efficient storage of symbol sets and to create an appropriate interface between symbology and the underlying database model.

254

A. Nass et al.

2- Approach & Methods We use ESRI's ArcGIS as our current development basis. Widely employed in the planetary community, this software is also used in the context of the USGS-led planetary mapping program. Thus, all generated components and conceptual approaches realized as part of this work have so far been software-dependent, i.e. proprietary. However, for the future we are looking at software-independent solutions with all map components converted into open formats.

2.1 Selection and implementation of the symbols As a reference source for compiling the symbol set required for planetary mapping, and more specifically for scientific maps with a geological and geomorphological focus, we used the Digital Cartographic Standard for Geologic Map Symbolization (FGDC 2006). This document contains detailed descriptions to visualize scale-independent symbols. However, since this standard was primarily developed for terrestrial mapping and includes a large number of symbols that are not required in planetary mapping, we concentrated mainly on chapter 25 – Planetary Geology Features. We complemented this planetary symbol set by adding a number of symbols from chapters 12 – Fluvial and Alluvial Features, 18 – Volcanic Features, and 33 – Suggested Ranges of Map Unit Colors , which gave us a singlesource collection of usable cartographic symbols to be distributed among the planetary mapping community (see Figure 1). This symbol set should to be linked to our data model framework to provide a more efficient way of managing, arranging and describing planetary mapping data. Conceivable details will be discussed in a later part of this article. For a proper treatment of undefined spatial objects and for handling any statistical data that may emerge in the course of planetary mapping we generated four additional symbol classes. (1) Vacant objects for individual graphical modifications and adjustments made by the mapper, and (2) uncertain spatial objects, to denote objects that cannot be described and assigned properly at the time of mapping. For representing thematic statistical data we defined symbols for (3) qualitative data, which can be varied in colour or shape, and for (4) quantitative data, which can be varied in size or brightness/saturation.

Requirements for Planetary Symbology in Geographic Information Systems

255

25.01 25.02 25.03 25.04 25.05 25.06 25.07 25.08 25.09 25.10 25.11 25.12 25.13 25.14 25.15 25.16 25.17 25.18 25.19 25.20 25.21 25.22 25.23 25.24 25.25 25.26 25.27 25.28 25.29 25.30 25.31 25.32 25.33 25.34 25.35 25.36 25.37 25.38 25.39 25.40 25.41 25.42 25.43 25.44 25.45 25.46 25.47 25.48 25.49 25.50 25.51 25.52 25.53 25.54 25.55 25.56 25.57 25.58 25.59 25.60 25.61 25.62 25.63 25.64 25.65 25.66 25.67 25.68 25.69 25.70 25.71 25.72 25.73 25.74 25.75 25.76 25.77 25.78 25.79 25.80 25.81 25.82 25.83 25.84 25.85 25.86 25.87 25.88 25.89 25.90

*

*

*

n.i.

*

25.91 25.92 25.93 25.94 25.95 25.96 25.97 25.98 25.99 25.100 25.101 25.102 25.103 25.104 25.105 25.106 25.107 25.108 25.109 25.110 25.111 25.112 25.113 25.114 25.115 25.116 25.117 25.118 25.119 25.120

v? 25.121 25.122 25.123 25.124 25.125 25.126 25.127 25.128 25.129 25.130 25.131 25.132 25.133 25.134 25.135 n.i. 12.05 12.06 12.07 12.08

n.i.

n.i.

n.i.

n.i.

18.41 18.42 18.43 18.44 18.45

33.010 33.030 33.050 33.070 33.0X0 33.057 33.07X 33.036 33.047 33.05X 33.A60 33.270 33.3X0 33.150 33.370 33.5X0 33.033 33.055 33.077 33.0XX 90.01 90.02 90.03

91.01 91.02 91.03

92.01 92.02 92.03

93.01 93.02 93.03

Figure 1 Symbols implemented in ArcGIS. Numbers are the reference number used in the FGDC standard (* refer to point symbols which were modified in absolute size in this graphic.

Three different methods are available in ArcGIS for the implementation (creation, modification and management) of individual cartographic symbols and symbol sets. 1. Modifying existing symbols: the fastest way to assign graphic properties to a spatial object is by selecting and modifying an existing symbol. For this, the mapper uses the symbol selector to access predefined symbols listed in categories, thus narrowing the search for a required symbol. The main advantage of this method is that it generates symbols directly and flexibly, and that it can be further refined on the user level to permit local feature representation and a local, project-based map design.

256

A. Nass et al.

2. Using styles organized in a style manager: technically, predefined colours, symbols and their specific properties as well as individual map elements are collected and stored in a binary file, one for each required symbol category. They can be chosen and applied by the mapper to represent particular map elements. Using this method meets symbolization standards, helps to promote consistency in the organization of maps and facilitates sharing symbol sets between different users. 3. Using Representation classes: representation classes allow the mapper to visualize data using a rule-based symbol structure. This structure is stored inside a geodatabase along with the data, and is organized at the feature-class level. Thus, symbols are stored with and related to the data and can easily be shared by the entire community. The benefit of this approach is that it permits deriving different map products from a single database because a single feature class can store multiple representations simultaneously. All these three methods are proprietary, thus limited to the ArcGIS environment, and consequently lack any portability between GI systems. Furthermore, although method (1) is easy to apply, it is not directly transportable between different users as data have to be converted into a specific binary layer file storing and organizing all properties. Although method (3) provides very good portability, handling and modification of symbols, the representation framework is quite complex and not easily accessible for the casual mapper. Thus, we decided to employ the style-definition method as it is the most accessible one so far, and the generated symbology style can easily be loaded for immediate use.

2.2 Portability of symbols To permit a most efficient use of the symbol catalogue it is important to make it easily exchangeable between different users/mappers, given the fact that the mapping community operates at highly distributed global locations. Symbol data files should therefore be distributable between users and easily importable within ArcGIS, i.e. transportable and independent of a mapper's working place. Furthermore, the data files should be designed in future for portability between different GI systems, too (e.g. as free and open software).

Requirements for Planetary Symbology in Geographic Information Systems

257

Regarding their transportability within ArcGIS, symbol data files can be exchanged using any of the three implementation methods, but each of them requires different amounts of time for project incorporation. A symbol catalogue designed in the form of a style file can easily be shared and offered for download via the Internet. Such a style file includes a definition for graphical visualization and a (meta-)description for each particular symbol. Where portability between different GI systems becomes an inevitable requirement, the handling of symbols is likely to be more complex. To achieve full portability of symbol sets, a number of new developments and methods are needed for converting and transporting symbologies into different file formats. As symbols are made up of vector-based definitions, the file and symbol description format must be vector-based as well to ensure that shapes and other (carto)graphic properties are described in a uniform and software-independent manner. The XML-based Scalable Vector Graphic (SVG) developed by the World Wide Web Consortium (W3C) and first presented in the year 2000 (WG3 2003, WG3 2008) seems to be an appropriate answer. Within an SVG script both, graphical attributes such as colours, type of line, stroke width or pattern as well as the object geometry of each individual object are described and stored in a consistent format, permitting cartographic vector-based map elements to be displayed on different systems (see Fig. 2). The principal structure of an SVG script is hierarchical, and the shapes of all graphical elements described and displayed by SVG are made up of simple basic elements (graphical primitive). In addition to their shape these elements can be further characterized by attributes such as filling and various types of boundary lines and text styles. Additional options provided by SVG scripts are element transformation and animation. As maps and map symbols deal with different topical spatial objects and different symbol geometries, description of symbol representations must also be variable, which (may) involve a number of difficulties. The SVGscripting for a point symbol, which represents, e.g. a local feature in small scaled maps, is static, and the symbol’s scale changes only with map scales. So an SVG script needs to be generated only once and can be used without further modifications. SVG scripts for linear or areal features are more complicated because both, the symbol's shape resp. graphical properties as well as the SVG script have to be adaptable. This means that the SVG script has to be split into two components. The first component has to statically describe graphical attributes for any particular symbol, e.g. colour (RGB, CMYK or HSV colour model), type of line (e.g. dashed or

258

A. Nass et al.

dotted, starting point, curvature and stretch), object orientation (geographic or arithmetic angle), stroke width (point or mm-declaration), text (font face and size in points) or pattern (vector-/ or image-(raster-) based description). The second component must include geometry information on each object, which stored within any particular vector-shape file. However, the exact physical implementation and assignment of symbols and object geometries is beyond the scope of this work as it is treated on the implementation level of a given GI system. A)

B)

Figure 2: A) Representation for knob or central peak (FGDC reference number 25.82) (FGDC, 2006). B) Generated script for representation in *.svg-format.

Requirements for Planetary Symbology in Geographic Information Systems

259

2.3 Linking symbology with the database model To ensure an efficient and standardized mapping workflow and uniform map results in the future, a direct link between the symbology and the underlying database model must be established. The exact way to do this and the amount of time and effort required to integrate symbologies into a GIS depends on the underlying data structure and on whether the framework is designed to handle and manage all the various data types needed for storing and accessing symbol sets. The only way to join up graphical information with a database within ArcGIS is, thus far, to use representations as described above. Our aim, however, is to explore the possibilities of generating a model that has maximum software flexibility as well as being extendable and generic. Consequently, the task is to find an adequate new approach to integrating the generated symbol catalogue into the database model. Managing symbologies within ArcGIS in the form of styles makes it easy to exchange and transport data between users. This method involves storing the symbol catalogue as a Microsoft DataBase file (*.mdb-format) in which each relation represents a way in which map elements are to be presented (line symbols, fill symbols, area patches). Each symbol element is referenced against a unique ID which is the relation’s primary key field. A symbol’s name and its associated category are stored as text strings in additional attribute fields. The symbol’s graphical information is more complex, which requires it to be stored as a Binary Large Object (BLOB). This data type allows storing and managing large binary multimedia data (e.g. raster graphics) dynamically. The current drawback is that there is no possibility so far to establish a direct link between symbologies and an underlying database model. However, the structure of a geodatabase could easily be expanded to include a style’s relational data organization by adding, e.g., associated relations. To determine, which is the most useful approach to implementing a symbology link into a database model we generated a number of different scenarios that are all geared to mapping at feature-class level (see Figure 3).

260

A. Nass et al.

Scenario 1

Example

FC RT Spatial Object Symbols id id (1,n) RT geometry ref_no PK Chapter symbol FK label (1,1) (1,n) id detail chapter_no FK no PK (1,n) label

Spatial Object id geometry symbol detail ... 82 ... Symbols id ref_no label chapter_no ... 25 ... 82 knob Chapter id no label ... 18 Volcanic ... 25 Planetary

2 FC Spatial Object id geometry chapter symbol detail

Subtypes code (chapter_no) description (chapter_label)

Spatial Object id geometry chapter symbol detail 25

Subtypes code/description Domains 12 Fluvial + Aluvial code (symbol_ref_no) 18 Volcanic Feat. 25 Planetary Geol. description (symbol_label) 33 Map-Unit Colors

Coded domain code/description 41 volc. fissure 42 ... Coded domain code/description 54 Depression 55 ...

3 RT FC Spatial Object (1,n) Symbol id PK id geometry chapter symbol FK (1,n) ref_no detail

Spatial Object id geometry symbol detail ... 513 ... Symbols id ref_no label chapter_no 25 82 knob 513 ...

Figure 3: Entity-Relation-model for scenario 1-3 with exemplary attribute tables within ArcGIS (FC stands for feature class, RT stands for relation table).

Scenario 1 is composed of two relations (i.e., tables (RT)) that are directly related to the feature-class (FC) attribute symbol. The first RT, Symbols, contains the GIS-managed object identifier id, the symbol reference number ref_no, as an artificial primary key (PK) composed of the FGDC symbol reference number and the appendix chapter number, a description of the feature resp. symbol label, and the attribute chapter_no that contains the appendix chapter as a foreign key (FK). The second RT Chapter has the attribute id, the appendix chapter number chapter_no, as primary key, and the label. Scenario 2 uses subtypes and domains to describe objects on the objectrelational level, permitting a convenient assignment of symbols on the

Requirements for Planetary Symbology in Geographic Information Systems

261

basis of a specific subtype. These subtypes are FC entities that share common attributes. For a direct assignment of symbols, the FGDC appendix chapter (chapter_no and chapter_label) represents the subtype. Once a selection of the chapter-no subtype has been made, this limits the possible values for the domains defined by the reference number (symbol_ref_no and symbol_label) of each symbol. Scenario 3 consists of a RT Symbol that is linked to the FC entry symbol, and carries the attribute‘s id (PK), chapter and ref_no. As some symbols need further editing, e.g. the assignment of an orientation for point features, we define an additional attribute field as pseudoBoolean with a value of 1 if the object’s symbol can/must be rotated. If this is the case, the mapper can use another attribute termed ‘angle’ for assigning values to symbol rotations. All the above scenarios deal with symbols on the basis of simplestructured relations that can easily be updated and modified as well as extended if required. Concerning symbol selection, all symbol descriptions are included as attribute value, which permits the mapper to navigate directly to any desired symbol. Thus, within all these scenarios the mapper no longer requires a printout of the FGDC standard as a lookup reference for the correct assignment of symbols, contrary to what is recommended e.g. in the Geologic Mapping Template developed by ESRI (ESRI Cartography Team 2009). By employing different subtypes, the mapper can pick the appropriate symbol by navigating in a hierarchical order through the respective (subtype control) attribute domains, which prevents the input of erroneous data and limits data inconsistencies. To illustrate how this the symbol implementation affects work on the user side, we include the example of an ArcGIS-based solution (see Fig. 4). In this instance, the user has to decide 1. if the spatial object is a geological or geomorphological surface type, 2. if the object is, e.g., a volcanic, fluvial/alluvial, aeolian unit type, and 3. if it belongs under the subtype of, e.g., dark-coloured or light-coloured ejecta. After this hierarchical selection and sub-selection the user has access to one particular or to a very confined range of symbols holding the relevant and predefined attributes.

262

A. Nass et al.

FGDC id

select surface type select unit type

select sub type Figure 4: Access to symbology within ArcGIS by using subtype-controlled attributes.

Regarding the possible symbol implementation choices presented so far it must be noted that none of the above scenarios describes an SVG-based solution in detail. As this is not the focus of this work and the topic is predominantly treated on the physical software level, nevertheless we would like to point out here some of the possibilities and requirements for an efficient handling of such a symbol implementation. An SVG-based description contains all the information (graphical representation and geometry) necessary to describe a vector graphic resp. a cartographic symbol in a standardized form. In order to handle an SVG script and its future implementation in a database model efficiently, we need to • evaluate methods and define efficient ways of storing SVG scripts within a relational database model, • find a way to address and physically access the SVG script in order to assign the defined graphical vector attributes to a particular object, • develop a possibility to split the SVG script into a static component for graphical symbol attributes and a flexible component for the object's individual geometry information, which can then be copied from the vector shape file automatically. An alternative option would be to store only the graphical symbol information as SVG script in an attribute value, while the geometry information is directly transferred from the vector shape file.

Requirements for Planetary Symbology in Geographic Information Systems

263

Another method to bind the graphical symbol information to the database model is a process known as XML shredding. Under this method, information is no longer stored in a single XML text file, but is rather split into individual XML-defined hierarchical values and stored as attribute values in the appropriate relations inside the relational database (see Figure 5). This option is suitable, if the SVG-script is split into hierarchical components, so that only the graphical attributes for the respective symbol are stored and the feature class' geometry is adopted from feature class attributes in parallel. The example in Figure 5 illustrates the shredding capability and theoretical possibilities of binding symbols to a given database model. The graphical information stored within the SVG script is now displayed in the form of relational attribute values. As one complex symbol is usually composed of several basic elements, the symbol relation also contains various attributes, e.g., colours, or stroke width. Another variant that exists besides general shredding is a process called partial shredding. This method might be useful in cases where not the entire SVG script is to be stored within relations and where a complex relational target schema becomes necessary to decompose the XML material. In our project this problem occurred e.g. with regard to the geometric information of cartographic elements/symbols.

264

A. Nass et al.

B)

A)

e_ID x 1 0 ... 2 ... 3

y width 0 595.28 ... ... ... ...

height 841.89 ... ...

CREATE TABLE elements( e_ID Integer, x Integer, y Integer, width Float(2,4), height Floar(2,4))

... ... ... ...

p_ID path fill 1 2 3

none none ...

stroke- strokewidth miterlimit #000000 1.0081 10 #000000 1.0081 10 ... ... ... stroke

CREATE TABLE path( p_ID :

d ... ... ...

Integer, Varchar(8), Varchar(8), Float(2,4), Integer, BLOB)

Figure 5: Exemplary xml shredding to link svg scripts into a relational database model (A) shows the svg script, and B) illustrates the splitting and entry transfer from svg-based text file to different relation tables).

Requirements for Planetary Symbology in Geographic Information Systems

265

4- Conclusion and Outlook In this paper we pointed out the requirements for planetary symbologies in geographic information systems. The main goals of our current project are to: 1. enable a standardized and easily accessible use of cartographic vectorbased symbols within ArcGIS 2. assign specific symbols to a particular spatial object/map element automatically if the spatial object is defined by context information in the form of descriptive attribute values, and 3. use all symbols software-independently. With respect to the discussed requirements we conclude with the following recommendations: • Solve SVG scripting challenges and difficulties on a conceptual level as long as no GIS-based implementation is realized. • Individual objects and even complete sets of mapping data could be efficiently arranged, managed, and stored by subtype-controlled symbology assignment. The main part of our current work is to review, edit and refine the requirements formulated above. We intend to focus on the following issues for the future: • Besides the standardized graphical display of symbols, we will formulate a number of additional requirements concerning the correct cartographic representation of planetary surface features, such as the still unsolved issue of symbol generalization within GIS. • We intend to develop alternatives for a standardized and softwareindependent storage of vector-based graphics. Symbology storage within a geodatabase requires creating further possibilities beyond SVG-shredding. • Beyond symbology, other components, such as a detailed and efficient description of (meta-)data at base data (raster) and map data (vector) level, will be implemented in ArcGIS and subsequently connected to our database model.

266

A. Nass et al.

References Astrogeology Science Center (2010) Map of the Planets and Satellites. http://astrogeology.usgs.gov/Projects/MapBook/. Accessed June 2010 Buchroithner M (1999) Mars map — the first of the series of multilingual relief maps of terrestrial planets and their moons. In: 19th ICA/ACI International Cartographic Conference, Ottawa, Canada, pp. 1–3 ESRI Cartography Team (2009) Geological Mapping Template. Technical Report ESRI FGDC – Federal Geographic Data Committee (2006) Digital Cartographic Standard for Geologic Map Symbolization. Federal Geographic Data Committee prepared by the U.S. Geological Survey (FGDC-STD-013-2006) Gehrke S, Wählisch M, Lehmann H, Albertz J, Neukum G, HRSC-Team (2006) Planetary Mapping with ‘Pimap’. In: 5th International Symposium Turkish-German Joint Geodetic Days Hargitai H, (2006) Planetary maps: visualization and nomenclature. In: Cartographica: The International Journal for Geographic Information and Geovisualization 41 (2), 149–164 LPI – Lunar and Planetary Institute (2006) Mars Map Catalog. http://www.lpi.usra.edu/resources/marsmaps/. Accessed March 2010 LPI – Lunar and Planetary Institute (2009) Lunar Map Catalog. http://www.lpi.usra.edu/resources/mapcatalog/. Accessed March 2010 Nass A, van Gasselt S, Jaumann R, Asche H (2010) Implementation of cartographic symbols for planetary mapping in geographic information systems. Planetary and Space Science. doi:10.1016/j.pss.2010.08.022. Online available Shingareva K, Krasnopevtseva B (2001) Venus map — the series of multilingual maps for terrestrial planets and their moons. In: 20th ICA/ACI International Cartographic Conference, Beijing, China, pp. 3279–3284 Shingareva K, Krasnpevtseva B, Buchroithner M (2002) Moon map — a new map out of the series of multilingual relief maps of terrestrial planets and their moons. In: Conference ‘GIS for Sustainable Development of Territories’. St. Petersburg, Russia, pp. 392–395. Shingareva K, Krasnopevtseva B, Leonenko S, Buchroithner M, Wälder O (2003) Mercury map — a new map out of the series of multilingual relief maps of terrestrial planets and their moon. In: 21st ICA/ACI International Cartographic Conference, Durban, South Africa, pp. 1551–1554 Shingareva K, Zimbelman J, Buchroithner M, Hargitai H (2006) The Realization of ICA Commission Projects on Planetary Cartography. In: Cartographica: The International Journal for Geographic Information and Geovisualization 40 (4), 105–114 US Geological Survey (2010) Complete Nomenclature List. USGS Astrogeology: Gazetteer of Planetary Nomenclature. http://planetarynames.wr.usgs.gov/ Accessed June 2010 van Gasselt S, Nass A (2010) Planetary Mapping – The Datamodel's Perspective and GIS Framework. Planetary and Space Science. doi: 10.1016/j.pss.2010.09.012. Online available van Gasselt S, Nass A (2011) Planetary Map Data Model for Geologic Mapping. AutoCarto 2010 CaGIS Special Issue. In press WG3 (2003) Scalable Vector Graphics (SVG) 1.1 Specification. World Web Consortium (RECSVG11-20030114/) WG3 (2008) Scalable Vector Graphics (SVG) Tiny 1.2 Specification. World Web Consortium (REC-SVGTiny12-20081222/)

Venus Mapping at small Scale: Source Data Processing and cartographic Interpretation Lazarev E, Rodionova J Sternberg Astronomical Institute, Moscow State University, Department of Lunar and Planetary Research, 119992, Moscow, Russian Federation, Universitetsky pr., 13 [email protected]

Abstract Through space technology it has become possible to study the planets of the Solar System in ever greater detail. This is particularly true for Venus, whose surface is hidden by a thick atmosphere so that the surface relief has only been observed using radar techniques. The new hypsometric map of Venus will improve and accelerate the study of its surface and the reliefforming processes there, give a precise and informative impression of the planet, and find wide use as an aid for students and scientists. The height contours were constructed using the Magellan altitude data, which is the most precise available. The main tasks in the production of the map were the development of an original hypsometric scale, the data processing, and the layout of the map.

1- Introduction Venus is among the brightest objects of the sky. Its brightness is explained by the reflection of the solar light from its thick, cloudy atmosphere. Because of this atmosphere, it is not possible to see the surface even from a spacecraft orbit about the planet. The mean radius of Venus is 6051.8 km. As other planets, Venus orbits the Sun in an anticlockwise direction, taking 225 days to do so. The period of its rotation about its own axis of 243 days was determined only in the 1960s with the early use of radar techniques. From the Earth-based radar observations two bright relief features were identified and named, from the Greek alphabet, Alpha and Beta. Using these features it was possible to discover the rotation period of the planet. Unlike other planets, where the orbital and rotational directions are the A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 267 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_17, © Springer-Verlag Berlin Heidelberg 2011

268

E. Lazarev and J. Rodionova

same, Venus rotates oppositely from its orbit – clockwise, that is why longitudes there are counted off from west to east. With the data received from the spacecrafts Pioneer Venus (Pettengill and Eliason 1980), Venera-15 and -16 (Rzhiga 1987) and Magellan (Batson et al. 1994; Wu and Howington-Kraus 1994), hypsometric maps of the planet have been created and increased in accuracy. In 2008 the Venus hypsometric hemispheres map at a scale 1:90M (figure 1) was compiled in Lunar and Planetary department (SAI) (Lazarev et al. 2009). This map was prepared on the base of heights the 64800 object points, having obtained from the Magellan altimetry data (Batson et al. 1994; Wu and Howington-Kraus 1994). The heights on Venus were referenced to a sphere of radius 6051.0 km (Lazarev and Rodionova 2006). In 2010 the new Venus hemispheres relief map at a scale 1:45M was compiled and published (figure 2). As compared with the former map the new one, source data and compiling techniques have some special features.

Figure 1: Venus hypsometric map (2008)

Venus Mapping at small Scale: Source Data Processing and cartographic Interpretation

269

Figure 2: Venus relief map (2010)

2- Data transformation and map compiling techniques For map compiling the sphere with the radius 6051,8 km was used as a reference surface (Seidelmann et al. 2007). The DTM consisted of more than 33 mln. object points and compiled on the base of the NASA spacecraft Magellan radar (SAR) data used as a source of height values. The size of radar spot was 5 to 5 km and in that way the distance between source object points was the same. For the map it would be enough to use Magellan radar DTM with the distance 0,07 degrees between object points, or 7,5 kilometer on the Venus equator. So we had to generalize the source DTM with minimum sacrifice of accuracy. The averaging process consisted of several stages. In the first, the polygon cells net was created for the whole Venus surface and superimposed with object points. The size of each polygon was 7,5 to 7,5 kilometers. Then the average height for each cell was calculated using the heights of each point having “fallen” in this cell (figure 3). Finally, latitude and longitude of cell centroid and new mean height were modified into coordinates of new

270

E. Lazarev and J. Rodionova

object point. The new DTM consists of more than 6 million object points, and it appears to be more than enough for map at a scale 1 to 45 million.

Figure 3: Source DTM processing. Using polygon cells net to calculate new object points with the distance between them 7,5 km.

The map was compiled in several stages. The DTM, of course, underlies. Then polygons were created using the spline interpolation method. The next was contours and hillshade to add a 3D effect and finally, feature names and other information. The location of central meridians of both Venus hemispheres were selected using the fact, that every 583 days Venus is positioned on lower conjunction relative to the Earth, i.e. when the planet are situated between the Earth and the Sun. In this position the hemisphere of Venus with the 320°E central longitude turns to the Earth. Accordingly 140°E longitude is the central for the other hemisphere. This situation will be observed at least 600 years (Burba 1996). Based on these conditions 320° and 140°E meridians are the central ones for the both hemispheres of the Venus relief map compiled in Lambert equal area azimuth projection. At the same time the choice of such central meridians is convenient for the largest Venus terra (Aphrodite Terra) representation, because this great relief feature are situated within the just one hemisphere entirely. The relief features names were taken from the Gazetteer of Planetary Nomenclature (http://planetarynames.wr.usgs.gov).

Venus Mapping at small Scale: Source Data Processing and cartographic Interpretation

271

3- Outlook and discussion As a result of data processing and compiling the Venus hemispheres relief map was created. Besides the main hemispheres, the small maps were compiled for the Venus Polar regions. For this map the special height color scale was used, where lowlands being below zero level represents by violet color. The precision of relief representation of the new map is much higher, than for the previous map having published in 2008 (figure 4).

Figure 4: Alpha Regio. Both Venus maps: comparison of relief representation

The DTM obtained can be used for creating different height histograms (figure 5) and for 3D modeling of Venus relief (figure 6) in different scales. In future it is planned to continue compiling relief and other thematic maps of terrestrial planets, their moons and moons of Giant planets in different scales using the developed approaches. At the same time 3D modeling and data processing will allow to make investigations of space bodies’ relief and compare them with each other in different ways.

272

E. Lazarev and J. Rodionova

Figure 5: Height histogram for both Venus hemispheres

Figure 6: 3D model of Alpha Regio

References Batson RM, Kirk RL et al. (1994) Venus cartography. JGR 99:21173–21182. Burba GA (1996) Cartographic aspects of Venus Global geologic mapping at 1:10,000,000 scale. Vernadskiy-Brown Micro 24 abstracts 11. Lazarev EN, Rodionova JF (2006) Automatic creation of the hypsometric map of Venus. BrownVernadsky Micro 44 abstracts 53-54.

Venus Mapping at small Scale: Source Data Processing and cartographic Interpretation

273

Lazarev EN, Rodionova JF (2009) Some features of the Earth, Venus and Mars global relief. Abstract book of International Conference on Comparative Planetology: Venus – Earth – Mars 78. Pettengill GH, Eliason E (1980) Pioneer Venus radar results - Altimetry and surface properties. JGR 85:8261-8270. Rzhiga ON (1987) Venera-15 and -16 spacecraft - Images and maps of Venus. COSPAR, IAU, Plenary Meeting, 26th vol. 7 12:269-278 Seidelmann PK, Archinal BA et al. (2007) Report of the IAU/IAG Working Group on cartographic coordinates and rotational elements 2006. Celest. Mech. Dyn. Astron. 98:155–180. Wu SSC, Howington-Kraus EA (1994) Magellan radar data for Venus topographic mapping.. Abstracts of the 25th Lunar and Planetary Science Conference 1519.

Planetary Nomenclature: a Representation of human Culture and alien Landscapes Henrik I. Hargitai1 and Kira B.Shingareva2 1

Cosmic Materials Space Research Group, Eötvös Loránd University 1117 Budapest, Pázmány Péter sétány 1/a., Hungary [email protected] 2

Moscow State University for Geodesy and Cartography, 105064 Moscow, Gorokhovsky per. 4, Russia [email protected]

Abstract Planetary nomenclature has two major functions: one is a cultural function: to commemorate people associated with space science; and other people, places or names representing human culture in general. These serve as the specific parts of extraterrestrial placenames. This function links extraterrestrial territories to our home planet, the Earth. The other function is to represent the true nature of the geographic (landscape) feature named. This second part of its name, the so called descriptor term is based the physical properties of the feature: its morphology (shape), and in some cases, its geological origin. The descriptor term is always in Latin, thus providing a link to medieval cartographic traditions and being neutral in a contemporary, international context. The system is standardized, having only one official form for each name. It is a problem, however, that no nation speaks Latin today; and the currently used names are mostly unknown for most of their readers. The traditions of the nomenclature (names of mythological characters, famous scientists, the use of Latin), which linked Earth and the celestial objects together in the last centuries, now alienate extraterrestrial landforms – the opposite effect from its original one. In this paper we investigate the evolution of the current Planetary Nomenclature Gazetteer and illuminate some disturbing problems related to the current practice of the use and development of the nomenclature. In the future additional nations are expected to have a more active role in A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 275 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_18, © Springer-Verlag Berlin Heidelberg 2011

276

H.I. Hargitai and K.B. Shingareva

Space Research; from cultures other than the Western (Old World) one; therefore it may be needed to reconsider some aspects of planetary nomenclature, especially its international nature which, as we show it, not as intercultural as it should be. We give some suggestions on how to improve the Gazetteer.

1- Introduction Planetary nomenclature is a representation of the culture of its creators. In this work we investigate the history and current state and policy of the nomenclature of planetary surface features. The objective of this work is to illuminate anomalies and new opportunities on the use of extraterrestrial placenames. We analyse planetary nomenclature with special focus on specific elements (designation), generic elements (descriptor term), methods of naming, and translation of names.

2- Evolution of the Planetary Nomenclature Planetary surface features in the pre-IAU era have been named by the scientists who personally studied those landforms by telescopic observations. The first observer of Mars, Huygens, only produced drawings without labels (1659), as Galilei did it in his drawing the Moon in 1609. Letters and numbers appeared in the following maps, like in those made by the German astronomer pair Beer and Mädler (1830) who labeled dark albedo features in lower case letters. This is a similar method to Thomas Harriot’s who labeled Lunar landforms with numbers in his first lunar drawings in 1609. John Philips has not named any features on Mars but distinguished two landform types: „sea” and „land”: these labels appeared as explanations on his map. The first truly labeled maps used names of contemporaries (scientists and kings), terrestrial geographical names, and abstract concepts. In 1645 Michael van Langern (Langrenus) differentiated land areas (Terra, Montes etc.), watery areas (Mare, Sinus etc.) and craters as the three main Lunar landform types, for craters having no generic term.

Planetary Nomenclature: a Representation of human Culture and alien Landscapes

277

He mainly used the names of scientists, members of catholic royalty, the nobility and catholic saints for craters; Terrestrial geographical names for Maria and other watery features, and desirable human qualities for Terrae (Whitaker 1999:45). This was the first time that an extraterrestrial landscape gained its own names. He was the first who discovered that “if the craters or seas had no names, the study of them would be practically impossible” (Loretta 1935). Hevelius did not follow his example, but likened the front side of the Moon to a map of Europe, Asia and Africa, naming the lunar features after terrestrial counterparts particularly with respect to the ancient world as known to the Roman and Greek civilizations (Hargitai 2006). The English astronomer Richard A. Proctor used the drawings of W. R. Dawes on his maps. Proctor named features (continents, lands, seas and oceans) after his contemporary and earlier astronomer colleagues who observed Mars: Philips, Dawes, Beer, Mädler, Secchi etc. Names like Dawes Sea can be therefore read as “the seas in Dawes’ drawing” etc. Naming planetary features after scientists have been a practice since Riccioli’s Lunar nomenclature (1651) which was a consistent and well developed system. He named lunar craters after philosophers and astronomers whose names were associated in some manner with the Moon. Names were arranged in chronological order from the North limb to the South, but they were also grouped by nationality, interests, teaching, and the like, with the more important names being assigned to the larger and more prominent craters (Whitaker 1971). He made the Moon “a cemetery of astronomers and pantheon of thinkers” (Moreux 1950:45). Riccioli kept the system of Terrae, Maria (and related landforms) and craters, but changed the specifics: names of Maria on his map were associated with water, while names of Terrae with soil. Since generic elements in Proctor’s map were shown in English, translations of Proctor’s popular book (Proctor 1870) included the maps with the generics also translated. Camille Flammarion also had his own nomenclature publicized widely as illustrations for the various, translated and improved editions of his book Astronomie Populaire or Terres du Ciel in the 1890s-1900s. He named features in French (continent, terre), using the same concept as Proctor, but extending and changing some names. A new Martian nomenclature has been developed in the 1870-1900s by the Italian Giovanni V. Schiaparelli. He named features after Biblical and ancient mythological places around the Mediterranean Sea. This was a new

278

H.I. Hargitai and K.B. Shingareva

concept, although Riccioli have already used few names of mythical heroes like Hercules on the Moon. Instead of naming features for contemporary persons or places, he has made Mars ancient and mysterious, and yet deeply rooted in our culture: not at all contemporary, but very old. Also for the first time, he related names of surface features to the name types of planets themselves: deities of the classical mythology. He introduced Latin in Martian nomenclature – Mare, Sinus – which may be attributed to him being Italian or that he used the traditions of the Lunar nomenclature, developed centuries ago when Latin was the common language of science. There were no Martian craters visible from the Earth but Schiaparelli did noticed a large system of canals which he also named after classical and mythological places and persons, so his system was coherent and uniform. His traditions have been followed by Percival Lowell, who created three new feature types for his canal nomenclature: single and double canals (with no generic) and oases (Aquae, Fons and Lacus) (1895). The system of canals was discarded only as late as in 1975, when analysis of Mariner 9 observations clearly and undisputably disproved their existence (Sagan and Fox 1975). An increasing number of telescopic observers continued developing Riccioli’s and Schiaparelli’s system for the Moon and Mars, respectively, in many cases naming the same feature differently at the same time. Another significant change is that by the 20th century, no Terra names were in use on the Moon. The resulting chaos was clarified by the newly formed IAU (International Astronomical Union) which compiled the first standardized catalog of „Named Lunar Formations” (1935) (Mary Andela Blagg and Karl Müller). Newly standardized names (especially those of mountains) were given in English language (e.g. Carpathian Mts). This marks a fundamental change in the naming process: from this time, names are not assigned directly by the observers but by an international commission. From 1959 on, Soviet scientists had the exclusive right to name newly observed features of the far side of the Moon, which resulted in a predominance of names of Soviet origin on the far side. Names were originally designated in Russian and have been instantly translated to English and other languages by the popular and news press (Kenny 1963). It is not by chance that IAU have decided to standardize all names in Latin form in 1961 (Sadler 1961:234). It was also decided that landforms can not be named after living persons; and their names should retain the original spelling.

Planetary Nomenclature: a Representation of human Culture and alien Landscapes

279

This marks a major change in the language of the planetary nomenclature, especially concerning the generic elements. Although it was only applied to the Moon that time, it re-established the 17th century style of uniform Latin naming. This may not be a coincidence that the United Nations Group of Experts on Geographical Names – which main task is to standardize Terrestrial placenames, i.e. selecting only one official form for international communication – was founded in 1959 (UGEGN 2001). Martian albedo nomenclature was clarified in 1958 when an ad hoc committee of the IAU chaired by Audouin Dollfus recommended the adoption of 128 names for albedo features (IAU 1960), to serve a safe base for the upcoming spacecraft missions. In the 1970s spacecraft missions experienced a “shortage of names of topographic features” and “classical lunar nomenclature prov[ed] to be outmoded, slow, and controversial” (C. Borkowski) which could be overcome only by appointing a commission for this task (UN 1971). The UN Ad Hoc Group of Experts on Geographic Names at its second meeting in New York (March, 10-20, 1970) adopted the recommendation to form a special working group on the names of extraterrestrial topographic features (UN 1970). It moved the rights of naming back to IAU, but stated that “further investigation should be performed by joint efforts of specialists in different sciences (astronomers, geographers, geomorphologists, cartographers and others)”. This has not been implemented as there is no such interdisciplinary expert group in the IAU WGPSN ever since. In 1973, for the first time, new names had to be assigned to surface features of a „new” planetary body: the then-existing names of the albedo nomenclature could not be matched with newly discovered topographic features. A commission had to decide how to proceed with naming concept, since more planetary bodies were about to be discovered in detail in the near future. The new topographic names were selected primary by the Mariner 9 team and US Geological Survey scientists in 1972, and adopted by the IAU in 1973 (Hartmann 2003). A working group of IAU and USGS astrogeologists chaired by Gerard de Vaucouleur developed a system for both specific and generic terms of the newly discovered Martian landscape in the early 1970s. They had to invent new generic terms for landforms not found on the Moon, and new themes for specifics. The decisions implied that naming policy will continue in five major lines:

280

H.I. Hargitai and K.B. Shingareva

1. the official language of the planetary nomenclature should be Latin, regarded to be neutral and international (as decided for the Moon in 1961) and by this, standardized, making translation unnecessary and difficult; including all future generic elements which may be necessary to introduce for previously unseen landform types. Making names’ meanings opaque gave grounds for later comments like the following one: “English is the lingua franca of the scientific world, virtually no one knows Latin, so it makes no sense - except to a pompous bureaucrat on an international committee - to invent new terms that won't be understood” (tychocrater 2007). Basically the same concern is expressed by W. K. Hartmann (2003): „For better or worse - probably worse – the mappers chose mostly Latin terms for topographic features, so the new Martian names can be opaque to modern readers.” 2. some places (craters on Mars) continued to be places of commemoration: named after scientists associated with Mars (following Riccioli’s Lunar tradition), but 3. themes should not be confined to physicists and astronomers, but a much wider range of scholarly and artistic figures (see the craters on Mercury) 4. other names should have a mythological (or in some places: literary) origin (following Schiaparelli’s Martian tradition), 5. the name bank generally should be as international as possible (a new idea): “it is important to make sure that the end result will be a nonprovincial distribution of nationalities, epochs, and occupations - a distribution that our great-grandchildren can be proud of” (Sagan 1976). These rules set the agenda for all future naming as the Working Group for Planetary System Nomenclature (WGPSN) formed at the 1973 IAU Sydney meeting. The strong intention towards a neutral and international system may have been attributed to several aspects: the two (English/French) official languages of IAU, the increasing international use of names (especially in popular and news press). A third reason might be the cold war situation in which Americans may have feared a Soviet dominance in naming territories discovered by future Soviet probes. In 1970, the IAU Working Group on Lunar Nomenclature used a more international approach in adopting more than 500 names, in the newly Latinized style.

Planetary Nomenclature: a Representation of human Culture and alien Landscapes

281

In 1973, in an atmosphere moving into international thinking, Carl Sagan proposed a new approach of naming: planetary nomenclature should represent not only science, philosophy and mythology of the western world, but the entire human culture. The first realization of this concept was to name Martian valleys for the word „Mars” in various languages, including Japanese, Hebrew, Arabic. Decisions of 1973 induced a flood of criticism over the new rules.

3- Some aspects of the planetary nomenclature

3.1 Names of scientists In the 1960s Gerard P. Kuiper mentioned that “there are continual proposals to add names of contemporary scientists to the crater names”. In 1955 it was suggested to name a lunar crater for Einstein. However Kuiper, who supervised Lunar nomenclature in the 1960s, suggested to keep the historical nature of it (Kenny 1963). Although the current system seems still to be a historical one, it is not any more. According to the present policy, “Commemoration of persons on planetary bodies should not normally be a goal in itself, but may be employed in special circumstances and is reserved for persons of high and enduring international standing. Persons being so honored must have been deceased for at least three years.” This rule means in practice that scientists leading or participating in space missions have a good chance to be commemorated as a crater name. Some features of Ida and Gaspra are named exclusively for “Galileo project participants”, on Eros for „scientists who have contributed to the exploration and study of Eros”, which also holds for large craters of Mars. “To the late G. P. Kuiper falls the ironic distinction of becoming the first individual to have been provisionally commemorated on no fewer than three planets, an excessive gesture that ignores the scrupulous care Kuiper took in seeing that the nomenclature of the lunar limb regions was properly revised” (Pike 1976). Names of other figures important for the American space research are also found on both Mars and the Moon (von Karman for example).

