TABLE OF CONTENTS
i
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA Table of Contents
Climate change and carbon dioxide: An introduction Charles D. Keeling
8273–8274
Tribute to Roger Revelle and his contribution to studies of carbon dioxide and climate change Walter H. Munk
8275–8279
Equilibration of the terrestrial water, nitrogen, and carbon cycles David S. Schimel, B. H. Braswell, and W. J. Parton
8280–8283
Potential responses of soil organic carbon to global environmental change Susan E. Trumbore
8284–8291
Global air-sea flux of CO2: An estimate based on measurements of sea–air pCO2 difference Taro Takahashi, Richard A. Feely, Ray F. Weiss, Rik H. Wanninkhof, David W. Chipman, Stewart C. Sutherland, and Timothy T. Takahashi
8292–8299
Characteristics of the deep ocean carbon system during the past 150,000 years: ΣCO2 distributions, deep water flow patterns, and abrupt climate change Edward A. Boyle
8300–8307
Direct observation of the oceanic CO2 increase revisited Peter G. Brewer, Catherine Goyet, and Gernot Friederich
8308–8313
The observed global warming record: What does it tell us? T. M. L. Wigley, P. D. Jones, and S. C. B. Raper
8314–8320
Possible forcing of global temperature by the oceanic tides Charles D. Keeling and Timothy P. Whorf
8321–8328
Spectrum of 100-kyr glacial cycle: Orbital inclination, not eccentricity Richard A. Muller and Gordon J. MacDonald
8329–8334
Can increasing carbon dioxide cause climate change? Richard S. Lindzen
8335–8342
Gases in ice cores Michael Bender, Todd Sowers, and Edward Brook
8343–8349
TABLE OF CONTENTS
ii
Tree rings, carbon dioxide, and climatic change Gordon C. Jacoby and Rosanne D. D'Arrigo
8350–8353
Geochemistry of corals: Proxies of past ocean chemistry, ocean circulation, and climate Ellen R. M. Druffel
8354–8361
A long marine history of carbon cycle modulation by orbital-climatic changes Timothy D. Herbert
8362–8369
Dependence of global temperatures on atmospheric CO2 and solar irradiance David J. Thomson
8370–8377
TABLE OF CONTENTS
iii
TABLE OF CONTENTS
iv
CLIMATE CHANGE AND CARBON DIOXIDE: AN INTRODUCTION
8273
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8273–8274, August 1997 Colloquium Paper This paper serves as an introduction to the following papers, which were presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held November 13–15, 1995, at the National Academy of Sciences, Irvine, CA.
Climate change and carbon dioxide: An introduction
CHARLES D. KEELING Scripps Institution of Oceanography, 10666 North Torrey Pines Road, La Jolla, CA 92037
© 1997 by The National Academy of Sciences 0027-8424/97/948273-2$2.00/0 PNAS is available online at http://www.pnas.org. Since the dawn of history, human beings have had the ability, superior to all other living beings, to exploit the Earth's environment to their own immediate advantage. For most of human history the consequences were only of local or regional significance. Over the past century, however, the rapid rise in population and the increasing intensity and scale of human enterprises have made it possible for humans to alter the Earth on a global scale. One important measure of human activity is the rate of utilization of energy. This rate has accelerated strikingly in the past hundred years because of rapidly increasing human population coupled with increasing per capita energy consumption. Is it possible that accelerating human activity has already caused globally significant environmental change, or is about to do so? One aspect of this question relates to possible human alteration of the Earth's climate, which is essentially the summation of weather and its variability. Although climate clearly varies with latitude and elevation and with physical and ecological features, such as deserts and forests, it once was considered to be constant over time. We now know, however, that weather does vary on long time scales and, therefore, that climate is variable. Climate has indeed varied profoundly, as evidenced by proxy records indicating a succession of ice ages and warm “interglacial” eras over the past million years. Proxy records also reveal climatic variability on time scales of hundreds to thousands of years. Long-term weather records even show evidence of significant variability over decades, which may be associated with climatic change. This short-term variability makes it difficult to separate out subtle climate changes that might be caused by accelerating human activities. Short-term climatic change was discussed recently in a National Research Council (NRC) workshop: Natural Climate Variability on Decade-toCentury Time Scales (1). By comparing past climatic conditions with recent ones, it was not clear whether human activities have altered climate or not. Better data and a better understanding of the causes of climatic variability are needed to decide this. Broadly speaking, climatic change is caused by exchanges of energy, momentum, and chemicals between the atmosphere, the oceans, and land surfaces. Oceanic and atmospheric circulation, turbulent mixing, photochemistry, and radiative transfer are all involved. These processes are mainly natural, but some, at least, are susceptible to human influence. Processes that involve the so-called greenhouse gases are probably the most critical candidates. These greenhouse gases, mainly carbon dioxide but including others such as methane, nitrous oxide, and halocarbons, enter the air mainly as byproducts of the combustion of coal, natural gas, and petroleum, and to a lesser degree through other industrial and agricultural activities. Their rates of emission into the air are roughly proportional to the global rate of energy consumption arising from human activity. Thus, as human population and per capita energy consumption have increased, concentrations of these gases have risen in nearly direct proportions to the product of both increases. As they build up, these gases trap radiation upwelling from the Earth's surface. The expected consequence is rising temperature at the Earth's surface unless some compensating process cancels out this tendency. Whether such compensation is occurring is presently a matter of debate. Carbon dioxide deserves attention as a greenhouse gas because it is indisputably rising in concentration. To understand what controls its abundance in the atmosphere, and hence its influence on the greenhouse effect, we must address all the processes that affect, and are affected by, its concentration in the atmosphere. These processes include its interactions with the chemically buffered carbonate system in seawater and with vegetation because of its vital role in photosynthesis. The sum of all processes affecting carbon on the Earth, and hence controlling the concentration of atmospheric carbon dioxide, is called the “carbon cycle.” We need to understand how the carbon cycle functions in order to know how human activities may affect carbon dioxide. Although the pathways of carbon through the global carbon cycle are understood in general, knowledge of the actual rates of change of the fluxes between the atmosphere, land, and ocean is less advanced. The annual anthropogenic carbon input to the atmosphere between 1980 and 1989 has been estimated (2) to include 5.5 ± 0.5 GtC (thousand million metric tons of carbon) from fossil fuel combustion and 1.6 ± 0.6 GtC from land-use change, yielding a total of 7.1 ± 1.1 GtC. Of this annual input, 3.3 ± 0.2 GtC remained in the atmosphere, and 3.8 GtC were removed. Oceanic uptake, related to carbonate buffering, is thought to account annually for about half of the removal. Regrowth of northern hemisphere forests has been estimated to account for perhaps 0.5 ± 0.5 GtC. The removal mechanisms of the remaining carbon, 1.3 ± 1.5 GtC per year, are uncertain. This residual term is commonly referred to as the “missing carbon sink.” It must be located and the uncertainty in the other individual terms in the global carbon cycle must be reduced if the extent of human impact on the carbon cycle is to be assessed reliably. Furthermore, a feedback mechanism exists whereby climate change may itself alter the carbon cycle. For example, widespread warming from increasing greenhouse gases may change the rates of uptake of carbon dioxide by the oceans globally and may alter gas exchange with vegetation. Even less is known about such feedback mechanisms than is known about the missing carbon sinks. To review progress in our understanding of the carbon cycle and climate, the National Academy of Sciences (NAS) supported the colloquium summarized in this volume. In planning for it, special attention was given to highlighting a portion of
CLIMATE CHANGE AND CARBON DIOXIDE: AN INTRODUCTION
8274
the illustrious career of the late Roger Revelle, a long-time NAS member and contributor to many NRC activities. The paper by Walter Munk that follows presents a detailed description of Revelle's career. My contacts with Roger Revelle, although less intimate than Walter Munk's, spanned nearly four decades, and left on me an indelible impression of one of the great figures in post-World War II science. Roger believed that science should be not only useful, but also enjoyable; that scientists should be held in high regard and should be allowed to follow their own leads in the quest for scientific knowledge with as much freedom as possible. Roger began his career as a chemist studying carbon in the oceans. During the years that I knew him he maintained an interest in the global carbon cycle, while making highly significant contributions in several other fields. It has been a great pleasure for me to have taken part in this colloquium on a subject in which Roger took keen interest throughout his career. He would have enjoyed attending, and we would have benefited from his wisdom as many of us did during his lifetime. With only our memories of him we have nevertheless tried to live up to his standards by steadfastly addressing important topics in a manner both useful and enjoyable. Fifteen refereed articles are included in this volume. Some of these papers review previous studies, while others present new data and analyses. All address topics in which Roger was keenly interested: (i) the extent to which climate is changing owing to both natural causes and human activities, (ii) whether these changes, in part, are long-term manifestations of increasing carbon dioxide, and (iii) how the oceans, terrestrial plants and soils, and atmosphere function in general as a necessary foundation for exploring the first two topics. The spirit of this offering is to advance knowledge so that all people will have a rational basis for dealing with environmental problems, especially those that mankind may have created. This mission is consistent with Revelle's optimistic belief that the human race, given the opportunity through enlightenment, will naturally serve its own best interests, and that people able to contribute to this enlightenment will do so zealously and unselfishly, as Roger did. Special thanks are owed to the committee that assisted me in planning this colloquium and in handling the review process: Peter Brewer, Ellen Druffel, Edward Frieman, Robert Knox, Walter Munk, Taro Takahashi, and Karl Turekian. In addition to funding from the NAS, the colloquium was supported by five federal agencies: the National Science Foundation, the National Oceanic and Atmospheric Administration, the Office of Naval Research, the U.S. Department of Energy, and the National Aeronautics and Space Administration. The planning committee members are grateful for this support and to Neil Andersen, formerly of the National Science Foundation, who played a key coordinating role. We also thank the National Research Council's Ocean Studies Board (OSB) and its staff, who helped us to make this colloquium a success, especially Ed Urban of the NRC for assistance in very many aspects of the preparation for the meeting and this volume of papers. Roger Revelle's significant positive influence on the NRC and OSB over many years was demonstrated by the NRC staff's enthusiasm for and dedication to this enterprise. 1. National Research Council (1996) Natural Climate Variability on Decade-to-Century Time Scales (National Academy Press, Washington, DC). 2. Intergovernmental Panel on Climate Change (1996) in Climate Change 1995: The Science of Climate Change, eds. Houghton, J. T., Meira Filho, L. G., Callander, B. A., Harris, N., Kattenberg, A. & Maskell, K. (Cambridge Univ. Press, New York), p. 17.
TRIBUTE TO ROGER REVELLE AND HIS CONTRIBUTION TO STUDIES OF CARBON DIOXIDE AND CLIMATE CHANGE
8275
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8275–8279, August 1997 Colloquium Paper This paper was presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held November 13–15, 1995, at the National Academy of Sciences, Irvine, CA.
Tribute to Roger Revelle and his contribution to studies of carbon dioxide and climate change WALTER H. MUNK Institute of Geophysics and Planetary Physics, University of California at San Diego, La Jolla, CA 92093-0225
© 1997 by The National Academy of Sciences 0027-8424/97/948275-5$2.00/0 PNAS is available online at http://www.pnas.org. I first came to Scripps Institution of Oceanography (SIO) in the summer of 1939, after completing my junior year at the California Institute of Technology. Roger Revelle was 30 years old, with the rank of instructor (long since abolished by the University of California), and a lieutenant junior grade in the Naval Reserve. Roger invited me to come along on an experiment to measure currents in the waters over the California borderland. The standard tool was an Ekman Current Meter; for every 100 revolutions of a propeller, a 2-mm ball is dropped into a compass box with 36 compartments, each corresponding to a 10° segment in current direction. The trouble was that the balls would not fall into the compartments. Roger was up all night doggedly fussing with the current meter until, at breakfast time, the release was functioning. This is my earliest memory of Roger. Scripps and the War Years After taking up geology at Pomona College under the legendary teacher Alfred “Woody” Woodford, followed by a graduate year at University of California at Berkeley, Roger came to Scripps in 1931 to study deep-sea muds. By 1936 he had completed his thesis, “Marine Bottom Samples Collected in the Pacific Ocean by the Carnegie on Its Seventh Cruise,” and stayed on as an instructor (Fig. 1). During his year at Berkeley, Roger married Ellen Clark, a grandniece of E. W. Scripps and Ellen B. Scripps, after whom the Scripps Institution of Oceanography was named. After a year at the Geophysical Institute in Bergen, Norway, he returned to Scripps. Roger went on active naval duty 6 months before the bombing of Pearl Harbor and stayed in the Navy for 7 years. He was instrumental in organizing the Office of Naval Research. In 1946 he was officer in charge during Operation Crossroads of the geophysical measurements taken during the atomic bomb tests at Bikini Atoll. None of the participants will ever forget this experience. For many years, Roger contributed to the understanding of the environmental effects of radiation and to questions of disposal of atomic wastes at sea (1). [Revelle contributed to the report in ref. 1 as Chairman of the National Academy of Sciences Panel on Biological Effects of Atomic Radiation (BEAR).] I suspect that Roger's participation for so many years, from 1958 to 1981, in the Pugwash Disarmament Conferences can be traced to the Bikini bomb tests. The Scripps Directorship After more than 40 years as a local marine station, Scripps Institution had agreed to undertake a program to study the disappearance of sardines from California waters (Fig. 2). This involved the commissioning of two vessels. Scripps Director Harald Sverdrup was anxious for Roger to return to La Jolla to succeed him as director of Scripps. Sverdrup (2) wrote, “regardless of the capacity in which you return here, you are the logical man to take charge . . . of the work at sea.” And Roger (3) agreed: Sverdrup's support for me as successor is also based upon the fact that I am practically the only person available who has had extensive experience at sea, in particular in the organization and carrying out of expeditions. He feels that Scripps must be, at least in part, re-oriented toward work on the high seas rather than the inshore and laboratory type of research which is being largely done at present. Sverdrup's statement “regardless of the capacity in which you return” was a reference to a developing opposition to Roger as the next director. One Scripps professor complained that Roger was too untidy to be trusted with administration (3), noting that he “just let everything pile up on his desk” and “was to easily diverted.” Again Roger agreed (3), referring to his own “obvious and numerous weaknesses, such as a tendency to procrastinate, to take on too many obligations, not to delegate authority, to be high-handed.” The outcome was that Carl Eckart was appointed director of Scripps in 1948, and Roger was appointed associate director with the expectation to succeed Eckart in a few years. It wasn't that easy! A 1950 letter to University of California President Robert Sproul (4), signed by more than half the Scripps faculty, states: We understand that the impression has been gained in some quarters that opposition is vanishing at Scripps Institution to Dr. Revelle as a candidate for Director. We assure you that whereas we have a high regard and friendship for him, we feel as strongly as before that his appointment . . . would not be in the interest of the institution. His recent administrative actions confirmed our conviction. Roger was appointed director in 1950. It is a tribute to Roger's disdain for pettiness that some years later one of the writers referred to Roger's “brilliant Directorship” (13). The Heady Expedition Days The era opened in 1950 with the Mid-Pacific expedition into the equatorial waters of the central Pacific Ocean. This was followed in 1952–1953 by an extended voyage to the South Pacific, which was called Capricorn. Both expeditions were led personally by Roger. It was discovered that only a thin veneer of sediments overlies the solid rock, that the heat flow through the sea floor is about the same as that on land, and that the flat-topped seamounts at a depth of 2,000 m had been volcanic islands less than 100 million years ago. All of this spoke for great mobility of the “solid” Earth. When Roger and his
TRIBUTE TO ROGER REVELLE AND HIS CONTRIBUTION TO STUDIES OF CARBON DIOXIDE AND CLIMATE CHANGE
8276
colleagues tried to core and dredge the Tonga Trench, the instruments came up battered and bent, and empty. If there were any sediments, they were sparse and thin. The observations could best be explained if the rocky sea floor was disappearing into the Earth along the axis of the trench (this is now called subduction). On Capricorn, Ronald Mason towed a magnetometer behind the vessel and recorded a complicated set of wiggles that no one could understand. Later Mason produced a map of the magnetic field under the sea floor showing stripes of normal and reverse magnetization.
FIG. 1. Roger Revelle as an instructor at Scripps (circa 1936). Photo by Eugene LaFond. [Reproduced with the permission of the SIO Archives, UCSD.] With hindsight, the evidence was all there for proclaiming the doctrine of plate tectonics. And when, 10 years later, the puzzle was put together, Scripps unfortunately did not play a leading role. Still, I think of the 1950s as the great era of the Institution. When Roger left in 1961, Scripps had a Navy bigger than that of Costa Rica. Greenhouse Even as he led the exploration of the Pacific, Roger was active for several years in promoting the International Geophysical Year (IGY). In 1956 he became chairman of the IGY Panel on Oceanography. That same year, Charles David Keeling joined the Scripps Institution staff to head the IGY program on Atmospheric Carbon Dioxide and to start the measurements at Mauna Loa and Antarctica. And that is why we are here 40 years later. Keeling credits Harry Wexler and Roger Revelle for insisting on the continuity of the measurements; such time series are few and far between and worth their weight in gold. In 1957, Roger and Hans Suess demonstrated that carbon dioxide had increased in the air as a result of the consumption of fossil fuels, in a famous article published in Tellus (5). Roger's interest in CO2 was to engage his attention for the rest of his life. In 1965, the President's Science Advisory Committee Panel on Environmental Pollution under Roger's leadership published the first authoritative report that recognized CO2 from fossil fuels as a potential global problem (6). Public opinion was influenced through a widely read article in Scientific American (7). Roger participated in the exploration of the atmospheric greenhouse problem from the 1950s, when it was a cottage industry for a few academics, to the 1990s, when global climate change involved industry and government on an international scale. He once estimated that he had spent 20% of his time keeping current with the issues. THE MOHOLE PROJECT In 1957 Roger and I were among a group that called themselves the American Miscellaneous Society (AMSOC). AMSOC promoted an attempt to drill through the ocean floor into the Earth's mantle. A test off Guadalupe Island successfully drilled through 200 m of sediments into the basalt in water 4,000 m deep, demonstrating the feasibility of “dynamic positioning.” This MOHOLE project (Fig. 3) eventually failed because of poor Washington management but led some years later to the successful Ocean Drilling Program. Ocean Leadership The U.S. ocean program was then firmly in the hands of three men: Maurice Ewing, Columbus Iselin, and Roger Revelle. There has not been a comparable ocean leadership since those days.* While Revelle served as a founding member of the National Academy of Sciences Committee on Oceanography (NASCO), the funds budgeted nationally for oceanography rose from $12 million in 1957 to $97 million in 1960. Roger played a major role in organizing the IGY and in forming the Scientific Committee for Ocean Research (SCOR) and the International Oceanographic Commission (IOC), and then served as Chairman of a joint IOC/SCOR Committee on Climate Changes and the Ocean. These organizations continue to play an important role in international oceanography. Roger enjoyed an international reputation as oceanographer in the 1950s but became better known to the greater scientific community and to the public through his work for the National Academy of Sciences as a science spokesman with broad knowledge of the environment. He worked very hard behind the scenes to frame the important scientific questions and then to secure the resources to answer them. Policy makers looked to him for a reasonable assessment of which scientific problems should take priority. Scientists sought his advice and support to focus research and get it funded. Congressman Emilio Daddario (8) has remarked on Roger's “combined experience, intelligence and good judgment about issues.” Building the University of California at San Diego (UCSD), 1954–1961 In parallel with these developments came the beginnings of the UCSD. No oceanographic program, Roger said, could maintain intellectual excellence for more than a generation without an attachment to a great university. The obvious site was some 1,100 acres of largely undeveloped public land just to the north of the Scripps Institution of Oceanography. Fortuitously, Roger's initiative coincided with a new master plan for the
*This may have changed; in the last several years, Admiral (U.S. Navy, ret.) James Watkins has become a recognized national spokesman for ocean affairs.