282

H.I. Hargitai and K.B. Shingareva

Another exception is possibility of naming features for living scientists, over 100 years of age. This rule went into effect when „accidentally” a crater was named for Charles Abbot, who was still alive – 101 years old.

3.2 Proprietary rights vs. alien landscape The system of IAU WGPSN rules have multiple consequences on how people perceive extraterrestrial territories. Using terms in Latin and continuing the practice of using mythological names and introducing international names: all these factors contributed to alienate the extraterrestrial landforms alien enough on their own. This way IAU have re-established the pre-classical antique concept according to which the heavens and the Earth is fundamentally different. During the 1719th century, planetary nomenclature did not show this difference: they used the same language and system as far as the landscapes permit – at least for a reader who was brought up in the culture of the Old World. Using Latin and opaque terms and names today separates the Earth from heavenly bodies again. It is good, because it separates the human Oikumene or living space from territories not belonging to our realm. It has a negative message, however, because it suggests that planetary surface features are very different from ours. Planetary geology does not completely agree with this claim. Earth is just another planetary body: surface features’ formation follow common rules. Planetary surfaces can even help a better understanding of our terrestrial landscape. Using personal names or names of terrestrial cities bring these landforms closer to us. When humans name a landform in a sense they also claim the area to be their property: by naming they may feel that it belongs to them. This is the domestication of alien landscapes. This is crucial from a psychological point of view for the astronauts working on an alien landscape (Hargitai et al. 2007). According to the present view, planetary landforms are named in order to make scientific communication easy about them. They also serve to commemorate famous scientist, artists, philosophers etc. But at the same time, all these people belong to a particular nation, country, where people rightly think that the area which bears a name of their fellow citizen, belongs to them more than other areas. Similarly, naming craters of Mars for a town implies deeply that the particular crater have a strong

Planetary Nomenclature: a Representation of human Culture and alien Landscapes

283

connection to the town of origin. It is a much stronger feeling than a commemorative name: these towns do exist and people live in there. Naming a town in the Age of Discovery also meant occupation of the land by the people from that place. When another nation occupies a territory, especially a town, they traditionally rename it. Naming craters for cities implies a certain proprietary right. It may also be regarded as brilliant PR method for calling the attention of the common people living in that town to the mere existence of planetary surfaces.

3.3 International or supranational All these themes deeply suggest a common idea: that these territories belong to us, the human race. It is also stated in the UN Treaties And Principles On Outer Space: space is a “province of all mankind”. While for a 20th century way of thinking or from a legal point of view this may be true, and nomenclature serves this idea well, a 21th century “politically correct” thinking may not agree with this. There are two ways of moving ahead from the present concept. One is following the way of thinking of Carl Sagan, who first proposed that names should be international. As a following step, not olny the origin of names should be international but the system of nomenclature should be multinational and their linguistic variants be equal. Latin language and the Roman alphabet is still a biased devotion to the culture of the Old World. In the 1960s, when Latin was introduced to be the unique language of planetary nomenclature, it has been thought to be universally acceptable and international, however, it did not take into account that most of human population uses a writing system other that Latin. An interesting and disturbing fact is that Soviet planetary mapping have never fully adopted the international system, not only because of political reasons, but simply because they use Cyrillic alphabet which requires a transformation of the Latin names (some of their Lunar catena names have never been accepted by IAU but were too important and symbolic for the Russians not to give them up). Russians kept the meaning of both elements, and translated them into Russian language and alphabet. IAU WGPSN has never stated that planetary nomenclature can exist in forms other than in Latin language and Roman alphabet, so all deviation for the official system can be regarded as an unofficial, informal if not inferior version. This also implies that nations which use other writing systems can’t express planetary names in their own writing system officially.

284

H.I. Hargitai and K.B. Shingareva

It is already and will be an acute problem in the future when Chinese, Indian, Japanese probes will discover new features on Solar System bodies. A comparison of Itokawa’s suggested name bank to the accepted one illuminates several problems originating from differences of the Asian and European way of use of language. If the planetary nomenclature wants to be truly international, i.e. the common nomenclature of mankind, it can’t have an exclusive way of expression which happens to match with the Western European / American system. It is true that until the recent years other (Asian or Arab) cultures did not show considerable interest in space research, but this has changed or is changing. This argumentation does not bring up anything that has not happened already: Russians use their own Cyrillic writing system and translate both elements of names, Japanese, Chinese, Koreans do the same. They do not use IAU’s official system, instead, they adopt this system and transform into their own. The plurality of languages is still with us: political developments did not follow those that created the USA – there is no World Government –, nor those which created the Arab world – united under the language of the Qur’an - , but current international alliances are EU-like: all languages (and writing systems) are or should considered to be equal. The Latin system can’t be a system of all mankind, it can only be regarded a variant used in international conversations, which is open for a standardized conversion to fit each nation’s traditions in its writing system and system of geographic names. Naturally, all variants should reflect the original IAU names, both specific and generic terms. IAU should help and encourage all nations to develop their own system which should be standardized and official variant of the international Latin system. For being able to create such systems, it is needed that IAU provide the following data in its Gazetteer: 1. the language of original form in addition to “ethnic or cultural group or country” 2. the name in its original form and original writing system 3. the original pronunciation of names (IPA and audio) Audio files have been already produced by ICA Commission on Planetary Cartography (Shingareva 2005, Hargitai and Kereszturi 2010). But the current position of IAU is that they do not want to include any pronunciation remarks “because there are many variations on the

Planetary Nomenclature: a Representation of human Culture and alien Landscapes

285

pronunciation of names, and the IAU does not endorse any particular one.” (Personal letter from an anonymous IAU member, 2010) As the IAU was able to successfully select one accepted witten form for each names (from the many variations), the same way one “original” pronunciation form could also be selected - which should be by a native speaker from the nation of origin of the particular name. Pronunciation can not and should not be standardized but such a service from IAU would help oral communication which is as important as the written one. Most names are transcribed using English rules; and in fact the method of transcription / transliteration used by IAU is not published. Sources of some names are simply letters and lists supplied by individual scientists participating in working groups which may be a questionable reference concerning proper transcription or transliteration (IAU2010b). Nations should not only be given the right of selecting names, but also to decide on how to use the Gazetteer: translate, transcribe or transliterate the names and terms. A major problem is that the same word or placename may be used in several forms in the IAU nomenclature (word for “hot”, word for “Mars”, or the name of a river in several languages): they can not be automatically translated. Retaining such sophisticated discriminations of names is absolutely necessary in any language or writing system variant. With or without such help, some nations will develop their systems (Hargitai et al. 2008 Hargitai et al 2010) but they may do it in an uncontrolled way. IAU should recognize one system for each language as the official language variant of the international Latin system.

3.4 Disharmony of names Themes for surface feature specifics generally have a connection to the name of the planet, satellite or asteroid. Mercury is the messenger god and god of commerce, thieves, and travelers. But instead of famous thieves and merchants, craters are named for artists, as originally suggested by Carl Sagan (Sagan 1976). It is not related, but a uniform theme and unique for Mercury. Using the original naming method, Mars should have been named for things related to Mars, the god of war. But generals along with contemporary politicians are exiled: “No names having political, military or religious significance may be used, except for names of political figures

286

H.I. Hargitai and K.B. Shingareva

prior to the 19th century” (IAU 2010a). This fact reflects a 20th century pacifist thinking of scientists. In fact, numerous achievements of space research were and are based on previous military research and development, or are a children of the cold war. People who named Mars after the god of war may be very surprised hearing that their famous generals can not have their craters on that planet. R. J. Pike suggested in 1976, after the IAU decisions in 1973 that IAU names should be changed: “Craters on Mars should bear the names of mythical heroes ... and military personalities” (Pike 1976). Contemporary names of towns for smaller Martian craters along with new Lacus names on Titan or some contemporary Valley names on Mars violate a most basic rule in international cartography: duplicate names should be avoided (for the Americans, however, it is an everyday experience to have duplicates)

3.5 Transparent decision-making A major problem is the opaque decision making process of IAU: it is never known why a particular name was selected. It can be inferred from the name itself, but especially in the case of towns, it is hard to deduce which late or contemporary astronomer or mission team member lives or was born in that city. Such data, i.e. a public reasoning for why a name was selected and why it was assigned to the feature at the particular location (for example in relation to its neighboring placenames), could justify the selection of the working group. Without such transparent justification is may easily be the case even today that some names are put onto planetary bodies for very personal reasons. Decisions by a working group of an international organization may seem to be fully democratic but since no justification is given to the public (to be available in the public domain) why particular descriptor terms, themes, or names have been selected (persons suggesting names, however, have to give justification to the working group). In fact the system more resembles an authoritarian one not differing from the personal working method of Riccioli who selected names carefully and to his best knowlegde.

Planetary Nomenclature: a Representation of human Culture and alien Landscapes

287

4- Conclusion and future directions Descriptor terms of the planetary nomenclature represent the true nature of the named object: they are based on their morphology, and to a limited extent, origin. Specifics are clearly representing human culture only. This is why the suggestion of naming craters after birds on Mercury (Pike 1976) was rejected (Sagan 1976). As placenames on our own planet do change, it would not be a surprise, that in spite of all efforts of making a standardized and international system, planetary surface features will be renamed in the future as our civilization, culture and way of accepted human thinking will move in other directions. One direction might be that names would represent the Earth as a planet shared with other living creatures; putting human culture to a wider context. But until that would happen, more names will have to be invented for features on the bodies to be visited in the future, from cometary nuclei to dwarf planets and may be exoplanets as well.

References Bryson B (2003) A Short History of Nearly Everything. Broadway Books: New York. Hargitai H, Kereszturi Á (2010) Towards the development of supplements to the Gazetteer of Planetary Nomenclature. EPSC2010-865. European Planetary Science Congress 2010 19 – 24 September 2010, Rome, Italy. Hargitai H, Császár G, Bérczi Sz, Keresztúri Á (2008) Földön kívüli égitestek geológiai és rétegtani tagolása és nevezéktana [Geostratigraphy and nomenclature of extraterrestrial planetary bodies, in Hungarian], Földtani Közlöny 138/4 Hargitai H, Kozma J, Kereszturi Á, Bérczi Sz, Dutkó A, Illés E, Karátson D, Sik A (2010) Javaslat a planetológiai nevezéktan magyar rendszerére. [Recommendation for a Hungarian System of Planetary Nomenclature, in Hungarian] Meteor csillagászati évkönyv 2010 pp 280-302 Hartmann WK (2003) A traveler’s Guide to Mars. Workman Publishing.) Hargitai HI, Gregory HS, Osburg J, Hands D (2007) Development of a Local Toponym System at the Mars Desert Research Station Cartographica volt. 42, no. 2 / Summer 2007 pp 179-187 Hargitai HI (2006) Planetary Maps: Visualization and Nomenclature. Cartographica vol. 41, no 2 / Summer 2006 pp 149-167 IAU (1960) Transactions of the International Astronomical Union, Moscow, August 12-20, 1958: Cambridge University Press, v. 10, pl. 1, p. 262. IAU (2010a): IAU Rules and Conventions. Gazetter of Planetay Nomenclature. USGS http://planetarynames.wr.usgs.gov/Page/Rules Accessed 2010 Oct. 3. IAU (2010b) Sources of Planetary Names USGS http://planetarynames.wr.usgs.gov/References Accessed 2010 Oct. 3. Hamill K (1963) Place-names on the Moon: A report. Names: Journal of the American Name Society, Vol. 12,. No. 2

288

H.I. Hargitai and K.B. Shingareva

Shingareva KB, Zimbelman J, Buchroithner MF, Hargitai HI (2005) The Realization of ICA Commission Projects on Planetary Cartography Cartographica, vol. 40, no. 4 /Winter 2005 pp 105-114 Eppa L (1935) Proposal for naming the rays of the lunar craters, Popular Astronomy, 43, No 1 (Jan 1935) Millman PM (1976) Topographic nomenclature on planetary bodies Icarus, Volume 29, Issue 1, September 1976, Pages 155-157 Moreux Th (1950), Étude de la Lune avec Dictionnaire Sélénographique, Nouvelle Ed. Paris Pike RJ (1976) Disharmony of the spheres: Recent trends in planetary surface nomenclature Icarus, Volume 27, Issue 4, April 1976, Pages 577-583 Proctor RA (1870) Other worlds than ours. London. Sadler, DH (1961) (ed) Proceedings of the Eleventh General Assembly. Commission pour l'etude physique des planetes et des satellites. Session administrative du 17 août 1961. IAU: Berkeley. Sagan C, Fox P (1975) The canals of Mars: An assessment after Mariner 9 Icarus, Volume 25, Issue 4, August 1975, pp 602-612 Sagan C (1976) On solar system nomenclature Icarus, Volume 27, Issue 4, April 1976, pp 575576 tychocrater (2007) Catena. http://the-moon.wikispaces.com/Catena Accessed Jun 20, 2007. UGEGN (2001) E/2001/INF/3 Item 8 of the provisional agenda http://unstats.un.org/unsd/geoinfo/UNGEGNMandate2001.pdf Accessed Dec 1 2010. UN (1970) Paper ESA/RT/C/GN/I, items 46-48, April, 29, 1970 UN (2002) United Nations Treaties And Principles On Outer Space. ST/SPACE/11 UN: New York. Whitaker E (1971) A Short History Of Lunar Nomenclature in: Gary L. Gutschewski, Danny C. Kinsler and Ewen Whitaker (ed). Atlas and gazetteer of the near side of the moon. Washington: Scientific and Technical Information Office, National Aeronautics and Space Administration. (NASA SP-241) Whitaker, EA (1999) Mapping and Naming the Moon: A History of Lunar Cartography and Nomenclature. Cambridge University Press

A New Version of the Multilingual Glossary of Planetary Cartography Kira Shingareva, Bianna Krasnopevtseva Moscow State University for Geodesy and Cartography 4, Gorokhovsky per. 105064, Moscow, Russia [email protected]

Abstract The necessity to use of the explanatory terminological dictionary in Planetary Cartography was clear just after organizing ICA Planetary Cartography Commission. It was discussed several ideas and stages by its compiling. There are described main stages by collection of Planetary Cartography Multilingual Glossary. The last version will be published in spring 2011 and presented in Paris on ICC11.

The Multilingual Glossary The main goal of the development of the explanatory terminological dictionary in planetary cartography is to initially prepare an English glossary version with subsequent translation into a number of other languages. It was necessary to elaborate a basic structure, to look for materials and present all the data according to this structure consisting of various groups of meanings and different languages. There exist disagreements and different interpretations both in terminology and in objects names. A preliminary version of the glossary, first issued for and presented at the ICC in Beijing in 2001, included about 150 terms, the number of which is constantly growing. It was the first attempt to collect these terms and give them some corresponding definitions. The Commission on Planetary Cartography has compiled a dictionary of terms frequently used on planetary maps as well as a list of terms in different languages, sometimes with very different meanings, that can be used to identify features on planetary maps. The dictionary has initially been written in English and also existed in Russian. These versions, however, existed in parallel without a real A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 289 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_19, © Springer-Verlag Berlin Heidelberg 2011

290

K. Shingareva and B. Krasnopevtseva

connection to each other. They do not offer the possibility to come at once from an English term to the corresponding term with the same meaning in Russian or in another language. The dictionary has to comprehend frequently used terms on planetary maps as well as a list of terms used in various languages that can be applied to identify features on planetary maps, sometimes having very different meanings. For the preliminary version some thematically distinct groups of terms were defined. They had to be obvious in the configuration of the dictionary in order to allow to make full use of it. At this point it is necessary to mention that the structure of the dictionary does not use the alphabetic order for the terms, at least not exclusively. This implies that the content was thematically divided and subdivided into six parts. Therefore it was necessary to add indices for the terms of each group and also to list the contents, with the terms of the respective group (section) in alphabetical order, separately for each part. The following parts have been defined: (I) The first (general) section contains the basic terms outlining a series of objects, directly connected with planetary cartography. It comprehends such terms as space, near space, far space, solar system, planet, satellite, moon, leading hemisphere, trailing hemisphere etc. (II) The second section defines the rigid-surface objects represented in planetary maps. These comprehend various kinds and types of general geographic and thematic maps, globes, solid terrain models of whole celestial bodies and parts of their surface. Here it turned out to be necessary to integrate a number of specifications and updatings to terminology. It is e.g. known that a globe is defined as the cartographic representation of the surface of a sphere, keeping geometrical similarity of the relief and a equality of the areas. However, for the creation of the Phobos Globe a tri-axial ellipsoid has been used. For asteroid modeling e.g., in general it is necessary to abstain from an analytical representation of the surface. (III) The third section contains the basic terms which describe cartographic products in general and are not restricted to extraterrestrial objects. They include such terms as scale, cartographical projection, conventional signature, etc. Certainly, these commonly known and defined terms are available in many glossaries, but here they also seem to be pertinent, so that the user does not have to waste time with additional searches in other sources.

A New Version of the Multilingual Glossary of Planetary Cartography

291

(IV) The fourth section (until present the most extensive one) represents the terms used for relief forms of extraterrestrial objects depicted in maps and having no terrestrial analogies. Particular explanations for the respective celestial bodies are given. At this point it is necessary to notice that by tradition these terms are taken from Latin. For the Mars e.g. there exist terms like chaos, domus or labyrinth, for Venus terms like terra, tessera or corona, on the Jupiter moons one finds maculas, faculas, lineas etc. However, practically on all planets lineaments can be found. (V) The fifth section includes the terrestrial terms used for relief forms of extraterrestrial objects which have terrestrial analogies. They include craters, ridges, canyons, plateaus, plains, etc. It is, however, necessary to note that for celestial bodies it is essential to emphasize e.g. the origin of craters (shock, volcanic, primary, secondary craters, etc.) through an explanatory term. (VI) The sixth section contains the terms which reflect some specific peculiarities of physical properties for celestial bodies, frequently depicted in their thematic maps. For example, albedo, isostatic anomaly, mascon, masmin, geoid, etc. belong to these terms. Already at the 2001 ICC in Beijing it became clear that this project is very close to another activity of the ICA Commission on Planetary Cartography, namely the compilation and printing of the multilingual map series of planets and their moons. These maps contain collateral information printed in 5 languages (English, German, Russian, French and Spanish). Therefore the same group of languages was selected for the new version of the Planetary Dictionary. Due to the decision to use the 5 languages mentioned above the collaborating teams were not satisfied with the previous concept anymore. From the beginning the guiding idea was to introduce English terms with their definitions in alphabetic order with affiliated progressive numbers together with their foreign equivalents arranged in 5 columns. Thus, the English list contains e.g.: Number 18. “Extraterrestrial territories”. Territories of a planet located off the Earth with a solid surface or separate segments of a rigid surface (5, 12, 23, 17). The numbers assigned to this word in the alphabetic lists in German, French, Spanish and Russian are 5, 12, 23, and 17, referring to the corresponding terms in the glossaries in other languages. There they are also given in separate lists of alphabetic order with reference to the numbers from the English list. In this case, in the German alphabetic list reads:

292

K. Shingareva and B. Krasnopevtseva

5. Ausserirdische Territorien, 18 (number 18 in English Glossary). It is envisaged to distinguish the language lists by different colours, be they of the lettering or of the paper. At the 22nd ICC in A Coruna, Spain, in 2005 this idea was presented and demonstrated. However, by then only the English and the Russian parts were ready. The other parts (in German, in French and in Spanish) were still planned for the future. Below some fragments of this version are given (Table 1 and 2). Order num- Words in English ber

alphabet

An explanation in English

18

extraterrestrial

(III)

territories 15, 12, 25, 38

Territories located off the Earth having a solid surface or separate segments of a rigid surface

23

facula

Light spot on the Jupiter moons’ surface

(V)

1 6, 43, 35, 124

Table 1 Fragments of the glossary in English

Order num- Words in German ber

alphabet

15

ausserirdische

(III)

Territorien 18, 12, 25, 38

16

facula

(V)

23, 43, 35, 124

An explanation in German Territorien, die sich ausserhalb Erde befinden und eine feste Oberflaeche oder einzelne Segmente einer festen Oberflaeche besitzen Heller Fleck auf der Oberflaeche des Jupitermondes

Table 21 Fragments of the glossary in German

In the wake of a discussion during the ICA Commission meeting in A Coruna it was concluded that we do not have enough adequate terms in the different languages, but that it is necessary to have all the possible equivalents mentioned. The English terms of “Mapping”, “Map-making” and cartography” have their particular meanings. In Russian we have “Cartography” as a general term and “Cartographiering” as a noun describing the mapmaking process. In German, ”Kartographie” is a general term covering any aspect and activity in relation with map-making and map-reading. In addition there exist the terms “Kartierung” for mapping in the field and

A New Version of the Multilingual Glossary of Planetary Cartography

293

“Kartographierung” as a more general term for map production. All these fuzzinesses have somewhat to be taken into account in an updated version of the dictionary. This new argument caused additional activities. Soon it became clear that the envisaged opus must not be an ordinary dictionary but a real multilingual glossary with interlingual cross-references. It implied that it was necessary to give an explanation for every term in every language. Such a structure would certainly be more substantiated. This is why now the following structure is proposed. The final version of the glossary has to consist of three columns for each language, namely the order number of the term proper, the alphabetic order of terms in definite language, the index numbers of the term in the other language lists and the explanation of this term in the respective language. This means that all five index numbers for English, German, French, Spanish and Russian have to be given. The numbers could be given in different colors. The alphabetic order of the terms is defined by the respective language. The present state of the glossary presents itself as follows. The newly prepared version includes five alphabetical lists in English, German, French, Spanish and Russian compiled according to the new concept. Each list contains about 350 terms. The meaning of each term is explained in every language and all the terms are cross-referenced to each other by ordinal numbers (index numbers). The new set-up of the contents is also given in separate parts or sections. The tables 3, 4, 5, 6, 7 reflect this new concept, where I is English, II is German, III is French, IV is Spanish, V is Russian. 39

Catena,

crater chain, or a string of closely spaced depressions

II.115, III.39, IV.20, V.358 40 41

Cavus, II.91, III.30, IV.25, V.188

steep slope depression with irregular form in the plan. Usually located in groups (in polar areas).

Chaos,

characteristic region of a destruction relief

II.35, III.44, IV.26, V.355 42

Chasma,

canyon, a steep slope linear depression

II.36, III.45, IV.27, V.117 43

Canyon,

see Chasma

II.127, III.52, IV.22, V.117

Table 3: English (English version)

115

Kette, I.39, III.39, IV.20, V.358

Eine Reihe der Krater, die nacheinander liegen

294

91

K. Shingareva and B. Krasnopevtseva

Hoehle,

Ein Loch, der unregelmaessige Form hat

I.40, III.30, IV.25, V.188 35

Chaos,

Eine Teil der ungeordneten Relief

I.41, III.44, IV.26, V.355 36

Chasma,

Sieh. Kluft

I.42, III.45, IV.27, V.117 127

Kluft, I.43, III.52, IV.22, V.117

Eine langfoermige Depression, die geradlinig oder oder kuvenfoermig ist.

Table 4: German (German version) 30

Cavus, I.40, II.91, IV.25, V.188

Une dépression avec une pente forte et un plan irrégulier. Normalement, on les trouve en groupe (dans les régions polaires).

39

Chainette,

Une chaînette de cratères ou des dépressions proches

44

Chaos,

I.39, II.115, IV.20, V.358 Une région caractéristique d'un relief détruit.

I.41, II.35, IV.26, V.355 45

Chasma,

52

Col,

Une dépression linéaire avec une pente forte.

I.42, II.36, IV.27, V.117 Voir chasma

I.43, II.127, IV.22, V.117

Table 5: French (French version) 20

Cadena, I.39, II.115, III. 39, V.358

Cadena de cráteres, o una hilera depresiones poco espaciadas.

25

Cavus, I.40, II.91, III.30, V.188

Depresión de fuerte pendiente con planta de forma irregular. Se suelen localizar en grupos (en áreas polares).

26

Chaos,

Región característica de un relieve de destrucción.

I.41, II.35, III.44, V.355 27

Chasma,

Ver.Cañón

I.42, II.36, III.45, V.117 22

Cañón,

Depresión lineal de fuerte pendiente

I.43, II.127, III.52, V.117

Table 6: Spanish (Spanish version)

117

Каньон I.42, II.36, III.45, IV.45

глубокая, крутосклонная линейная депрессия

A New Version of the Multilingual Glossary of Planetary Cartography

188

355

Котловина,

295

I.40, II.91, III.30, IV.25

крутосклонная депрессия неправильной в плане формы. Обычно К. расположены группами (в полярных областях)

Хаос,

характерный район разрушенного рельефа

I.41, II.35, III.44, IV. 26 358

Цепочка,

цепочка или линия кратеров

I.39, II.115, III.39, IV.20 117

Часма,

см. каньон

I.43, II.127, III.52, IV.22

Table 7: Russian (Russian version)

With the purpose to trigger discussion and to attract experts at an international level the present version of the Glossary will, until the end of 2007, be placed on the Internet on a site of the commission (http://www.nasm.edu/ceps/ica). Its continuous expansion and increase in detail, its further translation into other languages, and all the questions and notes concerning the definitions are thus put up for discussion. This is certainly useful to be done before the printing of the final version. At its first stage this work was carried out by scientists from Russia (K. B. Shingareva, B. V. Krasnopevtseva, MIIGAiK) and the USA (Jim Zimbelman, Smithsonian Institution) with the participation of experts from Germany (Manfred Buchroitner, Dresden University of Technology, Egon Dorrer, Munich University of Federal Forces). Later Egon Dorrer volunteered to take care of the German version of the Glossary, while Manfred Buchroithner prepared the French version. Later also scientists from Hungary (Henrik Hargetai, Elte University Budapest) and Canada (Philip Stooke, University of the Western Ontario) joined. Recently Spanish colleagues - Rufino Pérez Góm and Vázquez Hoehne, ETSI en Topografía, Geodesia y Cartografía – joined the team and prepared the Spanish version. We are trustful that this list of cooperating experts will continuously extend. The new edition of dictionary (May 2011) will be represented in Paris on ICC2011.

Cartographic Support of a New Landing Site for “Phobos-Grunt” Mission Kira Shingareva, Bianna Krasnopevtseva, Anatoliy Konopikhin, Konstantin Zeljkov, Sergey Dubov Moscow State University for Geodesy and Cartography 4, Gorokhovsky per. 105064, Moscow, Russia [email protected]

Abstract The main goal of the “Phobos-Grunt” spacecraft mission is the Phobos soil sample delivery to the Earth, and also the larger Mars moon study using a set of scientific equipments installed on the spacecraft board. One of the tasks for successful flight, especially a soft landing is correctly preselected a flat enough landing area with precise coordinates. There are described the main cartographic problems demand new high resolution images and enough ground control points.

New Landing Site Phobos and Deimos have been studied by means of ground based telescopic observations from the late 19th century up to the present day (Morley 1989, Lainey et al. 2005). Mariner 9 was the first spacecraft in Mars orbit (1971-1972) to deliver resolved images of Phobos and Deimos that revealed their surfaces. During the Viking missions (1976 to 1980), several close flybys of Phobos and Deimos were carried out and good area coverage of the Martian satellites for modeling of size and shape was obtained. During the ill-fated Soviet Phobos mission, though operating for a limited time span of 2 months only, unique multispectral surface observations became available. Almost 10 years after, Mars Global Surveyor had 4 encounters with Phobos revealing the nearside of Phobos at unprecedented resolution which made detailed photogeologic mapping possible (Thomas et al. 2000). More A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 297 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_20, © Springer-Verlag Berlin Heidelberg 2011

298

K. Shingareva et al.

recently, Mars Reconnaissance Orbiter camera HiRISE has returned highresolution 2-color images of the same hemisphere for mapping and interpretation of regolith spectral properties (Thomas et al. 2010). Also, observations of Phobos and Deimos from the surface of Mars became available by the cameras on Mars Pathfinder (Thomas 1999) and the MER rovers (Bell et al. 2005). These include nighttime observations of the two satellites, as well as their transits across the solar disk. Currently, Mars Express, owing to its elliptic orbit, is the only spacecraft to make close encounters with Phobos on a regular basis (see details further below). Analyses of radio tracking data from the flybys have been used to improve estimates of the Phobos mass. A large volume of images have been obtained, which formed the basis for new Phobos orbit, shape, and rotation models as well as maps. Data from the Omega spectrometer have revealed chemical constitutents of the surface soil. However, only limited amounts of the image- and spectral data have been analyzed as of the present day. At November 2009 the launch of “Phobos-Grunt” spacecraft was postponed on November 2011 because of technical reasons. This delay had some positive consequences too, namely there were obtained new Phobos pictures for its side opposite to Mars. It was selected for the area of choosing the spacecraft landing site position. The new material was appeared thanks the special surveying by ESA spacecraft “Mars Express” changing its orbit extra for this purpose. The main goal of the “Phobos-Grunt” spacecraft mission is the Phobos soil sample delivery to the Earth, and also the larger Mars moon study using a set of scientific equipments installed on the spacecraft board. That is why it is required the spacecraft soft landing at a pre-selected area on the Phobos surface. At this stage of the project two organizations (Lavochkin NPO and IPM RAN) have developed a scenario of flight. Russian comparative planetology scientists (GeoKHI RAN) have pre-selected potential landing areas, and Planetary Cartography Laboratory (MIIGAiK) has become responsible for mission geodesy and cartography support. It means it is responsible in front of solving the following main tasks: calculation of corrections to the coordinates of the landing site and then a variety of cartographic material on Phobos, which should be compiled on the basis of new high resolution images coming from the spacecraft “Phobos-Grunt”. DLR scientists using the results of previous surveys created a new network of 665 ground control points and a digital model of Phobos with high accuracy taking into account the revised satellite perturbed motion of Phobos. But the interesting landing area was still not covered by high-resolution images, and Viking spacecraft images was

Cartographic Support of a New Landing Site for “Phobos-Grunt” Mission

299

used for this new model in spite of their exterior orientation parameters are known with low accuracy. Summering up, we can to establish a fact that all of the models and their corresponding coordinate systems have a relatively accuracy that does not exceed 100 m. At the same time, the coordinates of the landing point should be determined with a precision not worse than 10 m. Thus, a team of MIIGAiK Planetary Cartography Laboratory has a complicate task: the creation a new coordinate system and the control points network, which should ensure the accuracy of landing at the given point. That is why the modeling on the base of new “Mars Express” pictures gave us a possibility to improve the last control point catalogue and the landing site coordinates too. Then the spacecraft should be transferred to the quasi-synchronous orbit with the orbital period, which is equal to the period of Phobos itself. The spacecraft in the Phobos coordinate system turns around the moon of Mars on a trajectory close to an ellipse. Phobos gravity provocates that the center of the ellipse has temporal oscillations, resulting in an unclosed trajectory. During this orbit driving the distance to the surface of Phobos will vary - from 30 to 70 km. Transmitting to the final orbit can be done through an intermediate orbit. The satellite will remain at this orbit about 30 days. This time will be used for the Phobos surface imaging with high spatial resolution. After processing the received images a digital terrain model (DTM) will be created and clarifed the choice of landing area. Landing on Phobos starts with the distance to the surface about 30 km and takes about 30 minutes. When the spacecraft approaches to the region of landing television monitoring together with laser altimeter, it will be used for selection a landing site size of ~ 10 m without dangerous roughness. In the process of landing it will be sending some images to the Earth. They will be having the improving resolution. After taking samples of Phobos soil the return module comes back to the Earth. After the flight module to go back the experiments on the surface of Phobos will be continued by the landing module. During this period, using images made by Viking and Mars Express spacecrafts, a control point network and a coordinate system will create on carrying out stereo images photogrammetric processing and Phobos DTM simulation. Thus it is possible to test all the work planned for quasisynchronous orbit. In addition, it is reached an agreement that the DLR German colleagues will provide additional photographic information on Phobos, which would be obtained by Mars Express spacecraft during the “Phobos-Grunt” spacecraft flight on the Earth – Mars trajectory.

300

K. Shingareva et al.

• Selection of control points and the photogrammetric processing of images; • Creation of a new coordinate system and the new control points with a high accuracy in the vicinity of the spacecraft intended landing using the results of processing; • Creating Phobos digital elevation models (DEM) with the increased accuracy in the vicinity of the expected spacecraft landing; • Conversion of the old landing place coordinates to the newly established system of coordinates. • The relevance of the different values were taking into account by analyzing today the most accurate Phobos DTM which was created by K.Willner (DLR). It is clear that the closest surface passing to the real surface of Phobos is the triaxial ellipsoid. The average height above the surface of Phobos triaxial ellipsoid with semi-axis a = 13.3 km; b = 11.1 km; c = 9.3 km is just over 107 meters. In relation to the sphere with radius of 11.0 km has an average altitude of 147 meters, but the maximum and minimum deviations are much larger. But further conclusions should be born in mind that the spacecraft provided a rectangular coordinate system and, therefore, some corrections to the landing site coordinates will also need to be in the form of ΔX, ΔY, ΔZ because of the coordinate system difference. Then it is preferable to use spherical coordinates for the transition area, since all the action at quasi-synchronous orbit will require maximum speed giving the possibility of easy and precise result control. In future a tri-axial ellipsoid will be better pass for working out various cartographic products. However, it should be noted that the current studies are ongoing and using of tri-axial ellipsoid with the calculation of not "geodesic" heights (i.e. perpendicular to the surface of the ellipsoid), but toward the center of the ellipsoid. This option is very useful, because the spacecraft during landing by using a laser altimeter will measure the distance which is not perpendicular to the surface of the reference ellipsoid, but toward the Phobos mass center. The system of coordinates will be chosen in such a way that its center was as close as possible to the center of Phobos masses and, ideally, coincide with it. Then it is obvious that the heights above the ellipsoid defined along the line toward the center of mass would be preferable. In addition, It can be also used a morphographic projection for compiling maps on the surveying results because they have already demonstrated a good quality for mapping celestial body of irregular shape.

Cartographic Support of a New Landing Site for “Phobos-Grunt” Mission

301

In conclusion, it should be noted that the creation of software and the choice of the reference surface have to be finished in time for launch in ТТТNovember this year. For about 11 months the spacecraft will be located on the Earth - Mars flight trajectory. During this period, using images made by Viking and Mars Express spacecrafts, a control point network and a coordinate system will create on carrying out stereo images photogrammetric processing and Phobos DTM simulation. Thus it is possible to test all the work planned for quasi-synchronous orbit. In addition, it is reached an agreement that the DLR German colleagues will provide additional photographic information on Phobos, which would be obtained by Mars Express spacecraft during the “Phobos-Grunt" spacecraft flight on the Earth – Mars trajectory.

References Bell JF et al. (2005) Solar Eclipses of Phobos and Deimos observed from the surface of Mars, Nature, 436:55-57. doi:10.1038/nature03437 Lainey V, Dehant V, Rosenblatt P, Andert T, Pätzold M (2005) New ephemeris of Phobos from MarsExpress close flybys, abstract, EGU05-A-07117 Morley TA (1989) A catalogue of ground-based astrometric observations of the Martian satellites, 1877 -1982, Astron. Astrophys. Suppl. Ser., 77:209-226. Thomas N, Britt DT, Herkenhoff KE, Murchie SL, Semenov B, Keller HU, Smith PH (1999) Observations of Phobos, Deimos, and bright stars with the Imager for Mars Pathfinder, J. Geophys. Res., 104: 9055-9068, Issue E4 Thomas N, Stelter R, Ivanov A, Bridges NT, Herkenhoff KE, McEwen AS (2010) Spectral Heterogeneity on Phobos and Deimos: HiRISE Observations and Comparisons to Mars Pathfinder Results, In LPSC Abstracts, 41: 2595+of LPI Tech. Rep Thomas PC, Veverka J, Sullivan R, Simonelli DP, Malin MC, Caplinger M, Hartmann WK James PB (2000) Phobos: regolith and ejecta blocks investigated with Mars Orbiter Camera images, J. Geophys. Res., 105: 15091-15106, Issue E6

Making GeoInformation from Image Analysis

Evaluation of the spatial Dynamics of Great Oran (Algeria) using spatial Imagery and GIS Fouzia Bendraoua, Ali Bedidi, Bernard Cervelle Université Paris-Est Marne-la-Vallée, Laboratoire Géomatériaux et Environnement, EA 4508, 5 Boulevard Descartes, 77454 Marne-La-Vallée cedex 2, France. [email protected], [email protected], [email protected]

Abstract The last decade has been characterised by a transfer of demographic growth from the city of Oran (Algeria) to its immediate peripheral areas, especially towards Bir El Djir and Sidi Chahmi to the east and towards Es Senia to the south. This paper attempts to evaluate this spatial growth over the period from 1991 to 2003 by characterising and quantifying the spaces that have become urbanised over this time frame. To achieve this, we have used panchromatic Spot images from 1991, 1998 and 2003 to measure this spatial dynamic and GIS to evaluate the built surfaces and their locations. This methodology has proven itself efficient for evaluating and monitoring spatial dynamics in a regional metropolis such as Oran, Algeria’s second largest city after the capital (Algiers), which is subject to strong land development pressure.

1- Introduction Following the mass departure of Europeans in the summer of 1962, Algerian towns underwent a period of considerable migratory turbulence. People living in the Algerian countryside continued to nourish the rural exodus to the city. Consequently, during the first years of independence, there were no major changes in the spatial growth of the cities. The city socially reorganised itself while remaining within the same urban perimeter. The first urban crisis goes back to 1954 and continued through the first years of independence as the city had not yet been able to organise and develop a sufficient economic base to integrate the population flows (Souiah 2003). A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 305 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_21, © Springer-Verlag Berlin Heidelberg 2011

306

F. Bendraoua et al.

As from the 1970’s, the urban landscape of Algerian towns and the city of Oran in particular underwent a change. The difficulty in finding housing began to make itself felt while, at the same time, the inherited housing stock was beginning to become saturated. The launching of economic development plans, especially the three-year plans (1970-73, 1974-1977), had very little impact on the extension and urbanisation of Oran. It was not until the first four-year plan (1980-1984) and the instigation of the Great Oran1 urban master plan (PUD – launched in 1973-1974 and approved in 1977), reinforced by the urban modernisation plan (PMU) in 1975, that the city began to experience significant spatial dynamics. These changes were expressed by the implementation of large-scale town planning operations that had the effect of fundamentally changing the city’s immediate peripheral areas. These operations simultaneously concerned housing (new urban housing areas – ZHUN – and subdivisions) and economic activities through the creation of industrial zones (ZI) and work and storage zones. However, the creation of large housing projects in the form of ZHUN was translated by a considerable land consumption estimated at 800 hectares for the five ZHUN in Oran laid out in a ring around the city (Smair 1989). The ZHUN procedure, although meeting the housing requirements that had continued to grow since the 1970’s, remains questionable in terms of the choice of locations, the attributed land surfaces and their characteristic lack of amenities. This is perhaps due to a lack of a global approach to the city’s development. Nonetheless, the ZHUN led to a deterioration in the quality of the environment (for example, playing fields were not provided with amenities or maintained) as well as an anarchic densification. The result was the complete opposite of what had been enacted by ministerial decision (concerning the ZHUN) and whose intention had been to remedy inefficient land use. The city intra-muros was saturated by the priority given to the ZHUN programmes while subdivisions, whether or not formal, found themselves as a result of this situation in the periurban space that was more vulnerable and where the land tension was not so acutely felt at that time. As a result, the number of unregulated slums increased and this increase was further 1

In this paper we call the smallest administration division in Algeria municipality. In Algerian urban planning “Le Groupement Urbain d’Oran (GUO)” is formed by 4 municipalities: Oran, Bir El Djir, Es-Senia and Sidi Chahmi. In this paper the GUO is called Great Oran.

Evaluation of the spatial Dynamics of Great Oran (Algeria) using spatial Imagery and GIS

307

accelerated by the government’s disengagement and the accompanying measures (open property market, subdivisions, assignation of State property, real estate development, etc.), which accentuated the difficulties faced by the most deprived to access housing and urban land. This resulted in a dual process: an urban sprawl in central and peri-central spaces and a differentiated spread over the city’s peripheral areas (Bendraoua 2005). This study is based on two complementary information sources, one being statistical and making use of four general population and housing censuses (GPHC), the other being spatial using satellite data to analyse and quantify the space urbanised in and around the city of Oran over the period from 1991 to 2003. A global approach was made possible by using geographical information systems (GIS) to cross-reference spatial imagery and related data.

2- Evolution of Great Oran through analysis of the various censuses An analysis of the four GPHC censuses carried out in Algeria in 1966, 1977, 1987 and 1998 reveals that most of the demographic growth taking place in Great Oran essentially took place in the city of Oran peripheral areas (Figure 1). In 1998, the Great Oran population attained 830,795 residents, being 147,911 more residents than in 1987 (ONS 1987) (ONS 1999).

308

F. Bendraoua et al.

Figure 1: Limits of Great Oran’s municipalities, suburban areas, Oran’s neighbourhoods and the third ring road. On the back the 2003 SPOT 5 images (© Cnes2003-distribution Spot Image)

While the population of the municipality of Oran continued to increase between 1966 and 1987 (Figure 2), the growth rate then stagnated for a decade (1987-1998) while the populations of the outlying municipalities (Bir El Djir, Es Senia and Sidi Chahmi) underwent a much greater growth rate (Figure 2). Oran’s urban sprawl essentially spread towards the peri-urban ring and essentially affected nearby suburban towns such as Bir El Djir, Sidi Chahmi and Es Senia, whose growth rates respectively achieved 11.73%, 10.14% and 4.06% between 1987 and 1998. While the towns of Bir El Djir and Sidi Chahmi recorded comparable demographic growth over the last four censuses (Figure 2), their spatial dynamics took different forms. The town of Bir El Djir evolved spatially and regularly from its connection point to the city and urbanised the eastern area of the Oran, while the spatial growth of Sidi Chahmi was more concentrated in the Nedjma metropolitan area whose annual growth rate attained an exceptional level of 20.38 % between 1987 and 1998.

Evaluation of the spatial Dynamics of Great Oran (Algeria) using spatial Imagery and GIS

309

Figure 2: Distribution of the population of Oran and the three surrounding municipalities according to the censuses carried out in 1966, 1977, 1987 and 1998

The fact that the municipality of Oran underwent fairly sustained demographic growth up to 1987 is largely explained by two types of migratory movements over different periods, one old and the other more recent. The older migratory movement is linked to the departure of the Europeans and the conquest of the city intra muros by the indigenous population. This process of appropriating the city produced, on the one hand, a considerable shedding of the social content of the city’s socio-spatial margins (Bendraoua and Souiah 2008) and, on the other hand, accentuated the rural exodus in the first years of independence. The second more recent type of migration is linked to the consequences of the post-colonial urban policy which, among others, was oriented towards the local decentralisation of activities and the creation of new facilities which in the past had been centralised within the city of Oran. It might be considered that the spatial sprawl in Oran took place over three clear-cut phases. • The first phase concerns the filling of the city’s third peripheral suburban ring by the construction of ZHUN which, in a discontinuous manner and running from east to west, occupied areas along the third ring road (figure 1) while avoiding certain protected areas such as farmland.

310

F. Bendraoua et al.

• The second phase extended the city by creating a major north-south axis towards Es Senia. This was dedicated to large socio-educational infrastructures (universities, institutes and colleges) and industrial zones. • The last phase marks Oran’s urban spread. Various types of outlying spaces were created: informal peripheral areas to the west of the city in the form of illegal slums (Les Planteurs, Les Amandiers, El Hassi, Haï Cheikh Bouamama), formal peripheral areas in the form of subdivisions and rehousing programmes to the east (Douar Sidi Bachir and Sidi Maarouf) and south-west (Douar Ain El Beida) of the city, and peripheral areas in the form of housing cooperatives (Chérif Yahia and the east part of Bir El Djir). There were also extensions to ex-colonial centres such as Es Senia and Sidi Chahmi. Has Oran’s spatial organisation remained unchanged since the 1990s or are we seeing a new dynamic? How has Oran’s spatial dynamic been expressed over the last two decades? Is it possible to identify its rhythm, modalities and characteristics? Is it possible to detect predominant spatial forms? Have these new urbanised areas modified Oran’s initial radioconcentric layout? We shall demonstrate the performance and the contribution made by GIS and remote sensing for studying the urban dynamics of Great Oran. The treatment sequence is broken down into two parts: remote sensing for treating and interpreting photographic satellite images, and GIS for vectorisation and spatial analysis of the built fabric. It should be noted that Oran’s spatio-temporal quantification for the period from 1991 to 2003 provides a rigorous framework for the various statistical surveys.

3- Extraction of the urban fabric: Methodology and Results

3.1 Presentation of satellite data To carry out this work, we used Spot satellite panchromatic images of Great Oran taken in 1991, 1998 and 2003. The 1991 and 1998 images have a 10 m resolution and that of 2003 a 2.5 m resolution.

Evaluation of the spatial Dynamics of Great Oran (Algeria) using spatial Imagery and GIS

311

Currently, certain satellite images can provide resolutions as detailed as aerial photographs with, as an advantage, a far greater spatial cover. This particularly applies to Spot 5 images (panchromatic resolution of 2.5 metres and multispectral resolution of 10 m covering 3,600 km²), Quickbird (with a panchromatic resolution of 61 cm and a multispectral resolution of 2.5 m with a coverage of 256 km2) and Ikonos images (panchromatic resolution of 1 metre and multispectral resolution of 4 metres with a coverage of 121 km2). However, as for all aerial photographs, the inconvenience lies in the cost of acquiring the latter images (Quickbird, Ikonos). This considerably curtails their use. This is why we decided to use panchromatic Spot images. On the one hand, they have the capacity of covering the city of Oran and its outlying metropolitan areas forming Great Oran and, on the other hand, the detailed level of their resolution. This particularly applies to Spot 5 images with a resolution of 2.5 m, being sufficient for a precise interpretation of urban objects (Bendraoua 2005). The interpretation of very high spatial resolution images, especially Spot 5 (panchromatic), represents a vital source of information for the identification of urban objects. The value of the latter for urban planning is selfevident, on condition that certain details be completed by site investigations. Our thesis research work (Bendraoua, 2005) revealed the relevance and usefulness of this type of data as a source of information in its own right for efficient land management. It also made it possible to check the satisfactory application and evaluation of town planning tools, a factor meeting the requirements of those involved and making decisions in the urban planning and development sector.

3.2 Pre-processing The study of the multi-temporal spatial development of Great Oran for the period from 1991 to 2003 was carried out using multi-date panchromatic Spot satellite images (1991, 1998, 2003) treated using GIS-ArcView. The initial pre-treatment of the images consisted in visualising and correcting the radiometry of the various images. We then extracted from the entire scene the space covering Great Oran and formed by the municipalities of Oran, Bir El Djir, Es Senia and Sidi Chahmi. To permit a good overlaying of images taken on different dates, we chose the Spot 5 image as a geographical referential for Spot 2 images (1991)

312

F. Bendraoua et al.

and Spot 4 images (1998). The 1991 image was geo-referenced with an average Root Mean Square error (RMS) of 0.8 pixels and that of 1998 with an RMS of 0.9 pixels. In order to evaluate the urban development of Oran up to 2003 at the same resolution as the Spot 2 and Spot 4 images, being 10 metres, the Spot 5 image was re-sampled at the same resolution. The re-sampling of the Spot 5 image at 10 m was carried out as follows: • For each 4 x 4 pixel tile in the SPOT 5 image (2.5 m of resolution), the values were averaged to obtain a pixel at 10 m. • The resulting image was filtered to obtain a “smoother” image and a visual appearance similar to those of Spot 2 and 4 images.

3.3 Which method to use for extracting urban objects from the images? New generations of sensors with very high level of definition of around one metre, led to various methods being developed to carry out analyses in the urban environment. The complexity of the urban environment introduces limits to the identification of objects in the city when simply using spatial imagery. In fact, whatever the methods used, be they automatic, semi-automatic or based on photo-interpretation, they all present limits to an optimal discrimination providing details concerning urban objects (Bendraoua 2005). However, for a sufficiently precise and detailed extraction (large scale) of useful urban objects in the general field of development and in urban planning in particular, the method based on photointerpretation remains the one that is best adapted (Lortic 2003). This method, because it is based on terrain knowledge and uses the richness of images with a high spatial resolution, can discriminate urban objects with a better level of assurance than the automatic method whose level of precision is not always satisfactory. In addition, the transfer of the methodology from one area to another is not always guaranteed (Weber et al. 1997). This has been confirmed by the urban applications for various cities such as Strasbourg, Geneva and Liège (Weber et al. 1997) or Bordeaux and Toulouse (Albert 2003).

Evaluation of the spatial Dynamics of Great Oran (Algeria) using spatial Imagery and GIS

313

Figure 3: Extraction of the built fabric from a 1998 Spot satellite image by photointerpretation

3.4 Extraction of the urban fabric The three satellite images were integrated into the GIS to measure the temporal development of Oran’s urban space between 1991 and 2003. To do this, we digitised the built fabric on each of the images. The urban fabric at different dates (1991, 1998 and 2003) was extracted by photointerpretation using GIS-ArcView (example of 1998 urban fabric extraction, Figure 3). For the photo-interpretation, we based ourselves on our knowledge of the area and the input specifications that we had defined on the basis of the French geographic institute’s topographical data base (IGN, 1999). For the initial approach, we avoided introducing any differentiation in terms of urbanised spaces and only used the difference between built fabric/non-built fabric to define forms and identity major trends. The built fabric at each of the dates was represented by a layer formed by polygons. In compliance with the chosen specifications, each polygon represents a built fabric unit (Bendraoua 2005). To measure the growth of the built fabric between two dates, we examined the different layers of the corresponding built fabric to obtain the urban extension between these two dates. We eliminated small surfaces that are much more the result of digitising errors (estimated at 5 m) than a real growth in the urbanised space. We added mathematical morphology

314

F. Bendraoua et al.

techniques to the process, corresponding to growth of the built fabric between two dates, by application of an erosion followed by an opening (Serra 1992). To do this, we used a five metre buffer zone corresponding to the digitising error. An example showing the extraction of the built fabric between 1998 and 2003 is provided in Figure 4. This work allowed us to quantify and geographically localise the built space and its growth over a number of dates (1991, 1998 and 2003). We shall now present and analyse the obtained results.

Figure.4: Extraction of the built fabric in Great Oran between 1998 and 2003. On the left, the built fabric in 1991 (light green) overlaps the built fabric in 1998 (dark blue)

3.5 Evaluation of the urbanised space in Great Oran between 1991 and 2003 The spatio-temporal growth of Great Oran is based on a comparison with Oran’s built fabric in 1991 (Spot 2). Evaluating the spatial growth of Oran and its peripheral areas between 1991 and 2003 was carried out by calculating the urbanised surface areas of each of the municipalities in Great Oran over two periods: 1991 - 1998 and 1998 - 2003 (Table 1). We have evaluated this spatial development by defining the growth rate over these

Evaluation of the spatial Dynamics of Great Oran (Algeria) using spatial Imagery and GIS

315

two periods (1991-1998 and 1998-2003) and the annual rate of growth for each of the municipalities (Table 2). Oran Es Senia Bir El Djir Sidi Chahmi Great Oran Built fabric in 1991

2631 725

549

438

Built fabric in 1998

3016 912

862

568

4343 5348

Built fabric in 2003

3281 1101

1047

743

6173

Built between 1991 and 1998 385

187

304

130

1005

Built between 1998 and 2003 266

189

195

175

825

Built between 1991 and 2003 660

376

499

305

1830

Table 1: Built surface area (in hectares) and its growth from 1991 to 2003 in the municipalities forming Great Oran.

Annual growth rate

Growth rate over the period

Municipality

1991-1998

1998-2003

1991-2003

1991-1998

1998-2003

1991-2003

Oran

2.0%

1.7%

1.9%

14.6%

8.8%

24.7%

Es Senia

3.3%

3.8%

3.5%

25.7%

20.7%

51.8%

Bir El Djir

6.5%

4.2%

5.5%

55.3%

22.9%

90.9%

Sidi Chahmi

3.8%

5.5%

4.5%

29.7%

30.8%

69.7%

Great Oran

3.0%

2.9%

3.0%

23.1%

15.4%

42.1%

Table 2: Growth rate and annual growth rate of the built fabric from 1991 to 2003 in the towns forming Great Oran.

4- Analysis of Great Oran’s urban dynamics from 1991 to 2003 The urbanised space of Great Oran in 2003 was estimated at 6,173 hectares, being an additional 1,830 hectares when compared to the urbanised space in 1991 (Table 1). Great Oran became increasingly urbanised over this period with annual growth rates of around 3%. The growth of urbanised space in Great Oran over the last decade has essentially occurred in the slum districts and outlying metropolitan areas. This has taken place in three main ways: The filling of residual voids within the city’s 1991 perimeter, essentially in the districts around the centre of Oran.

316

F. Bendraoua et al.

New buildings especially in the north-west of the city on the flank of the Murdjadjo Mountain (Douar Bouakeul, les Amandiers, El Hassi, etc.). After the analysis, our visit to these places showed that these new buildings are in the form of illegal slums. New urban extensions concentrated in outlying metropolitan areas. Our analysis has been completed by a visit to these places which showed that these urban extensions are in the form of one-family housing subdivisions for working and middle class population. These houses do not respect town planning standards (sites without land servicing and facilities). This particularly applies to the extensions to outlying metropolitan areas (Ain Beida, Nedjma, Es Senia, Sidi Maarouf, Sidi Bachir and Bir El Djir) (figure 5).

Figure 5: Urban and peri-urban dynamics in Great Oran between 1991 and 2003. In the background, the 2003 Spot 5 image (© Cnes 2003-Spot Image distribution)

Evaluation of the spatial Dynamics of Great Oran (Algeria) using spatial Imagery and GIS

317

When compared with the other towns, Bir El Djir, located in the eastern part of Great Oran, underwent the greatest spatial growth. It saw an annual growth rate of 6.5% between 1991 and 1998, and 4.2% between 1998 and 2003. The overall growth rate of this municipality for the two periods remains very high: 55.3% from 1991 to 1998 and 22.9 % from 1998 to 2003 (table 2). This latter estimate confirms Oran’s urban sprawl over the last decade and the transfer of urbanisation to the eastern part of Great Oran. Concerning this point, it should be underlined that the Great Oran planning programme in force (the PDAU) provides for a massive long term extension towards this area by 2015. The municipality of Sidi Chahmi holds second position in terms of spatial growth with an annual growth rate of 3.8% between 1991 and 1998 and 5.5% between 1998 and 2003. The municipality of Es Senia lies in third position with annual growth rates of 3.3% and 3.8% respectively for the 1991-1998 and 1998-2003 periods. The municipality of Oran is in last place with annual growth rates varying between 2% and 1.7% for the two concerned periods. The municipality of Oran has lower spatial dynamics than those characterising the outlying towns. The layering of the new extensions on the oldest image, being that of 1991, reveals that the urban dynamics of Great Oran between 1991 and 2003 are spatially expressed in two ways. The first is characterised by an extension to the urban fabric of the existing cores, as can be seen with the new extensions to the towns of Chérif Yahia, Ain El Beida, Es Senia, etc. The second spatial form is translated by ex-nihilo constructions in the municipalities of Bir El Djir and Sidi Chahmi (Figure 6).

318

F. Bendraoua et al.

Figure 6: (a) New urbanised space in Great Oran between 1991 and 2003. In the background the 1991 Spot image (© Cnes 1991 - Spot Image distribution). New creation of built fabric in Bir El Djir (b1: 1991, b2: 2003) and in Sidi Chahmi (c1:1991, c2: 2003). Extension of the existing built fabric in Chérif Yahia (d1: 1991 and d2: 2003)

In the most recent configuration, Oran is attempting to reclaim its initial radioconcentric layout. This is made easier by the opening of the fourth ring road which gives easier access to certain peripheral areas that had been protected in the past as they were used for farming.

Evaluation of the spatial Dynamics of Great Oran (Algeria) using spatial Imagery and GIS

319

5- Conclusion This study reveals that since the 1970s and following the implementation of urban planning policies, Great Oran has been subject to considerable land use pressures on its immediate surrounds. These particularly concern existing and newly created metropolitan areas. The analysis of recent urban growth over the period from 1991 to 2003, using spatial imagery and GIS, reveals a high level of spatial growth, equivalent to 42% of the total urbanised area determined in 2003. This new urbanisation of Great Oran is expressed through the massive extensions taking place in the outlying metropolitan areas and the proliferation of slums on the Murdjadjo mountain whose landtake is rapidly expanding into the forest estate. This slum habitat reveals the socio-spatial fringes present in the city. It is a phenomenon that damages the image of the city given that the buildings completely ignore all urban planning rules on sites that are supposed to be nonconstructible. Certain types of extension represent considerable problems and make it impossible to control spatial management and development. The non-compliance between the urban development master plan (PDAU) and what is in fact being urbanised reveals the lack of respect for town planning instruments and the lack of control over the spatial growth of Great Oran. In addition, since independence in 1962, no development plans, be they local or regional, such as the 1975 town planning master plan (PUD) or Oran’s 1994 Wilaya development programme (PAW) have ever been respected or fully implemented. We were able to observe the considerable difference between what is required by the PDAU and what has in fact been built. These malfunctions, which can be blamed on uncontrolled forms of growth, considerably affect the harmonious and sustainable development of the city. Long term forecasts for the development of Oran are being compromised by a premature consumption of the city’s land stock whose urbanisation was not programmed to take place before 2015. In addition, the present study provides a concrete illustration of the operational advantages provided by the treatment of satellite images combined with the use of GIS in the approach to and analysis of urban environments. These techniques have now proven themselves very useful in situations where there are no up to date maps or where statistics are rare or unreliable, a situation particularly applicable to a large number of developing countries.

320

F. Bendraoua et al.

References Albert P (2003) Le suivi de l’urbanisation des agglomérations : les capteurs à très haute résolution, une révolution ? L’exemple de Spot5. In : SFPT (ed) Colloque Pixels et Cités, Télédétection et photogrammétrie pour le développement en milieu urbain, edn. SFPT, France Bendraoua F (2005) Dynamique urbaine, instruments d’urbanisme et réalité terrain du groupement urbain d’Oran (Algérie) : Création d’un SIG et imagerie spatiale. Dissertation, Université de Marne la Vallée, France. Bendraoua F, Souiah SA (2008) Quand les pouvoirs publics produisent de nouvelles marginalités urbaines. Le cas des recasés de Nedjma à Oran (Algérie). Autrepart 45:173-190. Institut Géographique National (1999) BD TOPO® : Spécifications détaillées, Version 3.1. IGN Paris Lortic B, Couret D (2003) Panorama des réalisations et interprétations : variété des contextes Nord/Sud : Les cités vues de l’espace. In : SFPT (ed) Colloque Pixels et Cités, Télédétection et photogrammétrie pour le développement en milieu urbain, edn. SFPT, France ONS (1987) Armature urbaine 1987. In: Collections statistiques n°04 (ed) IIIème Recensement général de la population et de l'Habitat (RGPH), edn. Office National des Statistiques, Alger ONS (1999) Principaux résultats issus du sondage au 1/10°. In: Collection statistiques n°80 (ed) IVème Recensement général de la population et de l'Habitat (RGPH) 1998, edn. Office National des Statistiques, Alger Smair, A. (1987). Les nouveaux tissus urbains d'Oran. Paper presented at Les tissus urbains colloque international, Oran (Algeria). Souiah SA (2003) Les marginalités socio-spatiales. Dissertation, Université de Cergy-Pontoise, France. Weber C, Donnay JP, Collet C (1997) Reconnaissance des formes urbaines : transfert méthodologique Nord-Sud. In: Dubois JM et al. (ed) Télédétection des milieux urbains et périurbains, Collection Actualité Scientifique, edn. AUPELF-UREF, Montréal Weber C, Hirsch J, Serradj A (1997) Pour une autre approche de la délimitation urbaine : application à Strasbourg (France). In: Dubois JM et al. (ed) Télédétection des milieux urbains et périurbains, Collection Actualité Scientifique, edn. AUPELF-UREF, Montréal

Topographical Mapping at 1:50,000 Scale from Satellite Imagery using CARTOSAT-1 Takka El-hadi Institut National de Cartographie et de Télédétection 123, Rue Tripoli, Hussein-Dey, BP430, 16040 Algiers, Algeria [email protected]

Abstract High-resolution stereo imaging satellites are a valuable tool for topographic mapping and map updating. Cartosat-1 2.5 meter GSD stereo images were used by “Institut National de Cartographie et de Télédétection” (INCT), Algeria, in collaboration with the Space Application Centre, ISRO, India, in order to map an area of interest situated in the north western side of Algeria at 1:50,000 scale which has never been mapped since the colonial era. This paper presents an effective methodology, which starts from the data preparation and ground control points collection using Differential GPS survey to map creation. Within this project, Cartosat-1 stereo images limitations and potentialities in topographical mapping were assessed. The developed process has allowed us to produce three toposheets of 15’x15’. Although, filed verification has demonstrated that some punctual, linear and surface features could neither be extracted, nor interpreted, mapping at this scale using Cartosat-1 stereo images is possible. However, it requires more time in field survey comparing to the classical mapping process where aerial photos are used.

1- Introduction Topographic maps make a presentation of our surroundings, giving us a possibility to feature all visible territory elements in the same detail. The maps depict the relief, hydrography, vegetation, soil and subsoil, human settlements, road networks, and other facilities, which allow us to make a complex evaluation of the territory (INCT 2005).

A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 321 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_22, © Springer-Verlag Berlin Heidelberg 2011

322

T. El-hadi

Aerial photos have been traditionally used as the primary source for the topographic maps drawing. Satellite-based digital images open new horizons: cheaper repeated imagery, increased territory coverage and decreased relief distortions. Therefore, satellite-based imagery is used increasingly wider and may soon become the principle mean for topographic maps updating and production. The graphical accuracy of maps drawing and printing is accounted for (0.1 mm) (INCT 2005), thus, the images should have a spatial resolution of at least 5 meters for producing maps at a scale of 1:50,000 (Puissant and Hirsh 2004; Kshma and Sadhana 2005). Therefore, Cartosat-1 images with a pixel size of 2.5 meters can be used to draw up the quasi-totality of elements of 1:50 000 scale maps. Cartosat-1 satellite was built by ISRO mainly for mapping. The satellite was launched into circular (altitude is 618 km) near-polar sun-synchronous orbit on May 5, 2005. Cartosat-1 is equipped with two panchromatic cameras capable of simultaneous acquiring images of 2.5 meters spatial resolution. One camera is looking at +26 degrees forward while another looks at –5 degrees backward to acquire stereoscopic imagery with base to height ratio of 0.62. The time difference between acquisition of the stereopair images is approximately 52 seconds. The cameras are across-track-steerable to enhance the system productivity. The radiometric resolution is 10 bits, stereoscopic swath width is about 26 km while wide-field (using both cameras) mono swath is 55 km (NRSA 2006). This paper describes a practical methodology to produce 1:50,000 topographical maps using Cartosat-1stereo images. This approach include the pre-processing of images for GPS survey for ground control points collection (GCP), modeling of space imagery, DEM and Ortho-image generation and accuracy improvement of DEM.

2- Study area Covering an area of 30’x30’ span, the test site is bound between geocoordinates 34° 45’ N to 35° 15’ N and 1° 30’ W to 1° 00’ W. The study area forms part of Ghazaouet and Tlemcen and the general elevation of the area ranges from sea level to 1400 meters above sea level.

Topographical Mapping at 1:50,000 Scale from Satellite Imagery using CARTOSAT-1

323

3- Dataset used Imagery data provided by ISRO are Stereo Orthokit Product that consists of along-track stereo pairs, Rational Polynomial Coefficients (RPCs) for each image and product metadata. Images are geometrically raw but radiometrically corrected. Six orthokits were used to produce three topo-sheets of 15’x15’ (~23km x 27km). The examination of used Orthokits has shown that the image quality is good, with no undesirable radiometric defects. Imagery from the Fore band is seen to be much softer than from the Aft band, as it is illustrated in figure 1, but it is definitely sufficient for comfortable stereo viewing and feature extraction (Amitabh 2008).

Figure 1: Fore band & Aft band of port of Ghazaouet

4- Methodology The methodology is grouped into four stages. The first stage is related to GCPs collection (using Differential GPS) required for the refinement of RPCs for generating an accurate stereo model. The second stage is about performing block triangulation and DEM generation/ edition. Ortho images are also generated during this stage but they are not used in the rest of the workflow. The third stage consists in 3D feature extraction. The last one is about map creation. The details of these stages are as follows:

324

T. El-hadi

4.1 Collect of GCPs The methodology is presented in fig. 2. Prior to DGPS survey, all Aft bands of ordered orthokits are oriented using RPCs with positional accuracies in the order of 150 to 250 meters. A reconnaissance survey was carried out to locate the points identified on the images in such a way that the target locations were suitable for GPS deployment and measurements. After reconnaissance survey, 55 points were selected mainly road intersections. Using ArcGis, 55 identification sheets, as illustrated in fig. 3, were elaborated. Three dual frequency geodetic GPS receivers ASHTECH Z12 were used. They were used in synchronous mode. Two receivers were used for acquiring the GPS observations at the defined GCPs, one receiver was used as reference station at geodetic point. The GPS measurement campaign for establishing coordinates of GCP was for a total of 21 days duration. The cut-off angle was selected as 15° and epoch was set to 10 seconds. Geometric dilution of precision (GDOP) was better than 5 for most of the observation period. The precise coordinates of the geodetic point are used for the computation in post processing mode using BERNESE s/w. (06) orthokits ERDAS Orientation of all Aft-bands using RPC only (06) Oriented images ArcGis -import the Aft oriented bands -Identification of well distributed characteristic points -measure their approximate coordinates (55) Identification sheets of GCPs Collect of GCPs - three Geodetic GPS receivers - three Identification sheets of geodetic points GPS data BERNESE - DGPS computation post processing (55) accurate GCPs

Figure 2: Methodology of GCPs collection

Topographical Mapping at 1:50,000 Scale from Satellite Imagery using CARTOSAT-1

325

Figure 3: GCP identification sheet

4.2 Block Triangulation The methodology is presented in fig. 4. Raw satellite imagery contained in the orthokit product has large geometric distortion that is caused by various systematic and non-systematic factors (Titarov 2008). Interior and Exterior Orientations remove these distorsions. A project of block triangulation and stereo processing was created in LPS by defining the local cartographic projection, the geometric model of Cartosat-1 camera and importing the 12 images contained in 6 orthokits. Interior Orientation was computed for each image in the block using RPC. Exterior Orientation was computed using a total of (50) collected GCPs out of which (05) points were used as check points (refer to figure 5). Image matching performed between the fore and Aft images yielding a total number of 3477 tie points successfully created at a correlation coefficient better than 0.85. Triangulation was performed with 2nd order polynomial. The overall image RMSE achieved is 0.38 pixels. The overall ground RMSE of control points achieved is 2.62 meters in X-direction, 2.87m in Y-direction and 7.27m in Z-direction. The overall ground RMSE of checkpoints achieved is 3.04m in X-direction, 3.67m in Y-direction. This accuracy is sufficient for production of maps at 1:50.000 scale. A coarse DEM was generated at 10m grid interval. The correction of this DEM was performed in TerrainEditor and ProDTM. This corrected DEM was used in Ortho-image generation. We note that these intermediary products are used for other purposes and the 2D mode for features extraction

326

T. El-hadi

by on-screen digitization cannot be adopted in topographic mapping due to their poor accuracy for this kind of cartographic task. (06) orthokits and (55) GCPs LPS - Creation of the block - Interior and Exterior Orientations for all images in the bloc using RPC - Refinement of RPCs using GCPs - Generating of tie points -Improve RMSE -Triangulation

Coarse DEM LPS Terrain Editor ProDTM of Pro600 DEM editing Accurate DTM LPS Ortho-image generation

Ortho-images

Figure 4: Methodology of Block triangulation, DTM and Ortho-image generation

Figure 5: Control point and Checkpoint distribution over images and toposheets

Topographical Mapping at 1:50,000 Scale from Satellite Imagery using CARTOSAT-1

327

4.3 Feature extraction The resulted stereo model is used in feature extraction in stereo mode. The library of features to be extracted is a little bit different than the one which is usually used for aerial photography. We notice for example that high and very high voltage electrical network will not be extracted because we can not expect to see them in cartosat-1 images. The used library (non exhaustive list) is depicted in the table1. The used s/w in feature extraction are: LPS Stereo 9.1, Pro600 for LPS 9.1 and Microstation 8.0. The resulted DGN file were restructured using AcGis 9.2 s/w and exported to shapfile in order to be completed on the field using mobile mapping solution. Attributes and objects which were not visible on the images were brought from field and added to feature shapefiles. The final feature shapefiles were structured as GIS Database. An otho image overlaid with some extracted features is illustrated in figure 6. A visual comparison was made for contour lines generated automatically from an accurate DEM with contour lines extracted manually. This DEM was generated with 10m grid interval and it was corrected in TerrainEditor of LPS. It is important to notice that the process of correction has lasted more than two months for an area of 15’x15’ span. The visual result is shown at the real scale mapping in fig.7. The shift between the two kinds of contour lines was measured in X and Y direction. It was observed that this shift is within the range of 0m to 75m in hilly area and 0m to 200m in plane area. The automatic proceeding was rejected, after this experience as it was proven that it can not be adopted due to its poor accuracy for topographical mapping at 1:50,000 scale.

328

T. El-hadi

Cartographic features Communication line Highway, expressway

Cartographic features Construction Building

Road

Separated house

Unpaved track

Tight urban

railway

Muslim cemetery

Street and Alley

Ruin

Road en embankement

Aerodrome

Road under construction

Stadium Enclosure

Orography

Index contour

Barrage

Contour normal

dam

Contour thin Height spot

Hydrography Reservoir

Cliff

Permanent wadi

Embankment

Temporary wadi

Ridge line

Permanent water limit

Rocky area

Temporary water limit Sebkha limit

Vegetation

Forest

Coastal limit

Firebreak

Seguia

Vegetation limit

Foggara

Alfa limit

Gara

Vine limit

Basin

Parcel limit Arborous scrub Scattered scrub Orchard limit Trees line

Table 1: Library of features extraction (non exhaustive list)

Topographical Mapping at 1:50,000 Scale from Satellite Imagery using CARTOSAT-1

329

Figure 6: Ortho image overlaid with 3D extracted features

Figure 7: Comparison between manual extraction and automatic generation of contour lines

4.4 Field verification The field verification was completed using a mobile mapping solution (Tablet PC, hand held GPS and ArcPad s/w). The aim of this stage is mainly the confirmation of the extracted features with the ground observations and secondly to collect missing features and additional information

330

T. El-hadi

such as geographic names, the viability of roads and their administrative classification. Afterwards, all these in formations are integrated into a GIS Database. The field verification of the map of GHAZAOUET EAST has shown that the quasi totality of features on the ground has been extracted successfully, especially surface and linear features such as roads, vegetation, tight urban area, big buildings and hydrography. Nevertheless, Cartosat-1 imagery has shown some limits in identification of small features (punctual features in 1:50k map). The table 2 summarizes the results of field verification. Cartographic features

Number of features collected Number of features colfrom field lected from imagery

Reservoir

26

5

well

25

0

Source

17

0

Basin

18

2

Pumping station

13

0

Muslim cemetery

43

0

Table 2: Number of features collected from field and imagery

Moreover, during the image interpretation, some features were misinterpreted with other features. The field checking shows the results summed up in the table 3. Cartographic features Collected from field Extracted from imagery as Reservoir

12

Individual building

Basin

6

Individual building

Muslim cemetery

31

farms

Kouba-Marabout

39

Individual building

Small Mosque

40

Individual building

Ruin

8

Individual building

Non cart track

10

cart track

Table 3: Misinterpreted extracted features

Topographical Mapping at 1:50,000 Scale from Satellite Imagery using CARTOSAT-1

331

4.5 Map creation The cartographic edition was performed using LORIK s/w from the resulted GIS database. The conventional legend usually applied in 1:50,000 topographical mapping from aerial photo was used. Two topographical maps as final products are illustrated in figure 8 and figure 9.

Figure 8: Ghazaouet-West

Figure 9: Ghazaouet East

5- Results and conclusion Primarily, this study has assessed the capabilities and limitations of Cartosat-1 stereo images in topographical map production at 1:50,000 scale. It demonstrates that although some punctual features can not be extracted or could be misinterpreted as different features, the mapping at this scale using Cartosat-1 stereo images is possible; only it requires more time in field survey in comparison with classic mapping process where aerial photos are used. Secondly, the proposed methodology and the s/w and h/w framework allowed us to produce three effectual toposheets and may be adopted by other mapping entities. Finally, during this project, the alterna-

332

T. El-hadi

tive of automatic extraction of contours has been reviewed and rejected for the cartography at 1:50 000 scale.

References Amitabh B (2008) An integrated approach for topographical mapping from space using Cartosat1 and Cartosat-2 imagery, The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. http://www.isprs.org/proceedings/xxxvii/congress/4pdf/238.pdf. Accessed 10 June 2010 INCT (2005) Les spécifications cartographiques en usage à l’INCT, Le Bulletin Interne de l’Institut National de Cartographie et de Télédétection, Special edn. Kshma G, Sadhana J (2005) Enhanced capabilities of IRS-P6 LISS IV sensor for urban mapping, Current Science, 89:1805-1812 NRSA (2006) Cartosat-1, data user’s handbook. http://www.nrsa.gov.in/irs_documents/handbook/cartosat1.pdf. Accessed 7 January 2010 Puissant A, Hirsh J (2004) Télédétection urbaine et résolution optimale. Revue Internationale de Géomatique, 14 :403-415, doi :10.3166/rig.14.403-415 Titarov P S (2008) Cartosat-1 Stereo Orthokit Data Evaluation. http://www.racurs.ru/www_download /articles/cartosat_en.pdf. Acessed 15 May 2010

3D Topographic Mapping using TerraSAR-X Elevation Frank Teufel, Ernest Fahrland, Henning Schrader Astrium GEO-Information Service Claude-Dornier-Strasse 88090 Immenstaad, Germany [email protected], [email protected], [email protected]

Abstract The following paper describes the extraction of 1:25.000/1:50.000 topographic maps based on space borne SAR data derived from the German radar sensor TerraSAR-X. The developed mapping approach can be applied worldwide, even for small scale mapping, due to the high geometric accuracy as well as high reliability and imaging mode flexibility of the data acquisition of current SAR systems. The use of weather independent SAR imagery for object classification and Digital Surface Models (DSM) for extracting contour lines and for geolocating topographic and man-made objects was tested in the given approach. In the beginning, the background information for the development of the mapping process is described including a short touch on the current challenge for responsible authorities and institutions all over the world, especially in tropical regions, where closing the gap between non-existing or outdated maps and up-to-date topographic maps is aspired. Furthermore, the technical background of the used TerraSAR-X StripMap data, the details of the applied image acquisition scenario as well as the processing steps from raw satellite imagery to a source data set for feature extraction are described. An approach based on one sensor is presented; from the generation of a Digital Surface Model by using radargrammetric techniques through the correction of sensor based errors in the surface data to the subsequent orthorectification of the primary TerraSAR-X StripMap data will be presented. Subsequently, the derivation of a radargrammetric Digital Terrain Model (DTM) for generating contour lines that meet the requirements for the target scale of 1:25.000/1:50.000 is described. Processing steps for extracting topographic features in a combined processing in a 3D stereo working environment and the standard 2D working A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 333 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_23, © Springer-Verlag Berlin Heidelberg 2011

334

F. Teufel et al.

environment (based on TerraSAR-X StripMap data) are presented. This includes the optional use of ancillary data other than SAR. The description of the Quality assurance procedure description contains the presentation of tools developed to verify the quality of the image interpretation and technical data set quality. The visual example of a final 1:25.000 map sheet is given presented conclusively. The special application of change detection analysis based on SAR amplitude imagery for later map updates is described as enhancement of the one sensor approach. Finally, a conclusion and an outlook are given by presenting the future options of deriving upto-date topographic maps from space borne SAR Earth observation systems.

1- Background and objectives The acquisition of remote sensing data for topographic base mapping at scales 1:50,000 and 1:25,000 is often limited by unfavorable weather conditions such as permanent cloud cover or haze; particularly in equatorial regions. Even today, large regions remain unmapped; in other areas available maps date back to the 1940ies. The advantages of cloud-penetrating SAR systems carried by aircrafts have long been used for data acquisition in these areas, however, airborne data acquisition is expensive, time-consuming and, especially in border regions, often restricted by cross-national flight regulations. With TerraSAR-X, today's geometrically most accurate high resolution space borne SAR system, Astrium GEO-Information Services offers a new approach: It now enables a weather-independent, rapid and reliable data acquisition for large areas: a valuable basis for generating and updating 3D topographic base maps. In addition, future updates using SAR-specific change detection technologies reduce time and cost effort compared to conventional methods. A one-stop approach for SAR data acquisition, elevation model generation and map production is proposed in order to ensure a rapid availability of map updates or new base maps in high positional accuracy and thematic quality: TerraSAR-X data will be acquired in its high-performance mapping mode “StripMap”. Radargrammetric data pairs will be collected from two differ-

3D Topographic Mapping using TerraSAR-X Elevation

335

ent viewing angles in order to guarantee an area-wide survey without imagery information gaps caused by SAR characteristics such as radar shadow and to serve as input for a high quality elevation model extraction done with Astrium GEO-Information Services’ PixelFactory™. Map object interpretation is performed in a 3D environment (similar to the classical stereoscopic plotting), thus mitigating typical difficulties encountered in 2D radar interpretation as opposed to the interpretation of optical imagery. During the object capturing process, the TerraSAR-X ELEVATION DSM (Digital Surface Model) is edited to gather bare soil information, which leads to a Digital Terrain Model (DTM) required as an input for the generation of contour lines. The combination of high quality data and sophisticated techniques results in high-precision topographic map sheets in scale 1:25.000/1:50.000. The proposed concept comprises not only data acquisition, image processing and DSM generation, but all know-how necessary to set up the respective 3D Topographic Mapping production environment (technical consultancy, training, the licensing of the approved processing chain) which enables the customer to establish an independent map sheet production. The proposed concept's full comprehensiveness is exploited when map updates are generated, typically after 3 to 5 years: TerraSAR-X's consistent imaging geometry and the application of semi-automatic SAR based change detection algorithms. Operationally, the turn-key solution map generation as a full service is proposed. The key advantages of this concept: • rapid availability of country-wide image data and DSM, • ideal image and DSM pixel co-registration by one-shot acquisitions • field-work independent solution thanks to geometric accuracy of the system • precise topographic base maps including DSM and DTM, through 3D exploitation of required features, • cost-effective and rapid updates possible: consistent imaging geometry and sophisticated SAR-based change detection technologies. The application of space borne high resolution SAR data for topographic base mapping (particularly in large areas often covered by clouds) represents a readily available, fast, and cost-efficient alternative to current opti-

336

F. Teufel et al.

cal or airborne radar approaches, especially with regard to the increasing capabilities of upcoming space borne SAR missions.

2 Approach, methods & results Topographic base mapping at scales of 1:50,000 is a key challenge to the responsible authorities/institutions in many regions of the world: Particularly in equatorial/tropical regions, the reliable acquisition of data using space borne or airborne data sources is limited by frequent to even permanent cloud coverage and the fact that often vast areas need to be covered. To date, some remote areas around the globe have not been mapped at all; often, the available maps are from the mid of the last century. Since the early 1990ies, airborne interferometric SAR systems have been used for topographic mapping of large areas, providing weatherindependent SAR imagery for object classification and digital surface models for extracting contour lines and for positioning topographic as well as man-made objects. While such airborne campaigns usually provide excellent results, their drawback is a high data acquisition cost, substantial non-recurring cost generated by mobilization and demobilization of crew and equipped aircraft, and the often extensive time frames: Assuming areas of more than 300,000 km², airborne SAR imagery mapping projects typically require at least four years for completion.

2.1 Solution Today’s geometrically most accurate high resolution SAR system, TerraSAR-X, was successfully launched mid 2007 and has reliably been acquiring premium quality radar data since then. TerraSAR-X can provide high resolution image data that fulfils all object extraction requirements for topographic base mapping at the scale 1:50,000 for large area coverage. Compared to other existing SAR satellites, TerraSAR-X is superiorly suited for topographic mapping applications due to its high resolution, geometrical accuracy and flexible acquisition modes. In combination with its short access time, this results in an optimal combination of SAR data acquisition parameters. A digital elevation model for extracting contour lines is a substantial ele-

3D Topographic Mapping using TerraSAR-X Elevation

337

ment for topographic base mapping. In order to achieve the required geolocation accuracy a DSM (Digital Surface Model) is generated using radargrammetric TerraSAR-X data pairs and applying the respective radargrammetric techniques. In an additional processing step, the DSM is edited into a bare earth DTM (Digital Terrain Model).

Figure 1: General workflow topographic mapping

The described approach provides a solution of high quality, speediness in mapping and cost-efficiency: Based on TerraSAR-X, data acquisition for large areas can be achieved in a significantly shorter time frame than by airborne campaigns. Data sets become available successively throughout the project duration, rather than as one large bulk at the end of an airborne campaign, and map sheets can be produced locally (in country) while data acquisition is still ongoing.

2.2 TerraSAR-X data acquisition In the first phase of the outlined project, TerraSAR-X data in StripMap Mode (3m resolution) will be acquired instantaneously and as rapidly as possible. For reliable map interpretation and DEM generation, data will be acquired in four coverages: one radargrammetric pair in ascending and one pair in descending orbit. This ensures a complete survey without gaps in mountainous areas and guarantees a precise elevation model production.