TRIBUTE TO ROGER REVELLE AND HIS CONTRIBUTION TO STUDIES OF CARBON DIOXIDE AND CLIMATE CHANGE
8277
University of California, which foresaw the need to establish two new campuses in southern California.
FIG. 2. Gulf of California Expedition, 1939. (Left to Right) Erik Moberg, Roger Revelle, Seaman Andrew Boffinger, Richard Fleming (with binoculars), Machinist Bob MacDonald, George Hale, Lee Haines, Engineer Walter Robinson, Martin Johnson, and Loye H. Miller. [Reproduced with the permission of the SIO Archives, UCSD.] Roger had in mind a major university in the manner of The John Hopkins University or the University of Chicago, with a heavy concentration of graduate students. The plan ran into opposition by a 1956 University of California at Los Angeles (UCLA) review committee, which proposed that UCSD should be permitted to offer only lower division undergraduate courses at first, and only after a later review to add upper division courses, but not a graduate program. We pointed out that Scripps had been granting Ph.D. degrees when UCLA was still a teacher's college.
FIG. 3. MOHOLE project, aboard the CUSS I off Guadalupe Island in 1961. (Left to Right) John Steinbeck, Josh Tracey, Unidentified, William Riedel, Roger Revelle, Walter Munk, Gustav Arrhenius, and Willard Bascom, examining specimen. Photograph by Fritz Goro, Life Magazine (© Time Inc.).
TRIBUTE TO ROGER REVELLE AND HIS CONTRIBUTION TO STUDIES OF CARBON DIOXIDE AND CLIMATE CHANGE
8278
Roger (9) roasted them for trying to forestall competition: The Committees on Educational Policy . . . were experts at seeing clouds no bigger than a man's hand. It was clear to them that a new graduate school would draw money away from their own campuses; it might even attract outstanding scientists. . . . They thought it would be nice to have an undergraduate school at La Jolla, managed by a farm team of dedicated teachers, which could provide well-trained new graduate students for their own laboratories. Roger put in an enormous effort in recruiting faculty for his new school, originally housed on the Scripps Campus; among the recruits were Harold Urey, Joseph and Maria Mayer, Jim Arnold, David Bonner, Walter Elsasser, Martin Kamen, Bernd Matthias, and Bruno Zimm. The secret of Roger's recruiting success was not a secret at all. He put in a major effort to learn what these people really wanted to do and then went out to provide the opportunity for them to live their dream. But he was not unwilling to exercise some salesmanship. Judith and I were on the recruiting route, and Roger would end up at our home with an exhausted candidate in tow at cocktail time, saying, “And here is a typical faculty house.” There was at the time an unspoken covenant among La Jolla Realtors not to sell to Jews. Roger went out to break this covenant, realizing that it was incompatible with his vision of the university. He was somewhat aided by the fact that most members of the real estate community were not fundamentally opposed to enhancing their commissions. The chairman of the University of California Board of Regents at the time was the oil magnate Edwin Pauley. Pauley wanted the campus in Balboa Park in downtown San Diego. Roger wanted it next to Scripps. Regent Pauley had commissioned a study that concluded that the aircraft noise associated with Miramar Naval Air Station would make the site 20% more expensive than that at Balboa Park. Roger had gotten hold of a previous report by that same architect dealing with the location of Scripps Hospital (under the same flight pattern and even closer to the Air Station) which concluded that the noise would not appreciably increase the cost! Roger won that battle, but as it happened to King Pyrrhus of Epirus, he had won one too many. In 1961, when it came time to appoint the first chancellor, the Regents selected Herbert York. It was a major blow to Roger, reminiscent of the long delays in his appointment as Scripps Director. The Exile Roger determined that his continued presence on the campus would make it very difficult for Herbert York to function effectively, so he left for what turned out to be a 14-year exile. President Kerr appointed Roger to the meaningless position as Dean of Research for the University of California statewide. Roger next became Science Advisor to Secretary of Interior Morris Udall and then was appointed Richard Saltonstall Professor of Population Policy at Harvard University, a chair he held for just over a decade. Among his students was Benazir Bhutto and Albert Gore. Gore credited Roger with having aroused his interest in environmental problems. Many years later, during the 1992 presidential campaign, Gore was accused of having misrepresented Roger's position on global warming. The problem arose in connection with an article first published in the Cosmos Club Journal, “What to Do About Greenhouse Warming: Look Before You Leap.” The cautionary admonition “look before you leap” is uncharacteristically tame for Roger, and it is my contention that it represented more the views of the other authors, Fred Singer and Chauncey Starr. I cannot do justice to the many accomplishments during the exile era, but I need to mention one activity which goes back a long time: Roger's continuing love affair with India. Roger took many trips to India and served as an advisor to various government agencies on a broad range of topics, centered on food and population problems. I was amazed on a recent trip to learn of how many Indian careers and lives had been influenced by Roger.
FIG. 4. Revelle explaining core sample to roughneck aboard CUSS I in 1961. This is my favorite photograph of Roger; it shows his total attention to the person with whom he is speaking at the moment. Photograph by Fritz Goro, Life Magazine (© Time Inc.). Roger and Ellen thought of their Harvard years as some of the most fulfilling of their lives. Calling it an “exile” reflects my own provincial Scripps perspective. Coming Home In 1975 Roger returned to UCSD to become Professor of Science and Public Policy. For the next 15 years he taught courses in marine policy and population, and he continued to be active in oceanographic affairs. When in 1978 the American Association for the Advancement of Science (AAAS) decided to focus its international efforts on a few selected issues, Roger chaired the AAAS group that identified the build-up of heat-absorbing gases in the atmosphere as one such issue. As a result, the AAAS Board created the Committee on Climate, and Roger served as its chairman for a decade. The Committee was responsible for the first effort to identify the costs and benefits of increased atmospheric carbon dioxide. He received the National Medal of Science from President George Bush in 1991 for his pioneering work in the areas of carbon dioxide and climate modifications, oceanographic exploration presaging plate tectonics, and the biological effects of radiation in the marine environment, and studies of population growth and global food supplies. To a reporter asking why he got the medal, Roger (10) said, “I got it for being the grandfather of the greenhouse effect.” It is difficult to do justice to a man with such broad accomplishments. When questioned about his profession, Roger would reply “I am an oceanographer.” But this was hardly restrictive because he defined the profession of oceanography as whatever anyone at Scripps does. This has saved me on one occasion. During one of the chronic Scripps space
TRIBUTE TO ROGER REVELLE AND HIS CONTRIBUTION TO STUDIES OF CARBON DIOXIDE AND CLIMATE CHANGE
8279
shortages, Director William Nierenberg surprised Judith and me in a SIO laboratory as we were using laser pulses to remove encrustation's from a renaissance statue. “What are you doing?” Bill asked. “I am doing oceanography,” I replied.
FIG. 5. Roger in 1976. Photo by Glasheen. [Negative AN24/41906/484/34, UCSD Special Collections.] A Personal Appreciation Roger and I have collaborated on a number of papers: on a global compilation of the seasonal change in sea level, on an attempt to infer the melting of the Greenland ice cap from the slowing of the Earth's rate of rotation and the motion of the pole toward Greenland, and on a 1977 National Academy of Sciences report (11) in which we traced the partition CO2 among the atmosphere, ocean, and biosphere. Roger's way of working was anything but analytical; rather he followed a Sherlock Holmes procedure of eliminating one hypothesis after another. In doing his sums, he showed an accountant's revulsion for dropping nonsignificant digits. But my thoughts of Roger are not particularly related to these joint publications. He was my friend for 50 years. I remember weekends in the Revelle cottage in Julian, and sailing in the Aegean. I remember all-night sessions of Roger and Harry Hess at the Cosmos Club. I remember 9 months in the South Pacific, with a luncheon hosted by the Crown Prince, now King, of Tonga. I remember sleepless nights with Roger and John Steinbeck on the drilling ship CUSS I in Mexican waters prior to the demise of the MOHOLE Project. Toward the end of his life, Roger's health deteriorated; he walked in pain and with some difficulty. One year before his death, I was visiting John Knauss, then Administrator of The National Oceanographic and Atmospheric Administration (NOAA), to seek help for the Heard Island Expedition, when Roger unexpectedly showed up. He had walked the endless corridors of the commerce building to lend his silent support. During the expedition, when all the equipment was demolished in a gale on station in the South Indian Ocean, Roger sent a soothing message by fax: “Wish I were with you; and then I am glad I'm not.” Roger was upset by a critical news article on the Heard Island Expedition published in Science (14) and wrote a letter to the editor starting with the words: “Shame on you” (15). It was to be the last thing Roger published (Fig. 5). In an obituary for the Independent of London (12), the oceanographer Henry Charnock spoke for many of us when he noted that, “[f] or an informed view on earth science, and on its repercussions on the human predicament, he was in a class of his own.” Deborah Day at Scripps Archives is responsible for much of this material. 1. Revelle, R. (1956) The Biological Effects of Atomic Radiation: A Report to the Public (National Academy of Sciences, Washington, DC). 2. Sverdrup, H., Director of Scripps Institution of Oceanography. Letter to Comdr. Roger Revelle, Cosmos Club, Washington, DC, September 25, 1947. Roger Revelle Papers (MC 6), Box 2, Folder 10. SIO Archives, UCSD. 3. Revelle, R. Letter to Dean M. P. O'Brien, University of California, Berkeley, November 7, 1947. Roger Revelle Papers (MC 6), Box 2, Folder 10. SIO Archives, UCSD. 4. Fox, D., Hubbs, C., McEwen, G., Shepard, F. & ZoBell, C. Letter to Robert Sproul, President of the University of California, Office of the President, Berkeley, April 12, 1950. S. V. “Scripps Institution of Oceanography. Part I: Directorship 1947–50.” Bancroft Library, University Archives, University of California, Berkeley. 5. Revelle, R. & Suess, H. E. (1957) Tellus 9, 18–27. 6. Revelle, R., Broecker, W., Craig, H., Keeling, C. D. & Smagorinsky, J. (1965) Restoring the Quality of our Environment: Report of the Environmental Pollution Panel, President's Science Advisory Committee (The White House, Washington, DC), pp. 111–133. 7. Revelle, R. (1982) Sci. Am. 247, 35–43. 8. Daddario, E. Q. “The Revelle Impact,” Transcription of a speech delivered at the Scripps Institution of Oceanography, March 10, 1984, p. 3. Accession 84–14. SIO Archives, UCSD. 9. Revelle, R. “On Starting a University,” Manuscript prepared but not published by Daedalus, 1974, p. 3. Roger Revelle Papers (MC6A), Box 158, Folder 19. SIO Archives, UCSD. 10. Lister, P., “Revelle Awarded National Medal of Science ‘90,” San Diego Daily Transcript, June 27, 1990, p. 1A. 11. Revelle, R. & Munk, W. H. (1977) Energy and Climate (National Academy of Sciences, Washington, DC), pp. 140–158. 12. Charnock, H., “Professor Roger Revelle,” The Independent (London), August 5, 1991. 13. Day, D. “Memorandum of Conversation with Dr. Francis P. Shepard, July 27, 1981. SIO Archives, UCSD. 14. Cohen, J. (1991) Science 252, 912. 15. Revelle, R. (1991) Science 253, 118.
EQUILIBRATION OF THE TERRESTRIAL WATER, NITROGEN, AND CARBON CYCLES
8280
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8280–8283, August 1997 Colloquium Paper This paper was presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held November 13–15, 1995, at the National Academy of Sciences, Irvine, CA.
Equilibration of the terrestrial water, nitrogen, and carbon cycles
(climate/ecosystems/global change/nitrogen use efficiency/resource use efficiency) DAVID S. SCHIMEL*, B. H. BRASWELL* †, AND W. J. PARTON‡
© 1997 by The National Academy of Sciences 0027-8424/97/948280-4$2.00/0 PNAS is available online at http://www.pnas.org. ABSTRACT Recent advances in biologically based ecosystem models of the coupled terrestrial, hydrological, carbon, and nutrient cycles have provided new perspectives on the terrestrial biosphere's behavior globally, over a range of time scales. We used the terrestrial ecosystem model Century to examine relationships between carbon, nitrogen, and water dynamics. The model, run to a quasi-steady-state, shows strong correlations between carbon, water, and nitrogen fluxes that lead to equilibration of water/ energy and nitrogen limitation of net primary productivity. This occurs because as the water flux increases, the potentials for carbon uptake (photosynthesis), and inputs and losses of nitrogen, all increase. As the flux of carbon increases, the amount of nitrogen that can be captured into organic matter and then recycled also increases. Because most plant-available nitrogen is derived from internal recycling, this latter process is critical to sustaining high productivity in environments where water and energy are plentiful. At steady-state, water/energy and nitrogen limitation “equilibrate,” but because the water, carbon, and nitrogen cycles have different response times, inclusion of nitrogen cycling into ecosystem models adds behavior at longer time scales than in purely biophysical models. The tight correlations among nitrogen fluxes with evapotranspiration implies that either climate change or changes to nitrogen inputs (from fertilization or air pollution) will have large and long-lived effects on both productivity and nitrogen losses through hydrological and trace gas pathways. Comprehensive analyses of the role of ecosystems in the carbon cycle must consider mechanisms that arise from the interaction of the hydrological, carbon, and nutrient cycles in ecosystems. Global models of the terrestrial carbon cycle used in geochemical and assessment studies have generally lacked any serious representation of ecological processes or feedbacks and have been extrapolated into the future using a simple parameterization of the relationship between atmospheric CO2 and ecosystem carbon storage (1, 2). Biosphere models used to calculate surface water and energy exchanges in climate models commonly employ sophisticated representations of photosynthesis and respiration, but omit biogeochemical processes associated with the formation and turnover of organic matter (3, 4). Recently, process-based models for terrestrial biogeochemistry have been developed, based on theory linking climate, soil properties, and species- or growth form-specific traits to biogeochemical responses of plants and microorganisms. These models simulate the uptake and release of carbon in response to light, water, temperature, and nutrients (5, 6, 7, 8, 9 and 10). The roles of climate and nutrient limitations inherent in modern ecology (discussed in refs. 11 and 12) are important because the sensitivity of ecosystem models to climate change and increasing CO2 is strongly modulated by nutrients (13, 14). Response of modeled carbon storage to increasing CO2 and temperature is modified by increasing nutrient limitation (14, 15). Large-scale patterns in terrestrial primary productivity, soil carbon, and soil metabolism can often be explained from simple equations using climate parameters (precipitation, actual evapotranspiration, solar radiation) (16, 17, 18, 19, 20, 21 and 22). However, nutrients often limit terrestrial primary productivity in the sense that added nutrients lead to additional plant growth and carbon storage (15, 23). Current process-level models couple biophysical and biogeochemical limits to ecosystem processes explicitly (14, 24, 25). Recent work suggests that, in fact, biophysical and biogeochemical (nutrient) limitations to productivity and carbon storage may come into equilibrium with each other as ecosystems develop over time (24, 25). In this paper, we present a model-based analysis of the processes whereby water/energy and biogeochemical controls over ecosystem productivity and carbon storage converge, as a theoretical underpinning for the eventual quantitative analysis of terrestrial biogeochemical response to global change. Model and Methods In this study, we used the Century terrestrial ecosystem model, developed by Parton et al. (26) over the past decade. In the past few years the model has been extensively evaluated relative to observations along climate gradients (25, 27), at continental scales (25), globally (13, 28), and compared with remote sensing (12, 25). The model simulates the major pathways for water, carbon, and nitrogen exchange, including atmospheric and biological N inputs, and gaseous, combustion-related, and hydrological N losses (12, 13, 26, 28, 29, 30 and 31). Century explicitly partitions live biomass and organic matter (nonliving) into compartments defined by differing turnover times. For the live components, these correspond to leaves, fine roots, coarse roots, branches, and stems. For organic matter the model is based on isotopic and other evidence for multiple turnover times in detritus and soil organic matter (28, 32, 33). The model is integrated globally using gridded global climate, soils, and vegetation data sets with 0.5 degree resolution (24). Results (annual fluxes) shown in this paper are from a simulation of the Northern Hemisphere, using an updated version of Century (24). For this analysis the model was integrated using CO2 concentrations and nitrogen input rates deemed to be representative of the preindustrial biosphere. For example, nitrogen inputs from precipitation were simulated to be 30–50% lower than current levels in moderately to severely polluted areas (34). We did this to simulate, for diagnostic purposes, a nearly steady-state biosphere. Much recent evidence suggests
*National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307-3000; †Institute for Earth, Oceans and Space, University of New Hampshire, Durham, NH 03824; and ‡Natural Resource Ecology Laboratory, Colorado State University, Fort Collins CO 80523
EQUILIBRATION OF THE TERRESTRIAL WATER, NITROGEN, AND CARBON CYCLES
8281
that the forcing due to increasing N deposition and frequency of ecosystem disturbance over the past 50–100 years may have resulted in non-steady-state N cycles in much of the world (34, 35 and 36). Abbreviations: ET, evapotranspiration; NPP, net primary productivity. Equilibration of Nitrogen and Water Limitations In ref. 24 we argued that spatial patterns of biophysical and nitrogen limitation are correlated because carbon and nitrogen fluxes are both strongly influenced by water and energy availability. This mechanism of equilibration is evident in Century because the model simulates the inputs and losses of N, rather than being calibrated to observed ecosystem N stocks (25). The equations in Century governing nitrogen fluxes include biophysical and soil biogeochemical processes. Atmospheric inputs of N are directly linked to precipitation (wet deposition). Biological nitrogen fixation is influenced by soil N and C availability and is assumed to be correlated with annual evapotranspiration (ET). The correlation is based on information indicating high rates of N fixation in humid tropical and temperate rain forests, and generally lower rates in mesic and arid systems, although the biogeography of nitrogen fixation is poorly known (23). It is noteworthy that, globally, patterns of N inputs through all processes are poorly known, and given their importance, require much more study (37). N inputs, as expected (summing biological and atmospheric processes) are strongly correlated with annual ET (Table 1). Losses of nitrogen are controlled by soil moisture and water flux. Leaching losses of NO3 and dissolved organic N (DON) are directly controlled by the product of water flux and NO3/DON concentrations (28). Losses of N trace gases are linked to the rate of mineralization of NH4 and NO3 from organic matter, a rate that increases as temperature and soil moisture increase (28, 38). The proportional as well as absolute losses of gaseous N from inorganic N also increase with increasing soil moisture (30). Century simulates several pathways of N trace gas losses: the summed losses of N2, N2O, and NO from soil nitrification and denitrification are likewise highly correlated with ET (Table 1). This arises because of the strong first-order kinetic regulation of trace gas emissions with respect to soil inorganic N turnover. A key index of soil inorganic N turnover, N mineralization, is likewise strongly correlated with ET (Table 1). As noted in ref. 24, the correlation of N mineralization and ET, though strong, varies among ecosystem types, as is evident for other processes (see Fig. 2). Trace gas losses show similar patterns (data not shown), indicating ecosystem type-specific relationships between biophysical controls and N trace gas emissions, a factor not widely recognized (24). Spatial patterns of nitrate N leaching (data not shown) show strong dependence on ecosystem type, with many systems showing no or low losses; here we computed correlations for systems with non-zero leaching losses. Leaching losses are less directly related to ET, perhaps because ET is a poor predictor of available water below the rooting zone. Nitrate leaching is, however, strongly correlated with precipitation minus ET (P-E), which is related to the amount of water available for movement below the rooting zone (Table 1). Organic N leaching only occurs in a small fraction of grid cells (10%) and generally at low rates. It is poorly correlated with either ET or P-E (Table 1). The low leaching losses of N from many of the world's ecosystems in this simulation of a preindustrial biosphere are consistent with Hedin et al. (35), who suggested that undisturbed ecosystems may have very low losses compared with the bulk of extant ecosystems. The results indicate significant correlation between key fluxes in the nitrogen budget and biophysical controls, although ecosystem-specific processes such as organic N leaching add some variability to patterns of equilibration. Table 1. Correlation structure emerging from key linkages between mechanisms shown in Fig. 1, as implemented in the simulation described in Model and Methods NMIN NPP NINPUT NGAS NO3 DON NMIN — 0.90 0.54 — — ET 0.67 0.71 0.96 0.71 0.33 0.00 P-E — — — 0.74 0.05 — — 0.74 — — NINPUT NMIN, nitrogen mineralization; NPP, net primary productivity; NINPUTs, nitrogen inputs; NGAS, trace gas losses of N; NO3, nitrate leaching; DON, organic nitrogen leaching; ET, evapotranspiration; P-E, precipitation minus ET. All correlations shown are significant at P < 0.05 (except for ET vs. DON).