338

F. Teufel et al.

Figure 2: TerraSAR-X StripMap acquisition mode.

As an example, a full radargrammetric stereo coverage of Sumatra (300,000 km²) in ascending and descending path direction (4 complete coverage’s) with TerraSAR-X StripMap data requires 2415 scenes. To cover the Sumatra area 4 times an acquisition window of at least 8 months is required. It will be necessary to begin data acquisition in due time prior to the start of the production phase in order to ensure a tightened timeline of the mapping project.

Figure 3: Coverage analysis (ascending/descending stereo, 2415 TerraSAR-X StripMap scenes) for Sumatra.

3D Topographic Mapping using TerraSAR-X Elevation

339

2.3 TerraSAR-X Radargrammetric DSM As part of the map production process, Astrium GEO-Information Services will create its TerraSAR-X ELEVATION DSM using TerraSAR-X StripMap image pairs that meet radargrammetric geometry requirements. The radargrammetry technology allows height extraction applying the similar principle as in photogrammetry: the measurement of the deviation of objects in image pairs acquired from different viewing angles in order to obtain 3D-coordinates for pixels. The precise knowledge of the TerraSAR-X Satellite orbit and altitude allows extraordinary position accuracy for the TerraSAR-X imagery. TerraSAR-X has been calibrated and validated by the German Aerospace Center (DLR) during the satellite commissioning phase. In addition the accuracy has been confirmed by a number of international organizations and can be found in publications like (Ager and Bresnahan 2009). The orbit position accuracy is guaranteed to be better than 50 cm in across and along track. This world wide unprecedented satellite precision results in xyz without Ground Control Points (GCPs) in a range of 5 to 10 m applying TerraSAR-X based radargrammetry.

Table 1: Overall absolute vertical accuracy of TerraSAR-X Stereo DEM depending on slope

To achieve the required horizontal accuracy for topographical mapping at scale up to 1: 25,000 TerraSAR-X provides a sufficient solution without using of any GCPs.

340

F. Teufel et al.

2.4 DSM generation and editing The basis for the TerraSAR-X ELEVATION DSM is a complete coverage of the target area with overlapping image pairs of StripMap data. A second image pair taken from the opposite path direction will allow to fill gaps created by SAR-specific effects like shadowed areas. Radargrammetry is based on the matching of homologous points from two amplitude images of the same area, which have a different geometry. The radargrammetry process contains three general processing steps: • Set up of geometric stereo model • Raw DSM generation based on stereo matching • Merging of raw DSM for ascending and descending orbit directions. Output of the radargrammetry process is a raw DSM that still contains void areas and artifacts which will be subject to a subsequent DSM editing process. For a continuous and accurate representation of the Earth’s surface elevation, the raw DSM needs to be edited. In order to ensure that the TerraSAR-X Elevation is consistent, the editing is carried out in two main steps: The first editing step contains the correction of elevation errors caused by SAR specific characteristics or DSM processing (DEM level 1). After that the editing of the water bodies (hydro enforcement) will be performed if requested (DEM level 2). The following processing steps and rules applied to the raw DSM are: • Interpolation of small voids (smaller or equal 8 pixels) and blunders • Filling of large voids (larger than 8 pixels) and larger artefacts with ancillary DSM data (as available) • Global or local smoothing, if required • Additional manual editing, if any artefacts remain present in DSM data On request, the water bodies will be edited based on the following editing steps and rules: • Identification and extraction of any water body feature according to Water Body Editing Conventions and feature elevation measurement.

3D Topographic Mapping using TerraSAR-X Elevation

341

The extent of the water bodies is derived from corresponding TerraSAR-X image data considering the following feature parameters: • Permanent water bodies: lakes and reservoirs with an area larger or equal 25,000 m² are set to a single elevation • Double line drains: rivers and canals with a length of 300 m or larger along medial axis and a width of 90 m or larger perpendicular to the medial axis are flattened with monotonic flow • Tidal water bodies: the elevation height is set to 0 m • Islands with a minimum area of 12,000 m² or a height difference of more or equal to 15 m to the surrounding water surface • Coastal infrastructure features are removed, unless they comply with editing rules • Automatic inclusion of the water body data set into the DSM data set (including correction of shore pixels to ensure a hydrological consistency). The shoreline is always higher than the water surface elevation. • Additional manual editing of shorelines according to the morphology of the Earth’s surface derived from TerraSAR-X images Further editing towards a terrain model (DTM) is described in the following chapter.

2.5 3D Topographic Mapping 1:50,000 and DTM generation The extraction of topographic features is performed on a 3D Analytic Workstation (similar to the classical stereo-plotter) using the TerraSAR-X StripMap image data and the elevation model which was the output of the DSM editing process described in the previous chapter. The DSM is converted into a Digital Terrain Model (DTM): all unwanted elevations from the earth’s surface (e.g. vegetation or buildings) are removed to finally represent the bare Earth elevation. This is achieved by digitizing vector elements such as break lines, depression lines and areas indicating necessary height reduction and / or homogenous leveling of the surface (e.g. forests, built-up areas).

342

F. Teufel et al.

Figure 4: Digital Surface & Terrain Model (DSM & DTM).

Simultaneously to this process, the topographic objects (mainly linear objects such as hydrographic or the networks) are digitized using a 3D-GIS environment.

Figure 5: Main feature classes extracted in 2D & 3D GIS environment (red boxes)

The object and attribute extraction of single objects is based on TerraSARX StripMap data and according to the Object Feature Catalogue. If any additional data are available and useful (e.g. topographic maps) they will be integrated into the process. In order to work simultaneously on one map sheet and to avoid unnecessary effort in working unit edge matching the chessboard principle will be applied.

3D Topographic Mapping using TerraSAR-X Elevation

343

Figure 6: Chessboard principle: tile complementation starting with linear elements, followed by build-up & landcover area features

The extraction will be done as precise as required; generalisation is only permitted for specific feature classes. The displayed workflow is also independent from the remote sensing base data used. The object extraction scheme is given by the sequence of the geometry types: 1. Object extraction of linear object types 2. Object extraction of areas and points. At a first step all linear objects will be extracted to secure a first object network. In most cases area objects have to be coincident with line objects (e.g. built-up areas have to be connected to roads). Therefore all transport objects will be extracted first.

344

F. Teufel et al.

By regarding the different geometry object types (points, lines, areas) and the thematic focus a strict object extraction sequence will be applied to the process: 1. Hydrography 2. Man-made ground coverage, aviation information 3. Natural land cover.

Figure 7: Object extraction sequence

This thematic procedure is essential because there are relations between several object features existent who require a fixed sequence. All object feature classes existing inside a single working area have to be extracted by the operators. The attribute extraction for the objects follows the guidelines of the object feature catalogue and the extraction guide, framed by combination rules for attribute values and ranges. Feature extraction and attributing will be performed according customer specified standards. For interpretation of the TerraSAR-X StripMap data a comprehensive extraction guide is available to guarantee the capturing of

3D Topographic Mapping using TerraSAR-X Elevation

345

all required features. It contains a full description of the relevant feature and its attributes and gives interpretation advices in combination with the appearance in optical data by the possible export of a Google Earth *.kmlfile.

Figure 8: Example of the TerraSAR-X Extraction Guide

346

F. Teufel et al.

Table 2: TerraSAR-X main feature classes according FACC (Feature and Attribute Coding Catalogue) and reliability of classification using StripMap data (3m resolution) for topographic mapping at a scale 1:50,000 1 Reliability of classification is an average value and depends on the mapped area 2 Reliability for land cover features can be increased by using additional sources like archive optical data

3D Topographic Mapping using TerraSAR-X Elevation

347

In a further step, contour lines are derived from the finalized DTM. For mapping at scale 1: 50,000 major contour lines will be extracted every 250m (equidistance), accompanied by auxiliary lines every 25m. A minor smoothing algorithm will be applied in order to meet the cartographic expectations. After contour line extraction verification with water bodies and break-lines will be performed. Beside this standard approach every user specified parameters can be integrated into the process.

Figure 9: Working steps from raw DSM to a full scale topographic map as shaded relief (clockwise from upper left): Raw DSM; edited DTM; DTM with break lines and river lines extracted during editing; final layers of vector information

Additional data (such as optical archive imagery from Landsat and Google Earth) is not necessarily required, but integrated when available to support and verify the interpretation.

348

F. Teufel et al.

A sophisticated production management and reliable data administration of the spatial geo data base enables multi-user editing. Extraction of specific feature classes can thus be performed by specialized operators.

Figure 10: Final map sheet 1:25.000.

Figure 11: Final SAR map 1:25.000

3D Topographic Mapping using TerraSAR-X Elevation

349

2.6 Quality Assurance A reliable Quality Assurance (QA) concept is implemented throughout the process of production, with an emphasis to quality control procedures for the interfaces between different process building blocks, in order to monitor the production workflow and to guarantee a constant level of product quality according to the customer’s specification. As a fixed part of the work flow, certain staff will be responsible for the implementation, execution and documentation of QA procedures. The main quality assurance procedures for the final product can be separated into three steps: • • • •

Operator QA QA1: Final controlling of all processed working units QA2 3D: Plausibility check in 3D environment QA3: Final overall quality checks for extracted objects and attributes.

The first QA step is done by the operator himself to control the own work and support a learning procedure. This thematic check is performed on working unit level. Through several program routines and scripts, based mainly on special ESRI tools (developed by Astrium GEO-Information Services), a structured and scheduled working will be implemented in QA 1. These routines are covering typical GIS aspects like snapping on line and point features or topological relationship between distinct object feature classes. The results of the checks will be visualised (error tables) and corrected by the Operator. The QA1 will be executed by well experienced operators, who are not involved in the initial object extraction procedure. By using this independent quality it is secured, that every object inside the data set is checked by at least two persons. All detected aspects will be discussed with the operator to secure a learning effect and common understanding. The reported aspects are stored inside a specific database, to guarantee a fast and efficient quality check cycle. Beneath the standardised detection comments or assignments for the operator are integrated in a structured way. All executed quality control will be reported in check lists. The control mechanism contains visual and automated checks.

350

F. Teufel et al.

Figure 12: QA workflow with major QA breakpoints

Figure 13: Example error table for verification

After finalizing QA1, all working units will be combined physically into one single main data set. All processed working unit areas will be checked within a 3D environment for plausibility and information density to secure a harmonised data set.

3D Topographic Mapping using TerraSAR-X Elevation

351

At a closing verification step all geometries and attributes are being verified with Astrium GEO-Information Services/ESRI tool routines. The final overall quality procedure will be executed by GIS experts who are also a broad range of interpretation experience. All aspects are reported according to the QA1 procedure.

Table 3: Quality assurance steps regarding thematic content, topology and plausibility

2.7 Option: Future Map Updates Upcoming updates of the topographic maps will be facilitated by the SARspecific change detection technology and pixel precise fitting accuracy, which enables a rapid and accurate detection of changes in SAR-imagery from different acquisition dates. TerraSAR-X scenes of a specific location, captured with identical acquisition parameters (looking direction, orbit & incidence angle) will show exactly the same reflectance of objects on the ground. Any changes in land cover and infrastructure are highlighted when comparing the images and can be extracted using a semi-automatic approach (see figure 14). Such efficient update processes reduce the cost of up-to-date map generation significantly. The sustainability of these possibilities is ensured beyond the lifetime of TerraSAR-X and TanDEM-X by the future TerraSAR-X 2 missions

352

F. Teufel et al.

Figure 14: Change monitoring in urban environment; single TerraSAR-X acquisition (left, centre) with a six weeks difference in the capture date. The overlay of those scenes on the right highlights construction sites by showing coloured areas

2.8 TerraSAR-X Topographic Mapping Processing Chain Astrium GEO-Information Services’ processing chain for the 3D topographic mapping with TerraSAR-X StripMap data is an efficient and validated approach to fulfil the customer requirements rapidly while securing a proven and constant quality. The processing chain is based on TerraSARX raw data, which guarantees the high-precision geometry and contains DSM and DTM production as a necessary input for the 3D object extraction. It is an end-to-end solution fully aligned from data source to final map sheet in order to generate a precise and high quality product. The Astrium GEO-Information Services TerraSAR-X Topographic Mapping Processing Chain consist of all required software modules, job instructions, data handling and administration procedures, quality assurance steps, maintenance and management instructions to guarantee a recurred quality and performance even with varying staff.

3- Conclusion and future plans

3.1 Methods for DTM Generation TerraSAR-X radargrammetry generally applies the same technical principals as the well known photogrammetry. The DTM generation in dense forest is a general challenge. Since X-Band is reflected in the upper 50 cm

3D Topographic Mapping using TerraSAR-X Elevation

353

of the tree crowns, the elevation estimation of bare soil meaning the terrain requires interpolation between visual detected ground points. This approach, like in photogrammetry, leads to a ground level especially on dense forest which is not fully accurate due to the approximation. Another approach is the application of airborne based P-Band IFSAR for DTM generation. Unfortunately P-Band is also not reflected on the ground in dense forests. The majority of the P-Band reflection is from the tree trunks in approximately 2 meter height from the ground. It is preferably used for biomass retrieval via respective models. The model required to derive the DTM has also its uncertainties and is not to be considered as a more reliable method compared to TerraSAR-X based radargrammetry, especially when considering the additional effort of P-Band data acquisition and processing. The elevation requirements for 1:50.000 topographical maps, in general can be met by the TerraSAR-X derived DTMs.

3.2 Conclusion The presented approach based on the use of TerraSAR-X and a TerraSARX derived radargrammetric elevation model allows a quick and reliable topographic mapping of large areas, such as the complete territory of Sumatra/Indonesia, in a comparably short time frame. Thus, mapping contracts can be designed to available budgets upon short notice and with a comparably low planning effort. The proposed approach will deliver an initial topographic base map rapidly and, upon full completion, provide position accuracies satisfying current requirements for the respective scale using the high-precision elevation model. While the final product quality is comparable to today's airborne InSAR solutions, the price is considerably lower compared to airborne campaigns. With this concept, the high-precision topographic maps and DEMs will be available within 2 years. Further, there are a number of constraints associated with airborne SAR data collection and delivery compared to satellite data: The described space-based approach, compared to an airborne acquisition approach, enables following advantages: • • • •

Lower price compared to an airborne solution; Timely delivery of topographic map products; High quality, homogeneous DEM; Adequate accuracy for the addressed scales;

354

F. Teufel et al.

• Maximum cost-efficiency for map updates (key point for a long term mapping strategy).

Table 4: Comparison Airborne SAR / Spaceborne SAR

Further, reliable future map updates are ensured due to the envisaged continuity of the TerraSAR mission: TerraSAR-X, TanDEM-X and the upcoming sensor TerraSAR-X 2. Thus, a consistent mapping and update scheme can be provided: based on the same data sources, the implementation of real update techniques rather than new mapping is possible. This concept, enabling rapid availability of country-wide imagery, the coregistered pixel accurate high-precision elevation model and the resulting precise topographic base maps, make the application of space borne high resolution SAR data for topographic base mapping (particularly in large areas often covered by clouds) a readily available, fast, and cost-efficient alternative to current optical or airborne radar approaches.

References Ager T and Bresnahan P (2009) Geometric Precision in space radar imaging: results from TerraSAR-X, ASPRS 2009 Annual Conference, Baltimore, Maryland, USA, March 9-13, 2009

Terrain Mapping and Analysis

An Influence of Spatial Range of Input Data set on Terrain Relief form Classification Homogeneity for Glacial Area Małgorzata Wieczorek Department of Cartography, University of Wrocław, Poland [email protected]

Abstract Morphometric analyses performed on various areas and various methods are still very popular nowadays. There are both raster data and vector data analysed in this scope. In this article, there is a comparison described between two analyses of the same area. The input data was a 10 m × 10 m raster DEM representing SW part of Spitsbergen with actual size 15 km × 18 km. The data was characterized by 5 attributes: relative height, slope, aspect, and also plan and profile curvature. The first analysis was done on the full input data. For the other one, the input data was limited to the mainland only (excluding waters and subarea with glacier itself). In both classifications k-median method with Manhattan metric was used. The results were compared using statistical methods and visual evaluation. More detailed analyses were performed for two selected classifications. Finally, it is remarkable that suggested limitation of analyzed area improved homogeneity of result classes.

1- Background and objectives Classification of relief forms is still an interesting topic for many geomorphologists, geographers and people who work with spatial analysis. Frankl et al. (2009) analysed glacial landscape focusing on landforms. This analysis was based on DEM in TIN format and on interpretations of the stereographic aerial photographs. The main morphometric variables were slope gradient, valley profiles and slope orientation.

A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 357 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_24, © Springer-Verlag Berlin Heidelberg 2011

358

M. Wieczorek

Another paper about using DEM to determine relief forms was written by Chvatalova (2000). In this analysis the morphometric variables were: mean height, slope gradient, profile and plan curvature. Similar relief attributes used Jordan (2003) examining utility of digital geomorphometric method for tectonic geomorphology. He achieved it by applying trend analysis. This idea is similar to the statistical method proposed by Evans (1972). Very useful morphometric attributes from Wood’s dissertation (1996) were used by Jasiewicz and Hildebrand-Radke (2007) for relief form classification on lowland area. The simplest approach to object analysing is taking into account only one particular feature. It is quite easy to evaluate similarity and correlation between two given features using statistical appliances. Geographical space, including geomorphological one, is more complex, this is why it is always described by many features. Nowadays, when we have much more random access to digital data describing geographical surface, the problem is focused not on data accessibility, but on its interpretation in multidimensional space described by this data. To support this phenomenon, specialized geomorphological data management systems are being developed and Jedlicka (2009) described one of them. Considering the problem by analysing many variables at the same time, statistical methods might be helpful to solve it. Especially the ones that can be applied not only in single dimensional analyses, but also multidimensional analyses. The most popular here are methods that implement data mining including fuzzy k-mean algorithm, ISODATA algorithm, k-mean algorithm or SOM (Self-Organizing Maps). Morphometric analyses are used not only for continuous, but also for noncontinuous data, for instance irregular points. An example of such application is paper of Bishop (2009) where he uses taxonomy for statistical spatial analysis of landform patterns within GIS. An interesting approach was presented by Arrel et al. (2007), where classification procedure was run several times on the same area, but using DEMs with different resolutions of 50, 100, 200 and 400 m. These analyses were compared to each other and the influence of this contribution to classification results has been proved. Spatial analyses can be based on either vector or raster data. When we want to use raster data (e. g. DEM), a number of samples is determined by the area size and raster resolution. Adequate area selection is the most sig-

An Influence of Spatial Range of Input Data set on Terrain Relief

359

nificant for methodical research based on statistical methods. Wrong data selection can produce incorrect conclusions. How does terrain relief morphometric classification work if we reduce the analysed area if we split it into smaller parts, non-homogeneous in morphometric aspect? Analysed area is either rectangular or the boundaries are aligned to the actual boundaries of a given natural (watershed, geographical region) or artificial unit (countries, states). The Werenskiold glacier and its surroundings differ a lot in morphometric aspect. The surface of glacier is quite flat so with little elevation change and vertical curvature value mostly near zero. It is also very coarse so that aspect varies a lot. On the other hand periglacial area is characterised by greater values of all slopes, both curvatures and altitude change. In this article, the author describes an experiment that is based on comparing the classification of periglacial area containing glacial area itself to the classification of the same area but excluding the glacial area. The key questions were: does this exclusion influence the final classification result? If so, what is the strength of this influence, does the difference spread to all dimensions or is it accumulated in one parameter only?

2- Approach & methods Using the area of SW Spitsbergen (Svalbard) there was an experiment run as a comparison between results of two morphometric classifications that used k-median method. The main focus was put on the form shape, putting morphology and genesis aspect out of the scope. This area has a great variety of elevation. Variety of surface shape is different on glacier and mainland. The size of the basis sample area is 15 km × 18 km.

360

M. Wieczorek

Figure 1: Digital elevation model of area W1 – SW Spitsbergen.

The analysed data (Fig. 1) is the area of Spitsbergen limited by coastline on the South and West. On the North and East it is limited by the selected area size. The subarea for the second analysis was obtained by cutting out the part containing lakes and glacier (Fig. 2), which means such objects that are easy to identify on a satellite image. The full area is about two times greater than the subarea. DEM used in this analysis was made by NPI-Troms and University of Silesia – Sosnowiec, processed in 2005 by

An Influence of Spatial Range of Input Data set on Terrain Relief

361

Leszek Kolondra. Also, data for digitalization of glacier and coastline was obtained from the same source.

Figure 2: Area W2 – the same DEM as in W1, but containing mainland only.

The sole input data is DEM with resolution of 10 m × 10 m. From this data, 5 other data sets are derived by using proper transformations available in ArcGIS 9.2. The data obtained is: relative elevation (dH) evaluated in 5 × 5 raster frame, slope, aspect, plan curvature (plan) and profile curvature (profile). Then, all parameters were processed by statistical analy-

362

M. Wieczorek

ses. Descriptive statistics for both areas are presented in the table 1. It is important to pay attention to such variables as slope, aspect and relative elevation, that have asymmetric distribution for both areas (skeweness is greater than zero). Although the skeweness for these variables is greater for area W1, the distribution is still remarkable. Due to this lack of symmetry, k-median method was applied. For variables describing curvature, kurtosis reaches unordinary high value. Taking it together with percentiles, it can be noticed that 80% of observations are less variant than it could be expected if taking into account only the maximum and minimum values. Kurtosis for area W1 is a few times greater than the other one. It might suggest that excluding glacier subarea from the input data may improve classification homogeneity on the mainland. It is also remarkable that median value for parameters is greater in area W2 than in the area W1, which indicates that there is a big difference in distribution between these two areas. -

W1 dH

slope

N

W2

aspect

plan

profile

dH

slope

2 008 322

aspect

plan

profile

1 039 392

min

0,00

0,00

0,00

0,00

-1 -30,88 -17,84

q10

1,83

1,77

51

-0,51

-0,49

2,11

2,01

53

-0,65

-0,69

q25

2,96

3,26

111

-0,19

-0,15

3,84

4,41

129

-0,25

-0,23

13,55

13,69

189

0,03

0,03

19,44

19,29

195

0,08

0,01

mean median

0 -35,55 -42,49

6,33

7,18

197

0,00

0,03

15,50

16,97

208

0,03

0,05

q75

21,26

22,73

268

0,20

0,26

32,80

33,31

264

0,34

0,33

q90

37,18

36,37

317

0,58

0,60

42,76

40,03

315

0,85

0,72

max

86,57

68,34

360

36,32

47,68

86,57

65,01

360

17,88

34,52

std_dev

14,45

13,46

96,19

0,72

0,79

16,42

15,03

92,92

0,84

0,81

skeweness

1,33

1,03

-0,15

0,05

1,83

0,63

0,32

-0,33

1,02

-1,29

kurtosis

0,86

-0,23

-1,02 181,69 439,12

-0,65

-1,30

-0,76

18,41

22,75

Table 1: Descriptive statistics for area W1 and W2.

Aspect, as a directional variable, is not expressed in a linear scale, but circular. Due to this, there was the Manhattan metric used in analyses instead of common Euclidean (Wieczorek and Spallek 2008, Wieczorek 2009). To eliminate extreme particular values, median filter in a 5 × 5 frame was applied for all parameters (except aspect) (eq. 1- 4). This transformation meets the requirement that all morphometric variables must describe an area of the same size. Median filter is also more suitable for asymmetric

An Influence of Spatial Range of Input Data set on Terrain Relief

363

distribution than arithmetic mean or modal value, which is very important for such variables as slope or relative elevation. (1) hi = Me({dHi(1), …, dHi(25)}) (2) si = Me({slopei(1), …, slopei(25)}) ciprofile = Me({profilei(1), …, profilei(25)}) (3) (4) ciplan = Me({plani(1), …, plani(25)}) As far as aspect is concerned, a transformation function of the evaluation of a direction of mean vector (eq. 5) was applied in a frame with the same size as for other variables. ai = arc(mean vector({aspecti(1), …, aspecti(25)}))

(5)

The result of these transformations was a set of layers. It was then the subject of a final classification. The scope of values clearly decreased for variables transformed by single-step median filtration and did not change for aspect. This was the expected result (Wieczorek and Spallek 2008).

3- Results The classification of area W1 was described by Wieczorek (2009), where 4,…, 9 classes were used. In classification of area W2 there were 3–8 classes used. One result class (A) in area W1 was usually located in the sea, coast zone or inland waters. These subareas were not analysed within area W2. Therefore, the classification results of the W2 should be compared to results of area W1 with reduced number of classes. A classification result with the greatest average BSS value (between cluster sum of squares) was treated as the most optimal one. For area W1 the greatest value was obtained for k = 8 and only a bit smaller for k = 7. For area W2 the optimal classification was for k = 6 (Table 2). Shifted with one class, the difference between results from both areas was calculated. Such calculation gave the possibility of equality rate evaluation. The best value was observed for k = 4/3 (76%). The results differ the most for k = 6/5. Only 40% of input data was assigned to the same class. For k = 7/6 equality rate was almost 60% (see Table 3) the same as for k = 8/7. Considering BSS values only, the result of classification of W1 to 7 classes was compared to adequate one for W2, means 6 classes (Fig. 3). Classifi-

364

M. Wieczorek

cations for the second comparison were about to be chosen on the base of equality rate. The ones to 4/3 classes have the highest equality but the lowest BSS values. However, after a visual analysis in a map view, it can be noticed that such a varied terrain cannot be represented by such a small number of classes. Therefore, the 5/4 classes were selected (Fig. 4), because they also have high equality and BSS values are among the lowest ones. area W1

k

area W2

BSS

BSS/class

BSS

BSS/class

10,08

3,4

4

5,896

1,5

14,216

3,6

5

7,941

1,6

20,223

4.0

6

9,372

1,6

29,342

4,9

7

13,016

1,9

33,087

4,7

8

15,667

2,0

37,796

4,7

9

15,397

1,7

3

Table 2: BSS and average BSS values for area W1 and W2 depending on the number of classes (k). k for W1

k for W2

equality

4

3

76%

5

4

72%

6

5

40%

7

6

59%

8

7

59%

9

8

46%

Table 3: Classification equality rate between areas W1 and W2. k

W1 h

s

a

A

0,00

0,00

N

B

2,08

4,79 SW

W2 c

plan

profile

c

h

s

a -

0,00

0,00

-

-

-0,05

0,05

3,43

3,85 SW

cplan

cprofile

-

-

0,02

0,00

C

17,16

33,97 SW

0,14

0,06

33,41

33,63 SW

0,14

0,07

D

17,71

34,54

0,76

-0,27

35,95

35,87

S

0,77

-0,15

E

7,70

20,16 NW

-0,06

0,44

16,72

18,54

W

-0,07

0,40

F

2,24

4,78 SW

0,08

-0,04

G

16,60

34,26 SW

-0,34

0,12

36,85

36,28

SE

-0,32

0,13

-

-

28,75

30,14

N

0,41

-1,32

H

-

-

S

-

-

-

-

Table 4: Group centres comparison for classification into 7 and 6 groups.

-

-

An Influence of Spatial Range of Input Data set on Terrain Relief

365

Figure 3: Classification into 7 groups on area W1 (on the left) and 6 groups on area W2 (on the right). Grey colour represents area that was not compared.

As it can be seen in the figure, classes in area W2 are more consistent so it is easier to interpret them. On the base of maps of forms (Fig. 3 and 4) and group centres characteristics (Table 4 and 5) these classes can be interpreted as follows: Class A occurs on plain terrain like waters or seaside. Class B can be interpreted as slight slopes. It is the most varying class between two analysed areas. In area W1 such slopes are represented by mixture of classes A, B and F. In W2 these areas were represented consistently by class B. Steeper hillsides are described by classes C, D and G, where slope exceeds 30 degrees. These hillsides that have plan curvature nearly zero are mostly classified as C. Another kind of relief form, that is concave both in vertical and horizontal slice, can be identified as mountain ridge. All such terrains were classified as D on W1. These ridges that can be treated more as crests were allocated in a separate class (H) on W2. Horizontally convex part of hillside is a valley. This kind of relief form was grouped into class G.

366

M. Wieczorek

Class E is difficult to interpret. It occurs mostly as a boundary between hillside and plain or slight slopes. k

W1

W2

h

s

a

A

0,00

0,00

N

0,00

B

1,99

4,42

S

0,01

C

17,08

D

17,03

E

7,35

34,48 SW 33,46 SW 18,97

W

c

plan

profile

c

h

s

a

0,00

-

-

-

0,00

3,68

4,10 SW

-0,19

0,09

35,48

0,47

-0,14

-0,04

0,38

35,17

S

33,42

33,86

S

21,48

23,45 NW

cplan

cprofile

-

-

0,02

0,01

-0,06

0,07

0,64

-0,23

-0,07

0,43

Table 5: Group centres comparison for classification into 5 and 4 groups.

Figure 4: Classification into 5 groups on area W1 (on the left) and 4 groups on area W2 (on the right).

However, what is the data range variety like for particular variables or k values? Regardless of the number of classes, aspect takes values from the full scope in each analysed area. Curvature significantly increases its amplitude in W2 compared to W1 only in class B for k = 6 (Table 6). Other dimensions for all other classes have mostly similar amplitude to classes in W1 for k = 7. Nevertheless, standard deviation is usually lower in W2 classifications. This trend is more remarkable for k = 6 than for k = 4 (Table 6 and 7), which allows to deduce that classes in W2 classification are more homogeneous than in W1 case.

An Influence of Spatial Range of Input Data set on Terrain Relief

367

The differences between W1 and W2 in values ranges and standard deviations are similar for particular parameters in all classes. Therefore, none of the morphometric variables used cannot be indicated as the most significant for homogeneity increase. class B C D E G

area

max-min

std_dev c

plan

h

s

a

W1

41,3

34,2

361

20,5

W2

35,3

32,3

361

25,9

W1

82,3

59,8

360

W2

83,0

59,4

360

W1

84,4

62,4

W2

85,4

W1 W2

profile

c

cplan

cprofile

h

s

a

18,0

5,7

6,3

77,3

0,7

0,7

20,8

4,2

4,8

92,2

0,7

0,7

20,0

22,6

10,4

7,6

89,0

0,4

0,6

16,3

18,3

10,3

7,5

84,8

0,3

0,4

360

27,8

25,6

13,5

11,0

95,4

1,2

1,1

60,7

360

22,3

23,0

13,3

10,2

90,2

0,9

0,7

83,9

60,2

361

21,6

22,7

12,5

11,4 100,0

0,7

0,7

83,9

58,8

361

21,9

22,7

10,3

9,8

99,7

0,7

0,7

W1

82,2

64,6

360

48,8

52,4

11,8

9,2

89,4

0,7

0,6

W2

81,5

59,0

360

48,8

52,4

10,8

7,9

85,8

0,6

0,5

Table 6: Comparison of range and standard deviation between W1 and W2 in a given class and area context for classification into 7/6 groups. class B C D E

area

max-min

std_dev cplan

h

s

a

W1

50,2

42,4

361

22,3

W2

45,6

42,0

361

33,0

W1

81,4

64,5

360

W2

82,6

64,7

W1

84,9

W2

85,8

W1 W2

cprofile

cplan

cprofile

h

s

a

26,3

5,8

6,3

71,0

0,7

32,8

4,8

5,4

90,9

0,7

0,7

41,6

45,5

10,8

8,2

89,0

0,6

0,6

360

41,6

46,8

9,7

7,0

79,9

0,6

0,6

62,6

360

27,8

22,9

13,6

11,1

93,2

1,0

1,0

62,5

360

27,8

25,6

13,3

10,8

93,4

1,1

1,1

86,6

60,2

361

27,6

29,8

12,9

11,8

96,8

0,7

0,7

86,6

60,2

361

29,2

23,5

13,2

11,4 114,4

0,7

0,7

0,7

Table 7: Comparison of range and standard deviation between W1 and W2 in a given class and area context for classification into 5/4 groups.

4- Conclusion and future plans It is important for clustering that input data should be properly prepared. All areas like waters, forests and others not being in the scope of interest, should be excluded from the analysis. Good approach then is to use

368

M. Wieczorek

boundaries (natural or administrative) of a given phenomenon as boundaries for an analysis instead of using classic rectangular area. However, even if the input data contains areas that are not in the subject, classification should provide good results for more distinct relief structure. But minor ones might not be detected in such a case. This is possible, because areas with high elevation amplitude and continuous relief are clearly distinct in morphometric aspect. When considering terrain with low elevation amplitude with many hills and pits, the impact of such an exclusion of unimportant areas still needs to be analysed. Further planned researches on k-median method on this area are to analyse the influence of particular morphometric parameters on classification results.

Acknowledgments Author is grateful to North Polar Institute and University if Silesia – Sosnowiec for providing source data that was needed for performing this research.

References Arrell KE, Fisher PF, Tate NJ, Bastin L (2007) A fuzzy c-means classification of elevation derivatives to extract the morphometric classification of landforms in Snowdonia, Wales. Computer and Geosciences 33: 1366–1381 Bishop M A (2009) A generic classification for the morphological and spatial complexity of volcanic (and other) landforms. Geomorphology 111: 104–109 Chvatalova A (2000) Vyuziti digitalnich modelu reliefu v procesu gegrafickeho poznavani mistniho regionu. Pedagogicka fakulta Univerzity Jana Evangelisty Purkyne Evans IS (1972) General geomorphometry, derivatives of altitude and descriptive statistics. In: Chorley R (ed) Spatial Analysis in Geomorphology, Methuen and Co. London, pp 17–91 Frankl A et al. (2009) Use of Digital Elevation Models to understand and map glacial landforms – The case of the Canigou Massif (Eastern Pyrenees, France). Geomorphology 115: 78–89 Jasiewicz J, Hildebrandt-Radke I (2007) Zastosowanie cyfrowych modeli rzeźby do konstrukcji map geomorfometrycznych na obszarach nizinnych — propozycja metody. In: Rekonstrukcja dynamiki procesów geomorfologicznych — formy rzeźby i osady, Warszawa, OficynaWydawnicza Łosgraf, pp 239–244 (in polish) Jedlicka K (2009) Geomorphologic Information Systems. In: 24th International Cartographic Conference, 15-21 November 2009, Santiago, Chile [on-line] http://icaci.org/documents/ ICC_proceedings/ICC2009/html/nonref/11_6.pdf, Accesed 25 January 2010

An Influence of Spatial Range of Input Data set on Terrain Relief

369

Jordan G (2003) Morphometric analysis and tectonic interpretation of digital terrain data: a case study. Earth Surface Processes and Landforms 28: 807–822 Wieczorek M, Spallek W (2008) Identifying of relief forms with the use of statistical analyses. In: Żyszkowska W and Spallek W (eds) Main Problems of Contemporary Cartography 2008: Spatial analyses in cartography (in Polish). Wrocław: Uniwersytet Wrocławski, pp 47–62 Wieczorek M (2009) Werenskiold Glacier (SW Spitsbergen) — Morphometric Characteristics, In: 24th International Cartographic Conference, 15-21 November 2009, Santiago, Chile [online] http://icaci.org/documents/ICC_proceedings/ICC2009\html\ nonref\11_23.pdf, Accesed 25 NOvember 2009 Wood JD (1996) The geomorphological characterisation of digital elevation models. PhD Thesis, University of Leicester, UK, http://www.soi.city.ac.uk/ jwo/phd, Accesed 4 April 2008

Accuracy Assessment of ASTER GDEM in North Shaanxi Xin Yang, Guoan Tang, Wei Zhang, Shijie Zhu Key laboratory of Virtual Geographical Environment Nanjing Normal University Nanjing, Jiangsu Province, P.R.China 210046 [email protected]

Abstract For the preferred time-present, high accuracy and free availability, the ASTER GDEM becomes one of the most important global topographic data. Studies on accuracy assessment of ASTER GDEM are of great significance to its applications. However, the accuracy in specific area, like North Shaanxi province which is the main part of Loess Plateau of China, is not clarified ever before. In this paper, two main methods, namely root mean square error (RMSE) and contour matching difference (CMD), are used in accuracy assessment of ASTER GDEM. A RMSE map of ARSTER GDEM elevation in the North Shaanxi province is generated by comparing with the national DEMs with 1:10000 and 1:50000 map scales. The elevation error is higher in loess covered wind sand dune and Weihe river terrace area and is relatively lower in loess gully-hill area. However, the indicator of accumulate contour matching difference (ACMD) shows an inverse trend that most rugged area has a good contour matching result, compared with the bad contour matching result in the gentle relief area. In general the experiment indicates that the accuracy of Aster GDEM is similar to the national DEM with 1:50000 map scale, but greater with respect to topographic details.

1- Introduction Advanced Space-borne Thermal Emission and Reflection Radiometer Global Digital Elevation model (ASTER GDEM) is released on June 29, 2009. It covers almost 99% of Earth land surface, ranging from 83°N to A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 371 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_25, © Springer-Verlag Berlin Heidelberg 2011

372

X. Yang et al.

83°S, and considered to be the most comprehensive DEM data series regarding its resolution (30 meters). ASTER GDEM is generated by National Aeronautics and Space Administration (NASA，United States) and Ministry of Economy, Trade and Industry (METI，Japan) through adopting full automatic methods to establish Aster stereo-pair images, a procedure of cloud removing and elimination of residual outliers. ASTER GDEM draws more and more attentions recently for its preferred timepresent, high accuracy and free availability. It will be of great potential value to the researches and applications which are now challenged by the data accuracy of DEM. However, many research focus on DEM generation and accuracy from Aster stereo image data (Hirano et al. 2003, Kamp et al 2003, E et al 2009), and very few studies involve on the accuracy assessment and application feasibility of Aster GDEM. Although the initial verification by NASA and METI expressed that the elevation accuracy of Aster GDEM is 20 meters, it is merely a relatively general accuracy with respect to a global scale (http://www.ersdac.or.jp/). What is the accuracy of Aster GDEM in China and whether it changes across different landform? What can be concluded if it is compared with the national DEMs with the maps scale of 1:10000 and 1:50000? In this sense, the accuracy assessment of Aster GDEM is an urgent work. It could be helpful to understand the characters of Aster GDEM and to guide our application.

2- Test Areas and Data

2.1 Test Area In this paper, we choose 46 test areas which are distributed in the North Shaanxi province of China. North Shaanxi is the main part of the Loess Plateau in China, and typical landforms of Loess Plateau. From north to south, landforms in turn can be named as wind sand gully, dome-shaped hills, strip hills, loess Yuan (loess plain), broken loess Yuan and Weihe river terraces. The 8 test sites among all 46 sites are the typical landform plots which selected for contour matching.

Accuracy Assessment of ASTER GDEM in North Shaanxi

373

Figure1: Location of test areas in North Shaanxi province

2.2 Data and its preprocessing

2.2.1 Test data Aster GDEM data with test area are download from the website. We take 5 meter resolution DEM for validation. The DEMs are generated by the State Bureau of Surveying and Mapping (SBSM) through interpolating contours of 1:10 000 scaled topographic map. The accuracy of 1:10 000 DEM shows as table 1. When original maps or images are rather poor the accuracy of DEM can be lower to third level. For the high accuracy we assumed the 1: 10 000 DEM as verified data. In addition, the DEMs of 1:50 000 map scale, with 25 meters resolution, which also generated by SBSM, is matched as another reference dataset for comparison.

374

X. Yang et al.

All these DEMS are merged or clipped to rectangular shape with an area of about 100KM2 in each test area. Landform

Contour Interval in Topographic Map

Slope

Grid Interval（m）

RMSE of Grid Points（m） First Level

Second Level

Third Level

Flat land

1

25

5

5.0

6.7

10.0

Table 1: Accuracy of 1:10 000 DEM

2.2.2 Data preprocessing The DEMs from SBSM are produced with the projection of Gauss-Krüger and the coordinate system of Xi’an 1980. The ASTER GDEM, on the other hand, adopts coordinate system of WGS84. Hence, it needs efforts to ensure a correctly matching of all the DEMs. By using a seven-parameter method, ASTER GDEMs are projected to the Gauss-Krüger and coordinate system of Xi’an 1980. In addition, owing to the 3 degrees zoning of 1:10 000 scale of DEM, it is also needed to project them to the GaussKrüger with 6 degree zoning.

3- Methodology In the study of DEM accuracy, the commonly error indicators includes: root mean square error (RMSE), approximation error (Hu et al. 2007), terrain representation errors (Tang et al. 2000, Tang et al. 2001) and contour matching difference (Tang et al. 2007, Zhu et al. 2008) which is widely used in accuracy assessment of DEMs from different acquisition and generation methods. Hu et al (2003) assessed the accuracy of 1:10 000 and 1:50 000 map scale DEM by using high precision triangular points and field surveying points. The researchers in German compared the GTOPO30 data with the DTM from field survey, and presented the difference by use of mean error, RMSE and the maximum error (Denker 2004). In the aspect of SRTM DEM, most studies evaluated accuracy by comparing SRTM and GPS control points (Brown et al. 2005). As far as ASTER

Accuracy Assessment of ASTER GDEM in North Shaanxi

375

data is concerted, in addition to the preliminary studies from METI and NASA, the researches mainly focused on the DEM generation and accuracy from ASTER stereo-pair images (Hirano et al. 2003, Kamp et al. 2003). By the analysis of precious works, RMSE and contour matching difference indicators are used in this paper to evaluate the accuracy of Aster GDEM.