In Century the potential for carbon fixation increases as evapotranspiration increases via an equation that constrains primary production based on moisture available for transpiration (28). This equation integrates precipitation, energy, and soil hydrological constraints over the water flux in evapotranspiration. ET is linked to both precipitation, soil properties and radiation, as radiant energy is the driving force for ET. Thus, ET, which together with soil hydrological properties, controls the partitioning of soil moisture into runoff and fluxes back to the atmosphere or to depths below the rooting zone. Primary production also requires nitrogen to form organic matter meeting critical C/N ratios for wood, foliage, and roots. On an annual time scale most plant-available N is derived from nitrogen mineralization, which arises from organic matter turnover (decomposition); rates of N mineralization range from 0.2 to 30 g·m2·yr−1, greatly exceeding inputs in most cases. N inputs range from 0.5 to 1.5 g·m2·yr−1. Whereas N availability can vary substantially from year to year, the natural nitrogen budget changes on centennial time scales, as inputs and losses are small fractions of soil N stocks, which typically exceed 500 g·m2. As a consequence of the tight coupling of the water/energy fluxes and nitrogen budget in Century, strong correlations between ET, nitrogen availability, and net primary productivity appear in global Century simulations (see Fig. 2). The correlations arise because water and energy fluxes controls both carbon and nitrogen fluxes (Fig. 1). These fluxes of carbon and nitrogen are mutually interdependent through the dual requirements of nitrogen in the formation of organic matter and of the role of organic matter decomposition in nitrogen mineralization. As water flux increases, N flux increases (inputs and losses), and likewise, the potential for carbon fixation increases. As carbon fixation increases, the amount of the N flux that can be captured in organic matter increases. As more nitrogen is captured in organic matter, its subsequent turnover also contributes to plant available N, allowing more plant productivity. Thus, water/energy and nutrient limitation of plant primary productivity and ecosystem carbon storage tend to “equilibrate” in near-steady-state ecosystems, as illustrated by the spatial patterns of correlation in Fig. 2. The relationships of NPP and N availability with ET are modulated by other factors that influence turnover times. The relationships between NPP, ET, and N are modified by ecosystem type-specific factors that control resource use efficiencies. Effectively these are the carbon-to-nutrient stoichiometry of plants and microorganisms, and water use efficiency (or organic matter produced per unit water transpired) (Fig. 2). Ecosystems with wider C/N ratios in plant tissue have higher NPP per unit N mineralization (higher nitrogen use efficiencies). Systems with lower C/N ratios in leaf and root tissues have higher rates of N cycling per unit ET. C/N ratios reflect both plasticity in foliar and root composition, and more significantly, changes in allocation between high and low-N tissues (wood vs. leaves or roots). Although large-scale patterns arise from system-level interactions of the biogeochemical and hydrological cycles, substantial variation is induced by
EQUILIBRATION OF THE TERRESTRIAL WATER, NITROGEN, AND CARBON CYCLES
8282
species- and growth form-specific traits related to allocation patterns and C/N ratios. The correlation of variables in Fig. 2 indicates the extent to which NPP and nitrogen cycling are controlled by system-level dynamics: the vertical scatter indicates roughly the extent to which system-specific ecological traits influence NPP and nitrogen cycling.
FIG. 1. Schematic illustration of the coupling of water, nitrogen cycling, and carbon in ecosystems. Principle features of these coupled controls are that (i) water controls the inputs and outputs of nitrogen, (ii) increasing net primary productivity (NPP) allows more of the N flux through the system to be captured into organic matter, (iii) increasing organic N stocks allow for more N mineralization, supporting more NPP, and (iv) increasing precipitation both allows more NPP and more N cycling, thus water and nutrient limitation of NPP tend to become correlated. The model results are consistent with observations of large-scale correlations of NPP with direct or derived climate variables, but also with experimental evidence of nutrient limitation. The modulation of the water-carbon-nitrogen system by species- and/or growth formspecific traits implies that large-scale dynamics are influenced by population dynamics on time scales longer than the life spans of individual plants (years-centuries). The relationship of biogeochemistry to population dynamics is outside the scope of this paper, but see refs. 39 and 40. Conclusions We hypothesized that water and nitrogen limitations of NPP are correlated at steady-state because of the control of carbon and nitrogen fluxes by the water budget. We further hypothesized that these correlations arise because of the system-level structure of interactions among the water, carbon, and nitrogen cycles. In model simulations, the correlations between biophysical and nutrient limitations to NPP and carbon storage arise because both carbon and nitrogen fluxes (ecosystem inputs and outputs) are influenced by water and energy availability. This model-based analysis is consistent with the widespread reports of strong correlations of climate with ecosystem processes, suggesting dominant climate controls, and strong experimental evidence that nutrient additions can increase productivity and carbon storage. The tight correlations among N budgetary fluxes (Table 1) suggest that atmospheric trace gas composition may be affected by changes to either climate or N inputs via air pollution or fertilization. Predicted ecosystem behavior becomes more complex when biophysical and nutrient constraints are considered together as compared with purely biophysical formulations (24, 41, 42). The biophysical effects of temperature, moisture, and radiation on photosynthesis, plant respiration, and evapotranspiration can be simulated with relatively few ecosystem-specific controls (4). Nutrient cycling is additionally coupled to spatial patterns of N inputs (34, 35 and 36) and to patterns of plant allocation of carbon and nitrogen among roots, wood, and leaves. These allocation patterns, in turn, influence the distribution among long and short-lived compartments of living (e.g., wood vs. leaves) and detrital organic matter. As more of a system's organic matter becomes tied up in long-lived compartments, fewer nutrients are available for rapid recycling, and nutrient limitation becomes tied to processes with longer time scales, such as soil carbon turnover or tree mortality and wood decomposition.
FIG. 2. Results from an integration of Century for the Northern Hemisphere. Simulations used standard global climate, soils, and vegetation type distribution data sets and were carried out globally on an 0.5 × 0.5 degree grid. Points indicated in green are for forest ecosystems, yellow indicates grasslands, and black indicates “mixed” ecosystems that include both grasses and trees or shrubs (such as savannas). (a) The relationship between ET and NPP (r2 = 0.71). (b) The relationship between nitrogen mineralization and NPP (r2 = 0.90). Nutrient-mediated processes assume increasing importance as ecosystem behavior is considered on interannual and longer time scales because they can cause lagged responses to climate change and variability (24). Models such as those discussed by Sellers et al. (3) describe the behavior of the “fast” carbon-water-energy system (43); biogeochemical models add the consequences of slower processes such as soil carbon and biomass accumulation, and allocation patterns between leaves, roots, and wood. Whereas even biophysical models may have “memory” over one to two years through soil moisture storage, biogeochemical models can simulate lagged effects over de
EQUILIBRATION OF THE TERRESTRIAL WATER, NITROGEN, AND CARBON CYCLES
8283
cades through the decomposition of wood or soil organic matter. Here we identify “equilibration” of the nitrogen budget with water/energy and carbon fluxes as an additional process causing decadal and longer time scale behavior. Just as perturbations to the climate system can cause the nitrogen and carbon systems to respond (4, 24), perturbations of the N inputs from air pollution or fertilization will also cause long-lived ecosystem changes (34, 36, 44). Analyses of the past interannual variability of the carbon cycle, and of its potential future behavior, must consider mechanisms that act through the coupled water/energy, carbon, and nitrogen cycles. We acknowledge the assistance of Rebecca McKeown, Melannie Hartmann, and Hank Fisher with conducting and analyzing global Century runs, with special thanks to Becky for her exceptional effort in completing the global calculations for this paper. Dennis Ojima, Beth Holland, Alan Townsend, and Jason Neff all helped conceive of or design the model experiments. This research was supported by the National Aeronautics and Space Administration Earth Observing System Interdisciplinary Science Program, and by the National Center for Atmospheric Research. The National Center for Atmospheric Research is sponsored by the National Science Foundation. 1. Enting, I. G., Wigley, T.M.L. & Heimann, M. (1994) Future Emissions and Concentrations of Carbon Dioxide: Key Ocean/Atmosphere/Land Analyses (Division of Atmospheric Research, Commonwealth Scientific and Industrial Research Organization, Australia), Tech. Paper No. 31. 2. Siegenthaler, U. & Joos, F. (1992) Tellus B 44, 186–207. 3. Sellers, P. J., Bounoua, L., Collatz, G. J., Randall, D. A., Dazlich, D. A., Los, S. O., Berry, J. A., Fung, I., Tucker, C. J., Field, C. B. & Jensen, T. G. (1996) Science 271, 1402–1406. 4. Sellers, P. J., Dickinson, R. E., Randall, D. A., Betts, A. K., Hall, F. G., Berry, J. A., Collatz, G. J., Denning, A. S., Mooney, H. A., Nobre, C. A., Sato, N., Field, C. B. & Henderson-Sellers, A. (1997) Science 275, 502–509. 5. Farquhar, G. D., Von Caemmerer, S. & Berry, J. A. (1980) Planta 149, 78–90. 6. Melillo, J. M., Naiman, R. J., Aber, J. D. & Linkins, A. E. (1984) Bull. Mar. Sci. 35, 341–356. 7. Bloom, A. J., Chapin, F. S., III, & Mooney, H. A. (1985) Annu. Rev. Ecol. Syst. 16, 363–393. 8. Chapin, F. S., III, Bloom, A. J., Field, C. B. & Waring, R. H. (1987) BioScience 37, 49–57. 9. Nobel, P. S. (1991) Physicochemical and Environmental Plant Physiology (Academic, San Diego). 10. Running, S.W. & Nemani, R. R. (1991) Clim. Change 19, 349–368. 11. Schulze, E. D., De Vries, W., Hauhs, M., Rosén, K., Rasmussen, L., Tann, O.-C. & Nilsson, J. (1989) Water Air Soil Pollut. 48, 451–456. 12. Schimel, D. S., Kittel, T. G. F. & Parton, W. J. (1991) Tellus AB 43, 188–203. 13. Schimel, D. S., Braswell Jr., B. H., Holland, E. A., McKeown, R., Ojima, D. S., Painter, T. H., Parton, W.J. & Townsend, A. R. (1994) Global Biogeochem. Cycles 8, 279–293. 14. VEMAP Participants (1995) Global Biogeochem. Cycles 9, 407–438. 15. Schimel, D. S. (1995) Global Change Biol. 1, 77–91. 16. Leith, H. (1975) in Primary Productivity of the Biosphere, eds. Leith, H. & Whittaker, R. B. (Springer, New York), pp. 237–263. 17. Uchijima, Z. & Seino, H. (1985) J. Agric. Meteorol. 40, 43–352. 18. Sala, O. E., Parton, W. J., Joyce, L. A. & Lauenroth, W. K. (1988) Ecology 69, 40–45. 19. Potter, C. S., Randerson, J. T., Field, C. B., Matson, P. A., Vitousek, P. M., Mooney, H. A. & Klooster, S.A. (1993) Global Biogeochem. Cycles 7, 811–841. 20. Gifford, R. M. (1994) Aust. J. Plant Physiol. 21, 1–15. 21. Zak, D. R., Tilman, D., Parmenter, R. R., Rice, C. W., Fisher, F. M., Vose, J., Milchunas, D. & Martin, C. W. (1994) Ecology 75, 2333–2347. 22. Post, W. M., Pastor, J., Zinke, P. J. & Stangenberger, A. G. (1985) Nature (London) 317, 613–616. 23. Vitousek, P. M. & Howarth, R. W. (1991) Biogeochemistry 13, 87–115. 24. Schimel, D. S., Braswell, B. H., McKeown, R., Ojima, D. S., Parton, W. J. & Pulliam, W. (1996) Global Biogeochem. Cycles 10, 677–692. 25. Schimel, D. S., VEMAP Participants & Braswell, B. H. (1997) Ecol. Monogr., in press. 26. Parton, W. J., Schimel, D. S., Cole, C. V. & Ojima, D. S. (1987) Soil Sci. Soc. Am. J. 51, 1173–1179. 27. Townsend, A. R., Vitousek, P. M. & Trumbore, S. E. (1995) Ecology 76, 721–733. 28. Parton, W. J., Ojima, D. S., Cole, C. V. & Schimel, D. S. (1994) Quantitative Modeling of Soil Forming Processes (Soil Science Society of America, Madison, WI), pp. 147–167. 29. Schimel, D. S., Parton, W.J., Kittel, T. G. F., Ojima, D. S. & Cole, C. V. (1990) Clim. Change 17, 13–25. 30. Parton, W. J., Stewart, J. W. B. & Cole, C. V. (1988) Biogeochemistry 5, 109–131. 31. Ojima, D. S., Schimel, D. S., Parton, W. J. & Owensby, C. E. (1994) Biogeochemistry 24, 67–84. 32. Trumbore, S. E. (1993) Global Biogeochem. Cycles 7, 275–290. 33. Parton, W. J., Scurlock, J. M. O., Ojima, D. S., Schimel, D. S. & Hall, D. O. (1995) Global Change Biol. 1, 13–22. 34. Townsend, A. R., Braswell, B. H., Holland, E. A. & Penner, J. E. (1996) Ecol. Appl. 6, 806–814. 35. Hedin, L. O., Armesto, J. J. & Johnson, A. H. (1995) Ecology 76, 493–509. 36. Holland, E. A., Braswell, B. H., Lamarque, J.-F., Townsend, A., Sulzman, J. M., Müller, J.-F., Dentener, F., Brasseur, G., Levy, H., II, Penner, J. E. & Roelofs, G. (1997) J. Geophys. Res., in press. 37. Galloway, J. N., Levy, H., II, & Kasibhatla, P. S. (1994) Ambio 23, 120–123. 38. Holland, E. A., Townsend, A. R. & Vitousek, P. M. (1995) Global Change Biol. 1, 115–123. 39. Schimel, D. S. (1993) in Biotic Interactions and Global Change, eds. Kareiva, P. M., Kingsolver, J. G. & Huey, R. B. (Sinauer, Boston), pp. 45–54. 40. Pastor, J. & Post, W. M. (1986) Biogeochemistry 2, 3–27. 41. Pastor, J. & Post, W. M. (1993) Clim. Change 23, 111–119. 42. Bolker, B. M., Pacala, S.W., Bazzaz, F.A., Canham, C. D. & Levin, S. A. (1995) Global Change Biol. 1, 373–381. 43. Sellers, P. & Schimel, D. (1993) Global Planet. Change 7, 279–297. 44. Aber, J. D., Nadlehoffer, J. K., Steudler, P. A. & Melillo, J. M. (1989) BioScience 39, 378–386.
POTENTIAL RESPONSES OF SOIL ORGANIC CARBON TO GLOBAL ENVIRONMENTAL CHANGE
8284
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8284–8291, August 1997 Colloquium Paper This paper was presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held November 13–15, 1995, at the National Academy of Sciences, Irvine, CA.
Potential responses of soil organic carbon to global environmental change SUSAN E. TRUMBORE Department of Earth System Science, University of California, Irvine, CA 92697–3100
© 1997 by The National Academy of Sciences 0027-8424/97/948284-8$2.00/0 PNAS is available online at http://www.pnas.org. ABSTRACT Recent improvements in our understanding of the dynamics of soil carbon have shown that 20–40% of the approximately 1,500 Pg of C stored as organic matter in the upper meter of soils has turnover times of centuries or less. This fastcycling organic matter is largely comprised of undecomposed plant material and hydrolyzable components associated with mineral surfaces. Turnover times of fast-cycling carbon vary with climate and vegetation, and range from <20 years at low latitudes to 60 years at high latitudes. The amount and turnover time of C in passive soil carbon pools (organic matter strongly stabilized on mineral surfaces with turnover times of millennia and longer) depend on factors like soil maturity and mineralogy, which, in turn, reflect long-term climate conditions. Transient sources or sinks in terrestrial carbon pools result from the time lag between photosynthetic uptake of CO2 by plants and the subsequent return of C to the atmosphere through plant, heterotrophic, and microbial respiration. Differential responses of primary production and respiration to climate change or ecosystem fertilization have the potential to cause significant interrannual to decadal imbalances in terrestrial C storage and release. Rates of carbon storage and release in recently disturbed ecosystems can be much larger than rates in more mature ecosystems. Changes in disturbance frequency and regime resulting from future climate change may be more important than equilibrium responses in determining the carbon balance of terrestrial ecosystems. Soil carbon inventories and turnover rates are influenced by climate, vegetation, parent material, topography, and time, the fundamental state factors outlined by Jenny (1, 2). Studies attempting to understand the influence of a specific factor (e.g., temperature or moisture) on soil properties have found it useful to identify a suite of soils for which the factor in question varies whereas the others are held constant (1, 2, 3 and 4). This approach has been used successfully to look at the role of temperature (3, 5) and time (6, 7, 8, 9 and 10) on the turnover of soil C. Ecosystem models such as CENTURY (11, 12), CASA (13), or the Rothamsted model (14, 15) predict the sensitivity of soil C inventory and turnover to climate, vegetation, and parent material, but as yet few data exist to test these predictions. Parameterizations of decomposition used in these models are based on empirical fits to specific calibration sites and may not include enough basic understanding of the interaction between plant substrates and the soil environment to make successful predictions in different environments (16). The reservoir of soil carbon has been proposed as both a significant source and sink of atmospheric CO2. A soil source results when net decomposition exceeds C inputs to the soil, either as a result of human activities such as clearing forests for agriculture (17, 18) or because of increased decomposition rates due to global warming (12, 14, 19, 20). Net sinks of C in soils are postulated from the difference between net ecosystem C uptake and tree growth rates (21) or from presumed increases in net C inputs from CO2 or N-fertilization of plants (19, 20, 22, 23 and 24). In both cases, the magnitude and timing of the response depends on the amount of carbon in pools that respond quickly to changes in climate and vegetation, and to the time lag between fixation of C by plants and its subsequent release to the atmosphere during decomposition. This paper will describe recent approaches used to study soil C dynamics, and preliminary applications of these tools to the problems of soil C response to global environmental changes. The results indicate the importance of the global soil C pool to the global C cycle on interrannual to century time scales and suggest profitable areas for future research. The Nature of Soil Organic Matter (SOM) SOM is defined here as the nonliving component of organic matter in soil. The ultimate source of organic matter in soils is CO2 fixed by plants, including leaf litter, roots, and root exudates. The activity of soil fauna (especially fungi and microbial communities) metabolizes some of these substrates and transforms others into more resistant organic compounds (collectively referred to as humus). The stabilization and fate of organic matter residues is affected by the quality of the original plant substrate (25), and the physical environment in the soil [clay content and mineralogy (26), pH, O2 availability (27), formation and disruption of soil aggregates (28, 29 and 30)]. Carbon is lost from soil mostly as CO2 produced during decomposition of organic matter, though losses of carbon through leaching or erosion may be important when considering C balance in soils on long time scales (29, 31, 32). Organic matter in soils plays important roles in determining soil water-holding capacity and soil structure, and provides a long-term store of nutrients needed by plants. Thus changes in SOM will have important feedbacks on hydrology and plant productivity (33). SOM is difficult to study because it is a complex mixture of substances having turnover rates that range from days to millennia. The average global turnover time for soil organic carbon (to 1-m depth) was estimated as 32 years by Raich and Schlesinger (34), who divided the total C stock in soils by the average CO2 flux from soil (corrected for root respiration contribution). Turnover times varied from 14 years to 400 years for different ecosystems in their study. Radiocarbon measurements of bulk soil C, however, often show that the average age of C in soils is several hundred to several thousand years (35, 36, 37 and 38). Both results are explained if SOM contains components that turn over slower and faster than the several-decade average. Abbreviation: SOM, soil organic matter.