3.1 RMSE Assume the elevation of validation point is Zk, the elevation of the corresponding point in DEM is zk, then the RMSE is: n

RMSE =

∑(Z i =1

k

− zk )

2

n

(1)

In accuracy assessment of 1:10000 and 1:50000 map scale DEM, SBSM usually use 28 checking points, distributed in the middle and marginal parts of the map, to calculate the RMSE. In fact, 28 checking points are not enough to get a relatively objective RMSE result. Therefore, in this paper all ASTER GDEM grid cells are joined in the RMSE calculation to get a reliable result. By converting DEM grid cells to feature points which are regarded as checking points, the RMSE in each test area can be calculated.

3.2 Contour matching difference RMSE can only describe the dispersion between the checking point value and true value in a general way. It can not reflect the error on every single point. The method of comparing original contours and contours extracted from DEM is one of the most effective methods to estimate DEM accuracy. It has advantages of both excluding local gross errors and assessing an overall quality of DEM effectively (Hu et al. 2007).

376

X. Yang et al.

3.2.1 Definition Contour matching difference (CMD) is a quantities indicator describing the difference between original contour and regenerated contour. Through overlap between original contour and regenerated contour, some fine polygons along the original contour occur. The ratio of the total area with these fine polygons and the enveloping area of neighbouring two half-interval contours is called accumulative contour matching difference (ACMD). Figure 2 shows the definition of ACMD. For a single contour, ACMD can be expressed as: ACM D =

∑S

(2)

i

M

Where, Si is total area of fine polygons. M denotes area closed by neighboring two half-interval contours. For total test area, the calculation of ACMDt can be given as: AC M Dt =

1 m

∑

ACM D 2

(3)

Where, m is the number of contours. On one hand, as a statistics indicator, ACMD denotes the matching degree between original contour and regenerated contour. On the other hand, through introducing area ratio ACMD can reflect the relative difference of elevation error, which can present DEM quality effectively (Hu et al. 2007).

Figure 2: Conception of contour matching difference (From Tang et al. 2007)

Accuracy Assessment of ASTER GDEM in North Shaanxi

377

3.2.2 Contour consistency The method of ACMD need to meet the condition that the number of regenerated contours is well matched the number of original contours. However, owing to the difference of data acquisition and data precision, regenerated contours are hardly consistent with original contours. Zhu et al. 2008 proposed arithmetic to make them correspondingly (Zhu et al. 2008). Assuming a DEM has n number of original contours and m number of regenerated contours, there exist four type relationships between them: one to one, one to many, many to one, and many to many. Among these relationships the one to one relationship is ideal for handling contour consistency, only by comparing the number between original and regenerated contours. For the other three relationships it is obviously not enough to judge contour consistency merely with elevation. Contour consistency is a pair of contours whose elevation, location and shape are most similar to each other. Contour similarity can be measured by the following conditions: ① whether the elevation of contours are equal; ②whether enclosing rectangular of contours are intersect; ③the size of overlap area between the enclosing rectangular. The relationship of these three conditions is showed in Figure 3. These conditions, all together, form a collection of similarity measures from coarse level to fine level, which have an ability to narrow the judgments scope and help to determine the spatial consistency between two contours finally. The basic idea of the above arithmetic is as follows: ①Chose a contour from original contours as the basic contour, and select a contour with the same elevation from regenerated contours as candidate contour. ②Calculate enclosing rectangular of above two contours, then check the intersection condition of these two enclosing rectangular. If there is only one candidate contour and its enclosing rectangular intersect with basic contours’, then these two contours can be regarded as a pair of contours. Otherwise, to calculate the overlap areas of those enclosing rectangular in turn, and select the contour which have maximum overlap area as corresponding contour to basic contour.

378

X. Yang et al.

Same elevation Intersection of enclosing rectangle Overlap areas of enclosing rectangle

Figure 3: Relationship of contour similarity measurement

4- Result

4.1 RMSE of Aster GDEM in North Shaanxi province Based on the RMSE values on the 46 test areas, the RMSE map all over North Shaanxi is generated with IDW interpolation method. From the Figure 4 we can see that the RMSE has a trend of increase-decrease from north to south, and reaches the maximum value in Ganquan area (loess Liang-mao hills). In the north sand valley area, because of the low and contiguous dunes, the relative relief is gently, and the average RMSE is smaller, 11.9 meters. While in the middle of loess plateau, the densely deep gully and large relative relief, the average RMSE is 18.9 meters and maximum of 24 meters. Weihe river terraces and loess Yuan lies in the south part of loess plateau get the average of RMSE of 14.2 meters.

Accuracy Assessment of ASTER GDEM in North Shaanxi

379

Figure 4: RMSE of Aster GDEM elevation of North Shaanxi province

4.2 ACMD in typical test area A smaller ACMD value indicates a better matching of two contours. Tang et al (2007) pointed out that contour matching is very well when the ACMD is smaller than 0.05, denoting a high quality of DEM. While when ACMD is greater than 0.3, contour offset is lager, which means a low quality of DEM (Tang et al. 2007). Considering the discrepancy of spatial resolution and generalization method between Aster GDEM and SBSM DEM, it is not suitable here if the ACMD value is smaller than 0.05. The experiment shows that when ACMD is smaller than 0.1, contour matching is very well, and contour matching is rather poor when ACMD is greater than 0.3.

380

X. Yang et al.

Figure 5: Contour matching in typical landform areas

Table 2 shows that ACMD values of all four test areas are smaller than 0.1, showing a well matching with the contour of 1:10 000 scale, which means a high representation of Aster DEM to real terrain in these areas. ACMD values of Fugu and Yichuan area are a little greater, but still less than 0.11. In fact contour matching is satisfied in most part of Fugu area, exception for broad valley in west area with a relatively large mismatching. In Yichuan area contour matching between ASTER GDEM and 1:10 000 scale DEM is better than contour matching between 1:50 000 and 1:10 000 scale DEM. Besides, Aster GDEM can describe terrain variation in details. Mismatched contour exist in wide valley again in this area.

Accuracy Assessment of ASTER GDEM in North Shaanxi

381

Xianyang area locates in Weihe river terrace. There are too much differences of contour matching to calculate ACMD in the area of Xianyang. The Aster GDEM can describe more detailed information of topographic variety in gentle relief area, but it seems to have more data noise there. ACMD

Fugu

Yan’an Zhidan

Ganquan Huanglong Changwu Yichuan Xianyang

ASTER GDEM 0.108

0.077

0.070

0.094

0.078

0.155

0.104

-

1:50 000 DEM 0.058

0.072

0.049

0.104

0.041

0.139

0.087

-

Table 2: ACMD of ASTER GDEM and 1:50 000 scale of DEM in typical landform areas.

5- Conclusion Accuracy of ASTER GDEM is different across the North Shaanxi area. It is higher in loess covered wind sand dune and Weihe river terrace area and is relative lower in loess gully-hill area. However, the indicator of ACMD shows an inverse trend that most rugged area has a good contour matching result, compared with the bad contour matching result in the gentle relief area. Although Aster GDEM has a larger height error of single point in loess gully-hill area, it can present the relief characters effectively. Contour matching difference shows Aster GDEM have higher reliability in description of terrain structure. Contrarily, in spite of higher accuracy of single point in gentle relief area, its ability to express the terrain structure is inferior, owing to more noise of Aster GDEM there. Therefore, before using Aster GDEMs, we should take both elevation accuracy and description ability of terrain structure into consideration. In general, the accuracy of Aster GDEM is similar to the DEM of 1:50000 scale from State Bureau of Surveying and Mapping, and greater in terms of its ability to represent the topographic details. Further studies should be conducted in aspects of topographic parameters and feature extracted from Aster GDEM, for studying its quality and applicability.

382

X. Yang et al.

Acknowledgments This research is supported by National Natural Science Foundation of China (40901185, 40930531) and Excellent Thesis Cultivation Fund of Nanjing Normal University (2010ss0013, 2010bs0002).

References Denker H (2004) Evaluation of SRTM3 and GOTOP30 Terrain Data in Germany. Proceeding of GGSM 2004,IAG Porto,Portugal E D, SHEN Q, XU Y, et al(2009) High-accuracy topographical information extraction based on fusion of ASTER stereo-data and ICESat/GLAS data in Antarctica. Science in China Series D: Earth Sciences, 52(5): 714-722 Hirano A, Welch R, Lang H (2003) Mapping from ASTER stereo image data: DEM validation and accuracy assessment. ISPRS Photogrammetry & Remote Sense, 57: 356-370 http://www.ersdac.or.jp/GDEM/E/2.html HU P, WU Y, HU H (2003) Research of fundamental theory of assessing the accuracy of DEM. Geo-information Science 5(3):64-69 (In Chinese) HU P, YANG C, WU Y, et al (2007) New DEM theories, methods, standards and application. Sino Map Press, Beijing (In Chinese) Kamp U, Bolch T,Olsenholler J (2003) DEM generation from ASTER satellite data for geomorphometric analysis of Cerro Sillajhuay, Chile /Bolivia∥Proceedings of A SPRS 2003 Conference Anchorage, Alaska TANG G, ZHAO M, CAO D (2000) An investigation of spatial structure of DEM errors. Journal of Northwest University (Nature Science Edition) 30(4):349-352 (In Chinese) TANG G, GONG J, CHEN Z (2001) A simulation on the accuracy of DEM terrain representation. Acta Geodatica et Cartographic Sinica 30(4):361-365 (In Chinese) TANG G, TAO Y, WANG C (2007) Contour matching difference and its application in DEM quality assessment. Bulletin of Surveying and Mapping (7):65-67 (In Chinese) ZHU C, WANG Zhiwe, LIU H (2008) Accuracy evaluation model of DEM based on reconstructed contours. Geomatics and Information Science of Wuhan University 33(2):153-156 (In Chinese) Brown Jr C G, Sarabandi K, Leland EP (2005) Validation of the Shuttle Radar Topography Mission Height Data. IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 43(8):1707-1715

DEM based Terrain Factor of Soil Erosion at regional Scale and Soil Erosion Mapping Fayuan Li, Guoan Tang Key Laboratory of Virtual Geographic Environment, Ministry of Education Nanjing Normal University, Nanjing, 210046, P.R.China [email protected]

Abstract Taking 5 m grid resolution DEMs of 48 test areas in the Loess Plateau in northern Shaanxi as test data, this paper introduces a new terrain factor of soil erosion at regional scale, i.e. G. Results show that spatial distribution of G is related to the spatial distribution of the soil erosion intensity and G can be taken as a terrain factor of regional soil erosion in the Loess Plateau. Comparison analysis between G and LS shows G is more capable of playing the role as a regional terrain factor of soil erosion than that of LS, at least in the Loess Plateau. But G is not exact enough in evaluation processing of regional soil erosion, an integrated analysis combining the climate, vegetation cover, soil and water conservation management is demanded. For lack of measurement data at corresponding test area, quantitative analysis of correlation between G and soil erosion modulus is not reality. But based on a regional soil erosion model, a qualitative soil erosion intensity map is designed based on an overlaying calculation of G, vegetation cover, rainfall erosivity and soil antierodibility.

1- Background and objective Terrain is commonly accepted as one of the key factors leading to soil erosion. More works have been done on investigating the relations between erosion and its impacting factors i.e. vegetation cover, terrain, climate, soil, land use at the lever of slope and small watershed (Groundwater 2000; Woodward 1999; Sidorchuk 1999; Favis-Mortlock et al. 1998; Jong et al. 1999; Jiang et al. 1996; Yin et al. 1989). A series of quantitative models were set up on the basis of experiments as well as theoretical deduction, A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 383 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_26, © Springer-Verlag Berlin Heidelberg 2011

384

F. Li and G. Tang

such as Xinanjiang model, Jiangzhongshan model, Yangzisheng model, Liubaoyuan model and so on (Liu et al. 2001). Meanwhile, some international-advanced erosion models i.e. USLE, WEPP, LISEM, ANSWERS, AGNPS, were applied in soil erosion researches in the Loess Plateau. However, these models are available mostly only at surface slope lever or watershed lever. It could not be easily adapted to regional erosion studies. Hence, soil erosion model at regional scale is insistently demanded for the research of regional soil erosion process, evaluation and prediction, environment effect of soil and water conservation in the Loess Plateau. Since terrain factor is one of the most important parameters in regional soil erosion model, a rather comprehensive and overall terrain factor should be put forward, on which the contribution of terrain complexity and roughness in surface erosion could be investigated at a macroscopic level. DEM data availability and different state-of-the-art analysis technologies have extended the usage of DEM in many ways (Richard 2000; Zhou and Liu 2006). Some researches have applied DEM in their study of regional soil erosion or soil mapping (Jong et al. 1999; Pickup et al. 2001; Renschler et al. 2002; Dobos et al. 2001). But a rather comprehensive and overall terrain factor fitting for regional soil erosion in the Loess Plateau have not been put forward. Objective of this paper is to explore a regional terrain factors and its spatial distribution based on digital elevation models.

2- Data and methods

2.1 Study area description The Loess Plateau is the most severe soil and water loss area in the world because of the special geographic landscape, soil and climatic conditions, and long history (over 5000 years) of human activity. On-site soil and water loss has seriously depleted land resources and degraded the ecoenvironment in the Loess Plateau, while off-site sedimentation poses problems to waterways and reservoirs. For the Yellow River, 25 % of the sediment load deposits along the river bed in the lower reach which causes an annual rise in the river of 8 to 10 cm (Shi et al. 2000). The Yellow River is an ‘above ground river’ that is seriously threatening the security of its lower reach. Although management practices in the Loess Plateau have

DEM based Terrain Factor of Soil Erosion at regional Scale and Soil Erosion Mapping

385

helped decrease soil and water loss for several decades (Xu et al. 1994; Chang et al. 1996), severe geographical, soil and environmental conditions in the Loess Plateau still remain. Therefore, the control of soil and water loss and the improvement of eco-environments remain critical issues in China. Then, erosion modeling might be a useful tool to understand and predict erosion and to ultimately find ways to prevent it. Hence, the Loess Plateau in north Shaanxi province is chosen as the study area. Erosion of this area is the most intensive in the Loess Plateau, and here is the national key area to prevent soil and water loss. It covers an area of 92.5 thousand km2. The main landforms are Yuan (plain), Liang (ridge), Mao (hill) and various gullies, which construct a consecutive relief variation from south to north. Ground cracking is well-developed and gully density is high. Considering sample density and its representativeness, 48 test areas with different terrain complexity and roughness were selected (Figure 1).

China

2 448 m

323 m ●

test areas river

100

Figure 1: The distribution of the test areas

50

0

100 km

386

F. Li and G. Tang

2.2 Data Test data were the corresponding 5 m-grid resolution DEMs produced according the national standard of China. The 1:100,000 soil erosion intensity map of Shaanxi were used for a comparative analysis.

2.3 Definition of regional terrain factor (G) As we know, anything on the ground has the potential energy moving downslope, which is resulted by the component of gravity paralleled with ＝ • the downslope. It can be expressed as F●g●sinα (where g denotes gravity, F denotes its parallel component, α denotes local slope gradient). Under the same conditions (soil property, vegetation covering etc.), the bigger the sinα is, the stronger the potential energy of downslope moving will be. So at regional lever, the summation of F on every sample site could represent the driving force of surface erosion coming from the terrain itself. In terms of this principle, we define the driving force of soil erosion coming from the terrain as the terrain dynamic force factor (Td, Li and Tang 2006). Correlation analysis shows the Td is well positively correlated with terrain relief amplitude. So, Td can be thought as containing information of terrain dynamic force and terrain relief amplitude. Td based on grid DEM can be calculated with equation (1).

Td =

n

m

i =1

j =1

∑ ∑ sin α

i, j

(i × j )

(1)

Where i, j is the row and column of the grid DEM, αi,j is slope gradient at location i and j. Surface gradient are computed from the DEM with steepest descent algorithm. Moreover, erosion depends on slope as well as on upslope catchment area. Stream power index combines these two variables and it is used to model soil erosion. As specific catchment area and slope steepness increase, the amount of water contributed by upslope areas and the velocity of water flow increase, hence stream power index and erosion risk increase. Stream power index controls potential erosive power of overland flows, thickness of soil horizons, organic matter, pH, silt and sand content, plant cover distribution. Hence, we use Td’ to integrate the two factors (equation 2).

DEM based Terrain Factor of Soil Erosion at regional Scale and Soil Erosion Mapping

T d' =

n

m

i =1

j =1

∑ ∑ ln( CA

i, j

× sin α i , j )

387

(2)

(i × j )

Where CAi,j denote specific catchment area at location i and j. As to the terrain complexity, different kinds of parameter have been proposed to describe it, such as profile curvature, total curvature, selfγ

Max _ D ( direction

)

Max _ D

resemblance, fractal parameter ( ), SCI, spatial dihedral angle etc (Wang 2004; reference 9, 10). But there is no common understanding of it yet. This paper uses the slope variability (SOS) and aspect variability (SOA) to depict terrain complexity. SOS and SOA depict the variability of curvature along vertical and horizontal direction respectively. Tang (2000) has proved that the slope variability is positively correlated to terrain complex. Hence, using of SOS and SOA to depict terrain complexity is reliable. SOS can be derived by means of a calculation of ’slope of slope’ with the Arc-View spatial analysis tool and SOA can be derived by means of a calculation of ’slope of aspect’ with the same tool. Following an overall consideration of the three parameters, we propose the terrain factor of soil erosion at regional scale can be represented as:

G =

n

m

∑ ∑ ln( CA i =1

j =1

i, j

× sin α i , j × SOS

i, j

× SOA i , j )

(i × j )

(3)

Where CAi,j denote specific catchment area at location (i, j), SOSi,j denote slope variability at location (i, j), SOAi,j denote aspect variability at location (i, j).

2.4 Stable threshold area of G According to the definition of G, the G will vary following the sample area. Because of strong self-resemblance of the loess landform, the G will stable when sampling area is bigger than the threshold area just like the slope spectrum (Figure 2). Theoretic deduction shows that G has the same stable threshold area with the slope spectrum in a test area. Hence we can take the critical area of slope spectrum as the stable threshold area of G. The threshold area is tightly corrected with landform type, sample position and constrain condition for stable slope spectrum. In a certain area, a

388

F. Li and G. Tang

minimum threshold area can be calculated when sample area is able to represent whole geomorphic feature.

Sampling area

Slope spectrum

Figure 2: Slope spectrum getting stable while the sampling area is over critical area. If the slope spectrum is existed lie on if there is a critical area in which the slope spectrum can exist stably. Because the subsampled slope spectrum will present different shapes even in same study area when the sampling area less than the threshold area (Tang 2003).

2.5 Spatial prediction of G Spatial distribution of G is predicted using the ArcGIS geostatistical analysis tools. Predicting process is depicted as follows: 1) exploring the data distribution, if the data is not normally distribution, you should choose to transform the data to make it normal; 2) identifying global trends, the trends will be removed when modeling; 3) calculating distance and semivariance between all data and determining rational lag size; 4) establishing best-fit semivariance /Covariance modeling; 5) calculate the semivariance of each pair of points and that between each observation and prediction value, then build a matrix of these semivariance values; 6) calculate the weights of measured points around the prediction point; 7) multiply the weight for each measured value times the value, and add the products together to calculate the prediction value; 8) using crossvalidation to evaluate prediction accuracy, evaluation standard is following: Mean Error (ME) and Mean Standardized Error (MSE) should be close to 0, Average Standardized Error (ASE) should be close to Root Mean Square Error (RMSE) and Root Mean Square Standardized Error (RMSSE) close to 1.

DEM based Terrain Factor of Soil Erosion at regional Scale and Soil Erosion Mapping

389

3- Results and discussions Based on the DEMs data, G, Td, Td’, SOS and SOA of 48 test areas are calculated. Then the terrain factor G and its spatial distribution is predicted using the ArcGIS geostatistical analysis tools (Fig. 4). Prediction errors parameters are listed as following, ME = 0.0040; RMSE = 1.1260; ASE = 0.9493; MSE = -0.0012; RMSSE = 1.147. Figure 3 shows spatial distribution of G is obvious. The G varies from 4.741-9.760. Because of restriction of technology and measurement data, we can not get detail soil erosion modulus of corresponding test area. Hence, quantitative analysis of correlativity between G and soil erosion modulus is not reality. But comparison of regional distributions between G and soil erosion intensity is attainable. Fig.4 is the soil erosion intensity map of north Shaanxi produced by institute of soil and water conservation of C.A.S. According to value of soil erosion modulus, this map divides soil erosion intensity into six classes, i.e. faint intensity, low intensity, middle intensity, strong intensity, stronger intensity, the strongest intensity.

Legend

4.741~6.550 6.550~7.708 7.708~8.450 8.450~8.925 8.925~9.229 9.229~9.424 9.424~9.549 9.549~9.630 9.630~9.680 ●

test areas river

0

50

100 km

Figure 3: Map of spatial prediction of G in north Shaanxi province

Figure 3 and Figure 4 show that spatial distribution of G is related to spatial distribution of the soil erosion intensity. Along N-S trend, G value in the region of Kuyehe-river drainage area, Tuweihe-river drainage area, Suide County and Yanchuan County show the maximum of the North Shaanxi. Previous research have given the conclusion that this area is the

390

F. Li and G. Tang

center of intense soil erosion in the Loess Plateau with the soil erosion modulus over 20000 ton / (km2· ●a), and is the major source area of sediment load in the lower reach of the Yellow river. Along the W-E trend, from the east foot of the Baiyushan mountain, Zhidan County, Ansai County to Zichang County, it is another center of the soil erosion in the Loess Plateau, where the soil erosion modulus changed from 10000 to 15000 ton / (km2·a). But there are some inconsistencies between the two maps. For example, G of Huanglong area which lies to the east of the Yijun County is big, but erosion intensity here is faint. This is due to good vegetation cover condition and perfect soil and water conservation management. Another abnormal area is the north of Shaanxi, where soil erosion is the strongest yet because of the severe natural environment, such as long storm, endless desert, infrequent vegetation, and unconsolidated loess. But G of this area is small because of low relief. Although taking G as a regional terrain factor is appropriate.

Soil erosion intensity (t / k 2 )

Faint (●1000) Low (1000-2500) Middle 5000)

(2500-

Strong (5000-8000)

100

50

0

100

Figure 4: Soil erosion intensity map of northern Shaanxi (from Institute of soil and water conservation, CAS)

When we apply G to evaluate the regional soil erosion in the Loess Plateau, an integrated analysis combining the climate, vegetation cover, and soil and water conservation management is needed. According to the evaluation model provided by Li et al. (2008), we can set up a simple model: A ● Ra·Cb·Gc·Vd

（4）

Where A denotes evaluation coefficient of soil erosion intensity, R denotes rainfall erosivity, C denotes soil antierodibility, V denotes vegetation cover,

DEM based Terrain Factor of Soil Erosion at regional Scale and Soil Erosion Mapping

391

a, b, c, d denote weight coefficient. To avoid the impact of dimension, all factors are standardized with the way of maximum difference normalization method. Primary data of R were calculated by Wang (1996) according to 43 weather stations and hydrological stations which are mainly located in Shaanxi province. Primary data of C were on-the-spot observed by Jiang (1997). V is 1 km resolution NDVI index provide by CNES, this paper use the average of month 6, 7, 8, 9, 1998. G, R, C are interpolated as dada layers, and the converted into grid layers with 1 km resolution (Fig. 5-a, 5-b, 5-c). Then we can get A through data overlaying calculation (Figure 6).

a. rainfall erosivit

b. soil antierodibility

river

100

0

100 km

c. vegetation cover

Figure 5: Spatial distribution of rainfall erosivity, soil antierodibility and vegetation cover in northern Shaanxi province

392

F. Li and G. Tang

100

0

100 km

river

Soil erosion intensity low

high

● Figure 6: Soil erosion intensity map of northern Shaanxi●

Figure 6 shows that soil erosion intensity along Kuyehe-river drainage area, Tuweihe-river drainage area, Mizhi County and Suide County is the strongest, and then from north to south the soil erosion intensity changes gradually. With this way, we can get a simple model to evaluate soil erosion. However there still exit some errors, this is because the unsuitable sample density between the four factors. Considering the correlation between G and soil erosion modulus can not be quantitatively analyzed, comparison of G and other terrain factors used to evaluate regional soil erosion is realizable. The Revised Universal Soil Loss Equation (RUSLE) has long been used in the Loess Plateau to evaluate regional soil erosion (Liu 2001). LS factor is taken as the terrain factor in the model and many research works have been done on the validation of LS factor for soil erosion in the Loess Plateau (Jiang et al, 2005; Liu et al, 2000; Mou et al, 1983). Here we used the formula (4) proposed by Jiang (2005) to calculate the LS factor of 48test area.

DEM based Terrain Factor of Soil Erosion at regional Scale and Soil Erosion Mapping

LS = 1 . 07 × ( λ 22 ) 0 . 28 ×

393

(α 5 .16 )1 .45

(5)

Where λ denotes upslope length, α denote slope gradient. Regression analysis between G and LS shows that G is positive correlation with LS (Tab.1, Fig.7). The spatial distribution of LS factor show same pattern with that of G. But the spatial distribution of G is more similar with soil erosion intensity distribution than that of LS. So we are well-founded that G is more capable of playing role as a terrain factor for regional soil erosion than that of LS. df

● Regression analysis

SS

MS

1

72.5203

72.5203

Residual error

46

19.2210

0.4178

Sum

47

91.7414

F 173.5566

Significance F 3.20E-17 ●

Table 1: Result of variance analysis between G and LS in north Province Shaanxi

20 y = 2. 7201x - 12. 813

15

2

R = 0. 7905

10 5 0 4

5

6

7

8

9

10

Figure 7: Scatter map of G vs. LS in north Shaanxi Province

4- Conclusions DEM data availability and GIS-assisted processing of DEM data make it possible to find a new terrain factor for regional soil erosion based on DEMs. The G is proposed here. The spatial distribution of G is related to the spatial distribution of the soil erosion intensity. Generally speaking, the bigger the G, the stronger the soil erosion. But this is not exact the case,

394

F. Li and G. Tang

because of influence of soil property, climate, soil and water conservation management, and vegetation cover and so on. Hence, integrated analysis combining these factors is demanded when applying G to evaluate the regional soil erosion in the Loess Plateau. For lack of measurement data, we can not get detail soil erosion modulus of corresponding test area. Hence, quantitative analysis of correlation between G and soil erosion modulus is not reality. But comparison of G and other terrain factors used to evaluate regional soil erosion is realizable. Comparison analysis between G and LS shows G is more capable of playing role as a terrain factor for regional soil erosion that of LS, at least in the Loess Plateau. However, many detail work should be done to make sure this conclusion.

Acknowledgments Thanks for financially support from the National Natural Science Foundation of China (No. 40930531, No.40801148).

References Chang MD, Zhao GY and Tian AM (1996) Rural economic development and comprehensive control of small watershed on the loess Plateau. RES of SOIL WAT, 3(4): 88-92 (in Chinese) Chris SR and Jon H (2002) Soil erosion assessment tools from point to regional scales-the role of geomorphologists in land management research and implementation. GEOMORPHOLO, 47: 189-209 Cui Y, Li R and Yang QK (2006) Preliminary research on the patterns of regional soil erosion in view of GIS. J of Xi’an university of ART & S (natural science edition), 9(2): 5-8 (in Chinese) De JSM, Paracchini ML, Bertolo F et al (1999) Regional assessment of soil erosion using the distributed model SEMMED and remotely sensed data. CATENA, 37(3-4): 291-308 Endre D, Luca M, Thierry N et al (2001) A regional scale soil mapping approach using intergrated AVHRR and DEM data. JAG, 3(1): 30-42 Favis-Mortlock D, Guerra T and Boardman J (1998) A self-organizing systems approach to hillslop rill initiation and growth: model development and validation. IAHS publication, 249: 5361 Geoef P and Alan M (2001) Regional-scale sedimentation process models from airborne gamma ray remote sensing and digital elevation data. EARTH SURF, 26: 273–293 Groundwater P (2000) The influence of model resolution on rill development: A numerical modeling study. HYDROL PROC, 14(6): 2173-2205 http://www.geog.ntu.edu.tw/main/paper/R89228006/abstract.htm. Accessed 15 July 2010 http://www.itc.nl/personal/shrestha/DTA/terrain_par.htm. Accessed 11 June 2010

DEM based Terrain Factor of Soil Erosion at regional Scale and Soil Erosion Mapping

395

Jiang DS (1997) Pattern of soil erosion and management of the loess plateau. China WaterPower Press, Beijing (in Chinese) Jiang ZS, Wang ZQ and Liu Z (1996) Quantitative study on spatial variation of soil erosion in a small watershed in the Loess Hilly Region. J of soil erosion and soil conservation, 2(1): 1-9 Jiang ZS, Zheng FL, Wu M (2005) Prediction model of water erosion on hillslopes. J of Sediment Research, (4): 1-6 Li FY and Tang GA (2006) DEM based research on the terrain driving force of soil erosion in the Loess Plateau. in Geoinformatics 2006: Geospatial Information Science, edited by Jianya Gong, Jingxiong Zhang, Proc. of SPIE Vol. 6420, 64201W Li R, Yang WZ, Li BC (2008) Research and future prospects for the loess plateau of China. Science Press, Beijing (in Chinese) Liu BY, Nearing MA Shi PJ et al (2000) Slope length effects on soil loss for steep slopes. Soil Soc Am J, 64 (5): 1759-1763 Liu BY, Xie Y and Zhang KL (2001) Soil erosion prediction model. Chinese Science and Technology publishing house, Beijing (in Chinese) Mou JZ and Meng QM (1983) Preliminary research on prediction of rainfall soil loss. China SOIL WAT, (6): 23 - 27 (in Chinese) Richard JP (2000) Geomorphometry – diversity in quantitative surface analysis. PROG P GEO, 24(1):1-30 Shi H and Shao MG (2000) Soil and water loss from the Loess Plateau in China. J ARID ENV, 45: 9-20. Sidorchuk A (1999) Dynamic and static models of gully erosion. CATENA, 37(3-4): 401-414 Tang GA (2003) Slope Spectrum of the Loess Plateau and Its Spatial Distribution. Proceedings of 2003 annual meeting of the geographical society of China, Wuhan (in Chinese) Tang GA (2000) A research on the accuracy of digital elevation models. Science Press, Beijing Wang L, Tang GA, Liu XJ et al (2004) Terrain Complexity Index and Derivation of based on DEMs. Bulletin of SOIL WAT, 24(4): 55-58 (in Chinese) Wang WZ and Jiao JY (1996) Rainfall erosion of the loess plateau and sand transportation in yellow river. Science Press, Beijing (in Chinese) Woodward DE (1999) Method to predict cropland ephemeral gully erosion. CATENA, 37(2): 393-399 Xu TC, Gao RL and Wang DX (1994) Development and experimences of soil and water conservation on the Loess Plateau since the eighties. Bulletin of SOIL WAT, 14(4): 12-18 (in Chinese) Yin GK and Chen QL (1989) Characteristic index and statistical model of sediment yield in small drainage basins of Loess Plateau in China. Acta Geographica Sinica, 44(1): 32-45 (in Chinese) Zhou Q and Liu XJ (2006) Digital terrain analysis. Science Press, Beijing (in Chinese)

Analysis and Simulation for Application Fields : Urban Growth, Traffic, Epidemiology and Language Locations

Urban Growth Modeling with Road Network Expansion and Land Use Development Yikang Rui, Yifang Ban Geoinformatics, Department of Urban Planning & Environment, Royal Institute of Technology - KTH, Stockholm, Sweden [email protected]

Abstract Land use and transportation systems are considered as two most important subsystems determining urban form and structure, and are assumed to mutually influence each other overtime. To better understand the relationship between them, we build a simple dynamic model to simulate longterm urban growth instead of a static one. Our urban simulation combines vector road network growth with grid land use dynamics. Vector model has advantages in topological analysis and we use space syntax metrics to control road network growth by estimating road traffic flow, which is also adopted to calculated accessibility for land use simulation. The land use model includes five land use categories and two behavior sub-models: mobility model and location choice model. Our preliminary simulation results show similar land use patterns of how real cities grow.

1- Introduction Urban is a complex system, which consists of various interactive subsystems and is influenced by a variety of variables, such as governmental land policies, population growth, transportation infrastructure, market behavior, etc. Modeling and simulation is regarded as an important tool to study urban issues with numerous applications including mapping and visualization, urban planning, emergency response, entertainment and etc. (Vanegas et al. 2009 a). Among nine types of major urban subsystems (Wegener 2004), perhaps land use changes and transportation system are the two most important A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 399 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_27, © Springer-Verlag Berlin Heidelberg 2011

400

Y. Rui and Y. Ban

ones determining urban form and structure in the long term. Transportation networks and development of land use are assumed to mutually influence each other over time. Construction of a new road link or enlargement of an existing network would influence firm/household location choice, real estate development, land development and building density. At the same time land use changes in turn influence the demand for travel or the accessibility (Kelly 1994, Iacono et al. 2008). Three main traditional methods to study the reciprocal relationship between land use and transportation are surveys, empirical studies and mathematical models (Wegener, 2004). In recent years some microsimulation models are introduced, such as activity-based travel model, cellautomata model, and multi-agent model, using “bottom up” modeling frameworks. Many models have been developed, for exmaples, Ramblas (Veldhuisen et al. 2000), ILUMASS (Moeckel et al. 2003, Wagner and Wegener 2007), UrbanSim (Waddell et al. 2003, 2007), ILUTE (Miller et al. 2004), PUMA (Ettema et al. 2005). They use transportation networks to calculate travel times and costs, traffic assignment, congestion levels and etc. when simulating residents’ activity and travel patterns, firms’ location, flow of goods and services in the market, etc. While for some other models, such as ABLOoM (Otter et al. 2001), CityDev (Semboloni 2005), and Kou’s artificial urban (Kou et al. 2008), transportation is simply used as a road network to calculate the accessibility for different agents. When we intend to simulate long-term urban growth, one obvious disadvantage in these models is that transportation networks are assumed to be static. Currently a variety of methods have been developed to generate road networks. In geometric urban simulation, generating an underlying road network is critical to urban reconstruction. Parish and Muller (2001) used a procedural approach based on L-systems and road network templates to create a street map. Similar works were modified by Kelly and McCabe (2007) and Weber et al. (2009). Besides, Vanegas et al. (2009 b) generated a set of seeds, which are converted to intersections of the arterial road network, and then constructs arterial road segments. Sun et al. (2002) used a template-based model to generate a virtual traffic network. Esch et al. (2007) and Chen et al. (2008) adopted user-guided tensor fields as the input to generate street. Yamins (2002) used FindMax procedure to get a pair of points which there is the greatest need for new transport structure, and then builds a road between the points with least-cost function. Jiang (2007) used a transition potential map to select road nodes. Route is generated with minimum elevation difference or highest transition potential. A very different approach to road generation is the use of an agent based model.

Urban Growth Modeling with Road Network Expansion and Land Use Development

401

Lechner et al. (2004, 2006) and Watson (2007) applied extender and connector agents for the generation of tertiary and primary road networks. Most of land use models use static transportation networks, while most road generation methods use existing population density maps or current land use maps as input data. Because we intend to build a long term urban growth model, our idea is to combine two dynamics together to observe and study the relationship between land use change and transportation. Our urban growth model mainly consists of two parts: transportation network growth and land use changes. The major contributions of our works are: 1. We generate a dynamic urban model, instead of creating a static one. 2. We propose an urban simulation combining vector road network and grid land use. Vector model has advantages in topological analysis, especially for road network, while grid model is easy to visualize land cover changes and to hatch agents which could make individual choices and build behavioral sub-models. 3. Space syntax metrics are introduced for road network generation by selecting important roads according to predicted traffic flow, although these metrics have been widely used in network analysis and generalization. The paper is organized as follows. Section 2 introduces road growth method. Section 3 describes traffic flow computation, which is used to select road segments and calculate accessibility. Section 4 is the land use simulation. The implications of the experiments are discussed in section 5 before the paper draws to a conclusion in section 6.

2- Road growth Our Road network mainly consists of two types of roads: major roads and minor roads. Minor roads provide access to the major road network and service the local areas. Roads can be regarded as multiple agents at microscale, in which every road interacts with each other to form a connected network. We simply use two types of agents to generate road network. One is node; the other is road segment, which links two nodes. Road segments form individual roads according rules, which will be discussed later. Both major and minor road networks have the same data structure. Each node

402

Y. Rui and Y. Ban

has an attribute to distinguish major node from minor node. And road segments have attributes to tell which individual roads they belong to. Our road growth strategy is based on several previous works (e.g., Parish and Muller 2001, Kelly and McCabe 2007), with two differences. 1. We don’t use static population density map as input data, which they use and thus road networks in their paper are not generated in the same order as real city. 2. We add a traffic model to control road growth by selecting necessary road segments. Major roads are generated by starting from user defined city center or existing nodes. The heading h1 for a start node A is created firstly. Angle of deviation θ1 is decided by road patterns. The node A shoots a number of rays radically with a preset move step length L. Along each ray, elevation data is extracted. The direction with minimum elevation difference is chosen for continuing the growth (for example: segment AB). Calculate the angle between proposed segment (blue dash line in Figure 1, left) and existing segments (black solid lines) started from or ended to the node A. Choose the minimal angle θ2 and make sure it meets the threshold. Parameters could be used to control road pattern. Road patterns mainly consist of strokes, grids (raster), stars (radial and optional rings) and irregulars. Strokes are basic elements in grids and stars patterns (Marshall 2005, Heinzle et al. 2007). We implemented three different road patterns (i.e. grid, radial, and organic). Major road netwok applies radial and organic patterns, while minor road network applies a grid-like pattern, which simply means angle θ2 closes to 90 and no more than 4 segments are allowed to own the same node. Other parameters include move length L and density of intersection nodes. Similar to local constraints of Parish’s paper (2001) and legality tests of Weber’ paper (2009), proposed road segments should meet the following constraints: 1) Make sure proposed segment ends inside a legal area. If not, rotate the segment up to a maximal angle. Some areas, such as river, could be cross up to a specified length. 2) Check the relationship between proposed segment and existing nodes and segments. If this proposed segment intersects with an existing segment, generate a new intersection node E (Fig.1 Middle). A new segment AE or AC (only when segment EC is too short) is chosen. If the end node of the proposed segment closes to an existing node or segment, extend it to the existing node or segment (Figure1

Urban Growth Modeling with Road Network Expansion and Land Use Development

403

Right). If angles between segments are too small or segment BC is too short, we choose segment AC instead of segments AB + BC.

Figure 1: Left: Road growth from node A. Middle: Intersection test. Right: Snapping test

After constraints are executed for proposed road segments, we need to compute traffic flow for new segments, especially for major roads which apply organic pattern. The new road segments with high traffic flow would be confirmed as major roads.

3- Traffic computation Traffic model is built to select major roads and prove travel demands for land use simulation. Weber et al. (2009) pointed out that existing traffic demand models have some disadvantages. These models are too complex, need many external variables and data, and work with static cities. In order to compute traffic flow over time to show the evolution of major road network, they presented a simplified stochastic model which uses sampling to distribute a discrete number of trips in the road network. In this paper we propose a new method by using space syntax metrics to select important roads according to predicted traffic flow. In space syntax community, integration is a key measure, and can be measured at both local and global levels. For a node A in a space syntax graph, firstly we calculate the shortest distance fro others to node A. Here the distance is the length of the graph geodesic between two nodes and the maximum shortest distance is denoted by md. Then we count the number of nodes Nd with the distance d. Finally we get the following expression,

∑

m d =1

d × N d . If m =

1, we get the connectivity (degree) of node A. if m = md, it represents global integration, and if 1 < m < md, it is a local integration expression.

404

Y. Rui and Y. Ban

Local integration is the default measure of space syntax for traffic flow prediction. Usually we adopt two steps for the calculation, i.e. m is equal to 2. According to Jiang et al. (2007, 2008), point-based metrics assigned by summing into individual roads tend to have a much better correlation with traffic flow than line-based metric. Thus, we use node’s local integration to estimate the traffic flows. Although this assumption might not hold for precise predictions, they allow for displaying the general spatial distribution of traffic flows in a road network. First of all, we join some road segments into individual roads following a self-organized process based on three different join principles: every-bestfit, self-best-fit and self-fit. Jiang et al.(2008) pointed out self-best-fit principle is the best option for metric-flow correlation, and we use self-best-fit principle which means each segment only considers itself to find a best fit of its neighboring segments. Then we derive characteristic points from existing nodes to draw visibility graph, which shows how each characteristic point is visible to others (Jiang and Claramunt 2002). Angle deviation of two adjacent road segments and maximum distance a node could see should be considered. Thus connectivity values of characteristic points are derived. After computing each node’s local integration, we sum them into individual roads to describe the road traffic flow (Figure 2, b). We choose roads with high local integrations as major roads (Figure 2, c). A threshold is defined as the percentage of the maximum local integration. Confirmed major roads need to be modified to connect each other well as a network. We create a new link if one road end is near to another road (orange line in Figure 2, d). If an end part of one individual road, which is divided by an intersection node, is too short, we erase it or degrade it to a minor road (magenta dash line in Fig. 2, d). In the area serviced or surround by major roads, minor roads are created with a grid-like pattern (Figure 2, e). We also calculate the local integration of the whole road network (Figure 2, f), and find that proposed major roads generally maintain the framework of the whole road network according to the traffic flow, and some minor roads near city center also show heavy traffic flow. This phenomenon is quite obvious and common in a real road network.

Urban Growth Modeling with Road Network Expansion and Land Use Development

405

Figure 2: (a) is initial high density major road network. Background is water area and elevation map. (b) shows local integration of each individual road, dark red means high value while light is low one. (c) is major roads after selection. Modification is adopted in (d). (e) displays the whole road network. Light brown roads are minor roads. (f) shows local integration of the whole road network

4- Land use simulation Land use type determines demography distribution and influence travel demand, besides different land use types have different road patterns.

406

Y. Rui and Y. Ban

There are various types of land use that could be found in a city. We currently define five land development categories: residential (high-density and low-density), commercial (community-oriented and highwayoriented), industrial (heavy and light), transportation (road development) and parks (parks and forests), since they make up the majority of land use in most cities. Residential, commercial and industrial areas are together result in roughly 60% of urban land use (8: 1: 2). Percentage of area for transportation is 25% - 30%. And it is well know that the percentage of each category changes a bit at different stages of urban development.

4.1 Mobility model Our mobility algorithm moves a subset of land developer agents to new S = αD

t at time t, where α is a places in a period. Size of the subset is t small fraction and Dt is a set of developed land area, and it could be residential, commercial or industrial land use category. St is chosen at random, S ∪L → L

t +1 . It is rather placed in a waiting list Lt and removed from Dt. t t similar to employment mobility model and household mobility model of UrbanSim (Waddell et al. 2003, 2007), but we mainly focus on land use types. As the urban growth, there are more people and new lands are developed as urban area. We simply add these new demands as developer agents into the waiting list Lt.

4.2 location choice model For each developer agent a ∈ S , we choose a small random subset T of vacant places to be located. T = V–D, where V is the set of areas served by roads. The probability of location l ( i , j ) ∈ T to be selected is equal to the probability that the utility value at this location is greater than any other candidate locations (Li and Liu 2008). P ( a , l ) = exp(U ( a , l )) ∗ Con ( l = suitable ) / ∑ exp(U ( a , l ')) , l , l ' ∈ T . Con( ) is a constraint function, for example, some locations are forbidden to be transferred. Utility calculation depends on the various attributes. Transportation area could be easily derived from road network. Road width depends on road type and its traffic flow. Big parks often appear in rough terrain or flood

Urban Growth Modeling with Road Network Expansion and Land Use Development

407

plains. As far as the location of residential, commercial, and industrial development, many attributes would influence the decision. 1) Distance to center. Commercial developers prefer to city center, while industrial developers are opposite. 2) Terrain, including proximity to water and slope. Commercial and some industrial developers like to near water. Commercial developers prefer flat land. 3) Accessibility. Factories need convenient transportation so they are located near major road. Commercial area always has a higher accessibility than residential area. 4) Clustering. For example, residential developers intend to choose land close to existing residential area, because it reduces investment and development risk. 5) Neighborhood. Industrial development should far away from residential area, while commercial development should near residential area and market. For residential agents, U r = β Dw x Dw + β A x A + β Nr x Nr + β Ni (1 − x Ni )

For commercial agents, U c = β Dc x Dc + β Dw x Dw + β A x A + β Nc x Nc + β Nr x Nr

For industrial agents, U i = βT xT + β Dw x Dw + β Apr x Apr + β Ni x Ni + β Nr (1 − x Nr ) + β Nc (1 − x Nc )

where β represents relative importance weight for each attribute. xDw and x

= exp( − β d )

xDc represent proximity to water and city center. D , d is Euclidian distance. xT represents the terrain slope, i.e. flatness. xN represents the clustering and neighborhood influence of each land use type, which is simply calculated as the percentage of developed grids which have the same land use type in the neighborhood. where a could be r, c or i. xA represents the accessibility. x

x Na = ∑ N ( landuse = a ) /( ∑ N − 1)

x A = γ Apr x Apr + γ Asr x Asr = exp( −γ d )

,

. xAsr is the accessi-

bility to minor road network, Asr , and xApr is the accessibility to major road network. The novelty here is that we not only calculate the nearest road, but other roads in a threshold distance. Besides we use the estimated traffic flow as a weight parameter. It is clear that a place has a better accessibility if it could easily access to more important roads.

408

Y. Rui and Y. Ban

x Apr = ∑ (1 − exp( −γ ' LI )) ∗ exp( −γ '' d )

, LI is the value of local integration of the road being calculated. Fig. 3 shows the dynamics of accessibility maps.

5- Experiments and discussion In our experiments, the background data, i.e. basic land use and elevation data, is from the Greater Toronto Area. As described above, accessibility is not only related to road network, but also the traffic flow. Considering road network growth and traffic flow changes in each road, we have different accessibility maps for major roads in four periods. See Figure 3.

Figure 3: Major roads (white line) growth and accessibility (lighter means higher accessibility value) changes in four periods

Figure 4 shows one example of the long term development of a growing city. Initial parameters include water and elevation. Higher elevation is represented by increased luminance. Simulation starts with a given center. Residential areas are denoted by yellow, commercial are denoted by red, and industrial are represented by blue. Roads are transformed from vector data into grid data. Major roads are distinguished from minor roads in that they are wider and a deeper shade of gray.

Urban Growth Modeling with Road Network Expansion and Land Use Development

409

One challenge for urban growth modeling is the validation of the modeling result. Similar to other social or economic fields, the reality of urban development is hard to be reproduced, mainly because our observation usually comes from statistical surveys and entities in social systems have cognitive capabilities in contrast with natural science (Semboloni 2005). Utilized parameters in many large scale social models are calibrated through a lot of statistical data and not from direct observation. However in our behavior model, we use agents to simulate different types of entities, which interact with other entities. We could formalize the entities the same we observe in the reality. And parameters from observation, which describe entities’ choices, have a meaning not only to the model but also to the reality.

Figure 4: Simulation results of a growing city in three periods

Actually we do not intend to reproduce any existing city, but the typical urban patterns. We have some methods to judge our simulation results. One is to visually compare it with land use map of a real city. The other is to use measure metrics to describe land cover, such as spatial configuration, which refers to the distribution and clustering of land use (for example Houston, TX in Fig.16 in Lechner et al. 2006). Our simulation results show similar patterns of land use as that of real cities. Commercial areas

410

Y. Rui and Y. Ban

are drawn to city center and areas of high road density and close to the roads. Industry attempts to develop in urban periphery. With city growth, some are left, while some move to new city fringe. Residential areas tend to develop with the accessibility to road network and segregate themselves from industry. Major road network shows a skeleton of the growing city and demonstrate connectivity among the denser parts of the city. Note that some roads may be major roads in an early period, but are minor roads in the next period, depending on the traffic flow. It reveals road hierarchy changes well in a real world city.

6- Conclusion and future work We have presented a dynamic model to study long-term urban growth based on land use changes and transportation growth. Therefore our model mainly consists of two sub-models: vector road growth model and grid land use model. Our simulation spans a long time, so building a dynamic road network is necessary. We use space syntax to analysis road network, and local integration is calculated to predict traffic flow, which are used to select and confirm major roads, and calculate accessibility for land use simulation. Our accessibility maps are dynamic because both road network and traffic flow in the same individual roads change over time. Simulation results show some similar patterns of a real city, although our preliminary model is quite rough. Our model has several limitations and needs to be improved in the future. (1) Behavioral model used here is rather simple. We intend to add more elements, such as location choice model for household and employment. (2) A dynamic population distribution and density map should be calculated, which is very important for road growth.

References Chen G, Esch G, Wonka P, Müller P, Zhang E (2008) Interactive procedural street modeling. ACM Trans. Graph., 27(3) Esch G, Wonka P, Müller P, Zhang E (2007) Interactive procedural street modeling. In SIGGRAPH '07: ACM SIGGRAPH 2007 sketches, New York, NY, USA. ACM

Urban Growth Modeling with Road Network Expansion and Land Use Development

411

Ettema D, de Jong K, Timmermans H, Bakema A (2005) PUMA multi agent modelling of urban systems. In: 45th Congress of the European Regional Science Association, Vrije Universiteit Amsterdam Glass KR, Morkel C, Bangay SD (2006) Duplicating road patterns in south african informal settlements using procedural techniques. In Afrigaph '06: Proceedings of the 4th international conference on Computer graphics, virtual reality, visualization and interaction in Africa, pages 161-169, New York, NY, USA. ACM Press Heinzle F, Anders K, Sester M (2007) Automatic Detection of Patterns in Road Networks: Methods and Evaluation. In: Proc. of Joint Workshop Visualization and Exploration of Geospatial Data, Stuttgart, vol. XXXVI -4/W45 Iacono M, Levinson D, El-Geneidy A (2008) Models of transportation and land use change: a guide to the territory. Journal of Planning Literature, 22(4), pp. 323-340 Jiang B, Claramunt C (2002) Integration of space syntax into GIS: new perspectives for urban morphology, Transactions in GIS, 6(3), 295-309 Jiang B (2007) A topological pattern of urban street networks: universality and peculiarity, Physica A: Statistical Mechanics and its Applications, 384, 647-655 Jiang B, Zhao S, Yin J (2008) Self-organized natural roads for predicting traffic flow: a sensitivity study, Journal of Statistical Mechanics: Theory and Experiment July, P07008 Jiang Z (2007) The Road Extension Model in the Land Change Modeler for Ecological Sustainability of IDRISI. ACMGIS'07, Seattle, WA Kelly ED (1994) The transportation land-use link. Journal of Planning Literature 9(2): 1, 28-45 Kelly G, McCabe H (2007) Citygen: An Interactive System for Procedural City Generation. In Proceedings of GDTW 2007: The Fifth Annual International Conference in Computer Game Design and Technology, pages 8-16, Liverpool, UK Kou X, Yang L, Cai L (2008) Artificial Urban Planning: Application of MAS in Urban Planning Education. 2008 International Symposium on Computational Intelligence and Design, 349353 Lechner T, Watson B, Ren P, Wilensky U, Tisue S, Felsen M (2004) Procedural modeling of land use in cities. Technical report, Northwestern University Lechner T, Ren P, Watson B, Brozefski C, Wilenski U (2006) Procedural modeling of urban land use. In SIGGRAPH '06: ACM SIGGRAPH 2006 Research posters, New York, NY, USA. ACM Lechner T, Moeckel R, Schwarze B, Spiekermann K, Wegener M (2007) Simulating Interactions Between Land Use, Transport and Environment. 11th WCTR, 24-28 June 2007, UC Berkeley Li X, Liu X (2008) Embedding sustainable development strategies in agent-based models for use as a planning tool. International Journal of Geographical Information Science, Vol. 22, No. 1, January 2008, 21-45 Marshall S (2005) Streets & Patterns. Spon Press, Taylor & Francis Group, New York Miller JE, Hunt DJ, Abraham JE, Salvini PA (2004) Microsimulating urban systems. Computers, Environment and Urban Systems 28:9-44 Moeckel R, Spiekermann K, Schurmann C, Wegener M (2003) Microsimulation of land use, transport and environment. Paper presented at the 8th International Conference on Computers in Urban Planning and Urban Management, May 27-29, Sendai, Japan Parish Y, Müller P (2001) Procedural Modeling of Cities. In Proceedings of ACM SIGGRAPH 2001, ACM Press, E. Fiume, Ed., 301-308 Otter HS, van der Veen A, Vriend HJ (2001) ABLOoM: Location behavior, spatial patterns, and agent based modelling. Journal of Artificial Societiesand Social Simulation, 4 (4), published online Semboloni F (2005) Multi-agents simulation of urban dynamic. Proceedings of XII CUPUM. Ed. London Sun J, Baciu G, Yu X, Green M (2002) Template-Based Generation of Road Networks for Virtual City Modeling. VRST ’02, Hong Kong

412

Y. Rui and Y. Ban

Vanegas CA, Aliaga DG, Wonka P, Muller P, Waddell P, Watson B (2009 a) Modeling the appearance and Behavior of Urban Spaces. COMPUTER GRAPHICS forum. Volume 28, Number 2, pp.1-18 Vanegas CA, Aliaga DG, Benes B, Waddell PA (2009 b) Interactive design of urban spaces using geometrical and behavioral modeling. ACM SIGGRAPH Asia 2009. SESSION: Urban modeling. Article No.: 111 Veldhuisen J, Timmermans H, Kapoen L (2000) RAMBLAS: a regional planning model based on the microsimulation of daily activity travel patterns. Environment and Planning A 32 (3), 427-443 Waddell P, Borning A, Noth M, Freier N, Becke M, Ulfarsson G (2003) Microsimulation of Urban Development and Location Choices: Design and Implementation of UrbanSim, Networks and Spatial Economics, vol. 3, no. 1, pp. 43-67 Waddell P, Ulfarsson G, Franklin J, Lobb J (2007) Incorporating Land Use in Metropolitan Transportation Planning. Transp. Res. Part A: Policy and Practice, 41, 382/410 Wagner P, Wegener M (2007) Urban Land Use, Transport and Environment Models. DISP, 2007, 170, 3, 45-56 Watson B (2007) Real and virtual urban design. In SIGGRAPH '07: ACM SIG-GRAPH 2007 courses, pages 167-228, New York, NY, USA. ACM Weber B, Müller P, Wonka P, Gross M (2009) Interactive Geometric Simulation of 4D Cities. EUROGRAPHICS 2009. Volume 28, Number 2 Wegener M (2004) Overview of Land-use Transport Models. Chapter 9 in David A. Hensher and Kenneth Button (Eds.): Transport Geography and Spatial Systems. Handbook 5 of the Handbook in Transport. Pergamon/Elsevier Science, Kidlington, UK, 2004, 127-146 Xie F, Levinson D (2009) Modeling the Growth of Transportation Networks: A Comprehensive Review. Netw Spat Econ (2009) 9:291-307 Yamins D, Rasmussen S, Fogel D (2003) Growing Urban Roads. Networks and Spatial Economics, 3: 69-85

Conception of a GIS-Platform to simulate urban densification based on the analysis of topographic data. Anne Ruas1; Julien Perret1; Florence Curie1; Annabelle Mas2; Anne Puissant3; Gregorz Skupinski3, Dominique Badariotti3; Christiane Weber3; Pierre Gancarski4; Nicolas Lachiche4; Julien Lesbegueries4; Agnès Braud4 1

Laboratoire COGIT, Institut Géographique National 73 avenue de Paris 94165 Saint Mandé, France [email protected], [email protected], [email protected] 2

Département de Géographie, Université D'Orléans BP 6749 - 45067 Orléans cedex [email protected] 3

LIVE Université de Strasbourg 3 rue de l’Argonne 67000 Strasbourg France [email protected], [email protected], [email protected], [email protected] 4

LSIIT Boulevard Sébastien Brant 67412 Illkirch Cedex France ; [email protected], [email protected], [email protected], [email protected]

Abstract The aim of our research is to analyze the evolution of urbanization and to simulate it on specific areas. We focus on the evolution between 1950 and now. We analyse the densification by means of comparing temporal topographic data bases created from existing topographic data base and maps and photo from 1950. In this paper we present how a simulation works - which input data are used, which functions are used to densify the space and how the simulation works, is tuned and run - the densification method for each urban block illustrated with results, the method used during the project to build the required knowledge for simulation and we conclude and present the main research perspectives. The methods are implemented on a dedicated open source software named GeOpenSim. A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 413 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_28, © Springer-Verlag Berlin Heidelberg 2011

414

A. Ruas et al.

1- Introduction In this paper we present the current state of a research project named GeOpenSim funded by the French research funding agency (ANR) from December 2007 to March 2011. The aim of GeOpenSim is to build an Open Source GIS platform to analyze the evolution of urbanization and to simulate it on specific areas. Compared to other researches on urban growth and simulation (Couclelis 1997; Batty 2007), we do not use cellular automata but vector topographic agents (such as buildings, streets and urban blocks) and we use historical data to build evolution rules. In our research, simulation should be considered as a powerful tool to build and to consolidate our own understanding on urban morphological evolution. It is not a predictive tool. At final state the GeOpenSim platform will be under an Open Source license to allow geographers to set and test their own evolution rules on real topographic data. Practically we wish to see whether and how we can develop models and methods to analyze the landscape at different dates and to derive rules of evolution from these analyses. The platform is meant to be open to different criteria according to the richness of our data and knowledge. For example it should be able to simulate building densification of a urban block with or without planning rules, such as the fact that during a certain time period this urban block can only be devoted to housing. Moreover, each simulation should generate a new state described by the parameters of this simulation. By the way, these states can be compared to facilitate again our visual analysis and understanding. During the project and for practical reasons, our experiments were limited to two different parts of France (around Orleans and around Strasbourg), from 1955 to 2008. Of course experiments can be performed on different regions with different dates provided where vector topographic data are available. We have also limited our experiments on building and street densification and reconstruction because of time constraints (see below). In the paper we only present the building densification process. The research team is composed of researchers from COGIT laboratory and university of Orleans, specialized on the analysis of vector topographic data and agent modelling, researchers from the geographical department of

Conception of a GIS-Platform to simulate urban densification

415

the University of Strasbourg specialized in urbanization process and researchers from the computer science department of the University of Strasbourg, specialized in machine learning techniques. The present paper is structured in the following way: • In section 2 we present how a simulation works : which input data are used, which functions are used to densify the space and how the simulation works, is tuned and run, • In section 3 we present the densification method for each urban block illustrated with results, • In section 4 we present the method used during the project to build the required knowledge for simulation, • In section 5 we conclude and present the main research perspectives.

2- How does simulation work?

2.1 The global process The main process is what we call ‘densification or reconstruction’. We name densification the action of adding buildings (or streets) on an area that may or may not already contain buildings (or streets). We name reconstruction, the action of removing the existing buildings, redefining urban blocks boundaries and then proceeding to a densification on the new free space. The simulation in itself follows a predefined sequence that consists in choosing the area where the simulation will take place, the initial and final dates of the simulation period, the rule bases used to perform the simulation, the kind of building and street patterns that will be used during the simulation. Then, the system uses its rule bases and some spatial analysis to define which areas will change, with what kind of patterns and up to which level of densification. Last, the simulation runs and the result

416

A. Ruas et al.

is stored in the system with the parameters of the simulation (see figure 1). Of course the process requires that information already exist in the platform: the initial vector database at a specific date, some rules of evolution and patterns to densify the space. 1

C hoos e the area

2

S elect initial and final date for s imulation

3

C hoos e your evolution rules (default or s pecified)

What kind of evolution ? What rhythm ? Including urban cons traints ?

4

C hoos e the kind of patterns (default or s pecified)

Which patterns are favored ?

5

C ompute the characteris tic of the final s tate

6

S imulate the dens ification

7

Where?

F rom when to when ?

S imulate the dens ification according to urban block attractivenes s

Which urban blocks to dens ify Up to which dens ity level ? With what kind of pattern ?

S tore the res ult

Figure 1: Sequences to run a simulation

Thus a simulation follows 4 macro-steps: the spatio-temporal specifications (steps 1 and 2 in figure 1); the selection of the parameters of the simulation (steps 3 and 4 in figure 1); the simulation itself (steps 5 and 6 in figure 1) and the storage of the results for comparison purposes (step 7 in figure 1). Two strategies of simulation are presented in the following subsection 2.2. The quality of the simulation depends on three factors: • our capacity to analyze the entire urban area on which we make simulation in order to detect the best candidates (i.e. the best urban blocks) to densify or to reconstruct (figure 2),

Conception of a GIS-Platform to simulate urban densification

417

• the quality of the methods of densification that add streets and buildings not only onto free spaces but also on urban blocks that already have urban properties (see section 3), • our capacity to build relevant rules of evolution (see section 4). In a defined urban area, composed of a set of urban blocks, from this date to this date, which blocks are good candidate to change? In case of densification, how does it work ? - with same kind of buildings ? - with another kind of buildings ? Which one ? - up to which density ? In case of reconstruction - where might be the new urban block limit ? - what kind of buildings ? .. mixed or uniform types ? - up to which density ?

Figure 2: Where and how to densify an urban area?

2.2 The simulation based on agent modelling We have chosen to use the agent paradigm to control the simulation process. As for generalisation purposes, an agent is a geographical object that acts to reach its goals. It can be either meso - urban blocks – or micro – buildings (Ruas 1999). We test two different strategies to densify an urban area: • The first strategy is composed of two steps: We use a rule base to identify which urban blocks are good candidates to change and up to which density (step 5 in figure 1), then we use an agent based paradigm in order to let each urban block densify itself according to its characteristics and to its goals (step 6 in figure 1)

418

A. Ruas et al.

• The second strategy (steps 5 and 6, right side in figure 1) is under conception. It tries to follow an entire agent paradigm: each urban block has an attractiveness index to define itself as a good or bad candidate for densification. If it is a good candidate, it then follows the same process than the other strategy. For both strategies an urban block agent uses its own method 1/ to characterize itself, 2/ to compare its properties to its goals properties 3/ to select an appropriate method to densify or to reconstruct itself according to its properties and its goals 4/ to densify or reconstruct itself and 5/ to trigger building agents for spatial readjustment purposes. A building agent, when it is activated, modifies its location (distance and orientation) towards its neighbouring buildings and street. The result is different for each block for three reasons: 1/ each urban block has different goals (box 5 figure 1) 2/ the densification method integrates the properties of each urban block 3/ the densification method integrates some random decisions to ensure a certain diversity. It appears that the difficulty is neither to build an agent based approach nor to build a method of densification (see section 3) but to identify relevant knowledge to distribute evolution goals over a large set of urban blocks. This is due to the fact that urban growth depends on many factors including some for which we do not have any information. We thus have to propose methods that mix urban properties and unknown factors transformed into probability. This aspect is developed in section 4.

3- The method to densify urban blocks As previously stated, the densification methods are essential. The principle is 1/to choose a kind of building and 2/ to locate a set of them in a more or less regular way (building distribution). Of course as town planner, the location of the new buildings integrates the shape of the urban block and the location of the already existing buildings. Basically two strategies exist and are proposed. The first one consists in extending the existing building pattern up to the density goal. To do so, the system analyses the properties of the existing buildings and reproduces them. The second strategy

Conception of a GIS-Platform to simulate urban densification

419

consists in adding a specified new pattern in the urban block. To do so we developed a library of patterns. Each pattern describes the properties of the buildings and their spatial distribution.

3.1 Building a relevant pattern library and method for building location Our goal at this step is to build a generic pattern schema with relevant attributes that can describe a large diversity of real patterns, a schema that is used to automatically add buildings on an existing space. The method developed to locate the new buildings into an existing urban block is described in (Curie et al 2010). We first identified a draft data schema that contains attributes that describe building shape, size and distribution and we tried to describe a set of well known building patterns with this schema. We made simulations of different patterns, we showed the results to geographers specialized in town planning and we improved gradually our patterns according to their comments. The data schema today used to describe patterns is presented in figure 3. We defined a classical library of building shapes (square, rectangle, U, L, T, stairs, etc.) following different research such as Mackaness and Rainsford (2002), and in order to obtain better visual results we set that a building pattern may be composed of two different kinds of shapes with different frequencies (figure 3). For the spatial distribution described by the distance to the road and distance between buildings the values can be fixed, or can follow a Gaussian law or can be randomly chosen between the minimum and maximum value. During the project we proposed three types of patterns: individual houses, collective buildings and industrial buildings. More precisely we set 6 kinds of patterns because of their frequencies in the French landscape between 1950 to today. These patterns have been defined by means of a visual analysis of existing maps and databases and improved by means of geographers’ expert knowledge and knowledge from literature (Lacoste 1963; Panerai et al 2001; Gauthiez 2003; Allain 2004; Arnaud 2008). The process of conception is iterative: patterns were proposed and used to simulate urban densification or reconstruction; results were shown to our

420

A. Ruas et al.

experts and comments were used to improve the patterns. The 6 patterns conceived are the following: 1. Two patterns for individual houses: • Individual and not planned houses (called spontaneous) often complete an existing house pattern. Houses are mostly built along streets or road network • Individual planned houses are built during a limited time frame according to a house planning program which may include new access streets. Houses look the same, their locations are organised in a regular way along the streets. The point is to maximise the number of houses. 2. Three patterns for blocks of flats (collective buildings) : • The Large-BF were very frequently built in France during the 50ies , these buildings are rectangles, from 300 to 900m², with high elongation value (between 0.1 and 03). • The VeryLarge-BF corresponds to important housing programs in the 60ies in France during an important housing crisis. These buildings are composed of an important quantity of lodgements (> 1000). A set of buildings are built together, clustered in the centre of an urban block. Each building is either a rectangle with important elongation and with an area up to 800 or even 1000m², or square, smaller (between 300 to 400m²) and very tall. • The small-BF appears in France in the 80ies and is still used today in housing programs. Buildings are smaller (between 300 to 1000m²) not so high (often between 3 or 4 floors) and shapes are more original, including stairs. 3. One industrial pattern composed of large buildings, with compact shapes. The diversity of shapes is very important and as well as the rhythm of densification: some densifications are slow and regular while others are very quick.

Conception of a GIS-Platform to simulate urban densification

421

Figure 3: the data schema of building pattern

The method to add buildings in an area is described in (Curie et al 2010). To sum up for individual houses, houses are added one by one along the street network. For Large or Very large blocks of flats, the process seeks free space according to building size and shape. The shape, size and position of each house or building is computed from the parameters of each pattern (figure 3) including a range of values and a method (Gaussian law or Random) to compute the values (for example the size) inside this range.

3.2 Experimental results and comments The densification has been tested on real datasets to evaluate and improve the quality of the pattern and of the object location. During the project we created historical databases at different dates on different areas. The method to create these historical databases has been presented in (Perret et al 2009). To sum up, we used recent topographic databases, we copy cut them in the past, and we used scanned maps or orthophotographies to

422

A. Ruas et al.

‘down-date’ the data. As the main process is urban growth, we mainly removed houses, buildings and streets and created few of them. This historical dataset is essential to build evolution rules and to test our simulation. In figure 4 we simulate a house densification within an urban block from 1976 to 1989 and we compare it to the reality. This example illustrates the fact that we still have to be able to add dead-ends within urban blocks.

Figure 4: Experimental results for the densification with planned individual houses

Figure 5 illustrates a densification of a heterogeneous urban block with non planned individual houses. In reality the houses built in 1989 close to the isolated house in 1956 took the same orientation than this house whereas in our simulation, the new houses follow the orientation of the street.

Reality 1956

Reality 1989

Simulation from 1956

Figure 5: Experimental results for the densification of an urban block with spontaneous housing

Of course the simulation is never meant to look like the reality but each difference is full of meaning either on the reality or on the difficulty of producing relevant results.

Conception of a GIS-Platform to simulate urban densification

423

4- Building relevant knowledge from historical data From the beginning of the project the aim was to build relevant knowledge related to urban growth from existing data but also to propose relevant methods for building knowledge on other areas, on other time frames. We know that urban evolution depends on a large set of factors including time period, duration, existing patterns, urban plans, land readjustments and many other elements such as population growth. The first assumption of our study is that if there is a process of reconstruction and densification within an area, even if we do not know all information, the densification is never random and that some areas are more likely to change than others. For example if the density of an urban block is very high, changes are rather unlikely, but some buildings might be reconstructed. On the other hand, the best candidates for densification are empty or nearly empty areas, close to dense or semi-dense areas. Discussions with geographers and urban planners enlighten the fact that the densification or reconstruction depends on the type of urban blocks and on the location of the blocks with respect to the city center and the rural area. In order to build rules such as “If an urban block that has such and such properties might change in this way during this time period”; we decided first to learn how to classify automatically urban blocks according to a predefined classification (section 3.1). The second step is then to learn evolution rules such as 1-“this type of urban blocks, with such and such properties, are good candidates for densification” which answers to the question which urban blocks are good candidates for densification? and 2-“this type of urban blocks tends to change in such a way” which answers to the question How to densify an area ? With what kind of pattern?

4.1 Automated urban block classification based on supervised learning techniques Urban classification is often produced by image analysis with visual interpretation and field checking. It is of course a very good but time consuming process. Alternatively image processing can be used to reduce

424

A. Ruas et al.

the production cost (Donnay et al 2001; Blaschke et al. 2004). Supervised learning techniques can be used for that purposes (Bauer et al. 2001). In our case, we have to classify urban blocks at different dates, for which we do not have necessary images and obviously we cannot perform field completion. We have thus chosen to classify urban blocks from morphological properties computed from vector representation only. This method has already been tested for generalisation purposes in (Boffet 2001). The aim of our approach is to develop a classification method useful to study morphological evolution process, which can be easily performed on different areas, for different time frames. Thus we decided to use supervised learning techniques, to generate decision rules for classification and beforehand to provide a module of data labelling necessary to generate learning examples and that will be used to provide new classification rules on demand (Lesbegueries et al 2009). The first step of our research has been to visually analyse urban blocks and to define a classification that would be relevant for our research subject. This classification has to distinguish individual and collective buildings as well as the density since we are looking for the good candidates for reconstruction or densification. Several meetings and many discussions between geographers were necessary to reach a complex compromise as the only real validation could only occur after the evaluation of simulations on large datasets. Our current classification is our ‘working classification’. It will have to be validated in the future. Seven classes have been defined: Class 1- Continuous urban fabric (city center), Class 2- Discontinuous urban fabric with individual houses, Class 3- Discontinuous urban fabric with housing blocks (blocks of flats), Class 4- Mixed urban fabric (including individual housing (type 2) and housing blocks (type 3)), Class 5- Mixed areas (including types 1, 2, 3 and 6),

Conception of a GIS-Platform to simulate urban densification

Class 6- High density of specialised areas (including commercial, hospital or scholar buildings),

425

industrial,

Class 7- Low density of specialised areas (containing few or no buildings). For learning purposes, datasets have been manually classified by geographers with the data labelling module developed during the project. The labelling process was very interesting to define each class and its boundary. Of course some urban blocks are always in between two classes. Because the project focused on suburban areas and the learning process requires a number of samples large enough for each class, the class 1 was not studied. In a parallel way, we proposed a large set of measures on buildings and urban blocks. Classification tests showed that the most relevant measures were the following: • Area, elongation and convexity of the buildings; • Building orientation and relative orientation of a building towards its closest street; • Number of buildings, total area of buildings, density of buildings inside an urban block. The classification algorithm used is TILDE (Blockeel and De Raedt 1998). It is a supervised and symbolic algorithm that produces decision trees. It has a better expressivity than C4.5 (Quinlan 1995). Several tests have been applied on Strasbourg area on historical data over the fifty last years. To assess the classification we performed cross validation and we also classify new and unused samples. The quantitative evaluation is between 75 and 80 %. The results depend on the quantity of samples used per class and the type of class, knowing that some classes such as individual housing have much more samples than very mixed areas (class 5). The decision tree allows detecting rules such as: • If the block density < 0.05 Then the urban block is in class 7

426

A. Ruas et al.

• If density > 0.05 and number of buildings (whose area < 105m²) < 3 and number of buildings (whose area < 185m²) < 1 and average-building area < 1205m² Then the urban block is in class 3. This analysis allows finding relevant measures at the same time as threshold values. The results were also visually analysed to provide a qualitative evaluation. What we considered as the best rules according to quantitative and qualitative evaluation were applied on different areas and again visually evaluated (figure 6).

Figure 6: Urban blocks classification applied to 5 dates

Even imperfect, this classification is very interesting as it allows us to see urban evolution through a new perspective. It also allows studying the transition and stability of urban blocks along time period (section 4.2). Other criteria such as the distance to the city centre or the fact that an urban block looks like its neighbouring blocks are under evaluation to improve the rules.

Conception of a GIS-Platform to simulate urban densification

427

4.2 Learning transitions In order to make the simulations as realistic as possible, we need statistics about, on the one hand, the transitions themselves and on the other hand, about the most frequent kinds of urban block evolutions. With this second information, we can study each kind of evolution in order to define it according to characteristics of urban blocks involved in the evolution. These analyses will help identify the best candidates to change depending on the nature of urban blocks and their previous states. To do both analyses we have developed a new tool, named iVisualize (see Figure 7), that allows: • to calculate the number of occurrences of each type of transitions (e.g., “Low density of specialised areas” to “High density of specialised areas”) or no transition (e.g., “Road” remains “Road”) according to a given length (i.e., number of dates taken into account) and a minimum threshold, • to highlight classes of evolutions using learning algorithms.

Figure 7: Urban blocks transitions and evolutions analysis tool

428

A. Ruas et al.

While the first aspect has not been a particular problem, the second one requires choosing the type of learning algorithms to be used. It is very difficult and time consuming to define all examples needed for supervised algorithms. In fact, we choose to use unsupervised approaches and in particular distance-based algorithms (i.e., the Kmeans algorithm and the hierarchical agglomerative one) to highlight classes of evolutions. To apply these algorithms, in agreement with the geographers, we defined a new semantic distance between classes of urban blocks (from 0 for two objects from a same type to 4 for a “Continuous urban fabric” object and a “Low density of specialized areas” object, for example). Finally, to apply the Kmeans algorithm, we use the similarity measure called dynamic time warping (DTW) proposed by (Sakoe et al. 1971) and (Sakoe et al. 1978) and the global averaging method for dynamic time warping (DBA) introduced in (Petitjean et al. 2010). The first experiments carried out have produced promising results. The clusters of evolutions extracted by the algorithms have been studied by the experts and seem to correspond to thematic evolutions. These experiments have also enabled the experts to validate the semantic distance.

5- Conclusion and future work Developing a GIS platform to study and to simulate urban evolution is very challenging. During this three years long research project, many results have been obtained. We developed: • a data model to represent historical topographic data with explicit link between objects at different times (Perret et al 2009), • a module to create historical data-base from today topographic data base and old maps or orthophotographies (Perret et al 2009), • an agent engine that uses rules and densification methods to simulate urban growth, with the registration of each simulation for comparison purposes (section 2),

Conception of a GIS-Platform to simulate urban densification

429

• several methods of densification that add buildings on free or semi-free areas with different building patterns (section 3), • a building pattern library that can easily be enriched (section 3), • a module to classify urban blocks based on supervised learning techniques associated with a labelling module to classify new blocks with new measures (section 4.1), • a first set of evolution rules that focuses on the transitions of urban classification through time (section 4.2), • a first evaluation module that compares states. If the construction of relevant knowledge is a real difficulty, the results are improving and the association of different techniques for knowledge acquisition is promising. Current research aims at completing and improving the evolution rules by means of expert knowledge and statistics. We also intend to integrate urban planned constraints to study different kinds of urban evolution. A first module will be available in May 2011 on the GeOxygène opensource platform. It should allow to enlarge the community of researchers testing their own urban evolution rules.

References Allain R (2001) Morphologie urbaine - Géographie, aménagement et architecture de la ville. Armand Colin Arnaud J-L (2008) Analyse spatiale, cartographie et histoire urbaine. Parenthèses Batty M (2007) Cities and Complexity: Understanding Cities with Cellular Automata, AgentBased Models, and Fractals. MIT Press Bauer T, Steinnocher K (2001) Per parcel land use classification in urban areas applying a rulebased technique. In: GeoBIT/GIS 6 :24-27. Blaschke T, Burnett C, Pekkarinen A (2004) New contextual approaches using image segmentation for object-based classification. In: De Meer, F., de Jong, S. (Eds.), Remote Sensing Image Analysis: Including the spatial domain. Kluver Academic Publishers, Dordrecht, pp 211-236. Blockeel H, De Raedt L (1998) Top-down induction of first-order logical decision trees. Artificial Intelligence, vol. 101. Boffet A (2001) Méthode de création d'informations multi-niveaux pour la généralisation cartographique de l'urbain, http://recherche.ign.fr/labos/cogit/telechargementCOGIT.php PhD Thesis, University of Marne la Vallée, France

430

A. Ruas et al.

Couclelis H (1997) From cellular automata to urban models: new principles for model development and implementation. Environment and Planning B: Planning and Design 24: 165-174 Curie F, Perret J, Ruas A (2010) Simulation of urban blocks densification. Proceedings of the 13th AGILE International Conference on Geographic Information Science, Guimarães, Portugal Donnay J-P, Barnsley MJ, Longley PA (2001) Remote Sensing and Urban Analysis. Taylor and Francis, London and New York Gauthiez B (2003) Espace urbain : vocabulaire et morphologie. Editions du patrimoine. Lacoste Y (1963) Un problème complexe et débattu : les grands ensembles. Bulletin de l’association des géographes français n° 318-319. Lesbuegueries J, Lachiche N, Braud A, Puissant A, Skupinski G, Perret J (2009) A platform for spatial data labeling in an urban context. International Opensource Geospatial Research Symposium. Mackaness W A, Rainsford D (2002) Template matching in support of generalisation of rural buildings in: Richardson D and van Oosterom P (eds) : Advances in Spatial Data Handling. Berlin: Springer 137-152. Panerai P, Castex J, Depaule J (2001) Formes urbaines : de l’îlot à la barre, Parenthèses. Perret J, Ruas A, Boffet-Mas A (2009) Understanding urban dynamics: the use of vector topographic database and the creation of spatiotemporal databases. International Cartographic Conference - ICC 2009, Santiago, Chile. Petitjean F, Ketterlin A, Gançarski P (2010) A global averaging method for dynamic time warping, with applications to clustering, Pattern Recognition, doi:10.1016/j.patcog.2010.09.013 to appear. Quinlan J R (1995) C4.5 : Programs for machine Learning, Morgan Kauffmann. Ruas A (1999) Modèle de généralisation de données géographiques à base de contraintes et d'autonomie. http://recherche.ign.fr/labos/cogit/telechargementCOGIT.php - PhD Thesis, University of Marne la Vallée, France. Sakoe H, Chiba S (1971) A dynamic programming approach to continuous speech recognition, Proceedings of the Seventh International Congress on Acoustics vol. 3:65–69. Sakoe H, Chiba S (1978) Dynamic programming algorithm optimization for spoken word recognition, IEEE Transactions on Acoustics, Speech and Signal Processing 26(1):43–49.

Realistic Road Modelling for Driving Simulators using GIS Data Guillaume Despine, Caroline Baillard Siradel SAS [email protected], [email protected]

Abstract In this paper an approach is proposed for creating realistic models of existing roads adapted to driving simulation. Unlike most previous work, based on generic construction rules, urbanism patterns or sociological behaviour, our approach aims at reproducing existing road networks. First a data model based on a multi-layered graph is presented. This model can handle the three representation levels required by traffic simulation: the road network, the graphical level and the traffic level. In the second part of the paper, a method for modelling existing roads is proposed. The novelty of the approach is the use of existing road data bases to automatically create a virtual environment (3D model and traffic organisation) close to ground truth. An existing 3D GIS data base provides accurate information about road axes, ground and building 3D geometry, whereas a navigation road database helps refining the model by providing clues on network topology and traffic rules. The resulting virtual road system reproduces the real world and has been successfully interfaced with a driving simulator.

1- Introduction In the game and driving simulator areas, a lot of work has focused on automatic road generation for virtual environments. It is often based on generic growth models and it can integrate construction rules used by real road builders. All these studies offer a high degree of realism but they do not aim at reproducing the real world.

A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 431 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_29, © Springer-Verlag Berlin Heidelberg 2011

432

G. Despine and C. Baillard

On the contrary the collaborative project PlatSim1, driven by the ECA Faros2 Company, aims at giving potential drivers the opportunity to train within a true urban environment. For this purpose, realistic models of existing cities must be created. These models must not only describe the road geometry (width, lane number, junction shapes), but also the correct traffic rules (forbidden manoeuvres, reserved lanes, etc) and the corresponding road markings. The Siradel Company is partner of this project. As a geographical data producer, Siradel own many GIS high-resolution urban databases worldwide. Our role in the project is to provide driving simulators with realistic and compliant city models. This paper describes an automatic modelling approach making use of existing GIS data. This paper is organized as follows. In section 2 some related work is presented. A data structure especially designed for traffic simulation is introduced in section 3, followed by the description of our road modelling process in section 4. Finally some results and possible improvements are presented in section 5.