POTENTIAL RESPONSES OF SOIL ORGANIC CARBON TO GLOBAL ENVIRONMENTAL CHANGE
8285
Much of the work in understanding and modeling soil C is the study of relative decomposition rates for various organic substrates and organic matter byproducts. While organic matter may actually be a continuum, best represented by a distribution of turnover times (39), continuous distributions are difficult to constrain with field measurements. There is general agreement that SOM contains at least three identifiable C pools: root exudates and rapidly decomposed components of fresh plant litter (“active” pool); stabilized organic matter that persists in soils over several thousands of years (“passive” pool); and a poorly defined “intermediate” or “slow” C pool that has turnover times in the range of years to centuries (Fig. 1). I will use these terms in this paper, and in addition will refer to fast-cycling C as the combined active and intermediate pools (all nonpassive C). The mechanisms through which climate or land-use change may influence soil carbon cycling include a combination of biotic controls that affect the amount and resistance to decay of C added to soils by plants, and physical controls, including the area and chemistry of surfaces for stabilization of organic matter, and the availability of oxidants to decomposers. These mechanisms operate on vastly different time scales. Plant productivity and soil microbial communities respond to shifting climate conditions within hours (41), whereas changes in production and respiration on an annual basis are linked to interrannual variability in climate (42, 43). Persistent changes in climate can lead to species changes over time scales of decades to centuries (44), with concomitant shifts in plant production and litter quality (45). Soil minerals, forming over millennial time scales, control how much of the soil carbon is chemically/physically protected from decomposition (46, 47, 48 and 49). Finally, ecosystem disturbance caused by fire, flooding, tree blow-down, etc. will cause drastic changes in plant production and soil conditions for decomposition that far exceed those associated with interrannual variability. Thus the general design of experiments to understand soil C dynamics should depend on sampling existing environmental and disturbance gradients as well as manipulations to test soil forming factor effects on the amount and turnover times of C in active, intermediate, and passive C pools. Tools to Study SOM No single, satisfactory method yet exists to separate soil C into components with different turnover times. Instead, the dynamic makeup of soil C is deduced using many constraints, including: physical and chemical fractionation of organic matter, field and laboratory decomposition studies, measures of C fluxes into and out of the soil, measurements of 14C in soils sampled at various times before and after the peak of atmospheric nuclear weapons testing, changes in the 13C content of SOM after a vegetation change from plants with C3 to C4 photosynthetic pathway, and measurements of changes in the total amount of C in soils of different age or after disturbance. Each of these tools is suitable for determining different time scales of soil C dynamics, and much confusion may result from mixing of terms related to turnover times (active, intermediate, or passive) with those derived from measured soil parameters. For example, the increase in bomb 14C, or changes in 13C in SOM during the first several decades after conversion of a C3-dominated forest vegetation to C4-dominated pasture vegetation may be used to determine the amount and average turnover time of C that is fast-cycling (nonpassive). Fast-cycling C may be further comprised of active and intermediate pools with different turnover times, but isotopic techniques are not suitable for distinguishing this split.
FIG. 1. Conceptual model of SOM dynamics used in this paper (after ref. 40). Active Carbon Pools. Field decomposition studies involve both observation of loss of native plant litter and differences in decomposition of a common litter substrate at different sites. Isotope labeling studies usually follow specific compounds or compound classes (such as amino acids or carbohydrates), but also may follow the fate of below-ground C allocation, as in pulse-labeling studies. Although some incubations have been followed over periods of more than 10 years (26, 50), most are designed for shorter time periods. These studies are used to provide multiple rate constants for decomposition models like those of Jenkinson and Raynor (15). Soil-respired CO2 is produced either by metabolic root respiration or by decomposition of fast-cycling SOM pools (Fig. 1). In addition, CO2 fluxes measured at the soil surface are most likely dominated by decomposition in near-surface layers (including surface detritus). With these caveats, CO2 flux measurements provide an estimate of total decomposition C flux from the soil, if correction for root-respiration can be made (34). Radiocarbon measurements of respired CO2 may help distinguish sources of respiration experimentally. Root respiration and CO2 derived from decomposition of C with very rapid turnover should have ∆14C close to atmospheric values for the year of sampling, whereas CO2 derived from slow C pools will have higher or lower ∆14C values, depending on the average turnover time. Field measurements of decomposition and CO2 flux often are complicated by seasonal differences in the quality and quantity of detrital material (for example, in temperate zones, deciduous leaves are shed in autumn, but may be responsible for a pulse of decomposition the next spring). Laboratory incubations, where soil is separated from roots, allow observation of the CO2 evolved from decomposition alone (51). Long-term incubations (1 year) allow estimation of the relative contributions of active and slow pools to soil CO2 flux (5). However, the magnitude of laboratory fluxes of C may not be easily mapped to field conditions. Experiments manipulating soil temperature to determine the net response of soils (19) or soils plus vegetation (52) to increased temperature are now in progress in several places. These treatments have been in place in most cases for <10 years, and preliminary results are reflective of adjustments in the fast-cycling pools. Passive Soil Carbon and the Depth Distribution of 14C. The existence of passive SOM is deduced from radiocarbon measurements and the fact that some C3-plant-derived organic matter persists in disturbed soils even after a century or more of cultivation with C4 plants. In general, the amount and 14C content of SOM decrease with depth in soil profiles, indicating the increasing importance of passive organic matter with depth in the soil (37, 53).
POTENTIAL RESPONSES OF SOIL ORGANIC CARBON TO GLOBAL ENVIRONMENTAL CHANGE
8286
The 14C age of passive organic matter may give clues to the mechanisms that cause C to persist in soil over millennial time scales. Three approaches to isolating the properties of passive C from the bulk soil have been used. The first, proposed by O'Brien and Stout (37, 54) and used by Harrison et al. (22) uses a 14C measurement from deep in the soil as an estimate of the passive pool 14C content. This approach attributes decreases in C and 14C with increasing soil depth to a decrease in the abundance of fast-cycling C pools to near-zero at depth, while assuming that the abundance and properties of passive C remain constant over the same depth interval. A second approach toward determining the 14C content of passive SOM is based on operationally defined physical and chemical fractionation of the organic matter. Several investigators (36, 38, 53, 55, 56, 57 and 58) have shown that repeated extractions of SOM with acids and bases of increasing strength will leave a residue with a lower 14C/12C ratio than the starting material (a density separation is performed first to remove relatively intact vascular plant material, which can represent up to 80% of the C in soil A horizons). A third way to estimate the turnover rates in the passive fraction is based on a chronosequence approach (9, 59, 60). A chronosequence is a series of soils of different age, but developed on the same parent material with the same vegetation and climate (for example, soils formed on a series of lava flows or river terraces of different age). During the first several hundred years of soil formation, C accumulation in chronosequences is rapid as vegetation is assumed to establish steady state with fast-cycling soil C components. Accumulation of passive C over millennial time scales in soils (perhaps as authigenic soil minerals are formed and provide new surfaces for organic matter stabilization) is presumably responsible for long-term C increases. Increases in the amount of passive C over time should cause a steepening of the 14C-depth curve (6, 61). Goh et al. (8) reported decreasing 14C contents in deeper soil horizons with increasing soil age in a chronosequence developed on aeolean sand. Tools to Study the Intermediate Pool. The intermediate pool of soil C, with turnover times that vary from years to decades, is the largest fraction of organic matter in most soils. It is also the most difficult to study, as it is itself almost certainly made up of fractions with different turnover times. Ecosystem models such as CENTURY and the Rothamsted model subdivide fast-cycling soil C into readily decomposable and resistant plant debris, as well as an intermediate C pool. In temperate forests soils with low clay content, relatively undecomposed plant debris (the low-density fraction of SOM) can represent up to 80% of the C in A horizon SOM (3) and have turnover times of years to several decades. In tropical soils, low-density organic matter has turnover times of < 5 years and makes up roughly 10– 30% of the total SOM in the upper 10 cm of soil. Mineral-associated (high-density) C that is hydrolyzable in acids and bases, makes up about 50% of the A horizon C, and has turnover times of 20–30 years (62). While it is tempting to map the hydrolyzable C onto the models' intermediate C pool, operationally defined extraction methods give little practical insight into the mechanisms that are stabilizing this organic matter fraction.
FIG. 2. (Left) Change in ∆14C in atmospheric CO2 and in two homogeneous C pools with turnover times of 5, 10, 50, and 70 years. (Right) Effect of passive organic matter on ∆(∆14C) values in bulk SOM. Failure to account for 10% of the carbon in a passive fraction that does not change in 14C content between 1959 and 1992 will result in an underestimate of the increase in fast-cycling ∆14C and therefore of turnover time in that fraction. Thus it is important to assess whether passive components such as charcoal are present in presumed fast-cycling pools (such as low-density carbon). The Bomb 14C Tracer. The incorporation of 14C produced in the early 1960s by atmospheric thermonuclear weapons testing (bomb into SOM during the past 30 years provides a direct measure of the amount of fast-cycling (active + slow) SOM (5, 22, 35, 37, 63, 64, 65, 66 and 67). The most straightforward approach, which compares 14C measurements of SOM sampled before 1960 with contemporary samples from the same location, is summarized in Fig. 2 (62, 64, 68). A large increase in 14C content over the past 30 years indicates that significant portions of the SOM are exchanging carbon with atmospheric CO2 on decadal and shorter time scales. Harrison et al. (22) modeled the 14C increase with time for SOM measured over the past 30 years and reported in the literature (for mostly temperate soils) and obtained an average turnover time of about 25 years for fast-cycling C. Work in other locations has shown that turnover times may be either faster or slower than this average value. Trumbore et al. (3) used the bomb 14C technique to determine how the turnover time of fast-cycling organic matter varied along an elevation transect in the Sierra Nevada mountains of central California. Temperature decreases with elevation along the transect, though vegetation and annual precipitation also vary. C inventory in soil A horizons is predominantly in low-density fractions (<2.0 g/cm3; 50–80% of the total SOM), with another 10–30% in the hydrolyzable fraction of organic matter in the 2.0 g/cm3 fraction. Comparison of the 14C content of these fractions isolated from 0–20 cm soil 14C)
POTENTIAL RESPONSES OF SOIL ORGANIC CARBON TO GLOBAL ENVIRONMENTAL CHANGE
8287
collected in 1959 and 1992 (Fig. 3) shows turnover is fastest in low-density material and slowest in nonhydrolyzable organic matter. Translating the 14C increases into turnover times according to procedures in Fig. 2 for the low-density + hydrolyzable organic matter gives turnover times of 8–12 years at low elevation (highest temperature) and 60 years at high elevation (low temperature). Fig. 3 also shows that the residue after repeated hydrolysis of mineral-associated organic matter in acids and bases still may contain some fast-cycling carbon (as evidenced by an increase in 14C in the past 40 years).
FIG. 3. Increases in ∆14C between 1959 (1964 for lowest elevation site) and 1992 for (a) low density (<2.0 g/cm3), (b) hydrolyzable (2.0 g/cm3), and (c) nonhydrolyzable (2.0 g/cm3) portions of SOM from an elevation transect in the Sierra Nevada (3). The fraction of organic matter with the fastest turnover time will have the highest pre-1963 14C values and show the largest increase in 14C between 1959 and 1992 (a) while the fraction with slowest turnover has lowest initial 14C values and shows little increase in 14C over the past 30 years (c). Without prebomb samples for comparison, 14C measurements of contemporary soils may yield a very misleading picture of soil C dynamics. For example, the bulk 14C contents of the 0–20 cm layer of tropical soils sampled in the early 1990s range from −70% (Hawaii; ref. 5) to +210% (Manaus; ref. 67). The average turnover times of fast-cycling components of these soils, derived from 14C and other measures (see below), is similar (10–20 years). The amounts and 14C contents of carbon in the passive carbon pool are responsible for bulk 14C differences between these sites. The Hawaiian soils are developed on basaltic lava (Andisols) and stabilize large amounts of old carbon in noncrystalline minerals (6), whereas the soils in Manaus have very high percentages of sesquioxides that do not stabilize organic matter as well. Based on 14C measured on bulk SOM alone, one could mistakenly assume very slow turnover rates for the Hawaiian soils. While this is true for average soil C, the average is not a useful number for predicting the response of Hawaiian soils to climate or land-use change. Townsend et al. (5) report soil C turnover times for fast-cycling C along an elevation-based temperature transect of Andisols on the slope of Mauna Kea, near Lapahoehoe, Hawaii. No prebomb archived soil samples were available from these sites. Therefore, a suite of constraints, including 14C measurements, were used to determine turnover times of active, slow, and passive organic matter. The 13C changes in SOM accompanying C3 forest to C4 pasture conversion from three sites reflect the partitioning of fast-cycling (75% of the total SOM) vs. passive (25%) carbon pools in pastures that were roughly a century old. The application of this method is much the same as looking for the increase in bomb 14C in soils; carbon turning over on decadal and shorter time scales will reflect recent isotopic changes in C sources, whereas passive carbon pools will not (26, 69, 70, 71 and 72). Townsend et al. (5) estimated the breakdown of fast-cycling C into active and intermediate C pools using long-term (1 year) incubations—the ratio of CO2 evolution after 1 year to the initially observed CO2 evolution was assumed to equal the ratio of CO2 contributed from decomposition of intermediate C pools to that from active C pools. Estimates of passive pool 14C content were made in two ways: using the method of O'Brien and Stout (37) or calculating from a knowledge of the soil age (15,000 years) and assuming constant accumulation of passive C since the time of soil formation (73). Intermediate pool turnover times then were constrained to reproduce 1990 14C values and CO2 fluxes. Trumbore et al. (62) use a different set of constraints with 14C measurements to deduce the amount and turnover time of fast-cycling C in tropical forest soils in eastern Amazonia. Estimates of annual carbon input to the soil were derived from measurements of litterfall and fine root biomass. The 14C content of CO2 measured in the soil air space was greater than that of atmospheric CO2 for the year in which it was measured (73, 74), indicating that a large portion of the carbon contributing to soil respiration was fixed from the atmosphere by plants over the past several years to decades. The 14C content of CO2 evolved from SOM decomposition can be used as an additional constraint to break fast-cycling organic matter into active and intermediate pools. C Inventory Changes with Land Use. Changes in soil carbon accompanying human or natural disturbance also provide clues about C dynamics in soils. Decreases in soil C inventory associated with cultivation on average may be of the order of 10–40% of initial C inventories (18), but may be smaller and difficult to measure in some areas (62). Changes in 13C abundance in SOM after a shift in vegetation with C3 to C4 photosynthetic pathway (or vice versa) can be used to identify fast-cycling organic matter. These methods rely on accurate
POTENTIAL RESPONSES OF SOIL ORGANIC CARBON TO GLOBAL ENVIRONMENTAL CHANGE
8288
measures of C inventory change over time (26, 69, 70, 71 and 72), and a knowledge of the 13C content of new and old vegetation C inputs to soil. Comparison of 14C in paired undisturbed and disturbed ecosystems can be a sensitive measure of net C storage or loss over the past 30 years (62, 75). Soils that are accumulating C will have higher ∆14C values compared with an undisturbed ecosystem (presumed at steady state), whereas soils that are losing C will have lower 14C than the undisturbed site (73). Prediction of the Response of Soil C to Global Change Total C Inventory in Soils. Estimates of the amount of C stored in soils fall in the range of 1,200 to 1,500 Pg C (46, 76, 77). These estimates are based on relatively few soil C inventory data from many important regions, especially the tropics and northern wetlands. Global estimates are extrapolated using correlations of SOM content with vegetation/climate regime (46, 76) or by soil mapping unit (77). The apparent overall agreement between these estimates masks large uncertainties at the regional scale. For example, moist and wet tropical forest profiles reported in Schlesinger (76) have C inventories ranging from 6.5 to 20.5 kg C m−2 (average of 10.4 kg C m−2). Post et al. (46) use values of 11.5 and 21 kg C m−2 as averages for tropical moist and wet forests, respectively. Davidson and Lefebvre (78) discuss potential uncertainties involved in estimating C stores from large-scale soil map units. Present estimates of global soil C stocks are based on the upper meter of soils, without reference to what fraction of that organic matter is in active, intermediate, or passive pools. Sombroek (79) points out that many tropical soils extend deeper than 1 m, and adding deep C inventories can increase soil C stocks in the tropics markedly. The question of what depth should be the cutoff for counting soil C stores really depends on where and how much of the C is in fast-cycling pools, and how much of the passive soil C pool may be inherited in the case of sedimentary parent material. Ecological models that include cycling of SOM generally assume that only the 0- to 20- or 0- to 30-cm layer of soils contains significant amounts of fast-cycling organic matter. Estimates of fast-cycling SOM pools derived from these models (12, 13) range from about 250 to 530 Pg C globally. The assumption that organic matter deep in soils is not important in C cycling is not always correct. The presence of live roots to depths greater than 15 m in 40-m deep soils of seasonal tropical forests in eastern Amazonia (74) drives a fast carbon cycle to depths of at least 8 m (62, 74). Although bulk C concentrations and 14C contents of SOM at 1-m depths are low, estimates of turnover from fine-root inputs, CO2 production, and the 14C content of CO2 produced at depth show that up to 15% of the carbon inventory in the deep soil has turnover times of decades or less. In these soils, the amount of fast-cycling soil carbon between 1- and 8-m depths (2–3 kg C m−2, out of 17–18 kg C m−2) is significant compared with the amount present in the upper meter of soil (3–4 kg C m−2 out of 10–11 kg C m−2) (62). Predicted Changes in Soil C Due to Increased Plant Production. The simplest approach to predicting the response of soil C changes in plant productivity is to assume that the size of C pools will adjust to a new steady state with increased in C inputs. Fast-cycling pools will adjust most rapidly to new steady state conditions. Harrison et al. (22), assuming a β (growth-enhancement) factor of 0.35 and turnover time of fast-cycling C of 25 years, calculated that increased C inputs to soils should have resulted in net storage of 0.5 Gt C/yr in fast-cycling C in the 1980s (with an additional 0.1 Gt C/yr added to surface litter). This approach assumes no change in decomposition rates due to changes in the quality of those plant substrates, an assumption that has yet to be tested. The changes in C inventory predicted by Harrison et al. (22) are almost impossible to measure directly (62, 80). For example, even a predicted 0.5 Gt C/yr sink integrated over 30 years increases the total soil C pool only 1% (the fast-cycling pool by 3%). C inventory measurements, plagued by small-scale heterogeneity and comparatively large analytical uncertainties (for example, in bulk density measurements), cannot distinguish such small increases on local to regional scales. Thus, other methods must be discovered to determine whether C accumulation of this magnitude is actually occurring in soils. After CO2 fertilization, the most often-cited mechanism leading to a terrestrial carbon sink is enhanced nitrogen availability to plants due to enhanced nitrate deposition (19, 20, 22, 23, 81). Nitrogen availability also will be increased if there is net decomposition of SOM, because SOM is a major source of nitrogen for plants. Although nitrogen mineralization is accompanied by CO2 efflux, the transfer of nitrogen from soil to plants would result in net carbon sequestration because the C:N ratio of plants is roughly 10 times the ratio in SOM. Decomposition rates are predicted to increase as a result of warmer temperatures (see below). Response of Fast-Cycling Soil C to Climate Change. The amount of organic matter in soils is positively correlated with moisture and negatively correlated with precipitation (2, 46). Carbon turnover rates for combined fast-cycling C pools discussed above from the Sierra Nevada, Hawaii, and Paragominas, Brazil are plotted against mean annual temperature in Fig. 4. Although these soils differ in vegetation cover, precipitation (greater than 90 cm/yr with one exception), and soil parent material, a strong relation between turnover time of the fast-cycling C pools and mean annual (air) temperature is observed. This suggests that temperature is a major control of the turnover rate of fast-cycling soil C, in accord with recent modeling and data-based studies (5, 12, 20, 34, 50, 82, 83 and 84). The overall gradient with latitude from the tropics to boreal forests is in accord with trend derived from local elevation gradients. Fig. 4 also plots temperature-turnover time relations predicted by the CENTURY model (12) and a literature review of surface litter decomposition rates (51). Turnover times derived from CENTURY are longer than values derived from field measurements, and show a less steep relation to temperature. The relation derived by Kirschbaum (51) plotted on Fig. 4 is a literature summary of the rates of decomposition derived from laboratory incubations of surface litter. These data suggest turnover of surface detrital pools is faster and has greater temperature dependence than fast-cycling organic matter in soils. Mean annual air temperature is plotted in Fig. 4, but obviously soils (especially surface detritus) experience seasonal and diurnal temperature extremes that may exceed those of the air above. Both vegetation inputs and decomposition rates are expected to increase with increases in temperature (all other factors remaining equal). Soils become an amplifying feedback if the temperature dependence of decomposition is steeper than that of plant productivity (e.g. ref. 5). At present, this is still an open question. Kirschbaum (51) summarized data from incubation experiments for various types of plant litter and found that Q10 values for decomposition ranged from about 2.5 at 20°C to about 4.5 at 10°C. Data from field-based studies using climosequences (summarized in Fig. 4) suggest a Q10 for fast-cycling SOM of around 3, but are too few (particularly in tropical regions) to distinguish between cases with constant and changing Q10 values over a wide range of temperatures. Schimel et al. (12) summarize modeling and data-based predictions of C release from SOM for a 1°C increase in global mean annual temperature. The calculated net release ranges from 11.1 to 33.8 Pg C once a new steady state is reached. The lowest value is predicted by CENTURY, which accounts for feedbacks between plant productivity and release of nutrients that accompanies net SOM decomposition. In N-limited plant
POTENTIAL RESPONSES OF SOIL ORGANIC CARBON TO GLOBAL ENVIRONMENTAL CHANGE
8289
communities, the release of N through increased decomposition in a warmer world could further stimulate plant growth, and even cause net sequestering of C as nutrients are transferred from soils (low C:N ratios) to plant reservoirs with higher C:N ratios (19).
FIG. 4. Turnover times for fast-cycling fractions for a number of sites plotted against mean annual temperature: (a) lowdensity C in the Sierra Nevada transect (3); combined active plus slow C from the Hawaii temperature transect (5); lowdensity C from a mature seasonal tropical forest in eastern Amazonia (60); (b) using low-density plus hydrolyzable carbon for Sierra and Brazil sites. All data represent approximately the upper 20 cm of soil. Also shown are predicted turnover time-temperature relationships estimated from the CENTURY model and from the literature review of Kirschbaum (51). The CENTURY curve is based on model results reported in Schimel et al. (12) assuming the temperature dependence reported in that paper, and combining pools assuming 5% of the C to 0–20 cm is contained microbial biomass pool, 20% in the detrital pool, and 50% in the slow pool. One consequence of all studies of C turnover as function of temperature is that soils already should be a net source of C to the atmosphere due to increased decomposition rates because of documented temperature increases over the past century. This present net C loss presumably would offset hypothetical increases in soil C stores from increased plant production due to CO2 fertilization. The rate at which steady state is approached when fast-cycling C pools respond to a temperature change demonstrates the importance of recognizing differences in turnover rates of soil C between ecosystems. Trumbore et al. (3) estimate the response after changes in net primary production in tropical, temperate, and boreal forest soils. According to these calculations, interrannual variations in net primary production may cause significant interrannual variability in soil-atmosphere CO2 fluxes, with largest and fastest response in tropical regions. Moisture changes. Soil organic carbon inventories increase regionally from arid to wet environments (2, 46), whereas incubation and field decomposition studies show increases in decomposition rates with added moisture in aerobic soils. Decomposition rates for organic matter decrease dramatically under waterlogged conditions. Low C inventories in arid soils may be due to decreased C inputs, faster C turnover, or differences in the quality of plant material. As yet, the relative contributions of these factors are unknown. Disturbance. Soils accumulate organic carbon over at least the first 10,000 years of development (1, 9). Schlesinger (7) showed that the rates of C accumulation in young (thousands of years old) soils are an order of magnitude lower those needed for soils to be a significant C sink for anthropogenic CO2. The changes in C inventory over millennial time scales in soils most likely are caused by the development of secondary soils minerals with stabilization on new mineral surfaces (6). More rapid rates of C accumulation and loss may occur over shorter time scales as the large component of fast-cycling soil C responds to disturbance such as a change in vegetation. As discussed above, soils may lose a significant portion of their carbon after cultivation; these changes represent a loss of fast-cycling C rather than passive C pools (30, 33, 62). Work in boreal forest ecosystems (85) shows high rates of net C sequestration in regrowing surface moss layers in the decades after stand-killing fire events. In these systems, decomposition in mosses is so slow as to be less important than periodic fire at controlling the status of these soils as net C sinks. Thus, particularly in regions where decomposition rates are slow, changes in fire frequency linked to climate or to land use (fire suppression) may be the ultimate control of regional status as a C source or sink. The same may be true in other fire-dominated ecosystems, such as chaparral (68), where organic matter accumulates rapidly between fire events. Wetland areas store organic matter that is climatically stabilized; it is decomposing slowly only because of a lack of oxidants. Warming and draining of wetlands results in a loss of this stored organic matter (86, 87). In northern wetland soils (which contain an estimated one-fifth of total global soil C reservoir), this could lead to large positive feedbacks in warming. Conclusions Climate, vegetation, parent material, and time all affect the processes controlling accumulation and decomposition of organic matter in soils as has been known for decades. The quantification of these insights, however, is still in the beginning stages. Of the roughly 1,500 Gt of C stored in organic matter in soils, an estimated 250–530 Gt resides in C pools with turnover times of decades or less. Models extrapolating the response of the fast-cycling soil C pool on the global scale show its importance in affecting atmospheric CO2. A great deal of work remains to be done to improve our assessments of the role of soils in the global C cycle. Some areas of particular importance are: (i)
Tying soil C pools with different turnover times to specific organic compound classes or structures found in soils, to understand how soil environment and plant litter quality interact to stabilize organic matter. (ii) Better understanding of C accumulation and turnover in wetlands and how these will be affected by climatic and land use change. Present ecosystems models usually exclude wetland C pools from global studies. (iii) Investigation of the role of landscape-scale disturbances, such as fire, in switching soils from net sinks to net sources of C.
POTENTIAL RESPONSES OF SOIL ORGANIC CARBON TO GLOBAL ENVIRONMENTAL CHANGE
(iv) (v)
8290
Manipulations (field incubations) to look at specific responses of coupled soil and vegetation pools to changes in CO2, N deposition, or climate change. Most manipulations presently are taking place in temperate ecosystems; short-term manipulations may yield larger signals in the tropics, where turnover times for fast-cycling soil C are shorter. For larger scale responses over longer timescales, more work needs to be done along naturally occurring environmental gradients, especially those that may have been previously studied and may have archived soils for isotopic analysis.
Data presented here are the result of collaborative work with Eric Davidson, Dan Nepstad, Oliver Chadwick, Ron Amundson, Jennifer Harden, Alan Townsend, Margaret Torn, and Peter Vitousek. John Southon, Michaele Kashgarian, Jay Davis, and Ivan Proctor at the Center for Accelerator Mass Spectrometry at Lawrence Livermore Laboratory continue to provide valuable collaboration by measuring 14C. I thank the A. W. Mellon Foundation, the National Science Foundation, and the National Aeronautics and Space Administration's Mission to Planet Earth Terrestrial Ecology Program for support. 1. Jenny, H. (1941) Factors of Soil Formation (McGraw–Hill, NewYork). 2. Jenny, H. (1980) The Soil Resource: Origin and Behavior (Springer, New York). 3. Trumbore, S. E., Chadwick, O. A. & Amundson, R. (1996) Science 272, 393–396. 4. Vitousek, P. M., Turner, D. R. & Kitayama, K. (1995) Ecology 11, 189–203. 5. Townsend, A. R., Vitousek, P. M. & Trumbore, S. E. (1995) Ecology 11, 721–733. 6. Torn, M. S., Trumbore, S. E., Chandwick, O. A., Vitousek, P. M. & Hendricks, D. M. (1995) EOS 76, 73. 7. Schlesinger, W. H. (1990) Nature (London) 348, 232–234. 8. Goh, K. M., Rafter, T. A., Stout, J. D. & Walker, T. W. (1976) N. Z. J. Soil Sci. 27, 89–100. 9. Harden, J. W., Sundquist, E. T., Stallard, R. F. & Mark, R. K. (1992) Science 258, 1921–1924. 10. Chadwick, O. A., Kelly, E. F., Merritts, D. M. & Amundson, R. G. (1994) Biogeochemistry 24, 115–127. 11. Parton, W. J., Schimel, D. S., Cole, C. V. & Ojima, D. S. (1987) Soil Sci. Soc. Am. J. 51, 1173–1179. 12. Schimel, D. S., Braswell, B. H., Holland, E., McKeown, R., Ojima, D. S., Painter, T. H., Parton, W. J. & Townsend, A. R. (1994) Global Biogeochem. Cycles 8, 279–294. 13. Potter, C. S., Randerson, J. T., Field, C. B., Matson, P. A., Vitousek, P. M., Mooney, H. A. & Klooster, S. (1993) Global Biogeochem. Cycles 7, 811– 842. 14. Jenkinson, D. S. (1990) Philos. Trans. R. Soc. London B 329, 361–368. 15. Jenkinson, D. S. & Raynor, J. H. (1977) Soil Sci. 123, 298–305. 16. Davidson, E. A., Agren, G., Daniel, O., Emeis, K.-C., Largeau, C., Lee, C., Mopper, K., Oades, J. M., Reeburgh, W. S., Schimel, D. S. & Zepp, R. G. (1995) in What Are the Physical, Chemical, and Biological Processes That Control the Formation and Degradation of Nonliving Organic Matter?, ed. Sonntag, R. G. Z. A. C. (Wiley, New York), pp. 305–324. 17. Houghton, R. A., Hobbie, J. E., Melillo, J. M., Moore, B., Peterson, B. J., Shaver, G. R. & Woodwell, G. M. (1983) Ecol. Monogr. 53, 235–262. 18. Davidson, E. A. & Ackerman, I. L. (1993) Biogeochemistry 20, 161–194. 19. Melillo, J. M., Kicklighter, D. W., McGuire, A. D., Peterjohn, W. T. & Newkirk, K. M. (1995) in Global Change and Its Effects on Soil Organic Carbon Stocks, ed. Sonntag, R. G. Z. A. C. (Wiley, New York), pp. 175–190. 20. McGuire, A. D., Melillo, J. M., Joyce, L. A., Kicklighter, D. W., Grace, A. L., III, B.M. & Vorosmarty, C. J. (1992) Global Biogeochem. Cycles 6, 101–124. 21. Wofsy, S. C., Goulden, M. L., Munger, J. W. & Fan, S. M. (1993) Science 260, 1314–1317. 22. Harrison, K., Broecker, W. & Bonani, G. (1993) Global Biogeochem. Cycles 7, 69–80. 23. Townsend, A. R., Braswell, B. H., Holland, E. A. & Penner, J. E. (1996) Ecol. Appl. 6, 806–814. 24. Thompson, M. V., Randerson, J. T., Malmström, C. M. & Field, C. B. (1996) Global Biogeochem. Cycles 10, 711–726. 25. Melillo, J. M., Aber, J. D. & Muratore, J. F. (1982) Ecology 63, 621–626. 26. Balesdent, J., Wagner, G. H. & Mariotti, A. (1988) Soil Sci. Soc. Am. J. 52, 118–124. 27. Bunnell, F. L., Tait, D. E. N., Flanagan, P. W. & Cleve, K. V. (1977) Soil Biol. Biochem. 9, 33–40. 28. Tiessen, H. & Stewart, J. W. B. (1983) Soil Sci. Soc. Am. J. 47, 509–514. 29. Oades, J. M. (1995) in An Overview of Processes Affecting the Cycling of Organic Carbon in Soils, ed. Sonntag, R. G. Z. A. C. (Wiley, New York), pp. 293–304. 30. Cambardella, C. A. & Elliott, E. T. (1994) Soil Sci. Soc. Am. J. 58, 123–130. 31. Trumbore, S. E., Schiff, S. L., Aravena, R. & Elgood, R. (1992) Radiocarbon 34, 626–635. 32. Schiff, S. L., Aravena, R., Trumbore, S. E. & Dillon, P. J. (1990) Water Resourc. Res. 26, 2949–2957. 33. Tiessen, H., Cuevas, E. & Chacon, P. (1994) Nature (London) 371, 783–785. 34. Raich, J. W. & Schlesinger, W. H. (1992) Tellus 44, 81–99. 35. Anderson, D. W. & Paul, E. A. (1984) Soil Sci. Soc. Am. J. 48, 298–301. 36. Campbell, C. A., Paul, E. A., Rennie, D. A. & McCallum, K. J. (1967) Soil Sci. 104, 217–224. 37. O'Brien, B. J. & Stout, J. D. (1978) Soil Biol. Biochem. 10, 309–317. 38. Scharpenseel, H. W., Ronzani, C. & Pietig, F. (1968) in Comparative Age Determinations on Different Humic-Matter Fractions (International Atomic Energy Commission, Vienna), pp. 67–74. 39. Bosatta, E. & Agren, G. I. (1991) Am. Nat. 138, 227–245. 40. Townsend, A. R. (1993) Ph.D. thesis (Stanford University, Stanford, CA). 41. Linn, D. M. & Doran, J. W. (1984) Soil Sci. Soc. Am. J. 48, 1267–1272. 42. Harte, J., Torn, M. S. & Jensen, D. B. (1992) in The Nature and Consequences of Indirect Linkages Between Climate Change and Biological Diversity, eds. Peters, R. L. & Lovejoy, T. E. (Yale Univ. Press, New Haven, CT), pp. 325–343. 43. Goulden, M. L., Munger, J. W., Fan, S. M., Daube, B. C. & Wofsy, S. C. (1996) Science 271, 1576–1578. 44. Torn, M. S. & Fried, J. S. (1992) Clim. Change 21, 257–274. 45. Vitousek, P. M., Turner, D. R., Parton, W. J. & Sanford, R. L. (1994) Ecology 75, 418–429. 46. Post, W. M., Emanuel, W. P., Zinke, P. J. & Stangenberger, A. G. (1982) Nature (London) 298, 156–159. 47. Mayer, L. M. (1994) Geochim. Cosmochim. Acta 58, 1271–1284. 48. Mayer, L. M. (1994) Chem. Geol. 114, 347–363. 49. Chadwick, O. A., Kelly, E. F., Merritts, D. M. & Amundson, R. G. (1994) Biogeochemistry 24, 115–127. 50. Jenkinson, D. J., Adams, D. E. & Wild, A. (1991) Nature (London) 351, 304–306. 51. Kirschbaum, M. U. F. (1995) Soil Biol. Biochem. 27, 753–760. 52. Harte, J., Torn, M. S., Chang, F.-R., Feifarek, B., Kinzig, A. P., Shaw, M. R. & Shen, K. (1995) Ecol. Appl. 5, 132–150. 53. Scharpenseel, H. W., Becker-Heidmann, P., Neue, H. U. & Tsutsuki, K. (1989) Sci. Total Environ. 81/82, 99–110. 54. O'Brien, B. J. (1986) Radiocarbon 28, 358–362. 55. Goh, K. M., Stout, J. D. & O'Brien, B. J. (1984) N. Z. J. Soil Sci. 35, 69–72. 56. Scharpenseel, H. W., Schiffmann, H. & Becker-Heidmann, P. (1984) Radiocarbon 26, 367–383. 57. Scharpenseel, H. W. & Becker-Heidmann, P. (1992) Radiocarbon 34, 541–549. 58. Trumbore, S. E., Bonani, G. & Wolfi, W. (1990) in The Rates of Carbon Cycling in Several Soils from AMS 14C Measurements of Fractionated Soil Organic Matter, ed. Bouwman, A. F. (Wiley, New York), pp. 405–414. 59. Harden, J. W., Trumbore, S. E. & O'Neill, K. (1997) J. Geophys. Res., in press. 60. Schlesinger, W. H., Reynolds, J. F., Cunningham, F. L., Huenneke, L. F., Jarrell, W. M., Virginia, R. A. & Whitford, W. G. (1990) Science 247, 1043–1048. 61. Amundson, R. A., Chadwick, O., McFadden, L., Trumbore, S., Wells, S. & DeNiro, M. (1994) Earth Planet. Sci. Lett. 125, 385–405.
POTENTIAL RESPONSES OF SOIL ORGANIC CARBON TO GLOBAL ENVIRONMENTAL CHANGE
8291
62. Trumbore, S. E., Davidson, E. A., deCamargo, P. B., Nepstad, D. C. & Martinelli, L. A. (1995) Global Biogeochem. Cycles 9, 515–528. 63. Goh, K. M., Stout, J. D. & Rafter, T. A. (1977) Soil Sci. 123, 385–391. 64. Hsieh, Y.-P. (1993) Soil Sci. Soc. Am. J. 57, 1020–1022. 65. Harkness, D. D., Harrison, A. F. & Bacon, P. J. (1986) Radiocarbon 28, 328–337. 66. Jenkinson, D. S., Harkness, D. D., Vance, E. D., Adams, D. E. & Harrison, A. F. (1992) Soil Biol. Biochem. 24, 295–308. 67. Trumbore, S. E. (1993) Global Biogeochem. Cycles 7, 275–290. 68. Trumbore, S. E., Vogel, J. S. & Southon, J. R. (1989) Radiocarbon 31, 644–654. 69. Balesdent, J., Mariotti, A. & Guillet, B. (1987) Soil Biol. Biochem. 19, 25–30. 70. Cerri, C. C., Eduardo, B. P. & Piccolo, M. C. (1991) in Use of Stable Isotopes in Soil Organic Matter Studies (International Atomic Energy Agency, Vienna), pp. 247–259. 71. Desjardins, T., Andreux, F., Volkoff, B. & Cerri, C. C. (1994) Geoderma 61, 103–118. 72. Balesdent, J. & Mariotti, A. (1996) in Mass Spectrometry of Soils, eds. Boutton, T. & Yamasaki, S. (Dekker, New York), pp. 83–112. 73. Trumbore, S. E. & Druffel, E. R. M. (1995) in Carbon Isotopes for Characterizing Sources and Turnover of Nonliving Organic Matter, ed. Sonntag, R. G. Z. A. C. (Wiley, New York), pp. 7–21. 74. Nepstad, D. C., de Carvalho, C. R., Davidson, E. A., Jipp, P. H., Lefebvre, P. A., Negreiros, G. H., Silva, E. D. D., Stone, T. A., Trumbore, S. E. & Vieira, S. (1994) Nature (London) 372, 666–669. 75. Harrison, K. G. & Broecker, W. S. (1993) Science 262, 725–729. 76. Schlesinger, W. H. (1977) Annu. Rev. Ecol. Sys. 8, 51–81. 77. Eswaran, H., Berg, E. V. D. & Reich, P. (1993) Soil Sci. Soc. Am. J. 57, 192–194. 78. Davidson, E. A. & Lefebvre, P. A. (1993) Biogeochemistry 22, 107–131. 79. Sombroek, W. G., Nachtergaele, F. O. & Hebel, A. (1993) Ambio 22, 417–426. 80. Post, W. M., Anderson, D. W., Dahmke, A., Houghton, R. A., Huc, A.-Y., Lassiter, R., Najjar, R. G., Neue, H.-U., Pedersen, T. F., Trumbore, S. E. & Vaikmae, R. (1995) in What Is the Role of Nonliving Organic Matter Cycling on the Global Scale, eds. Zepp, R. G. & Sonntag, C. (Wiley, New York), pp. 155–174. 81. Schimel, D. S., Braswell, B. H., McKeown, R., Ojima, D. S., Parton, W. J. & Pullman, W. (1996) Global Biogeochem. Cycles 10, 677–692. 82. Buol, S. W., Sanchez, P. A , Kimble, J. M. & Weed, S. B. (1990) in Predicted Impact of Climate Warming on Soil Properties and Use, ed. Kimball, J. M. (American Soil Association, Madison), Special Publication No. 53, pp. 71–82. 83. Raich, J. W. & Potter, C. S. (1995) Global Biogeochem. Cycles 9, 23–36. 84. Townsend, A. R., Vitousek, P. M. & Holland, E. A. (1992) Clim. Change 22, 293–303. 85. Trumbore, S. E. & Harden, J. W. (1997) J. Geophys. Res. Atmosph., in press. 86. Goreham, E. (1991) Ecol. Appl. 1, 182–195. 87. Billings, W. D. (1987) Q. Sci. Rev. 6, 165–177.