2- Related work Two kinds of approaches for automatic road modelling can be distinguished: those which create fully generic road networks, and those creating road surfaces resembling the real world.

2.1 Generic road modelling (Parish and Müller 2001) have proposed a model based on L-systems to produce a fully generic road network. Predefined patterns can lead to a concentric road network (like in European cities) or to rectangular shapes (like in US cities). Moreover, an elevation map can be used to constrain street and highway directions. (Bakolas 2007) has improved this model by involving standard city development rules: whereas in the former model the city structure is fully 1

PlatSim is a collaborative project sponsored by the "Images & Réseaux" innovation cluster and the Bretagne Region (www.images-et-reseaux.com) 2 www.ecafaros.com, driving and flight trainer simulators

Realistic Road Modelling for Driving Simulators using GIS Data

433

described by the street network, Bakolas proposes a competitive influence of buildings and streets, based on growth models through time. This model is well adapted for European cities. These approaches are interesting in terms of network modelling but they do not consider roads as 3D geometrical structures or as traffic flows (street width, number of lanes, crossroad description, etc.). For these reasons these models are not appropriate for traffic simulation. (Willemsen et al. 2006) have proposed a road model especially dedicated to traffic simulation. All paths (vehicles, bikes and pedestrians) are represented by a ribbon. This ribbon is a 3D space curve attributed with width and surface normal at each point. At intersections, the curvilinear system is replaced by Cartesian coordinates and lanes are replaced by "corridors". Each corridor knows its neighbouring connected corridors. (Carles 2000) has also presented a complete model for traffic simulation. He has pointed out the need for a multilevel model in which the graphical 3D model is completed with a symbolical description of the environment. This description gives immediate access to information constraining the traffic, avoiding a wasteful interpretation of the graphical environment. His model guarantees consistency between the graphical and the symbolic levels.

2.2 Realistic road modelling A lot of research has been done on the automatic extraction of road models in urban or suburban areas using aerial or satellite images (Mena 2005), (Grote 2007), (Ravanbakhsh 2007), (Tournaire 2009) but the resulting information is not appropriate for driving simulation occurring at ground level. Recently, some studies based on Mobile Mapping Systems have been published (Soheilian 2007), (Shi 2008), (Tarel 2009), but automatic solution are not mature yet. Moreover the acquisition cost is still very high. A third kind of approach consists in using existing GIS data. (Décoret and Sillion 2002) have proposed a method to automatically construct a street graph from building footprints. First a Voronoï diagram is used to determine the graph topology (nodes and arcs). This step is insensitive to the detailed shape of the footprints. The second step is a geometric adjustment which places arcs at suitable positions between surrounding

434

G. Despine and C. Baillard

buildings. The third step is the generation of the street geometry based on the graph and a parametric street model. (Elberink and Vosselman 2006) have worked on the 3D modelling of roads by fusing 2D maps and lidar data. They present an "intelligent" road reconstruction system dealing with real infrastructures. By combining lidar and 2D vector maps, they can reconstruct a 3D topographic model and extrapolate hidden polygons (bridges, tunnels). The study from (Haunert et al. 2005) is the closest to our approach. They use a GIS to compose a virtual world with buildings, road axes, attributes and events. The information needed to create a simulator scene is extracted from this world along a user-defined path. In this model, driving surface rendering is managed by the simulator and intersections are simply considered as events along predefined routes. Unlike these approaches, we consider than crossroads are the key of road modelling and traffic flows, and we have paid special attention to reliably reproducing the crossroad organisation.

2.3 Data models RoadXML3 and OpenDrive4 are two standard open file formats. They both provide a complete description of the road handling the road physical aspect and the logical relationships affecting the traffic. The geometry is mainly based on mathematical functions such as cubics or clothoids, and the road surface has a parametric description. Our model is close to these formats with the following specific differences: • rather than a parametric description, the layers have a geometric description based on polylines, which fits better with the input GIS data and the representation model used by the simulator ; • the graph structure is the core of our model and all our layers derive from it.

3 4

http://www.road-xml.org/ http://www.opendrive.org/

Realistic Road Modelling for Driving Simulators using GIS Data

435

3- Presentation of the road data model

3.1 Three representation levels Road modelling for traffic simulation involve data of very different types. In the literature, at least two representation levels are generally distinguished: the logical level and the graphical level (Carles and Espie 2000), (Kaussner et al. 2002). We propose a graph-based representation with three levels (see Figure 1).

(a) The network level : road geometry, topology and attributes

(b) The traffic level: lane organisation

(c) The graphical and physical level

Figure 1: The three representation levels of the roads for driving simulation

• The first level is the road network. It describes how road axes are arranged, the main traffic rules and the priority system at crossroads. • The second level is the traffic level. Each lane is geometrically described with its direction and its connections to neighbouring lanes at intersections. This level completes the traffic and priority rules from the first level. • The third level manages the graphical and the physical aspects of the virtual world. It enables the generation of the 3D scene. It also constrains the vehicle behavior by defining authorized driving surfaces and obstacle areas. It contains the textured road surfaces, the road signs, the road markings, as well as the ground, the buildings and all visible objects. The network level is the core of the structure. It contains all the key information and interacts with the other two levels (see Figure 2). Importantly, the consistency between levels is guaranteed. Any change in the network

436

G. Despine and C. Baillard

level leads to a change in the graphical and the traffic levels. For instance, if a speed attribute is updated from 70 to 50 kph, then the corresponding road sign in the graphical level is also updated. If a road is removed from the network then the graphical level and the traffic levels take it into account.

3.2 Multi-layered graph model One natural way of designing a road structure is to use graphs, which can store both geometrical information and topological relationships. Our basic graph structure is defined as follows: • The Nodes represent connections; each Node knows all its ingoing and outgoing Arcs. • An Arc is directed and has an initial and a final Node; between these two Nodes, the geometry is described by AncillaryPoints. • A Domain represents a surface that is encompassed by a set of directed Arcs; An Arc knows which Domain is on its left or on its right. The three representation levels of the roads are derived from this basic graph structure (see Figure 2). The RoadNetwork is the graph of the road axes. The classes Section and CrossPoint respectively derive from the classes Arc and Node. A Section is the atomic road element. It has a constant number of lanes and constant attribute values. The RoadGraphic is the graph used for the graphical and physical level. The Domains are used to represent road surfaces, side-walks and any other planar object. The Edge class derives from the Arc class; it represents the road lateral borders (left or right) and their position relatively to the central axe. Tracks and InsertionPoints respectively locate linear objects (road markings, fences) and punctual objects (road signs, furniture). The RoadTraffic is the graph used to represent the traffic level. The Lane and JunctionLane classes are derived from the Arc class. The geometry of every Lane object is stored with driving attributes. The Lane orientation indicates the driving direction. At crossroads the JunctionLane enables a vehicle to fork off from a Lane to an other.

Realistic Road Modelling for Driving Simulators using GIS Data

437

Figure 2: The three representation levels derived from the graph structure. Each structure is represented with its corresponding inherited classes. The bold arrows indicate full data structure "inheritance". Rounded boxes group inherited classes with their mother class.

As shown in Figure 2 this description is completed with a set of links between the three data structures. Every Section is linked to its associated Lane objects, and every CrossPoint is linked to its associated JunctionLane objects. The Section and CrossPoint objects are also linked to the corresponding surfaces in the graphical level. This construction insures data consistency all along the model life. This model also facilitates requests such as "What is the type of the road arriving on my left ?" or "If the speed limit changes, what road signs must be introduced in the model ?".

438

G. Despine and C. Baillard

4- Automatic road modelling process The model described above encodes all the complementary description levels involved in traffic simulation. This section explains how to automatically instantiate this model using realistic data.

4.1 Input data In this study two kinds of data are used: 3D GIS data and GPS navigation data: • The GIS data produced by Siradel include a DTM (Digital Terrain Model), building footprints and heights, main vegetation areas and road axes. The road axes are approximately located at the middle of the road. The road and building geometry have a metric accuracy. However the semantic information related to the roads is generally limited to the road category, and the neighbouring relations between roads are not explicitly described. • GPS navigation data bases, like Open Street Map or TeleAtlas data, provide 2D road axes associated with relevant driving attributes, such as the average speed, the direction of circulation and the forbidden manoeuvres. Although the road topology is correct, the geometry lacks accuracy and building information is generally not available. It is important to note that the road explicit geometry (width, lane layout) is never provided and must be deduced during the modeling stage.

4.2 Method overview Firstly the network level is instantiated using the GPS map road axes and the driving attributes (see section 4.3). Secondly the traffic level is set by creating and connecting the lanes according to the road attributes (see section 4.4). Finally the graphical and physical level is established by defining the correct road profile, and by inferring realistic road markings consistent with the two other levels (see section 4.5). It is completed with additional objects provided by the GIS data base, such as buildings or trees.

Realistic Road Modelling for Driving Simulators using GIS Data

439

4.3 Creation of the road network (Network level) First the two input datasets are registered, then driving attributes are interpreted to provide complete information on the road geometry and topology. At each node the geometry and attributes of the incoming arcs are analyzed in order to detect eventual continuity between road sections. Finally, the set of resulting polylines and attributes are stored in the graph of the road network. These steps are detailed in the following subsections.

4.3.1 Road registration Navigation road data bases provide reliable information about the road network topology and the driving attributes. However the road position is generally not consistent with the GIS data. The GIS road geometry is preferable to the navigation data because it provides accurate 3D information and it is consistent with the building footprints. Hence, a registration procedure is required. The position of the GPS car map is refined according to the GIS geometry, using a graph matching algorithm (Zhang 2009).

4.3.2 Attribute interpretation GPS car map attributes provide various information like driving directions, road classes, estimated speed limits, authorized or forbidden manoeuvres, street names, etc. However they do not include any information about the road surface geometry (road width, number of lanes, road elevation) or crossroad organisation. A set of rules is therefore defined to interpret attributes and find the most likely geometry. The interpretation process takes place at two different levels: it is used to design the street geometry, and to design crossroads via the road continuity concept. Road geometry. The road geometry is designed according to a set of very simple rules. Basically, the available speed limit and the road typology are used to infer the number of lanes and their width. For instance, if a primary road is one-way, with a speed limit of 60 kph and a bus lane, then it is assumed that the road section is associated to three lanes, and that these lanes are rather wide.

440

G. Despine and C. Baillard

Road continuity. Continuity is a very important concept. It is used to associate two “similar” road sections at an intersection according to geometric and semantic criteria. More precisely the matching criteria are the number of lanes, the circulation direction, the speed limit, the street name and the geometrical alignment. A crossroad can then be regarded as a set of crossing traffic flows rather than independent branches connected to the same point. Continuity is not mandatory (in some cases it cannot be established) but it helps understanding the crossroad organization (see Figure 6) and incidentally the traffic rules. Note that the alignment is not the main criterion, as it can be seen in Figure 6a. (Boichis et al. 2000) highlights the importance of crossroads as a structuring element in road design. In generic virtual environment, crossroads usually follow simple pattern: X-cross (4 perpendicular branches), T-cross, or Y-fork. However in a real world, a wide variety of cases can be found (see Figure 3). This diversity requires the modelling process to work with no prior knowledge on intersection shapes.

(a) Shifted axes

(b) L-shaped main street

(c) K-shaped intersection

Figure 3: Illustration of crossroad complexity

Intersections with shifted branches are a particular issue because they cause superimposition effects during the lane and surface generation step. Therefore they are simplified as shown in Figure 4: the two junction points are merged into a single point connecting all the branches, and the initial shape is preserved thanks to two additional points inserted on the modified road sections. Intersections at roundabouts are also important, because they are associated to specific priority rules and they involve lanes with identical direction. The roundabouts are therefore explicitly identified by extracting and connecting circular road axes.

Realistic Road Modelling for Driving Simulators using GIS Data

441

Figure 4: Modelling intersections with shifted branches. On the left, the crossroad is considered as two T-cross junctions. On the right, the two intersections have been merged.

4.3.3 Creation of the road network graph The set of resulting polylines and attributes are stored into a roadNetWork Graph object. Each connection point defines a CrossPoint. The parts of polylines located between two CrossPoints define Sections (part of road axes), and the topological relationships are set. Continuous road Sections are linked into directed Road objects.

4.4 Lane creation and connection (Traffic level) The traffic level is instantiated by creating and connecting lanes according to the computed road attributes. The lanes are derived from the road axes by shifting every section sidewards by a distance +/-d. The right lanes are given the same direction as the original section whereas the left lanes are reversed, except on roundabouts and one-way streets. The directed lanes are stored as Lane objects. At road intersections, LaneJunctions objects are created to connect lanes and indicate authorized manoeuvres for vehicles. In simple crossroads (roads with 2 lanes) every input lane is connected to every possible output lane belonging to a different road section. However on major roads with multiple lanes, turning right or left is not necessarily allowed on every lane. The road continuity is then used to determine which lane corresponds to authorized turns (see Figure 5 and 6):

442

G. Despine and C. Baillard

• Two continuous road sections are associated to the "straight" direction, even if they are not aligned. Vehicles can follow this direction on any lane. • Right or left turns are defined according to the "straight" direction. Only the right lane (respectively left lane) allows to turn right (respectively left). For each lane pair defining an authorized turn, a JunctionLane object is created as a parabolic arc, as shown in Figure 5.

Figure 5: Lane connection at crossroads. An input lane is matched to an output lane, and a JunctionLane is created, allowing a vehicle to turn right.

(a) Continuity is based on lane comparison

(b) Turning maneuvers are forbidden in A and B

Figure 6: Examples of crossroads generation. Continuity is essential to understand crossroad organization and possible connections between lanes.

4.5 3D modelling (Graphical level) When the network and the traffic levels have been defined, the graphical level is finally instantiated. It is performed within four main steps, de-

Realistic Road Modelling for Driving Simulators using GIS Data

443

scribed in the following paragraphs: creation of the road edges, determination of the road 3D profile, addition of realistic and consistent road markings, and finally introduction of additional objects such as buildings or trees.

Figure 7: Road surface creation: (a) road edges are created according to the road width, (b) consecutive edges are connected, (c) edge connections are smoothed, (d) surfaces are created and stored as road or crossroad area.

Creation of the road edges: the road edges are computed by shifting the road sections of a distance +/- d, where d is the half of the road width. Then the resulting edges are connected as shown in Figure 7. Finally, the resulting road edges are smoothed in order to guarantee a smooth drive within the simulator. Refinement of the transversal profile. When the road surfaces have been created, the street gutters, curbs, and sidewalks are easily located by shifting the road edges (see Figure 8). Different road profiles have been defined and are applied according to the road category. Creation of road markings. Road markings are created according to simple rules. At each intersection a pedestrian crossing is created. Discontinuous lines are created to separate two lanes with identical direction or two lanes belonging to a two-lane road. Continuous lines are created in case of a multiple-lane road. Note, this step is highly dependent on the country, and there should be rules defined for every target country.

444

G. Despine and C. Baillard

Priority and road signs. When a major road is involved in a crossroad, a traffic light system is created. If roads are of minor importance the priority system is a “stop” or a “giveaway”. In both cases the road continuity helps to define the organisation of the various traffic flows. Introduction of additional objects. The DTM is inserted into the model, paying particular attention when connecting to the 3D road edges. In order to increase realism, buildings and trees are also inserted.

Figure 8: Definition of a road profile

5- Experimentation and results

5.1 Input data The method has been tested on a small city in the suburb of Lyon in France. The input data consist of a navigation database provided by TeleAtlas, and the GIS 3D data produced by Siradel using aerial imagery. The Siradel base is one meter accuracy and contains a DTM, building footprints and elevations, trees and road axes. As previously mentioned Siradel roads have a better positioning accuracy and it was necessary to preliminary register the TeleAtlas roads to them. An extract of the input data is shown in Figure 9a.

5.2 Result of 3D modelling An extract of the 3D modelled roads is shown in Figure 9(b). 939 streets were created in less than 5 minutes with a non optimized code. The resulting data base was successfully interfaced with the Eca Faros EFX simulator. Figure 10 shows details on complex crossroads: intersections with shifted branches are correctly processed as are the roundabout. The road

Realistic Road Modelling for Driving Simulators using GIS Data

445

continuity is correctly interpreted even in complex case. When comparing the created virtual world and the orthophoto of the same area, it can be seen than the virtual world is close to the real world although the lane number and the road width is initially unknown and no image information is used.

(a)

(b)

Figure 9: Extract of Lyon data base. (a) Ortho-photo of the suburb of Lyon with road axes from TeleAtlas and Siradel. (b) Generated roads and 3d scene integrated into the driving simulator.

(a)

(b)

(c)

(d)

Figure 10: Details of the Lyon base. (a.) input polylines from the GPS navigation map; (b.) traffic lanes in red and the carriageway in blue; (c.) resulting 3D scene in the simulator; (d.) corresponding orthoimage.

446

G. Despine and C. Baillard

It can be observed in Figure 10 that some minor information is missing like a few zebra crossing or isolated trees. In the future this information will be derived from the corresponding ortho-image. Finally, additional objects like secondary street furniture, mail boxes, bins, should also be sparsely introduced at strategic places in order to improve the immersion feeling. Figure 11 shows some views of the simulator running with the created database enhanced with generic façade textures.

Figure 11: Views of the simulator running with the resulting database. Roads are completed with additional Siradel data like trees and buildings

6- Conclusion In this paper we have proposed an automatic solution for creating virtual 3D road models for driving simulators. Unlike most previous work based on procedural rules, urbanism patterns or sociological behaviour, our approach reproduces existing road networks.

Realistic Road Modelling for Driving Simulators using GIS Data

447

First a structured road model is proposed, consisting of three representation levels and well adapted to driving simulators. Then a 3D modelling solution is presented, derived from available standard road data bases. The road complete geometry and the traffic rules are derived from existing road axes and road attributes. Although the input information is incomplete, the generated road model contains all the elements necessary to simulate traffic. The method has been tested on a real data set and shows very promising results. The produced data base has been successfully interfaced with the driving simulator used in the PlatSim project, allowing the user to train in an environment close to the real world. The work described in this paper is a part of a more general study involving various data sources, like aerial photographs, Digital Elevation Models, terrestrial images and terrestrial laser point clouds. The purpose is to automatically and reliably extract most information required by the driving simulators, in order to give the user the opportunity to train under real conditions.

References Bakolas R (2007) Virtual Urbanity: A parametric tool for the generation of virtual cities. Master’s thesis, University College London, Barlett School of graduate Studies Boichis N, Viglino J, Cocquerez J (2000) Knowledge based system for the automatic extraction of road intersections from aerial images. In Int. Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. 33, Amsterdam, the Netherlands Carles O, Espie S (2000) Multi-level environments modelling for road simulation. In Driving Simulation Conference Décoret X, Sillion F(2002) Street generation for city modelling. In Architectural and Urban Ambient Environment Elberink O, Vosselman G (2006) 3d modelling of topographic objects by fusing 2d maps and lidar data. In Proceedings of the ISPRS TC-IV Intl symp. on : Geospatial databases for sustainable development, pp. 199–204 Grote A, Butenuth M, Heipke C (2007) Road extraction in suburban areas based on normalized cuts. In PIA07. Int.l Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, 36 (3/W49A) Haunert J.H, Brenner C, Neidhart H (2005) Using a gis system for the generation of driving simulator scenes. In Advances in Transportation Studies Kaussner A, Mark C, Krueger HP, Noltemeier H (2002) Generic creation of landscapes and and modelling of complex parts of road networks. In Driving Simulation Conference Europe Koberstein J, Peters H, Luttenberger N (2008) Graph-based mobility model for urban areas fueled with real world datasets. In Simutools ’08: Proceedings of the 1st int. conf. on Simulation tools and techniques for communications, networks and systems & workshops (ICST, Brussels, Belgium, Belgium, pp. 1–8

448

G. Despine and C. Baillard

Mena J, Malpica J (2005) An automatic method for road extraction in rural and semi-urban areas. Pattern Recognition letters 26, 9, 1201–1220 Parish YIH, Müller P (2001) Procedural modeling of cities. In SIGGRAPH ’01: Proceedings of the 28th annual conference on Computer graphics and interactive techniques (New York, NY, USA, ACM, pp. 301–308 Ravanbakhsh M, Heipke C, Pakzad K (2007) Road junction extraction from high- resolution aerial images. In PIA07.Int. Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, 36 (3/W49A) Soheilian B, Paparoditis N, Boldo D, Rudant J (2007) Automatic 3d extraction of rectangular roadmarks with centimeter accuracy from stereo-pairs of a ground-based mobile mapping system. In Int. Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. XXXVI, Part 5/C55, pp.22-27, Padua, Italy Shi Y, Shibasaki R, Shi Z (2008) Towards automatic road mapping by fusing vehicle-borne multi-sensor data. In The Int. Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing Tarel JP, Bigorgne E (2009) Long-range road detection for off-line scene analysis. In Proceedings of IEEE Intelligent Vehicle Symposium (IV’2009), pp. 15–20. http://perso.lcpc.fr/tarel.jean-philippe/publis/iv09.html Tripp JL, Mortveit HS, Hansson AA, Gokhale M (2005) Metropolitan road traffic simulation on fpgas. In FCCM ’05: Proceedings of the 13th Annual IEEE Symposium on FieldProgrammable Custom Computing Machines, pp. 117–126. Tournaire O, Paparoditis N (2009) A geometric stochastic approach based on marked point processes for road mark detection from high resolution aerial images. ISPRS Journal of Photogrammetry and Remote Sensing 64, 6, 621–631. Zhang M (2009) Method and implementation of road-network matching. PhD, Intitute of Photogrammetry, University of Munich.

Modeling and Mapping Traffic-Congested Corridors for Statewide Decision Support Jeong C. Seong1, Habtewold Kassa2, David Choi1 1

Department of Geosciences, University of West Georgia 1601 Maple St. Carrollton, GA 30118, USA [email protected], [email protected] 2

Office of Planning, Georgia Department of Transportation 600 West Peachtree St. Atlanta, GA 30308, USA [email protected].

Abstract University of West Georgia and the Georgia Department of Transportation performed a joint research project on identifying traffic-congested corridors and chokepoints in the State of Georgia. Two different methods were applied to identify congested corridors – volume/capacity ratio (VCR) and volume per lane (VPL). The 2008 annual average daily traffic dataset was used in this analysis with linear referencing and dynamic segmentation in a geographic information systems environment. Even though the two methods show similar congestion patterns in many parts along arterial and interstate highways, results also show significantly different patterns in other parts. For example, the VCR results show relatively less congestion than the VPL results, and choke points are quite intermittent in the VCR results while the VPL results show them rather contiguous.

1- Background and Objectives Atlanta is the major commercial and transportation hub of the southeast United States and its international airport is one of the busiest in the world. Atlanta’s economy is led by manufacturing, retail and service. Atlanta is home to many major corporations including Coca-Cola where transportation plays a major role. Atlanta, Georgia recently has experienced significantly increasing traffic congestion, according to the Urban Mobility ReA. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 449 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_30, © Springer-Verlag Berlin Heidelberg 2011

450

J.C. Seong et al.

port published by Texas Transportation Institute (Shrank et al. 2010). The 2010 report, particularly, shows an increasing trend of travel time index values over the last some decades. For statewide transportation planners and decision makers, pin-pointing congested bottle-necks geographically is indispensible to developing long-term strategies and short-term solutions. Two different methods have been used for analyzing the spatial pattern of congested corridors (Saka 2009). One is to use static mapping methods with empirical or simulated data, and the other is to use real-time mapping. An example of the former approach is the congestion corridor analysis performed by the Richmond Area Metropolitan Planning Organization (RAMPO 2008). The analysis used actual traffic speed measurements in order to categorize two congestion levels: ‘Impaired’ for 5-20 miles per hour (mph) below posted speeds, and ‘Congested’ for 20+ mph below posted speeds. The latter approach, for example, is found at Google’s live transportation map (Brandt and Solyanik 2008). In the Google’s real-time traffic map, green is used for 50+ mph, yellow for 25-50 mph, red for less than 25 mph, red/black for stop and go, and gray for no data available. The NaviGAtor (http://www.georgia-navigator.com/realtimetraffic.html) is another example of using the real-time mapping with the Google traffic technology. As shown in Figure 1, various traffic layers are available and users may control the layer display. Even if real-time maps provide invaluable information to travelers, static maps are still important in long-term planning and decision making. In the State of Georgia, a pilot study was performed in 2009 by the Georgia Department of Transportation (GDOT) to identify 50 most congested corridors with traffic volume and design capacity of roads. This research aimed at advancing the pilot study by incorporating more accurate data and additional analyses.

Modeling and Mapping Traffic-Congested Corridors for Statewide Decision Support

451

Figure 1: NaviGAtor, a real-time mapping application developed by the Georgia Department of Transportation using Google maps.

2- Approaches and Methods The research team particularly applied two different methods to identify congested corridors – volume/capacity ratio (VCR) and volume per lane (VPL). In both VCR and VPL, the annual average daily traffic (AADT) was used as the traffic volume. VCR calculates the capacity values based on functional classification from the engineering design perspective – i.e. 10,000 vehicles for minor arterial street, urban principal arterial, rural minor arterial and rural principal arterial; and 20,000 vehicles for rural interstate principal arterial, urban principal arterial, and urban free-

452

J.C. Seong et al.

way/expressway. A four lane interstate highway, accordingly, has a total capacity of 80,000 vehicles per day (20,000 vehicles x 4 lanes). The other approach, VPL, does not consider the engineering design capacity. Instead, it considers speed limits. For example, a 70 mph lane would provide 1.75 times more capacity than a 40 mph lane. VPL is calculated as [VPL = (AADT x MaxSpeedLimit) / (Lanes x SpeedLimt)]. In the equation, MaxSpeedLimit indicates the maximum speed limit in the area of study. In the case of Georgia, it is 70 mph. The 2008 AADT data was used along with linearly-referenced roadway characteristics (RC) dataset with the courtesy of GDOT. ArcGIS™ by ESRI was used to analyze and visualize congested corridors. Figure 2 shows an example of the attribute table of the reference road data layer. Two fields – Begin_M_ft and End_M_ft – were used for linear referencing and dynamic segmentation. The LOS field indicates the level of service calculated with the VCR method.

Figure 2: An example of AADT line data layer with necessary fields joined and calculated.

3- Results Figures 3 and 4 show congestion corridors modeled with VCR and VPL methods, respectively. The legend of Figure 3 shows the level of service (LOS) indicated by VCR. The greater the VCR, the more congested the road. For example, the VCR of 1.0 means that the measured traffic volume is equal to the designed engineering capacity. Any number beyond 1.0 implies congestion. In the map, we categorized congestion corridors into three: Extremely High, Very High, and High. In the case of Extremely

Modeling and Mapping Traffic-Congested Corridors for Statewide Decision Support

453

High, actual traffic volume is two times larger than its designed engineering capacity. In the case of Figure 4, the numbers in the legend indicate VPL, particularly speed adjusted; therefore, there is no upper bound theoretically. We used 20,000 vehicles per lane as the cut value between high congestion and moderate/low in order to compare with the designed engineering capacities of roads used in the VCR calculation. Then, the “Extremely High” and “Very High” categories were calculated with the same ratios used in VCR. In the case of arterial interstate highways and expressways, VCR shows relatively less congestion than VPL. Choke points are quite intermittent in VCR while VPL shows them rather contiguous. VCR and VPL show extremely high congestions along GA-400 at the north of I-285 perimeter. Along I-85, the highest congestion corridor appears above the Hartsfield International Airport in both figures. It is very interesting that the downtown connector appears as ‘High Congestion’ in VCR, while it appears as ‘Very High Congestion’ in VPL. The downtown connector indicates the 8.4-mile Interstate highway section passing the Atlanta downtown area where I-75 and I-85 go together. It is also noticeable that VCR and VPL show relatively lighter congestion corridors along I-20. It implies that the traffic congestions along I-20 are most likely caused by merging/diverging and weaving into traffic while travelers around downtown are looking for their exits as well as crashes, disabled vehicles or road works. I-285 perimeter shows a striking difference between two results. VCR shows a disconnected, less congested pattern. A highly congested corridor appears at the southeast. VPL, however, shows very high congestion at the southeast of I-285 and high congestions on the east and west sides of I285. Both results show very similar patterns in the Lakewood freeway at the southwest Atlanta and the Stone Mountain freeway. Many arterial roads that are connected to interstate or express highways also show a similar pattern between two results; however, more arterial roads are congested in VCR. Particularly, the northwest Atlanta area shows a big difference. Beyond metro Atlanta, there were several places showing congestions in VLP such as I-75 in Macon, Savannah, Columbus and Augusta. VCR, however, showed congestions in most mid to large cities in Georgia.

454

J.C. Seong et al.

Figure 3: Congestion corridors in metro Atlanta, estimated with the volume/capacity ratio.

Modeling and Mapping Traffic-Congested Corridors for Statewide Decision Support

455

Figure 4: Congestion corridors in metro Atlanta, estimated with volume per lane considering speed limits.

456

J.C. Seong et al.

4- Conclusions and Future Works In this research, two different methods were applied to analyze road traffic congestion corridors in Georgia. The 2008 GDOT datasets were analyzed and visualized. Two approaches showed significant differences; however, the most congested corridors on GA-400 and I-85 above the Hartsfield international airport appeared in similar patterns on both maps. I-285 showed large differences. VCR revealed more congestion in arterial roads and in mid to large cities in Georgia. This research showed that VCR and VPL estimate congestion corridors similarly in many places; however, striking differences also appeared between two methods. Decision makers will benefit from the results in planning short-term solutions and long-term strategies. We plan on validating the results with field checks in the very near future. We also plan on incorporating more variables such as exits, traffic signals, curvature, and speed-limit calibration in estimating choke points.

Acknowledgments The authors wish to acknowledge the Georgia Department of Transportation and the Atlanta Regional Commission for the support of traffic and GIS datasets.

References Brandt E, Solyanik S (2008) New Ways to Beat Traffic with Google Maps. http://googlelatlong.blogspot.com/2008/04/new-ways-to-beat-traffic-with-google.html. Accessed 14 February 2011. Rampo (2008) Congestion Management System Technical Documentation. Technical Report for the 2031 Long-range Transportation Plan and Congestion Management System Update. Richmond Area Metropolitan Planning Organization. http://www.richmondregional.org/Publications/Reports_and_Documents/CMS_07-0708/CMP.pdf. Accessed 14 February 2011. Saka AA (2009) Statewide GIS Mapping of Recurring Congestion Corridors. Maryland State Highway Administration Project Report . http://www.morgan.edu/Documents/ACADEMICS/SOE/ntc/Statewide_Saka_0809.pdf. Accessed 14 February 2011. Schrank D, Lomax T, Turner S (2010) Urban Mobility Report 2010. Texas Transportation Institute. http://tti.tamu.edu/documents/mobility_report_2010.pdf. Accessed 14 February 2011.

How to Map out the Routes of Walkers in a Forestry Environment Considered to be of Risk ? The Case of Human Exposure to Lyme Borreliosis in the Forest of Sénart (Île-de-France, France) Vincent Godard, Christelle Meha, Olivier Thomas Université Paris 8, UMR ENEC, Saint Denis, France; [email protected]

Abstract Well known in rural areas, Lyme borreliosis poses a new problem of public health in heavily urbanized spaces. While it is too early to say whether there is an emergence (or a re-emergence) of Lyme borreliosis in the Îlede-France region, however, it is necessary to take into account exposures and risk factors related to this disease, particularly for urban populations which are sometimes poorly informed about the risk of Lyme borreliosis. For instance, we need to do some research on the contacts between the routes that people follow in the forest and the spaces and environments considered to be of risk (e.g. the most suitable habitat for ticks). Indeed, as an essential component of risk assessment, the human factor, which is characterized by certain types of behaviours, such as the mode of penetration and the frequency of visits of endemic foci, is too often overlooked in studies on Lyme disease ecology. So, we have therefore carried out a number of inquiries on people who are doing some physical activities in the pilot site of the Forest of Sénart in 2009-2010. Such inquiries have recorded, by mapping, the routes followed by users, and by questionnaires, the socio-demographic parameters that are associated to each of them. Another section of the present study is carried out, at the moment, to estimate tick abundance in the Sénart forest (ticks sampling from the vegetation and from deer populations). It will then be possible to find how the structure of a forest area is likely to favour the contact between forest users and vectors of the disease; the outcome being the spacialization of this contact and to study ways in which it is possible to minimize risk via the landscape and design. The results show a forestry contrast between attractive and repulsive areas. However, the declarative method of itineraries reported by maA. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 457 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_31, © Springer-Verlag Berlin Heidelberg 2011

458

V. Godard et al.

nual mapping must be compared with passive method to elucidate the behaviour of some walkers. In fine, this study deals with the issue of society’s vulnerability in relation to environmental health risks and looks particularly at how to manage the public’s use of forests in the context of an “emerging” health risk.

1- Introduction Lyme borreliosis, or “Lyme disease”, is a zoonotic infectious disease, though not contagious, due to Borrelia burgdorferi sensu lato (sl), that can be transmitted to humans by the bite of ticks infected by this complex bacterium. Ticks become infected when they feed on birds or mammals that carry the bacterium in their blood. Lyme borreliosis is widely spread in the northern hemisphere, with probably a little more than 50,000 cases in Europe (Perez-Eid 2001). It is, at the moment, the most frequent vector-borne disease in the whole area of distribution. From an epidemiological point of view, the risk of transmission of this infection to humans is related to the abundance and infection prevalence of its main vector, namely ticks of Ixodes ricinus species complex in Western Europe (Anderson et al. 1986) and to different types of human behaviour. Generally, this disease is contracted by human beings in forestry environments where ticks are preferentially found (Gilot et al. 1994). To know how hikers, bikers, pickers and forest professionals are exposed to this disease, it is tempting to observe their spatial habits. By mapping routes described by socio-demographic data collected in investigation questionnaires, combined with those related to tick density, it should be possible to represent the areas of contact, e.g. exposure. In this study, we intend to present a map representing the exposure of the non-professional population in Sénart Forest (Essonne and Seine-et-Marne, France). This little periurban forest (3, 200 ha), just south of Paris (Figure 1), is very highly visited with more than three million visitors per year estimated 10 years ago (Maresca 2000). To carry out a Lyme borreliosis risk map, several investigations have been started simultaneously in a pluridisciplinary research program, first of all by collecting ticks. The program was under the responsibility of the Institut Pasteur’s Centre National de Référence (CNR) for Borrelia, already responsible of sampling surveys undertaken in 2008 and 2009. The next sur-

How to Map out the Routes of Walkers in a Forestry Environment Considered to be of Risk ?

459

vey is due in 2011 in order to ensure effective follow-up (diachronic analysis). Infection rates (around 11 percent) in Sénart Forest are high and might present a risk for the numerous visitors of this site (Cornet and Ferquel 2009). If they prove to be stable in space and time (at least shortly) for the forest range, then the density of ticks for the forest parcel or vegetation-type group will have to be compared to the density of forest users. Such attendance, in particular the behaviour of walkers will be the second study of this program. A third part, guided by the Department of Infectious and Tropical Diseases of the Centre Hospitalier Intercommunal of Villeneuve-Saint-Georges (CHIV), will investigate the consequences of the disease in about fifty cities around the Forest of Sénart.

Figure 1: Map of localisation of Sénart Forest (general view) Sources: extract of Scan 25 IGN, 2003

How to map out these routes to estimate the human penetration in spaces and environments which may be considered to be of rick? Is the collection of information, on a car park, after doing an activity such as hiking or biking, reliable enough? Is human memory able to remember a route which is likely to last several hours? Is there a sampling method which does not make the use of memory? Finally, whatever methods of measurement and survey are used to track users through our study zone, which methods of data analysis should be chosen to assess human exposure to ticks?

460

V. Godard et al.

2- Background and objectives

2.1 Epidemiological context Well known in rural areas, due to its endemic character in certain regions of the east and the center of France, Lyme borreliosis poses a new problem for public health authorities in urbanized areas, due to high human concentration and the polarizing ability of periurban large forest ranges. The spatial spreading of the sanitary problem near large cities might lead to an outburst of the cases recorded by general practitioners and clinicians. The fear about an increase in Lyme borreliosis risk, though not yet found, is expressed significantly in the Île-de-France region, especially around the Forest of Sénart (Méha et al. 2010) according to information gathered by health practitioners during the last years. An incidence survey is now being carried out to estimate the risk level in this area knowing that updating the data is a matter of priority for an efficient prevention policy of the disease (Vassallo et al. 2000). That is part of why this forest located 22 km to the southeast of Paris and 7 to 8 times smaller than Fontainebleau Forest, was chosen as a pilot site for our study. Located at the interface between health and environment, the problem defined explained the need to identify and describe risk dummy variables linked to the Lyme borreliosis ecology. As the present knowledge makes it impossible to identify exactly these variables suitable for Ixodes ricinus ticks, it should be important, in a first stage, to refine our knowledge about the explanatory variables or predictors of tick distribution in order to locate potential foci of disease transmission. Our aim is to better understand the spread of the disease and to link it to the behaviour and habits of forest users. In addition, anticipating potential disease outbreaks in infested areas is based on the need to consider exposure and risk factors.

2.2 Purposes: to map out and feed the database Too often overlooked in studies on Lyme disease ecology, the human factor, which is characterized by certain types of behaviours, such as the mode of penetration and the frequency of visits of endemic foci, is an essential component of risk assessment. Therefore, the analysis of contacts

How to Map out the Routes of Walkers in a Forestry Environment Considered to be of Risk ?

461

between the itineraries that people adopt in forests and the spaces and environments considered to be of risk constitutes a privileged avenue of study. As an aside, it should be pointed out that Sénart Forest has been, for several decades, the object of internal and peripheral improvements related to the public’s use (Moigneu 2005). Furthermore, the spatial design of a forest, which takes into account elements such as accessibility and frequentation, especially in suburban contexts, could favour contact between forest users and vectors of the disease. Consequently, there is a need to study these spatial dynamics (and to model this contact), as well as to study ways in which it is possible to minimize risk via the landscape and design. Our study deals with the issue of society’s vulnerability in relation to environmental health risks and looks particularly at how to manage the public’s use of forests in the context of an emerging health risk. To spatialize the human-tick contact, a map of tick density is required and must take tick activity periods into account: generally, ticks cannot be seen during the winter months, because they are hibernating, while on the other hand, the number of questing ticks collected from the vegetation tends to increase significantly during the early-spring as a result of warmer temperatures. All stages (larvae, nymphs and adults) are active during the spring and summer months. However, ticks are extremely sensitive to desiccation and require a relative humidity of at least 80 percent throughout the year, so that they are confined to areas where a good cover of vegetation and a mat of decaying vegetation are present (Gray 1998). In addition, tick abundance may vary significantly from one year to the next according to late spring and early summer rains (McCabe and Bunnel 2004). It is the subject of a study by the Institut Pasteur’s CNR for Borrelia. This is different in focus from the present task, however, and it will not be discussed further. Instead, the following paragraphs are aiming at providing an overview of the procedure used to feed the database about the routes and sociodemographic characteristics associated to each of them. Then, according to the qualities and shortfalls of the sampling procedure, the statistical and spatial treatments that is possible to be applied in our data set will be described.

462

V. Godard et al.

3- Approaches and methods In order to investigate exposures and risk factors linked to Lyme borreliosis, surveys regarding the number of visitor going to the Forest of Sénart were carried out during the years 2009 and 2010, mainly on visitors aged 18 years and over. The people surveyed were a sample of volunteer adults, whether or not inhabitants of the Île-de-France region, obtained by a systematic sampling method (ask the nth person once we have finished completing the questionnaire). They were interviewed face-to-face, via a standardised questionnaire annotated for the investigators, and notably questioned about their habits in the forest that might be linked to an increase of Lyme borreliosis risk. In addition, the survey period took places during periods of high visitor density in the forest and during the seasonal activity period of the Ixodes ricinus tick. Such investigations have enabled us to describe exactly their socio-demographic characteristics, the places they attended and their behaviour within the forest. All data collected will then be analysed statistically. On a specific plan, forest users have been asked to describe in detail their trajectory through the study area, on the day of the survey. To test the protocol of data collection, a pre-investigation phase was carried out during 2009 on Sundays over several week-ends. Methodological results were shown and discussed in an international workshop held on the theme of “geo-simulation and mathematical modelling for zoonotic diseases”, coorganized by the Public Health Agency of Canada and York University, which took place in Toronto (Godard et al. 2009). Following the resulting confrontations, improvements regarding the survey method have been made and special attention has been paid to correctly identify time spent overall in the forest, the different paths that were used (including “offtracks”), number of halts and time spent in specific area and specific behaviour possibly linked to increased exposure to ticks (and potentially to Lyme borreliosis). Modelling and simulations of human behaviour (by multi-agent systems) in the forest are being carried out in cooperation with researchers of Laval University (Québec, Canada); as a result, it is essential to describe very precisely the behaviour and habits of forest users. The method developed to vectorize trace marked on a paper map (A3 black and white printing of Top25 2415 OT on the Forest of Sénart itself), is primarily a digitalizing by scannerizing, and then a vectorizing on screen. Two types of information are obtained from this operation: firstly the distribution of walkers within the whole area of the study, secondly the forest attendance, section by section. So the result of the vectorizing is a

How to Map out the Routes of Walkers in a Forestry Environment Considered to be of Risk ?