GLOBAL AIR-SEA FLUX OF CO2: AN ESTIMATE BASED ON MEASUREMENTS OF SEA–AIR PCO2 DIFFERENCE
8292
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8292–8299, August 1997 Colloquium Paper This paper was presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held November 13–15, 1995, at the National Academy of Sciences, Irvine, CA.
Global air-sea flux of CO2: An estimate based on measurements of sea–air pCO2 difference TARO TAKAHASHI*, RICHARD A. FEELY†, RAY F. WEISS‡, RIK H. WANNINKHOF§, DAVID W. CHIPMAN*, STEWART C. SUTHERLAND*, TIMOTHY T. TAKAHASHI ¶
AND
© 1997 by The National Academy of Sciences 0027-8424/97/948292-8$2.00/0 PNAS is available online at http://www.pnas.org. ABSTRACT Approximately 250,000 measurements made for the pCO2 difference between surface water and the marine atmosphere, ∆pCO2, have been assembled for the global oceans. Observations made in the equatorial Pacific during El Nino events have been excluded from the data set. These observations are mapped on the global 4° × 5° grid for a single virtual calendar year (chosen arbitrarily to be 1990) representing a non-El Nino year. Monthly global distributions of ∆pCO2 have been constructed using an interpolation method based on a lateral advection–diffusion transport equation. The net flux of CO2 across the sea surface has been computed using ∆pCO2 distributions and CO2 gas transfer coefficients across sea surface. The annual net uptake flux of CO2 by the global oceans thus estimated ranges from 0.60 to 1.34 Gt-C·yr−1 depending on different formulations used for wind speed dependence on the gas transfer coefficient. These estimates are subject to an error of up to 75% resulting from the numerical interpolation method used to estimate the distribution of ∆pCO2 over the global oceans. Temperate and polar oceans of the both hemispheres are the major sinks for atmospheric CO2, whereas the equatorial oceans are the major sources for CO2. The Atlantic Ocean is the most important CO2 sink, providing about 60% of the global ocean uptake, while the Pacific Ocean is neutral because of its equatorial source flux being balanced by the sink flux of the temperate oceans. The Indian and Southern Oceans take up about 20% each. Measurements of the atmospheric CO2 concentration indicate that it has been increasing at a rate about 50% of that which is expected from all industrial CO2 emissions. The oceans have been considered to be a major sink for CO2. Hence the improved knowledge of the net transport flux across the air–sea interface is important for understanding the fate of this important greenhouse gas emitted into the earth's atmosphere (1, 2, 3, 4 and 5). A number of different approaches has been used for estimating the role of the oceans as a CO2 sink, yielding a wide range of estimates for the CO2 uptake flux (3). Most commonly used are ocean–atmosphere CO2 cycle models. In these models, ocean circulation is modeled using various schemes ranging from one-dimensional box-diffusion models (4, 5) to three-dimensional ocean general circulation models (6, 7 and 8), and biological processes are assumed to be invariant with time and are not explicitly described. These “perturbation” models yield an oceanic uptake of about 2 Gt-C·yr−1 (= 2 × 10 15 g of carbon·yr−1), which corresponds to about 35% of the current industrial CO2 emission rate of about 6 Gt-C·yr−1. On the basis of temporal changes of the 13C/12C ratio in atmospheric and oceanic CO2, the annual oceanic uptake of atmospheric CO2 has been estimated to be 1.6 ± 0.9 Gt-C·yr− 1 (2, 9). Tans et al. (1) combined the meridional gradient of atmospheric CO2 concentration and the net CO2 flux over the northern oceans to constrain the CO2 budget and concluded that the net uptake of atmospheric CO2 by the global oceans is 1 Gt-C·yr−1 or less and that the northern terrestrial ecosystem is a major CO2 sink. Using an advanced atmospheric general circulation models that includes a more realistic description of turbulent mixing near the ground, Denning et al. (10) concluded that an even greater biospheric CO2 sink is needed in the northern hemisphere. In these studies, the mean annual meridional gradient of atmospheric CO2 was assumed implicitly to be zero during the pre-industrial period. This assumption, however, has been questioned by Broecker and Peng (11) on the basis of the meridional gradient of oceanic CO2 concentrations estimated for the preindustrial time. Since global ∆pCO2 data were summarized by Tans et al. (1), the number of observations has been increased significantly as a result of many recent field programs in the United States as well as those in Canada, France, Japan, the Nordic countries, and the United Kingdom (12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37 and 38). In this paper, we present monthly ∆pCO2 distribution maps in 1990 for the global oceans computed using these new data and give estimates for the net CO2 flux over the global oceans obtained using the ∆pCO2 distributions. The magnitude of errors in the mean ∆pCO2 values thus estimated has been estimated and discussed. Measurements All of the data used in this study are the results of direct pCO2 measurements made using air–seawater equilibration methods (12, 17, 39, 40). Although the CO2 concentration in equilibrated air was determined using various types of infrared gas analyzers and gas chromatographs, all these measurements are calibrated against the reference CO2/air mixtures of known CO2 concentrations that were determined by C. D. Keeling (Scripps Institution of Oceanography) using his manometric method. The pCO2 value in seawater was obtained using the CO2 concentration, the water vapor pressure at the equilibration temperature and the pressure of equilibrated gas (which is commonly equal to the barometric pressure). It was corrected to the seawater temperature using the difference between in situ and equilibration temperatures and the isochemical temperature effect on seawater pCO2 of 4.23% C−1 (31). The atmospheric pCO2 was obtained using the local mean
*Lamont–Doherty Earth Observatory of Columbia University, Palisades, NY 10964; †Pacific Marine Environmental Laboratory, National Oceanic and Atmospheric Administration, Seattle, WA 98115; ‡Scripps Institution of Oceanography, University of California at San Diego, La Jolla, CA 92093; §Atlantic Oceanographic and Meteorological Laboratory, National Oceanic and Atmospheric Administration, Miami, FL 33149; and ¶Ames Research Center, National Aeronautics and Space Administration, Moffett Field, CA 94035
GLOBAL AIR-SEA FLUX OF CO2: AN ESTIMATE BASED ON MEASUREMENTS OF SEA–AIR PCO2 DIFFERENCE
8293
value of CO2 concentrations in dry air measured concurrently with the seawater pCO2, barometric pressure, and water vapor pressure at seawater temperature. The sea–air pCO2 difference, ∆pCO2, was computed by subtracting it from the oceanic pCO2 value. Although the non-ideal behavior of CO2 gas due to CO2–CO2 as well as CO2–N2–O2–H2O molecular interactions has been estimated by Weiss and Price (41), its effect is about 1 µatm in the concentration range of CO2. Furthermore, since the corrections are similar for the air and seawater pCO2, the non-ideal effect cancels due to the differencing for ∆pCO2. Therefore, CO2 has been treated as an ideal gas. Abbreviation: SST, sea surface temperature. The pCO2 database assembled for this study consists of about 250,000 individual measurements made during about 250 expeditions. Many of the observations used have been published in scientific journals and in technical reports (12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37 and 38; 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56 and 57). In addition, many unpublished data in the archives of the authors at the Lamont–Doherty Earth Observatory, the Scripps Institution of Oceanography, the Pacific Marine Environmental Laboratory (National Oceanic and Atmospheric Administration), and the Atlantic Oceanographic and Meteorological Laboratory (National Oceanic and Atmospheric Administration) have been used. Adjustments and Resampling of the Data In assembling the global data set for ∆pCO2, the original measurements have been processed through the four steps which are discussed below. Averaging of the Underway pCO2 Data. There are two sources of pCO2 data in surface waters: those obtained for discrete water samples at each hydrographic station and those obtained semicontinuously (several times per hour) using underway systems. Because the number of the latter measurements far exceeds the number for discrete water sample measurements, the latter would statistically overwhelm the former. To prevent this, we have computed a 6-h mean value (e.g., a mean over 100 km if ship's speed is 8 knots) for underway measurements and counted this mean with an equal statistical weight as a discrete measurement. This averaging scheme has been shown to represent a spatial variation of pCO2 in seawater even in areas of strong gradients such as the equatorial Pacific. Exclusion of the Equatorial Pacific Data for El Nino Periods. The objective of this study is to obtain a representative distribution of global ocean ∆pCO2 during non-El Nino periods. Although El Niño Southern Oscillations (ENSO) events could affect a wide range of global meteorological and oceanographic conditions including those in the Southern Ocean (58, 59), the extent of its effects on the carbon chemistry beyond the equatorial Pacific belt has not been documented. Therefore, we have assumed that the effects are limited to the equatorial Pacific between 10°N and 10°S and removed the equatorial Pacific data from the data set for the following El Nino periods, which have been identified on the basis of the Southern Oscillation Index (less than −1.5) and sea surface temperature (SST) changes (NINO3 and NINO4) (59): March 1972–March 1973; May 1976–March 1977; June 1982–June 1983; August 1986–July 1987; October 1991–May 1992; October 1992–October 1993; and April 1994–February 1995. Through these two processes described above, the original 250,000 individual measurements were reduced to about 16,500 data points; their spatial distribution is shown in Fig. 1. Although the global ocean appears to be well covered over the 12-month period with the exception of the southern Indian Ocean, monthly data distributions (February and August are shown as examples) show large oceanic areas without measurements. Normalization of ∆pCO2 to the Year 1990 With the exceptions of well-studied areas such as the western North Pacific, the available observations are not sufficient to resolve the interannual variability of ∆pCO2 over the global oceans. Therefore, our approach is to combine all the observations onto a single virtual calendar year (chosen arbitrarily to be 1990). However, since the mean atmospheric CO2 concentration has increased by about 30 ppm from about 326 ppm in 1972 to 356 ppm in 1994, the secular increase in atmospheric CO2 must be taken into consideration when ∆pCO2 data are assembled to represent a single year. In subtropical gyres, vertical mixing of surface layer waters with subsurface waters is limited due to the strong stratification. Hence, as the surface water takes up more atmospheric CO2, the mean pCO2 in surface waters tends to increase with a rate similar to the atmospheric CO2 increase. This has been demonstrated over the Sargasso Sea (60) and over the western North Pacific between 3°N and 35°N along the 137°E meridian by Inoue et al. (21). R.A.F. (unpublished data) analyzed the surface water pCO2 data obtained in the equatorial Pacific along 140°W during the years 1984, 1988, 1990, and 1995, and observed that the surface water pCO2 values have been increasing at a mean rate of 1.3 ± 0.5 µatm·yr−1, which is consistent with the atmospheric rate of about 1.8 ppm·yr−1. This implies that the strong CO2 source in the Pacific equatorial belt is caused by the warming of recently ventilated, colder subsurface waters derived primarily from depths about 100–300 meters. Hence ∆pCO2 values for the temperate gyre and equatorial areas are nearly independent of the year of measurements and thus, those measured in different years may be treated as though those were measured in the same year.
FIG. 1. Locations of ∆pCO2 measurements used in this study. (a) All sample locations, 1960–1995; (b) those in the month of February; and (c) August during the same years.
GLOBAL AIR-SEA FLUX OF CO2: AN ESTIMATE BASED ON MEASUREMENTS OF SEA–AIR PCO2 DIFFERENCE
8294
In contrast to the warm water regime, cold high-latitude surface waters located poleward of the subpolar front are known to convectively mix with deep waters seasonally and thus reflect the chemical characteristics of deep waters, in which anthropogenic effects are diluted. In Fig. 2, the surface water pCO2 data observed in 1974–1979 at Weather Station “P” (50°N, 145°W) in the northeastern subarctic Pacific by Wong and Chan (36) are compared with the 1985–1989 observations by the Lamont–Doherty Earth Observatory group (31, 50) and with the 1986 observations by the Pacific Marine Environmental Laboratory group (25). This shows that, while the atmospheric CO2 concentration increased by about 15 ppm over the 10-year period, an increase in the oceanic pCO2 has not been detectable. Therefore, the sea–air pCO2 difference has increased with time approximately at the mean annual rate for atmospheric CO2 increase, and ∆pCO2 measured in a different year must be corrected for atmospheric CO2 changes that occurred between the time of measurement and the reference year of 1990. Since the annual rate of increase in surface water CO2 depends on the vertical flux of subsurface waters and the duration of exposure time to the atmosphere, surface water pCO2 values in some regions where vertical mixing is weaker and/or air–sea ventilation time is longer may increase at a slower rate. However, relevant observations are not yet available in other subpolar and polar oceanic regions. Therefore, to normalize the observed ∆pCO2 values to the reference year of 1990, we assumed two cases in this study: (i) ∆pCO2 in the cold water regimes increased at the same mean annual rate as atmospheric CO2 and (ii) it increased with one-half the rate of atmospheric CO2. These are named respectively as “full” and “half” atmospheric CO2 increase cases in Table 1 and Table 2. These two normalization schemes are applied to the ∆pCO2 values observed poleward of the subpolar front which are located approximately along 40°N in the North Pacific, 45°N in the North Atlantic, 50°S in the South Atlantic, South Indian, and western Pacific (west of 180°) oceans, and 60°S in the eastern South Pacific Ocean (between 60°W and 180°). The CO2 concentration in the marine air at a given time and location over the global oceans has been computed using an equation which has been fitted to the mean annual data summarized in refs. 1, 61, and 62 with better than 1 ppm. Because seasonal variation in atmospheric CO2 concentration is represented in the observed ∆pCO2 value by the concurrent measurements of atmospheric and oceanic pCO2, only the effect of the secular increase is considered for correction.
FIG. 2. The pCO2 values observed by Wong and Chan (36) during 1974–1979 in surface waters at or near the Weather Station “P” in the northeastern North Pacific (50°N, 145°W) are compared with more recent measurements by the Lamont–Doherty Earth Observatory during 1984–1989 (31, 53) and the Pacific Marine Environmental Laboratory during 1986–87 (25). The pCO2 values are normalized to a constant temperature of 10.0°C to remove the effect of temperature changes. The solid line indicates a sine curve fitted to the 1974–1979 data. Resampling of the Data onto a 4° × 5° Grid. For use in the computational scheme of this study, the ∆pCO2 values normalized to the reference year of 1990 must be resampled onto the computational grid which consists of 72 grid points in longitude by 41 in latitude for each of 365 days in time. This amounts to a total of about 750,000 pixels excluding land areas. Although observations made in some areas are sufficient to allow a finer grid locally, the global data set assembled for this study is limited in spatial and time resolutions to allow a finer grid over the entire global oceans. A weighted average, inversely proportional in both space and time, is used to transform about 16,500 ∆pCO2 data onto the 4° latitude × 5° longitude × 365 days grid. Observations within ±4° latitude, ±5° longitude, and ±1 day from the pixel center is utilized to compute a mean pixel value. This liberal resampling algorithm gives sparse data more representation. Computational Method for Time–Space Interpolation The data set which has been discussed above represents measurements made at locations and times that were dictated by ship tracks without mathematical regularity. A computational method is needed to interpolate these observations discretized in a 4° × 5° × 365 days grid to construct global ∆pCO2 distribution maps. For this purpose, a finite-difference algorithm based on a lateral two-dimensional transport model developed by Takahashi et al. (63) is used. In the interpolation scheme, all pixel values based on observations are explicitly satisfied, and those in pixels that have no observations are computed. Transport Equation and Boundary Conditions. The concentration of a chemical property throughout the oceans is governed by advective transport, eddy diffusion, internal sources, and sinks and exchange with the surroundings at the sea floor, the ocean margins, and the air–sea interface. In an interpolation scheme, in which the observations are satisfied, the effects on the surface water property of internal sources and sinks, exchange with atmosphere and upwelling of deep water are considered inherently imbedded in the observed data. Accordingly, the short-term behavior of surface water properties may be approximated using a lateral two-dimensional transport model without sink/source and exchange terms. The transport equation used for interpolation is: dS/dt = K 2S − (`S/`x Vx + `S/`y Vy) and 2S = ` 2S/`x2 + ` 2S/`y2, where S is a scalar quantity, Vx and Vy are the local surface water advective velocities, and K is the eddy diffusivity. The implied boundary condition is that there is no material transport across the sea–land interface: on an orthogonal grid, ` S/ ` x = 0 and ` S/ ` y = 0 along the boundaries. The computational domain is joined at the prime meridian, so that the oceans are freely interconnected in the east–west domain. Singularities at the poles are avoided by the presence of the Antarctic continent in the south (80°S) and the polar ice cap in the north (84°N) which is assumed to be a land mass. As discussed earlier, data collected in various years are mapped onto a single, virtual calendar year of 1990. Therefore, the computations are numerically joined across the December/January border ensuring continuity of solutions in both space and time. This allows solutions to be obtained iteratively over a single year period. The finite–difference algorithm required must possess some unusual properties. Due to the temporal and spatial connectivity of the computational matrix, it is essential that the
GLOBAL AIR-SEA FLUX OF CO2: AN ESTIMATE BASED ON MEASUREMENTS OF SEA–AIR PCO2 DIFFERENCE
8295
influence of observed data propagates at the same computational speed both forwards and backwards in time: an average of forwardcentered time step about time (t − 1) and backwards-centered time step about time (t + 1) is utilized. The spatial derivatives are secondorder accurate and are central differences: `S/`x = (Sx+l − Sx−1)/2∆x and ` 2S/` x2 = (Sx+l − 2Sx + Sx−1)/(∆x)2. The time increment, dt, is chosen to be 1 day to maintain numerical stability under both diffusion and advection (63). The recursion relationship is applied over all active nodes (where no observation is available) until convergence is obtained. The convergence criterion is the net aggregate change (R) in the value of S between successive iterations: R = Σ abs[S(x, y, t)i − S(x, y, t)i−1], where x = 180°W through 180°E, y = 80°S through 84°N, and t = day 1 through day 365 and S(x, y, t)i represents the value of the scalar function, S, at a node representing a given position and time at the iteration, i. The solution is considered converged when changes in R fall below 0.1%. Typically, it takes several thousand iterations for the convergence criterion to be satisfied. Although the solutions give daily distributions, monthly mean distributions have been computed and used for the flux estimates. Advective Field and Eddy Diffusion. For the computation of advective transport, we use the mean monthly surface water advective fields (4° × 5° resolution) of Bryan and Lewis (64), which have been kindly provided by the Geophysical Fluid Dynamics Laboratory (National Oceanic and Atmospheric Administration, Princeton, NJ). The advective field is changed each month in our computation. While the seasonal variability is small in the Pacific and Atlantic Oceans, currents in the northern Indian Ocean undergo a complete reorganization as a result of monsoon conditions. The lateral eddy diffusivity of surface waters has been estimated to be in a range of 1,000 to 3,000 m2·sec−1 (see refs. 65, 66 and 67) and is assumed in this study to be a constant value of 2,000 m2·sec−1. Preloading of the Field. To obtain converged solutions rapidly and to constrain the solutions in polar regions where data are very scarce, initial ∆pCO2 values for those pixels that have no measurement are assigned to be equal to the monthly zonal mean of the observed ∆pCO2 values in each of the polar (90° to 54°), subpolar (38° to 54°), temperate (18° to 38°), and tropical (18° to 18°) zones. In addition, the polar pixels with ice fields are assumed to have seasonally varying ∆pCO2 values that are consistent with a limited number of observations made in polynyas and adjacent to ice fields (29, 34, 44, 51, 52): for the Arctic pixels with ice, −150 µatm for August, −75 µatm for September, −50 µatm for October, and −35 µatm for the remaining 9 months; for the Antarctic pixels with ice, −85 µatm for February, −10 µatm for January and March, +5 µatm for December and April, and +30 µatm for the remaining 6 months. The CO2 flux is assumed to be zero when a pixel is covered with ice, the flux estimates are not sensitively affected by the choice of preset values. A constant temperature of −1.9°C is assigned to all pixels with ice. Starting with these preloaded field, the interpolation computation was performed using a time step of 1 day. The results obtained after 2,000 iterations are presented in this paper. Tests for the Interpolated Values Comparison with Climatological SST. To evaluate the validity of our results, the SST values that have been computed using the measurements made concurrently with ∆pCO2 are compared with the climatological SST compiled by Shea et al. (68). Consistent with the procedures used for ∆pCO2 interpolation, the temperature field was preloaded using mean monthly values for temperature values measured concurrently with ∆pCO2 in each climatic zone. In Fig. 3a, the difference between the mean monthly SST observed in 4° × 5° pixels and the corresponding climatological SST are plotted against the climatological SST; and in Fig. 3b, the difference between the interpolated value and the corresponding climatological SST is plotted. Fig. 3a shows that the measured temperatures are consistent with the climatological SST yielding a mean difference of −0.07°C (n = 5,780) with a rms deviation of ±1.5°C. The small mean difference indicates that our temperature observations are consistent with the climatological SST.
FIG. 3. Comparisons between (a) the mean monthly climatological SST data (68) and the mean monthly temperatures measured concurrently with ∆pCO2; and (b) the mean monthly climatological SST data and those interpolated for pixels with no measurements. The solid and dashed horizontal lines indicate the mean and rms deviation values, respectively. Fig. 3b shows that our interpolated temperatures are, on average, warmer than the climatological SST by 0.43 ± 1.89°C. This may be taken as a measure of errors attributable to our computational method and limited time–space database. The temperature dependence of pCO2 in surface waters varies regionally and seasonally from about +3% C−1 in temperate gyres to about −6% C−1 in polar oceans (31), and is about 3.5% C−1 (in the absolute magnitude) on the average over the global oceans. This gives that the SST error of 0.43°C corresponds to a ∆pCO2 of 5 µatm. If a mean global sea–air CO2 exchange rate of 19 mol·m−2yr−1 (69), which is consistent with the gas transfer formulation used by Tans et al. (1), is assumed, this error would correspond to a flux error of about 1 Gt-C·yr−1 or 75% of the net global ocean CO2 flux estimated in this study. Because of the sparseness of time–space coverage, the uncertainty in the global ocean ∆pCO2 depends critically on the availability of observations especially in data poor areas. The error could be reduced significantly by additional seasonal observations especially over the southern mid- and high-latitude oceans. Sensitivity to the Interpolated ∆pCO2 Values. The sensitivity of the interpolated ∆pCO2 values to the number of observations made along ship's tracks has been tested by eliminating a measurement from every 10 observations. The interpolated values obtained with 90% of the full database have been compared with those computed using 100% of the database. The effect of this reduction on the interpolated ∆pCO2 is small with a mean difference of −0.01 ± 2.2 µatm and is independent of latitude. Therefore, the estimated ∆pCO2 values over the global oceans are not sensitively affected by the number of measurements made along ship's tracks. Distribution of the CO2 Sink/Source Over the Global Oceans Distribution of ∆pCO2. The mean monthly distribution of sea–air pCO2 difference, ∆pCO2, over the global oceans is
GLOBAL AIR-SEA FLUX OF CO2: AN ESTIMATE BASED ON MEASUREMENTS OF SEA–AIR PCO2 DIFFERENCE
8296
tabulated in Table 1 for the “full” and “half” effects of the atmospheric CO2 increase; and its geographical distribution for the months of February and August 1990 is presented in Fig. 4 a and Fig. 4 b, respectively. During the month of February, the following areas are strong sources for atmospheric CO2 (positive ∆pCO2 values): (i) the northwestern subarctic Pacific (due to winter upwelling), (ii) the entire equatorial Pacific (upwelling), (iii) the eastern subtropical South Pacific (seasonal warming), (iv) the tropical and subtropical South Atlantic (upwelling and seasonal warming), (v) the western equatorial Indian Ocean, and (vi) a few patchy areas near Antarctica (local upwelling). The CO2 sink areas include (i) the temperate gyre areas in the North Pacific and North Atlantic (seasonal cooling), (ii) subpolar and polar areas of the South Atlantic and South Pacific (photosynthesis), and (iii) south and east of New Zealand (photosynthesis). In August (Fig. 4b), the strong source areas include (i) the central and eastern equatorial Pacific (upwelling), (ii) the North Pacific temperate gyre (seasonal warming), (iii) the northwestern North Atlantic (warming of the Labrador Current), (iv) Arabian Sea (upwelling), and (v) patchy areas near Antarctica (upwelling). The strong sink areas are as follows: (i) the northern North Atlantic and adjacent subpolar seas (photosynthesis), (ii) the northwestern North Pacific (photosynthesis), and (iii) the temperate regions of the south Indian, Pacific, and Atlantic Oceans (cooling). Over the subpolar and polar Atlantic, Pacific and Southern Oceans, the surface water pCO2 increases during winter due to winter convective mixing, whereas it decreases during summer due to photosynthesis in stratified photic waters (31). Table 1. Sea–air pCO2 difference (µatm) nomalized to 1990 with the effect of full and half atmospheric CO2 increase for five zones Latitude Effect Jan. Feb. March April May June July Aug. Sept. Oct. Nov. N of Full −15.9 −10.3 −13.7 −18.6 −28.5 −44.9 −50.4 −51.7 −51.3 −46.3 −39.8 50°N Half −11.9 −6.2 −9.6 −14.9 −25.1 −41.4 −46.4 −46.9 −47.2 −42.0 −35.4 14° Full −16.3 −18.6 −19.9 −20.2 −17.4 −9.0 +2.8 +7.6 +4.7 −1.8 −8.3 N-50°N Half −15.9 −18.3 −19.5 −19.8 −17.0 −8.7 +3.2 +8.0 +5.3 −1.1 −7.8 14° F/H +26.7 +27.9 +25.9 +28.6 +28.8 +26.1 +25.0 +23.0 +22.3 +21.9 +23.1 S-14°N 14° Full −3.0 −1.0 −0.9 −2.9 −6.8 −9.5 −11.2 −11.9 −11.2 −10.4 −8.9 S-50°S Half −2.6 −0.6 −0.6 −2.7 −6.5 −9.2 −10.8 −11.4 −10.7 −9.9 −8.4 S of 50° Full −13.6 −15.0 −13.0 −6.6 −4.5 −3.4 −3.0 −1.5 −3.3 −5.4 −7.3 S Half −12.0 −13.2 −12.3 −6.1 −4.1 −2.8 −2.5 −1.1 −2.9 −5.0 −6.9 Global Full +0.7 +1.1 +0.4 +1.1 +0.3 −0.1 +1.3 +1.6 +0.8 −0.4 −0.8 Mean Half +1.3 +1.7 +0.8 +1.4 +0.7 +0.3 +1.8 +2.1 +1.4 +0.3 −0.2
Dec. −29.1
Annual −34.5
−25.0 −13.6
−30.5 −9.2
−13.1 +24.6
−8.7 +25.3
−6.8
−7.1
−6.3 −11.0
−6.6 −7.9
−10.2 −0.8
−7.1 +0.4
−0.2
+1.0
The global mean values are area-weighted averages.
Table 1 shows that the mean annual ∆pCO2 values (area weighted) for the global oceans are small (+0.4 + 1 µatm), as are many of the monthly mean values (−0.8 +2.1 µatm). This means that the global oceans are, as a whole, nearly in equilibrium with atmospheric CO2, although they are locally out of equilibrium by as much as 30%. This suggests that the oceanic uptake of CO2 depends sensitively on the wind speed distribution especially in areas such as subpolar oceans where large negative ∆pCO2 and high wind speeds prevail. Computation of the Net CO2 Flux Across the Sea Surface. The net CO2 flux (F) across each 4° × 5° pixel area is computed using F = A × E × ∆pCO2, where A is a pixel area, E is mean monthly gas transfer coefficient across the sea surface, and ∆pCO2 is mean monthly sea–air pCO2 difference evaluated for each pixel. Since the E value for CO2 gas is not well known, we consider the following three different functions of wind speed (W): (i) a linear wind speed dependence based on the 14C budget in the atmosphere and oceans (1, 69): E (CO2 mol·m−2·yr−1·µatm−1) = 1.6 × 10−2 [W(m·sec−1) − 3.0] and E = 0 when W ≤3.0 (m·sec−1); (ii) a relationship formulated by Wanninkhof (70) for long-term wind (i.e., his equation 2): E = 1.13 × 10−3·W2; and (iii) a three linear-segment wind speed dependence formulated by Liss and Merlivat (71): E = 4.8 × 10−4 × W for 0 ≤ W ≤ 3.6 m·sec−1, E = 8.3 × 10−3 (W − 3.39) for 3.6 ≤ W ≤ 13 m·sec−1, and E = 1.7 × 10−2 (W − 8.36) for W ≥ 13 m·sec−1. While i yields the fastest gas transfer rate representing the upper limit, iii yields rates about 50% of the first, representing the lower limit, and ii yields rates about 15% smaller than the first. It is therefore important in the future to improve our understanding of the CO2 gas transfer rate across the sea surface over a wide range of turbulences near the air–sea interface. In this paper, the mean monthly wind speed data compiled for the global oceans (72) are used to compute the CO2 gas transfer rate constant, and the results of these three relationships are presented and compared. Net Flux of CO2 Over the Global Oceans. Fig. 5 shows the distribution of annual CO2 flux and indicates that the equatorial belt of the Pacific and Atlantic is a major CO2 source,
FIG 4. Mean monthly distribution of ∆pCO2 (µatm) for (a) February 1990 and (b) August 1990, estimated using the assumption of full atmospheric CO2 increase. The positive values indicate that the ocean is a source for atmospheric CO2, and the negative values indicate CO2 sinks. The pink lines indicate edges of ice field.
GLOBAL AIR-SEA FLUX OF CO2: AN ESTIMATE BASED ON MEASUREMENTS OF SEA–AIR PCO2 DIFFERENCE
8297
whereas temperate areas of the global oceans and the subarctic Atlantic are major sinks. Table 2 shows that the global ocean CO2 uptake flux ranges from 0.6 to 1.34 Gt-C·yr−1 depending on the gas transfer formulations used. This is about 2 times as large as those estimated by Tans et al. (1), whereas it is about 30–70% of the estimates of about 2 Gt-C·yr−1 obtained by perturbation models. It, however, overlaps with 1.6 ± 0.9 Gt-C·yr−1 based on 13C/12C distribution in seawater (2, 9). The mean ocean flux for each of the five latitudinal zones is compared in Table 2 with the results of Tans et al. (1). The northern ocean CO2 uptake estimated by this study is about twice as large as that of Tans et al. (1). This is partially due to more negative ∆pCO2 values resulting from correcting the older data to 1990, and partially due to improvements in the database and time–space interpolation method. While a time resolution of 4 months was used in Tans et al. (1), 1 month is used in this study, allowing a better representation for pCO2 drawdowns during spring phytoplankton blooms. The equatorial CO2 source flux estimated in this study is about 30% smaller than that by Tans et al. (1). This is a result of increased observations especially in the western and central equatorial Pacific by the Japanese (19, 20, 21) and U.S. (13, 16, 25) investigators, respectively. The Southern Ocean CO2 source flux of +0.5 Gt-C·yr−1, which was proposed by Tans et al. (1) to satisfy the observed meridional gradient of the atmospheric CO2 concentration, cannot be supported by the oceanic observations used in this study. Net CO2 Flux Over the Oceanic Basins. The equatorial zone (14°N to 14°S) of the Pacific is the major oceanic CO2 source area emitting (+0.35 to +0.79 Gt-C·yr−1) about 5 times as much CO2 to the atmosphere than the corresponding areas in the Atlantic and Indian Oceans. On an annual basis, the Atlantic (north of 50°S) is the most important CO2 sink (−0.42 to −0.85 Gt-C·yr−1) because of the temperate and subarctic oceans being strong CO2 sinks and the equatorial belt being a weak source. On the other hand, the Pacific (north of 50°S) as a whole is nearly neutral (+0.04 to −0.02 Gt-C·yr−1) because of the strong equatorial source nearly balancing the strong sinks in the temperate areas. The Southern Ocean (south of 50°S) is a strong sink (−0.15 to −0.3 Gt-C·yr−1), which is comparable to the uptake flux for the temperate oceanic areas. Each of the Pacific, Atlantic, and Indian basins shows different north–south contrasts. Excluding the Southern Ocean, south of 50°S, the magnitude of the uptake flux for the southern temperate Pacific is similar to that for the northern temperate Pacific. On the other hand, because of strong sink areas in the subarctic oceans, the northern Atlantic sink is about 4 times as strong as the southern temperate Atlantic. On the basis of the flux estimated using the gas transfer coefficient of Tans et al. (1), the north–south flux asymmetry is about 0.6 Gt-C·yr−1 if the Southern Ocean (south of 50°S) is excluded or about 0.45 Gt-C·yr−1 if the Atlantic sector of the Southern Ocean is included. This is comparable to the preindustrial north–south ocean transport of 0.6 Gt-C·yr−1 in the Atlantic proposed by Broecker and Peng (11).
FIG. 5. Mean annual net CO2 flux over the global oceans (in 1012 grams of C per year for each pixel area) computed for 1990 using the gas transfer coefficient formulated by Wanninkhof (70). The effect of full atmospheric CO2 increase is assumed for normalizing observed ∆pCO2 values in high latitude areas to the reference year of 1990. Areas covered with ice (i.e., the poleward of the pink lines in Fig. 4 a and Fig. 4 b) are assumed to have zero sea–air CO2 flux. Sources of Errors. The flux estimates are subject to errors from the following five independent sources: (i) the gas transfer coefficients, (ii) the wind speed variability, (iii) the normalization of observations to the reference year of 1990, (iv) the interpolation of limited observations, and (v) skin temperature effect. (i) (ii)
The estimated flux values, which range from 0.60 to 1.34 Gt-C·yr−1, depend on the choice of sea–air CO2 gas transfer formulations. Hence the error is of a systematic nature and may be reduced if the gas transfer coefficient is better understood in the future. Because the gas transfer coefficient increases with wind speed at an increasingly faster rate, gas exchange rates estimated using mean monthly wind speeds tend to be smaller than those estimated using high-frequency wind speed data with a large wind speed variability. Hence, our flux estimates are likely to represent a minimum value. The mean monthly gas transfer coefficient thus obtained may be underestimated relative to that obtained using high frequency wind data by about −10% on the average, although it could be off by as much as −51% locally depending upon the frequency distribution and magnitude of wind speed variations (73). Because the variation of wind speed is coupled with that of ∆pCO2 through the interactions between turbulent mixing in upper
*Based upon ∆pCO2 observations by Tans et al. (1). †Estimated using the atmospheric general circulation model of GISS to satisfy the hemispheric difference in the atmospheric CO2 concentration, industrial emission, and uptake of CO2 by the northern hemisphere oceans for various scenarios for biosphereic CO2 sources and sinks.