463

“spaghetti” network. You can observe in figure 2 an extract of the database representing fifteen people who has visited the study area.

Figure 2: Itineraries of a sample of 15 walkers in Sénart Forest (spring 2009) Sources: Extract of Scan 25 IGN, 2003

Figure 3: Process of data integration From Thomas (2010)

464

V. Godard et al.

In order to account the number of walkers who have followed a path (or a trail), all the itineraries must absolutely be superimposed over this path (or this trail) in such a way as to obtain a perfect aggregation of all routes; and then, a topologic treatment was carried out using ArcGIS by ESRI®. In tandem with this work, the questionnaires were integrated in our database. From a script written in VBA, data can be sorted according to various criteria (hiker/jogger, man/woman, with dog/without dog…) per each section; the data integration process is summarized in figure 3. This operation makes it possible, for example, to produce a map of the number of male joggers without dog per section (Figure 4).

Figure 4: Number of routes obtained among a sample of joggers interviewed in 5 different car parks (summer 2009). From Thomas (2010)

4- Results There are very few studies looking at the frequency of visits to a forest site and especially the routes of interviewed people. By comparison, urban areas seems to have been more often submitted to analysis of pedestrian (Foltête 2006), motor cars (Genre-Grandpierre et al. 2006), or even public transport (Genre-Grandpierre 2006) movements than forest areas. One of

How to Map out the Routes of Walkers in a Forestry Environment Considered to be of Risk ?

465

the main differences in both the sampling method being used and data treatment is certainly the possibility for people to free themselves from the need to use the path network when they walk through the forest. Accordingly, the results obtained may then be altered both by the method of sampling and the way to data transcribe information. When observing a sample of routes in our study zone (cf. fig. No 2), it is surprising to note the low number of forest users who leaves the paths to walk through the understory, e.g. “off-tracks”. So, most of the people follow the linear paths which converge to star-shaped crossroads (legacy of a royal hunting forest). If a family with a baby buggy prefers such linear roads instead of take a walk through the understory is understandable, but that the majority of users do the same, this requires us to verify our investigation method. Even though the forest is small compared to others, would all the visitors of Sénart Forest then have a singular behaviour? We will return to this subject later.

Figure 5: Extract of a sample of itineraries around the Faisanderie car park in Sénart Forest, showing an HP “off-tracks” (spring 2009). Sources: Extract of Scan 25 IGN, 2003

In order to support our analysis, one of the few examples of an “off-tracks” (HP) is highlighted in figure 5. Our type of investigation by a posterior mapping would it be involved in the “off-tracks” underestimation? There is no evidence that the potential bias introduced by a posterior statement is expressed by such practices. However, we thought we need to test other ways to track routes travelled to later trace them onto maps, such as, for instance, the use of GPS (Global Positioning System) navigation devices. This method is obviously one of the most advantageous ways to do it.

466

V. Godard et al.

Several studies have already been carried out in this way. One of them (Colas 2007) which took place in Fontainebleau Forest gave lessons to improve our methodological framework for mapping human exposure to tick populations (and potentially to Lyme borreliosis). As a way to utilize their full potential, a short methodological overview should be made. Wider public GPS are entrusted to forest users who, once they have accomplished their walk, are interviewed using questionnaires in order to supplement information about the routes they adopt through the forest. After they return the GPS device used for the walk on the day of the survey, GPS are downloaded later, and all the data are gathered and stored in a GIS database. The segments of a same trace are then aggregated into a unique route and aberrant records are discarded (Colas 2007). Two results from this test are shown in the figures 6 and 7 for comments.

Figure 6: Extract of a sample of itineraries recorded by GPS (Hautes Plaines, Fontainebleau) From Colas (2007)

Figure 7: Extract of a sample of aggregated itineraries to estimate the attendance of an area (Hautes Plaines, Fontainebleau) From Colas (2007)

Figure 6 enables to visualize some areas with high densities of GPS traces. This figure clearly indicates that the majority of forest users have followed

How to Map out the Routes of Walkers in a Forestry Environment Considered to be of Risk ?

467

well-established paths on their walks. But we also see very clearly that only one GPS trace has been recorded out of a well-establish path; in other words, this means that a user has walked through the understory (cf. the arrow and question mark, fig. No 6). After the data have been processed (aggregation of validated routes and the deletion of abnormal traces) in order to obtain an intensity of attendance per section, we only retain the main traces. Marginal traces of the people who are taking a walk out of the paths might be lost from the database or, at least, from the mapping, during the transformation (cf. the arrow and question mark, fig. No 7). However, this singular user poses an interesting problem when we studied it from an epidemiological perspective. Indeed, he is probably more interesting than other users because he will certainly be more exposed to the vector tick than if he would have stayed on well-established paths where there are no herbaceous plants likely to harbor questing ticks. The absence of marginal traces may be interpreted as referring to the frame on which our investigation is based, as opposed to the passive attitude of the GPS-equipped user in the Forest of Sénart.

Figure 8: Traces in the supervised biological Reserve (RBD) of the Mare aux Pigeons, close to the pond. From Colas (2007)

468

V. Godard et al.

However, the wealth of information collected by the GPS devices may be exploited if GPS traces are not aggregated systematically and if solitary traces are not deleted because they are considered as abnormal. Figure 8 gives us an overview of what can be a realistic attendance of a forest area before aggregating all the routes. The access to the Supervised Biological Reserve (RBD) is allowed to public, but visits are not encouraged neither by forest planning nor by road signs to get inside this area: Walkers are also coming in larger numbers, but still small as compared to what happens around here. Then, that is quantifiable and may be combined with other information such as environmental information characterizing the forest parcel.

5- Conclusions and perspectives This short contribution focuses on some methodological tools as used in geography applied to epidemiology, taking into account the landscape as a global space of transmission of diseases and support of human activities. If data acquisition on routes that people follow in forests is not particularly original (Foltête 2006), their use in a health risk assessment approach constitutes an axis not very much represented in epidemiological studies. Nevertheless, an approach with a similar methodology has been developed by Benabderrahmane et al. (2009) consisting to propose the modelling of human exposure to alveolar echinococcosis in a recreational area such as the highly visited La Courneuve Park (Seine-Saint-Denis, France). Though the movement in a forest area is not subject to the same constraints as in an urban park, this survey in La Courneuve Park may have been lightly distorted by a reduced declaration of “off-tracks”. The inquiry carried out in Sénart Forest is probably more distorted, though it is not possible, at the moment, to confirm it. So we need to test another approach to record routes. As previously reported, surveys regarding the number of visitor going to the Forest of Sénart do not inform, except scarcely, on the trajectography of forest users. But when they did, it is only at the end of the visit (when people need to hurry back home!). Respondents often rely on their memories, which are neither totally reliable nor exhaustive. In addition, the willingness of the forest users, who try to remember their route on a paper map, acts also as a constraint. There is at least an experimental study in the Île-de-France context, precisely in Fontainebleau Forest (not far from our

How to Map out the Routes of Walkers in a Forestry Environment Considered to be of Risk ?

469

study area), which used lending GPS to forest users in order to automatically track your location (from the beginning to the end of a walk) inside the forest (Colas 2007). However, this test required the loan of 80 GPS units to the wider public. Another possibility of experimentation can be tested with the support of computer specialists who have recently joined our research team. The French population, especially urban, is more and more equipped with smartphones with integrated GPS chipset. Why do not we consider to test, with the support of volunteers, a travel sensor for measuring a travel distance and routes of forest users (via an onboard application)? By this way, we think we can eliminate some short-term memory lapses and thus obtain maps of human exposure to ticks (via the spaces and environments considered to be of risk).

Acknowledgements The present investigation is the result of an action supported by Île-de-France region in a contract PICRI and an answer to the project validated by the Conseil Scientifique of Maison des Sciences de l’Homme Paris Nord. We also thank Bernard CAUCHETIER for the data of MOS and ECOMOS of IAU-IdF and Sylvain DUCROUX for the data and ONF support. Special thank to Michelle GIBOIRE for translation, Philip CONRAD SCOTT for proofreading and suggestions, and students team (Camille DELAHAYE, Dimitri LE TORRIELEC, Marianne LIECHTY, Samuel MERMET, Juliette PINARD, Aurélien PONCE) for survey.

References Anderson JF, Doby JM, Courtamanac’h A, Hyde FW & Johnson RC, 1986. « Différences antigéniques entre des souches de Borrelia burgdorferi isolées d’Ixodes ricinus en Bretagne ». Méd. Mal. Inf., 16, pp. 171-175. Benabderrahmane M-C, Barruel L, Chenchouni H, Combes B, Godard V, Tolle F, 2009. « Use of remote sensing and GIS in the eco-epidemiological landscape characterization of health risk in the green space of urban territories: Study case of alveolar Echinococcosis in Paris region (France) ». European Landscapes in Transformation: Challenges for Landscape Ecology and Management. 70 years of Landscape Ecology in Europe. European IALE Conference 2009, Salzburg (Austria), July 12 – 16, 2009. Colas S, 2007. Pour un meilleur équilibre entre accueil du public et préservation des milieux naturels: Évaluation et comparaison d’actions menées en Forêt de Fontainebleau et à la New Forest (Angleterre) - Nancy : AgroParisTech - ENGREF, 89 p. + annexes. Cornet M, Ferquel E, 2009. Résumé du rapport annuel d’activité, CNR des Borrelia, Institut Pasteur, Paris, 58 p.

470

V. Godard et al.

Ferquel E, Garnier M, Marie J, Bernede-Bauduin C, Baranton G, Perez-Eid C, Postic D, 2006. « Prevalence of Borrelia burgdorferi sensu lato and Anaplasmataceae members in Ixodes ricinus ticks in Alsace, a focus of Lyme borreliosis endemicity in France ». Appl. Environ. Microbiol., 72, 4, pp. 3074-3078. Foltête J-C, 2006. Paysage et mouvement. De l'écologie aux déplacements urbains : éléments pour une identification des paysages préférentiels. Thèse de HDR. Université de FrancheComté, 223 p. Genre-Grandpierre C et Josselin D, 2006. « Dépendance à l’automobile, tension dans les mobilités et stratégies des ménages ». Cybergeo : European Journal of Geography, Sélection des meilleurs articles de SAGEO 2006, article 419, mis en ligne le 23 avril 2008, modifié le 04 juillet 2008. Http://cybergeo.revues.org/index17762.html. Consulté le 11 octobre 2010. Genre-Grandpierre C, 2006. « Qualité de l’offre et usage du transport public en milieu urbain ». Cybergeo : European Journal of Geography, Sélection des meilleurs articles de SAGEO 2005, article 376, mis en ligne le 05 juin 2007, modifié le 10 février 2010. Http://cybergeo.revues.org/index6736.html. Consulté le 11 octobre 2010. Gilot B, Guigen C., Degeilh B, Doche B, Pichot J, Beaucornu J-C, 1994. « Phytoecological mapping of Ixodes ricinus as an approach to the distribution of Lyme borreliosis ». In Lyme borreliosis (1994), Axford J-S, DHE Rees ed., Plenum Press Nex York, pp. 105-112. Gray JS, 1998. «The ecology of ticks transmitting Lyme Borreliosis ». Exp. and Appl. Acarol., 22, pp. 249-258. Godard V, Méha C, Benabderramahne M-C, 2009. « Modeling of human Exposure to Lyme disease Risk in a French forest Landscape ». in: CODIGEOSIM Workshop on geosimulation and mathematical modelling for zoonotic diseases, Toronto, Canada August 19th-21st, 2009 at York University. Organizing Committee: Dr. Bernard Moulin, Laval University, Quebec, Dr. Nick Ogden, PHAC, St-Hyacinthe, Dr. Peter Buck, PHAC, Ottawa, Dr. Jianhong Wu, MITACS Centre for Disease Modelling. McCabe G-J, Bunnell J-E, 2004. « Precipitation and the Occurrence of Lyme Disease in the Northeastern United States ». Vector-Borne and Zoonotic Diseases, 4, 2, pp. 143-148. Maresca B, 2000. La fréquentation des forêts publiques en Île-de-France. Etude réalisée dans le cadre de l’évaluation du contrat de plan Etat-Région 1994-1999, de l’Ile-de-France. Juillet 2000, 40 p. Méha C, Godard V, Gramond D, 2010. « Forêts et santé : identification d’indicateurs spatiaux de foyers épidémiologiques. Exemple de la borréliose de Lyme en forêt de Sénart ». In : Des milieux aux territoires forestiers : itinéraires biogéographiques. Mélanges en l’honneur de Jean-Jacques Dubois (sous la dir. de Galochet M et Glon E), Artois Presses Université, Collection géographie, pp. 223-235. Moigneu Th, 2005. Gérer les forêts périurbaines. ONF, Paris, 414 p. Perez-Eid C, 2001. « Déterminisme de distribution géographique des maladies transmises par les tiques ». Méd. Mal. Infect., 31, suppl. 2, pp. 184-187. Smith R, O' Connell S, Palmer S, 2000. Lyme disease surveillance in England and Wales (19861998). Em. Inf. Dis., 6, 4, pp. 404-407. Thomas O, 2010. Epidémiologie de la borréliose de Lyme en forêt de Sénart. Analyse du comportement humain. Mémoire du Master 2 pro G2M, Université de Paris 8 (sous la dir. De V. Godard), 31 p. [En ligne] URL : http://julienas.ipt.univparis8.fr/vgodard/pub/geomarke/memoires/09-0/thomas.pdf Vassallo M, Pichon B, Cabaret J, Figureau C, Perez-Eid C, 2000. « Methodology for sampling questing nymphs of Ixodes ricinus (Acari:Ixodidae), the principal vector of Lyme disease in Europe », Entomol. Soc. Am., 37, pp. 335-339.

Estimation of the Locations of the LanguageVersions of Wikipedia - a Case Study on Geographic Data Mining Tobias Dahinden Institut für Kartographie und Geoinformatik Leibniz Universität Hannover, Appelstraße 9a 30167 Hannover, Germany [email protected]

Abstract People write about things they believe to know, and in particular those things that are within the environment they live in. They also write in a language they know. Therefore, there is a relation between the individual local environment and the language used for the description. In this paper the areas of several languages are estimated according to the geographic footprint of the language versions of Wikipedia. These estimated language areas are compared to those represented in linguistic maps. The results of this comparison are presented for a subset of Germanic languages.

1- Introduction The online encyclopaedia Wikipedia is available in more than 260 languages. At least 186 of these languages have more than 1000 articles (as of January 2010). The English Wikipedia is the largest with about 3 million articles followed by the German Wikipedia with about 1 million articles. There are different versions for varieties of languages: you find a Wikipedia not only for English, but also Old English, Middle English, and Simple English. Similarly, there are more than 13 German dialects with their own Wikipedia (Wikipedia 2010b). Not every language version though refers to an independent language.

A. Ruas (ed.), Advances in Cartography and GIScience. Volume 2: Selection from ICC 2011, 471 Paris, Lecture Notes in Geoinformation and Cartography 6, DOI 10.1007/978-3-642-19214-2_32, © Springer-Verlag Berlin Heidelberg 2011

472

T. Dahinden

There are articles in the various Wikipedias which refer to the same entity. For such an entity, a so called Interwikilink (IWL) is used to point from one Wikipedia article to the other. For example, 51.3% of the articles in the German Wikipedia have a link to equivalent articles in the English Wikipedia, but 39.8% of the articles have no link to another Wikipedia (c.f. Table 3). Articles about a geographic entity contain coordinates besides the text description. About 10% of all Wikipedia articles have a coordinate. The coordinate refers to an important part of the feature. A point coordinate is always used even if the object is linear or laminar. In this article, we present a method to determine which geographic region predominates in a given language version. According to the first law of Geography by (Tobler 1970), that states that "near things are more related than distant things", it can be expected that the authors of Wikipedia write about objects that are within the environment they live in and perhaps know better. This implies that the predominant geographic region in a given language version corresponds to the region where the language of the authors of Wikipedia is spoken. This assumption is also supported by Figure 1: It shows all articles of the English Wikipedia (above) and all articles of the English Wikipedia without linkage to articles of the German Wikipedia (below) for the northern hemisphere as black dots. It is clearly visible that Germany and Austria are nearly white but other regions (such as France) are black. This happens despite the translation of articles (Wikipedia, 2009a). Further investigations on Wikipedia revealed some dependency between the articles of a language version and the entities within the region of the world where this language is spoken (Hecht and Raubal 2008).

Estimation of the Locations of the Language-Versions of Wikipedia

473

Figure 1: Footprints of the articles of the English Wikipedia with geographic context (black dots). Upper image: all articles; lower image: all articles but disregarding articles also occurring in the German Wikipedia. Equirectangular projection, 150° W – 150° E, 0° N – 70° N.

The following assumption shall be verified: The spatial distribution of the articles of a language version of Wikipedia correlates with the region, where this language is spoken. The paper is organized as follows: In the next section related work about research on Wikipedia as well as resources from linguistics are discussed. Section three explains the technical details of our method used for the verification of the assumption. In the fourth section, some selected results are presented. The paper concludes with an outlook on open problems.

2- Related works

2.1 Wikipedia as a resource for research People are allowed to alter Wikipedia articles by editing as they like. As such, a very personal bias or even something incorrect can be included. Using Wikipedia as a source of knowledge is always regarded as a problem of "Garbage in – Garbage out". In spite of this, there are several investigations about the quality of Wikipedia (Clauson et al. 2008; Bragues 2009).

474

T. Dahinden

A more reliable application than studying the primary (easily editable) content of Wikipedia is the retrieval of information from its structures and metadata. Examples include: analysis about the community (Kittur et al. 2007; Ortega et al. 2007), the topics covered (Halavais and Lackaff 2008), the centrality of the articles according to the link structure (Dolan 2008), and cultural differences (Pfeil et al. 2006). In GI-Science and Cartography there are studies using Wikipedia as well. For instance, the content of Wikipedia was used for location based services (Schöning et al. 2007; Paelke et al. 2010), abstract entities and objects with vague boundaries were located (Hecht and Raubal 2008; Dahinden 2009), information about the relevance of objects was retrieved (Dahinden and Sester 2009), and even the first law of Tobler mentioned before has been verified (Hecht and Moxley 2009).

2.2 Linguistic maps A comprehensive introduction to linguistics, which includes bibliographical references, can be found e.g. in (Fagan 2009). For this study, it is important to consider that languages have regional variations and that they change during time. In particular, they interact with each other, see e.g. (Samuels 1972 chapter 6.7). Thus, language areas change. There are plenty of maps showing where a certain language is spoken. Only three examples are mentioned here: • The World Atlas of Language Structures by (Haspelmath et al. 2005) describes 142 linguistic features with at least one map per feature. In this atlas 2559 languages (these are thought to be half of all languages; ibid. p. 3) are shown on the maps. • The dtv-Atlas Deutsche Sprache (König and Paul 2007) is a summary about linguistics. It contains 165 single maps drawn by a graphic artist. • The Sprachatlas des Deutschen Reichs by (Wenker 1888-1923) is based on a survey of about 50,000 questionnaires. It contains 1647 coloured maps. There are several definitions for language, dialect and language area. For this reason the maps may differ substantially. For example, the area where Low Franconian is spoken is attached to Central German (Wenker 1888-

Estimation of the Locations of the Language-Versions of Wikipedia

475

1923; Brockhaus 1894) as well as to Low German (Protze 1969). Luckily this problem can be ignored for this study.

3- Approach

3.1 General approach To estimate the language area, the articles of this language version of Wikipedia are analysed. If the article has a coordinate, the coordinate is taken as an indication that the language is spoken at this place. It has to be considered that the article could be about something very famous (like Eiffel Tower). To reduce the influence of such famous features which are referred by more than one article, a weight is assigned to the reference coordinate. The weight is taken as the multiplicative inverse of the number of Interwikilinks (IWL) in the article. We assume that a language is mostly spoken in those regions where the density of the weighted coordinates is high. Hence, a simple point pattern analysis by estimating the kernel density according to the weighted coordinates is calculated using a Gaussian kernel (Silverman 1986; Smith et al. 2007). The bandwidth is then estimated by a rule-of-thumb: the minimum of the bounding box around all associated coordinates is divided by 30 (ESRI Support Centre 2010). Certainly, there exist more sophisticated methods such as using a variable bandwidth, but the rule-of-thumb is sufficient here. Of course, it cannot be generalized. To determine whether the estimations are valid, the results are compared with a map showing language areas. For this comparison the maximum surface density is normalized to 1, the density is classified between 0 and 1 by steps of 0.01. For each class the covered region was determined. This region is compared to the area, where the linguistic map indicates that the language is spoken. This is done by calculating the difference between pixels that are correctly allocated and pixels that are wrongly allocated. This difference in area is normalized according to the size of the area under investigation. The progression between 0 and 1 is drawn in a diagram (c.f. Fig. 3 and 6).

476

T. Dahinden

A disadvantage of the kernel density estimation is the fact, that the density will never be zero. Therefore, it is not possible to determine the size of an unknown region. But it is possible to calculate the difference between the density of weighted coordinates and the density of all coordinates of Wikipedia (with a weight of 1). The outcome of this is a region where the language version dominates the remaining Wikipedias. Further, this difference in density technique eliminates the influence of the inhomogeneous distribution of the articles. Calculating the difference of these two densities, the inhomogeneity cancels itself out. The results of this method are shown in a map series (Fig. 5) and the estimations are compared with a linguistic map.

3.2 Languages and area under investigation Selection of languages: Due to lack of space only few results can be shown here. In this article the research is restricted to Franconian German, Alemannic German, and Austro-Bavarian. This restriction is based on three preconditions: 1. The language versions under investigation represent Wikipedia that have different sizes. 2. The language areas represent the difference in size of language areas. 3. The real conditions of this area are known by the author. Therefore German, Dutch, Luxembourgish, Ripuarian, Alemannic, Limburgian, and Bavarian Wikipedia are under investigation. Note: Dutch and Limburgian are Low Franconian languages, Ripuarian and Luxembourgish are Central Franconian dialects. The investigation area is limited to the band of 4° E – 20° E, 46° N – 54° N. Selection of a map for the comparison: There are numerous linguistic maps for the area under investigation. Regrettably, in the author's view, none of them is really satisfying: • The map grid, projection, and scale are not declared: (Protze 1969; Bausch 2002).

Estimation of the Locations of the Language-Versions of Wikipedia

477

• The maps are too small, in a small scale, or the quality is not satisfying: (Maurmann 1888-1923; König and Paul 2007; Wikipedia 2007) • The maps contain only a part of the investigation area: (Bausch 2002). • The languages are represented as points: (Haspelmath et al. 2005). • The maps are probably out-of-date: (Wenker 1888-1923; Brockhaus 1894). • The source is not apparent: (Wikipedia 2007). From these maps we select one according to the following four criteria: 1. The map contains a map grid. So it can be georeferenced without any difficulty. 2. The map should completely cover the German language area of Europe. It should not be restricted to a single country. 3. The digital version of the map must have a useful size and resolution. 4. The borders between language areas should appear very precisely. The only map that fulfils these four criteria is the map of (Brockhaus 1894). Unfortunately this map is more than 100 years old. It is not possible to determine whether the language areas have changed. Thus, the language areas shown in this map have to be compared to contemporary maps such as the language map in the German National Atlas (Bausch 2002). By comparison we found only small differences. Indeed, resettlements happened during and after World War II east of Germany (Ther and Siljak 2001). Therefore, dramatic changes in the language areas may be expected for this region.

3.3 Relations of language versions in Wikipedia For the analysis database-dumps of December 2009 were used (Mediawiki 2010) and seven Wikipedias are under investigation .Table 1 shows the name, and the number of articles of the three largest Wikipedias and the Wikipedias under investigation. In the database-dumps, the IWL are stored in a structure like [[abbr:title]], where "abbr" is the abbreviation of the Wikipedia. For example [[bar:zirich]] links to the article about "Zirich" (Zurich) of the Bavarian Wikipedia. In the dump the IWL of a single article are usually stored one by one, so it is very easy to parse them.

478

T. Dahinden

The fact that IWL exist, shows that there is some redundancy of information about the topics. A rough estimate on how similar the topics are can be made from Table 2. As an example, the French Wikipedia has 278346 articles without an IWL and 540971 articles with a link to the English Wikipedia. Disregarding the "none"-column, then Table 2 should by symmetrical. There are several reasons why this does not hold: First of all, the dumps are not synchronous. Moreover, the IWL point to articles which are redirected to another article or only to sections of articles. This means that more than one IWL can point to a single article. Table 3 shows the percentage of IWL. Tables 1 to 3 give a hint on how the Wikipedias are related by content. For example, the largest Wikipedia (English) has three times the amount of articles as the second largest (German). But only 55.5% of the articles have no IWL. Only about 10% of the articles in Wikipedia have a coordinate. This value can be increased to 18.2% using IWL. Table 1 shows the number of articles with coordinates. Extracting the coordinates from the dumps is rather laborious, because several templates to store coordinates are in use. Luckily the coordinates of all articles of all Wikipedias are accumulated in a gazetteer called Wikipedia-World (Kühn and Alder 2009). There were 649,504 coordinates on August 17, 2009. They are stored together with the titles of the article. Table 4 shows an extract of the gazetteer. Original name

English name

English

English

en

3119385

365185

Deutsch

German

de

1003360

158682

184159 18.4

Français

French

fr

894393

10698

137326 15.4

Nederlands

Dutch

Lëtzebuergesch Luxembourgish

Abbrev. # of articles # of coords 1 # of coords 2 % 498142 16.0

nl

577641

32937

197213 34.1

lb

28200

-

4391 15.6

Ripoarisch

Ripuarian

ksh

10625

-

Alemannisch

Alemannic

als

5541

-

1397 25.2

485

4.6

Limburgs

Limburgian

li

5350

-

1636 30.6

Boarisch

Bavarian

bar

2936

-

803 27.4

Table 1: Wikipedias under investigation, abbreviation, number (#) of articles, number of coordinates originated in this Wikipedia (coords 1), number of coordinates using IWL (coords 2), and percentage of articles with coordinates.

Estimation of the Locations of the Language-Versions of Wikipedia

none en

en

1728398

de

fr

nl

lb

479

ksh

als

li

bar

- 504526 517897 369303 16123 5537 3902 3095 1928

de

398842 515129

fr

278346 540971 312755

- 308162 236740 18649 5733 5000 3140 2423

nl

133643 385221 242588 253278

- 247475 19214 5056 3696 3368 1545 - 18395 4600 3613 4271 1834

lb

5161

17475

19274

20242

18620

ksh

3014

6072

6185

5205

4660

- 3093 1093 1269 3101

-

465

447

614

186

als

375

4125

5017

3811

3529

982

318

-

485

364

li

988

3558

3342

3500

4228

1262

604

493

-

290

bar

304

2267

2582

1735

1922

484

183

378

290

-

Table 2: Number of articles with IWL. (The abbreviations are named in Table 1.) none

en

nl

lb

ksh

als

li bar

- 16.2 16.6 11.8

de

fr

0.5

0.2

0.1

0.1 0.1

en

55.4

de

39.8 51.3

fr

31.1 60.4 35.0

nl

23.1 66.7 42.0 43.8

lb

18.3 62.0 68.3 71.8 66.0

ksh

28.4 57.1 58.2 49.0 43.9 29.2

-

4.2

5.8 1.8

als

6.8 74.4 90.5 68.8 63.7 17.7

5.7

-

8.8 6.6

18.5 66.5 62.5 65.5 79.0 23.6 11.3

9.2

- 5.4

li

bar

10.4

- 30.7 23.6 - 27.7 -

1.9

0.6

0.5

0.3 0.2

2.1

0.6

0.4

0.4 0.2

3.2

0.8

0.6

0.7 0.3

- 11.0

3.9

4.5 1.6

77.2 87.9 59.1 65.5 16.5 6.2

12.9 9.9

-

Table 3: Percentage of articles with IWL. The bold numbers are the maximum of each row, the italic number are the maximum of each column. (The abbreviations are named in Table 1.) Title_en

Title_de

lat

lon

Europe

Europa

46

7

NULL

Milchbucktunnel

47.385

8.537

Dagstuhl Castle

NULL

49.5311

6.8965

Table 4: Extract of the Wikipedia-World Database: Title_en is the title of the article in the English Wikipedia, Title_de is the title of the article in the German Wikipedia, lat = latitude, lon = longitude.

480

T. Dahinden

4- Results

4.1 Density estimation The language areas of seven Wikipedias were estimated using density estimation. In Figure 2 the density of three of them (Ripuarian, Alemannic, and Bavarian) are shown at the levels of 25, 50 and 75%. Further, the major language regions according to (Brockhaus 1894) are included.

Figure 2: Language area of Alemannic (als), Bavarian (bar) and Ripuarian (ksh). Black: 75%, 50%, and 25% of the maximum surface density. The background (grey) shows the countryborders and the borders of the distribution of the languages according to the map of (Brockhaus 1894). Equirectangular projection, 4° E – 20° E, 46° N – 54° N.

In Figure 2 the estimation of Bavarian is distributed all over the language area of Bavarian. In contrast the estimation of Alemannic doesn't cover the eastern part (where Swabian is spoken). Ripuarian is overestimated. The statistics in Figure 3 shows the difference of correct and false located pixels normalized by the map size. The larger the values, the better the estimation of the language area. Values of 1 denote that everything is correctly located. Values of -1 denote everything is wrongly located.

Estimation of the Locations of the Language-Versions of Wikipedia

481

Figure 3: Approximation of language areas: Difference of correctly and falsely detected pixels. The larger the value is, the better the estimation.

The statistics contains the following information: • The value at a level of 0 is the difference between the area where the language is not spoken and the area where the language is spoken. The larger this value, the smaller the language area. • The value at level 1 is the additive inverse of the value at 0. • The value at 0 for German is negative, because more than the half of the map represents German speaking area. • The best estimation for German is at a level of 0.46. At this value the difference between correct and false is about 58% of the map size. • The value for Dutch decrease monotone after a level of 0.07. • The approximation of Alemannic and Bavarian increases up to a level of about 60%. • The rather small areas of Luxembourgish, Ripuarian and Limburgian are estimated correct to more than 90% of the map area up to a level of 57, 59, and 77% respective. The estimation seems to be quite good for the smaller Wikipedias. But the language areas of larger Wikipedias (Dutch and German) are approximated rather badly. Yet, drawing the estimation of German on a map (Fig. 4) shows the main shape of the language area, but there are discrepancies between the optimal level in West and East. This finding is according to the expected negative influence of the language map’s age. Probably it is based on the resettlements mentioned before.

482

T. Dahinden

Figure 4: Estimation of German at density levels of 25, 50, and 75% compared to the German language area according to (Brockhaus 1894) (background). Equirectangular projection, 4° E – 20°E, 46° N – 54° N.

4.2 Difference of densities The density was estimated for the basic population of coordinates, normalized and then subtracted from the normalized density estimation of each language. This leads to bounded density estimation. Figure 5 shows the basic population as well as the difference of densities. The estimations are represented in dark. The advantage of this method is the fact, that the areas are limited. By comparing these maps with the background map of Figure 4, the result can be described qualitative: • The estimation for Luxembourgish, Ripuarian, and Limburgian are reasonable. • Alemannic is underestimated. The eastern part of this region (Swabia) is missing, and also a larger part in Switzerland. May be this is because Alemannic has four different dialects that cannot be mixed easily. People may prefer standard German. • Bavarian is underestimated. The populated areas around Vienna, Munich and Nuremberg are missing. • The language area of German is underestimated even worse. The populated areas: Berlin, Ruhr, Western Lower Saxony, parts of Bavaria, and Austria were not recognized. This could be based in the equal number of coordinates for the estimations: There are only 342 different coordinates while 63565 are similar.

Estimation of the Locations of the Language-Versions of Wikipedia

483

• In contrast Dutch is overestimated. The southern part (French) of Belgium is indicated to be Dutch, and even more surprisingly parts of Slovenia and Tyrol.

Figure 5: Density of the basic population of all coordinates of Wikipedia (all) and difference of the estimated densities of the languages with this density. The dark areas refer to the expected language area.

The differences of densities are classified between 0 and 1 by steps of 0.01. Again, the difference of correct to false pixels was calculated. The result is shown in Figure 6. The statistic contains the following information: • The value at level 0 is the difference between the area where the language is not spoken and the area where the language is spoken.

484

T. Dahinden

• The value at the level of 1 is not the additive inverse to the value at 0. • The difference between correct and false is always larger than 70% of the map area except for German. • The values for German increase notably. • Alemannic and Bavarian slightly increase. • Luxemburgish, Ripuarian, and Limburgian slightly decrease. • Dutch decreases notably.

Figure 6: Approximation of language areas according to the difference of densities method: The larger the value is, the better the estimation.

5- Summary and Outlook on Future Work The objective of the paper was to compare language areas with the distribution of the articles of the Wikipedia that use this language. The assumption of the correlation of these areas was not falsified. Admittedly, there are some negative trends estimating the Dutch language area. But the results for German, Luxembourgish, Ripuarian, Alemannic, Limburgian and Bavarian are very encouraging. There are still some issues that have to be considered: The gazetteer contains "only" about 650,000 coordinates. A lot of coordinates are missing (Wikipedia 2010a) and some others are obviously wrong (Wikipedia 2009b). There are even missing Interwikilinks. The numbers of pages and the structures of Wikipedia are still growing. The current study examined only coordinates associated to Interwikilinks of December 2009. The following should be considered in a next study:

Estimation of the Locations of the Language-Versions of Wikipedia

485

• To determine whether it is also possible to find articles with coordinates but without Interwikilinks, when using larger Wikipedias such as German and Dutch. • The weighting of the coordinate could be improved using e.g. the length of the articles. • Wikipedia contains templates which can be analyzed, e.g. the German Wikipedia can easily be used to display who is born in London (Alder 2009). It could be assumed that the birth place is related to the language people are speaking. The bandwidth of the kernel density estimation was estimated by a rough rule-of-thumb. There are more sophisticated methods, for instance variable kernel density estimation. Additional information could improve the estimation of the language area. They would also be useful when showing geographic information, e.g. as tag-clouds (Paelke et al. 2010). The following research using Wikipedia could be useful: Detection of relations between languages, e.g. Alemannic is a German dialect. It can be derived from Table 2 that the Interwikilinks represent rather geographic than linguistic closeness. But, related languages haven't got to be close in space, e.g. Pennsylvanian German is surrounded by American English. Therefore, the spatial closeness should be considered as a negative factor for the detection of these relations.

References Alder T (2009) WikiProjekt Vorlagenauswertung.

Accessed 4 May 2009, from

http://de.wikipedia.org/wiki/Wikipedia:WikiProjekt\_Vorlagenauswertung.

Bausch K-H (2002) Die deutsche Sprache - eine Dialektlandschaft. Bildung und Kultur. In: (eds) Bildung und Kultur. Leibniz-Institut für Länderkunde: 94-95. Bragues G (2009) Wiki-philosophizing in a marketplace of ideas: Evaluating Wikipedia's entries on seven great minds. MediaTropes eJournal 2(1): 117-158. Brockhaus FA (1894). Karte der deutschen Mundarten, Geogr.-artist. Anstalt, Leipzip. 14. Auflage. Clauson KA, Polen HH, Boulos MNK, Dzenowagis JH (2008) Scope, completeness, and accuracy of drug information in Wikipedia. Ann Pharmacother 42(12): 1814-1821. Dahinden T (2009) Localization of uncertain and fuzzy-bordered areas by geocoded articles of a knowledge repository. In: Garcia-Huidobro CJV (eds) Proceedings of the International Cartographic Conference, Santiago de Chile. Dahinden T, Sester M (2009) Categorization of Linear Objects for Map Generalization Using Geocoded Articles of a Knowledge Repository. In: Winter S, Tenbrink T (eds) Workshop on

486

T. Dahinden

Presenting spatial information: Granularity, relevance and integration, Conference on Spatial Information Theory, Aber Wrac'h, France. Dolan S (2008) Six degrees of Wikipedia. Accessed 29 May 2008, from http://www.netsoc.tcd.ie/~mu/wiki/. ESRI Support Centre (2010) ArcGIS Desktop Help: Kernel Density. Accessed 13 April 2010, from http://webhelp.esri.com/arcgisdesktop/9.3/index.cfm?id=6190&pid=6188&topicname=Kernel_Den sity.

Fagan SMB (2009) German: a lingustic introduction, Cambridge University Press. Halavais A, Lackaff D (2008) An Analysis of Topical Coverage of Wikipedia. Jounral of Computer-Mediated Communication 13(2): 429-440. Haspelmath M, Dryer MS, Gil D, Comrie B, Eds. (2005) World Atlas of Language Structures, Oxford University Press. Hecht B, Moxley E (2009) Terabytes of Tobler: Evaluating the First Law in a Massive, DomainNeutral Representation of World Knowledge. Spatial Information Theory, 9th International Conference, COSIT 2009, Aber Wrac'h, France, September 21-25, 2009, Proceedings 5756: 88-105. Hecht B, Raubal M (2008). GeoSR: Geographically Explore Semantic Relations in World Knowledge. AGILE Conf., Springer: 95-113. Kittur A, Chi E, Pendleton BA, Suh B, Mytkowicz T (2007) Power of the Few vs. Wisdom of the Crowd: Wikipedia and the Rise of the Bourgeoisie. In: (eds) 25th Annual ACM Conference on Human Factors in Computing Systems (CHI), San Jose, CA. König W, Paul H-J (2007) dtv-Atlas Deutsche Sprache. Mit 155 farbigen Abbildungsseiten., Deutscher Taschenbuch Verlag. Kühn S, Alder T (2009) WP : GEO / Wikipedia-World. Accessed 31 August 2009, from http://de.wikipedia.org/wiki/Wikipedia:WikiProjekt_Georeferenzierung/Wikipedia-World. Maurmann E (1888-1923) Karte der Deutschen Mundarten. Sprachatlas des Deutschen Reichs. Mediawiki (2010) Wikimedia downloads. Accessed 11 June 2010, from http://download.wikimedia.org/. Ortega F, Gonzales-Barahona JM, Robles G (2007) The Top-Ten Wikipedias, A Quantitative Analysis using WikiXRay. International Conference on Software and Data Technology: 4653. Paelke V, Dahinden T, Eggert D, Mondzech J (2010) Location Based Context Awareness Through Tag-Cloud Visualization. In: (eds) Joint International Conference on Theory, Data Handling and Modelling in GeoSpatial Information Science, Hong Kong. Pfeil U, Zaphiris P, Ang CS (2006) Cultural differences in collaborative authoring of Wikipedia. Journal of Computer-Mediated Communication 12(1). Protze H (1969) Die deutschen Mundarten. Kleine Enzyklopädie Deutsche Sprache. Samuels ML (1972) Linguistic evolution : with special reference to English. London, Cambridge University Press. Schöning J, Hecht B, Rohs M, Starosielski N (2007) WikEar: Automatically Generated Location-Based Audio Stories between Public City Maps. Adjunct Proceedings of the 9th International Conference on Ubiquitous Computing. Silverman BW (1986) Density estimation for statistics and data analysis. London, Chapman and Hall (CRC Press). Smith MJd, Goodchild MF, Longley PA (2007) Geospatial Analysis, Second Edition, Troubador Publishing Ltd. Ther P, Siljak A (2001) Redrawing Nations: Ethnic Cleaning in East-Central Europe, 19441948., Rowman & Littlefield. Tobler WR (1970) A Computer Movie Simulating Urban Growth in the Detroit Region. Economic Geography 46: 234-240. Wenker G (1888-1923) Sprachatlas des Deutschen Reichs. Marburg.

Estimation of the Locations of the Language-Versions of Wikipedia

487

Wikipedia (2007) Die deutschen und niederländischen Dialekte nach dem Jahr 1945. Accessed 28 April 2007, from http://de.wikipedia.org/w/index.php?title=Datei:Deutsche_Dialekte.PNG. Wikipedia (2009a) Translation. Accessed 27 December 2009, from http://en.wikipedia.org/wiki/Wikipedia:Translation. Wikipedia (2009b) WP : GEO / mehrere Artikel an einer Koordinate. Accessed 31 August 2009, from http://de.wikipedia.org/wiki/Wikipedia:WikiProjekt_Georeferenzierung/mehrere_Artikel_an_einer _Koordinate.

Wikipedia (2010a) Articles needing coordinates.

Accessed 3 January 2010, from

http://en.wikipedia.org/wiki/Category:Articles\_needing\_coordinates.

Wikipedia

(2010b)

List

of

Wikipedias.

http://en.wikipedia.org/wiki/List_of_Wikipedias.

Accessed

6

January

2010,

from