GLOBAL AIR-SEA FLUX OF CO2: AN ESTIMATE BASED ON MEASUREMENTS OF SEA–AIR PCO2 DIFFERENCE
8298
waters and biological activities, the covariance of wind speed and ∆pCO2 should be taken into consideration in future studies. (iii) To correct for interannual changes in the oceanic and atmospheric pCO2 values, the observed ∆pCO2 values have been normalized to a reference year 1990 as explained earlier. The flux values for high latitude oceans computed using the “full” atmospheric CO2 increase are about 10% greater in magnitude than the respective values obtained for the “half” effect (Table 2), and hence the error due to this effect is about ±5%. (iv) Errors due to the computational method has been estimated to be about 5 µatm for the mean global ∆pCO2. This corresponds to an error of up to 75% of the flux estimates made using a given gas transfer formulation. (v) ∆pCO2 depends on the skin temperature of surface ocean water (49, 74), while the pCO2 has been evaluated at the bulk water temperature in this study. The effect of skin layer cooling on the global CO2 uptake has been estimated to be 0.1–0.6 Gt-C·yr−1 (75). However, the skin temperature may be higher or lower than the bulk water temperature depending upon meteorological and oceanic conditions, and the measurements are limited in space and time. Therefore, its effect on the global ∆pCO2 and flux has been neglected in this study. Summary and Conclusions A database for the sea–air pCO2 difference, ∆pCO2, has been assembled using about 250,000 observations made between 1960 and 1995 during 250 expeditions over the global oceans. Observations made in the equatorial Pacific during El Nino events have been excluded. In light of the sparseness of observations over large oceanic areas, the multiyear data have been corrected and combined to represent a single reference year of 1990. These observations have been organized into 4° latitude × 5° longitude × 1 day pixels for 365 days, and interpolated in space and time using a computational scheme based on the diffusive and advective transport of surface water (63). On the basis of the global distribution of ∆pCO2 values thus computed, a global net ocean uptake of 0.60 to 1.34 Gt-C·yr−1 is obtained for 1990 using three different formulations for the gas transfer coefficient. This is similar to 1.6 ± 0.9 Gt-C·yr−1 estimated on the basis of 13C changes in the atmosphere and ocean (2), but is smaller than about 2 Gt-C·yr−1 based on various ocean–atmosphere perturbation models (3, 4, 5, 6, 7 and 8). However, it is greater than the estimates based on the atmospheric CO2 distribution and mass balance (1). The Pacific equatorial belt is the largest oceanic CO2 source to the atmosphere. The temperate oceanic areas of the both hemispheres are the most important sinks, and their uptake fluxes exceed those of high latitude oceans (poleward of the 50° parallel) by a factor of 2 to 3. Among the four ocean basins, the Atlantic Ocean (north of 50°S) is the strongest sink providing about 60% of the total global ocean uptake, whereas the Pacific (north of 50°S) is nearly neutral. The Indian and Southern Oceans contribute about 20% each to the global uptake flux. The uptake flux by the North Pacific is similar to that by the South Pacific, whereas the North Atlantic takes up 0.45 to 0.6 Gt-C·yr−1 more CO2 than the South Atlantic and Southern Ocean combined. This is comparable to the preindustrial oceanic transport of CO2 from the North to South Atlantic estimated by Broecker and Peng (11). Because the results of this study differ significantly from the northern ocean uptake of CO2 used in the analysis of the atmospheric and oceanic data by Tans et al. (1), a global analysis of the CO2 and carbon isotope data in the ocean, atmosphere, and biosphere must be made to evaluate their mutual coherence. This study has benefited from observations made by many international investigators. We thank the following scientists for their contributions: W. S. Broecker, A. W. Dickson, R. H. Gammon, S. S. Jacobs, Walker Smith, H. Ducklow, W. M. Smethie, D. Martinson, P. Schlosser, J. Sarmiento, D. Wallace, E. Garvey, N. R. Bates, A. H. Knap, T. D. Foster, C. S. Wong, C. D. Keeling, L. Merlivat, C. LeQuere, V. Garcon, C. Provost, W. Roether, H. Y. Inoue, K. Fushimi, A. Watson and J. E. Robertson. We also thank members of our technical staff who ran instruments at sea and on land and processed data in our respective laboratories: J. Goddard, S. Rubin, R. Esmay, F. A. van Woy, P. K. Salameh, P. P. Murphy, K. C. Kelly, L. S. Waterman, Matt Steckley and David Ho. This work was supported by a number of grants from the National Science Foundation, the U.S. Department of Energy, and the National Oceanic and Atmospheric Administration to T.T. at the Lamont–Doherty Earth Observatory and to R.F.W. at the Scripps Institution of Oceanography. R.A.F. and R.H.W. have been supported by the Climate and Global Change Program of the National Oceanic and Atmospheric Administration. We gratefully acknowledge the support and encouragement received from these agencies. This is contribution 5573 of the Lamont–Doherty Earth Observatory. 1. Tans, P. P., Fung, I. Y. & Takahashi, T. (1990) Science 247, 1431–1438. 2. Quay, P. D., Tilbrook, B. & Wong, C. S. (1992) Science 256, 74–79. 3. Siegenthaler, U. & Sarmiento, J. L. (1993) Nature (London) 365, 119–125. 4. Oeschger, H., Siegenthaler, U. & Gugelmann, A. (1975) Tellus 27, 168–192. 5. Broecker, W. S. & Peng, T.-H. (1982) Tracers in the Sea (Eldigio, Palisades, NY). 6. Bacastow, R. & Maier-Reimer, E. (1990) Clim. Dyn. 4, 95–125. 7. Sarmiento, J. L., Orr, J. C. & Siegenthaler, U. (1992) J. Geophys. Res. 97, 3621–3645. 8. Sarmiento, J. L., Murnane, R. & Le Quere, C. (1995) Philos. Trans. R. Soc. London B 343, 211–219. 9. Tans, P. P. Berry,J. A. & Keeling R. (1993) Global Biogeochem. Cycles 7, 353–368. 10. Denning, A. S., Fung, I. Y. & Randall, D. (1995) Nature (London) 376, 240–243. 11. Broecker, W.S. & Peng, T-H. (1992) Nature (London) 356, 587–589. 12. Chipman, D. W., Marra, J. & Takahashi, T. (1993) Deep-Sea Res. 40, 151–169. 13. Archer, D. E., Takahashi, T., Sutherland, S., Goddard, J., Chipman, D., Rogers, K. & Ogura, H. (1996) Deep-Sea Res. PArt II 43, 779–808. 14. British Oceanographic Centre (1994) BOFS North Atlantic Data Set (CD-ROM) (British Oceanographic Centre, Bidson Observatory, Birkenhead, Merseyside, U.K.). 15. Chipman, D. W., Takahashi, T., Breger, D. & Sutherland, S. C. (1992) Investigation of Carbon Dioxide in the South Atlantic and Northern Weddell Sea Areas (WOCE Sections A-12 and A-21) during the Meteor Expedition 11/5, January–March, 1990 (Lamont–Doherty Geological Observatory, Palisades, NY). 16. Feely, R. A., Wanninkhof, R., Cosca, C. E., Murphy, P. P., Lamb, M. F. & Steckley, M. D. (1995) Deep-Sea Res. Part II 42, 365–386. 17. Goyet, C., Beauverger, C., Brunet, C. & Poisson, A. (1991) Tellus B 43, 1–11. 18. Goyet, C. & Peltzer, E. T. (1994) Marine Chem. 45, 257–266. 19. Inoue, H. Y. & Sugimura, Y. (1988) Tellus B 40, 308–320. 20. Inoue, H. Y. & Sugimura, Y. (1992) Tellus B 44, 1–22. 21. Inoue, H. Y., Mastueda, H., Ishii, M., Fushimi, K., Hirota, M., Asanuma, I. & Takasugi, Y. (1995) Tellus B 47, 391–413. 22. Ishii, M. & Inoue, H. Y. (1995) Tellus B 47, 447–460. 23. Metzl, N., Poisson, A., Louanchi, F., Brunet, C., Schauer, B. & Bres, B. (1995) Tellus B 47, 56–69. 24. Murphy, P. P., Feely, R. A., Gammon, R. H., Harrison, D. E., Kelly, K. C. & Waterman, L. S. (1991) J. Geophys. Res. 96, 20,455–20,465. 25. Murphy, P. P., Kelly, K. K., Feely, R. A., & Gammon, R. H. (1995) Carbon Dioxide Concentrations in Surface Water and the Atmosphere during 1986–1989 NOAA/PMEL Cruises in the Pacific and Indian Oceans (Oak Ridge Natl. Lab., Oak Ridge, TN), Rep. NRNL/CDIAC-75, NDP-047.
GLOBAL AIR-SEA FLUX OF CO2: AN ESTIMATE BASED ON MEASUREMENTS OF SEA–AIR PCO2 DIFFERENCE
8299
26. Poisson, A., Metzl, N., Brunet, C., Schauer, B. Bres, B., Buiz-Pino, D. & Louanchi, F., (1993) J. Geophys. Res. 98, 22759–22778. 27. Poisson, A., Metzl, N., Danet, X., Louanchi, F., Brunet, C., Schauer, B., Bres, B. & Ruiz-Pino, D. (1994) in The Polar Oceans and Their Role in Shaping the Global Environment, eds. Johannessen, O. M., Muench, R. D. & Overland, J. E. (Am. Geophys. Union, Washington, DC), Geophysical Monograph 85, pp. 273–284. 28. Robertson, J. E., Watson, A. J., Langdon, C. Ling, R. D. & Wood, J. W. (1993) Deep-Sea Res. 40, 409–422. 29. Rubin, S., Goddard, J., Chipman, D. W. & Takahashi, T. (1996) Antarct. J. USA 31, 113–115. 30. Takahashi, T., Goddard, J. G., Rubin, S., Chipman, D.W. & Sutherland, S. C. (1993) Investigation of Carbon Dioxide in the Central South Pacific Ocean (WOCE Sections P-16C and P-17C) During the TUNES/2 Expedition of the R/V Thomas Washington, July–August, 1991 (Lamont– Doherty Earth Observatory, Columbia Univ., Palisades, NY). 31. Takahashi, T., Olafsson, J., Goddard, J., Chipman, D.W. & Sutherland, S. C. (1993) Global Biogeochem. Cycles 7, 843–878. 32. Wanninkhof, R., Feely, R. A., Atwood, D. K., Berberian, G., Wilson, D., Murphy, P. P. & Lamb, M. F. (1995) Deep-Sea Res. Part II 42, 387–409. 33. Watson, A. J., Robinson, C., Robertson, J. E., Williams, P. J. le B. & Fasham, M. J. R. (1991) Nature (London) 350, 50–53. 34. Weiss, R. F., van Woy, F. A. & Salameh, P. K. (1992) Surface Water and Atmospheric Carbon Dioxide and Nitrous Oxide Observations by Shipboard Automated Gas Chromatography: Results from Expeditions Between 1977 and 1990 (Scripps Inst. of Oceanography, La Jolla, CA), SIO Report 92-11, ORNL/CDIAC-59, NPD-044. 35. Japan Meteorological Agency (1994) WMO World Data Center for Greenhouse Gases, Sample Issue, CD-ROM (Japan Meteorological Agency, Tokyo). 36. Wong, C. S. & Chan, Y.-H. (1991) Tellus B 43, 206–223. 37. Wong, C. S., Chan, Y.-H., Page, J. S., Smith, G. E., & Bellegay, R. D. (1993) Tellus B 45, 64–79. 38. Karl, D. M., Tilbrook, B. D. & Tien, G. (1991) Deep-Sea Res. 38, 1097–1126. 39. Broecker, W.S. & Takahashi, T. (1966) J. Geophys. Res. 71, 1575–1602. 40. Wanninkhof, R. & Thoning, K. (1993) Marine Chem. 44, 189–204. 41. Weiss, R. F. & Price, B. A. (1980) Marine Chem. 8, 347–359. 42. Andrie, C., Genthon, C. & Merlivat, L. (1986) J. Geophys. Res. 91, 11741–11755. 43. Chipman, D.W., Takahashi, T. & Sutherland, S.C. (1986) Carbon Chemistry of the South Atlantic Ocean and the Weddell Sea: The Results of the Atlantic Long Lines (AJAX) Expeditions, October 1983–February, 1984 (Lamont–Doherty Geological Observatory, Palisades, NY). 44. Chipman, D. W. & Takahashi, T. (1990) Investigation of Carbon Chemistry in the Weddell Sea Area During the 1986 Winter Expedition of the F/S Polarstern: June 28, 1986–September 16, 1986 (Lamont–Doherty Geological Observatory, Palisades, NY). 45. Kelley, J. J., Jr. (1970) Limnol. Oceanograph. 15, 80–87. 46. Kelley, J. J., Longerich, L. L. & Hood, D. W. (1971) J. Geophys. Res. 76, 8689–8693. 47. Peng, T.-H., Takahashi, T., Broecker, W. S. & Olafsson, J. (1987) Tellus B 39, 439–458. 48. Roos, M. & Gravenhorst, G. (1984) J. Geophys. Res. 89, 8,181–8,193. 49. Smethie, W. M., Takahashi, T., Chipman, D. & Ledwell, J. R. (1985) J. Geophys. Res. 90, 7005–7022. 50. Takahashi, T., Goddard, J., Chipman, D. W., Sutherland, S. C. & Mathieu, G. (1991) Assessment of Carbon Dioxide Sink/Source in the North Pacific Ocean: Seasonal and Geographic Variability 1984–1989 (Lamont–Doherty Geological Observatory, Palisades, NY). 51. Takahashi T., Chipman, D., Schechtman, N., Goddard, J. & Wanninkof, R. (1982) Measurements of the Partial Pressure of CO2 in Discrete Water Samples during the North Atlantic Expedition, the Transient Tracers of Oceans Project (Lamont–Doherty Geological Observatory, Palisades, NY). 52. Takahashi, T. & Chipman, D. (1982) Antarctic J. USA 17, 103–104. 53. Takahashi, T., Chipman, D. W., Smethie, Wm. Jr., Goddard, J. Trumbore, S., Mathieu, G. G. & Sutherland, S. C. (1985) Assessment of Carbon Dioxide Sink/Source in the Oceanic Areas: The Results of 1982–84 Investigation (Lamont–Doherty Geological Observatory, Palisades, NY). 54. Takahashi, T., Olafsson, J., Broecker, W. S. Goddard, J., Chipman, D. & White J. (1985) J. Mar. Res. Institute Reykjavik, 9, 20–36. 55. Takahashi, T., Goddard, J., Sutherland, S., Chipman, D. W., & Breeze, C. (1986) Seasonal and Geographic Variability of Carbon Dioxide Sink/Source in the Oceanic Areas: Observations in the North and Equatorial Pacific Ocean 1984–1986 and Global Summary (Lamont–Doherty Geological Observatory, Palisades, NY). 56. Takahashi, T., Chipman, D. W., Goddard, J., Mathieu, G. & Ma, L.-M. (1990) in Sea–Air Interaction in Tropical Western Pacific, eds. Chao J.-P. & Young, J.A. (China Ocean Press, Beijing, PRC), pp. 511–539. 57. Takahashi, T., Goddard, J., Chipman, D. W., Sutherland, S. C. & Mathieu, G. (1991) Assessment of Carbon Dioxide Sink/Source in the North Pacific Ocean: Seasonal and Geographic Variability 1984–1989 (Lamont–Doherty Geological Observatory, Palisades, NY). 58. White, W. B. & Peterson, R.G. (1996) Nature (London) 380, 699–702. 59. Trenberth, K. E. & Hoar, T. J. (1996) Geophys. Res. Lett. 23, 57–60. 60. Takahashi, T., Chipman, D. & Volk, T. (1983) Proceedings: Carbon Dioxide Research Conference: Carbon Dioxide, Science and Consensus (U.S. Department of Energy, Washington, DC), Vol. 2, Rep. CONF-820970, pp. 123–145. 61. Keeling, C. D. & Whorf, T. P. (1994) in Trends '93: A Compendium of Data on Global Change, eds. Boden, T. A., Sepanski, R. J. & Stoss, F.W. (Carbon Dioxide Information Analysis Center, Oak Ridge Natl. Lab., Oak Ridge, TN), Rep. ORNL/CDIAC-65, pp. 16–26. 62. Conway, T. J., Tans, P. P. & Waterman, L. S. (1994) in Trend '93: A Compendium of Data on Global Change, eds. Boden, T.A., Kaiser, D. P., Sepanski, R. J. & Stoss, F. W. (Carbon Dixoide Information Analysis Center, Oak Ridge National Laboratory, Oak Ridge, TN), Rep. ORNL/ CDIAC-65, pp. 41–119. 63. Takahashi, T., Takahashi, T. T. & Sutherland, S. C. (1995) Philos. Trans. R. Soc London B 348, 143–152. 64. Bryan, K. & Lewis, L. J. (1979) J. Geophys. Res. 84, 2503–2517. 65. Bretherton, F. P. & Karweit, M. (1975) Proceedings of Symposium on Numerical Models of Ocean Circulation (Natl. Acad. of Sci., Washington, DC), pp. 237–249. 66. Thiele, G., Roether, W., Schlosser, P., Kuntz, R., Siedler, G. & Stramma, L. (1986) J. Phys. Oceanogr. 16, 814–826. 67. Jenkins, W. J. (1991) J. Phys. Oceanogr. 21, 1058–1061. 68. Shea, D. J., Trenberth, K. E. & Reynolds, R. W. (1992) J. Clim. 5, 987–1001. 69. Broecker, W. S., Ledwell, J. R., Takahashi, T., Weiss, R. F., Merlivat, L., Memery, L., Peng, T.-H., Jahne, B. & Munnich, K. O. (1986) J. Geophys. Res. 91, 10517–10527. 70. Wanninkhof, R. (1992) J. Geophys. Res. 97, 7373–7382. 71. Liss, P. S. & Merlivat, L. (1986) in The Role of Air-Sea Exchange in Geochemical Cycling, ed. Buat-Menard, P. (Reidel, Hingham, The Netherlands), pp. 113–127. 72. Esbensen, S. K. & Kushnir, Y. (1981) The Heat Budget of the Global Ocean: An Atlas Based on Estimates from the Surface Marine Observations (Oregon State Univ., Corvalis, OR), Climatic Research Institute Report 29. 73. Boutin, J. & Etcheto, J. (1991) Tellus B 43, 236–246. 74. van Scoy, K. A., Morris, K. P., Robertson, J. E. & Watson, A. J. (1995) Global Biogeochem. Cycles 9, 253–262. 75. Sarmiento, J. L. & Sundquist, E. T. (1992) Nature (London) 356, 589–593.
CHARACTERISTICS OF THE DEEP OCEAN CARBON SYSTEM DURING THE PAST 150,000 YEARS: ΣCO2 DISTRIBUTIONS, DEEP WATER FLOW PATTERNS, AND ABRUPT CLIMATE CHANGE
8300
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8300–8307, August 1997 Colloquium Paper This paper was presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held November 13–15, 1995, at the National Academy of Sciences, Irvine, CA.
Characteristics of the deep ocean carbon system during the past 150,000 years: CO2 distributions, deep water flow patterns, and abrupt climate change EDWARD A. BOYLE Department of Earth, Atmospheric, and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139
© 1997 by The National Academy of Sciences 0027-8424/97/948300-8$2.00/0 PNAS is available online at http://www.pnas.org. ABSTRACT Studies of carbon isotopes and cadmium in bottom-dwelling foraminifera from ocean sediment cores have advanced our knowledge of ocean chemical distributions during the late Pleistocene. Last Glacial Maximum data are consistent with a persistent high-ΣCO2 state for eastern Pacific deep water. Both tracers indicate that the mid-depth North and tropical Atlantic Ocean almost always has lower ΣCO2 levels than those in the Pacific. Upper waters of the Last Glacial Maximum Atlantic are more ΣCO2-depleted and deep waters are ΣCO2-enriched compared with the waters of the present. In the northern Indian Ocean, δ13C and Cd data are consistent with upper water ΣCO2 depletion relative to the present. There is no evident proximate source of this ΣCO2-depleted water, so I suggest that ΣCO2-depleted North Atlantic intermediate/deep water turns northward around the southern tip of Africa and moves toward the equator as a western boundary current. At long periods ( 15,000 years), Milankovitch cycle variability is evident in paleochemical time series. But rapid millennial-scale variability can be seen in cores from high accumulation rate series. Atlantic deep water chemical properties are seen to change in as little as a few hundred years or less. An extraordinary new 52.7-m-long core from the Bermuda Rise contains a faithful record of climate variability with century-scale resolution. Sediment composition can be linked in detail with the isotope stage 3 interstadials recorded in Greenland ice cores. This new record shows at least 12 major climate fluctuations within marine isotope stage 5 (about 70,000–130,000 years before the present). On time scales exceeding millennia, the ocean carbon system regulates atmospheric CO2 levels. The gas trapped as bubbles in polar ice cores shows that CO2 levels varied between 190 and 280 ppmV during glaciation cycles and, hence, that some aspects of the oceanic carbon system are naturally variable. Several ideas on how the ocean produces these changes in atmospheric CO2 have been advanced. Although many ideas have merit because they call attention to processes that regulate atmospheric CO2, it has proven difficult to come up with a widely accepted model for the dominant cause of glacial/interglacial CO2 variability. Despite this persistent sticking point, significant progress has been made concerning the distribution of metabolically regenerated CO2 throughout the deep ocean. The past state of the oceanic CO2 distribution is recorded by carbon isotopes and phosphorus-analog Cd as they are incorporated into shells of benthic foraminifera preserved in deep-sea sediments. Marine organic matter is enriched in 12C and Cd, so that surface waters are depleted in 12C and Cd because of their removal by plant growth, and deep waters are enriched in 12C and Cd as debris from those plants (and animals that eat them) decompose in deeper waters. From global data on the distribution of metabolic CO2 and its analogs, we can infer characteristics of thermohaline spreading patterns. This paper will first briefly review significant aspects of Last Glacial Maximum (LGM) ocean chemical distributions in major ocean basins, emphasizing major areas of agreement and discordance, and then tie this evidence together into a unifying hypothesis for LGM deep-ocean circulation patterns. The paper concludes with new evidence concerning century-scale variability of North Atlantic climate during the past 150,000 years (15 kyr). LGM Nutrient Distribution in Deep Waters of Major Ocean Basins: A Brief Overview Eastern Tropical Pacific. A consensus view has dominated the past decade. Foraminiferal LGM δ13C was −0.3% lower than the level in the Holocene in the benthic foraminifera Uvigerina and Cibicidoides wuellerstorfi in the eastern tropical Pacific. This shift has been interpreted as due to a whole-ocean shift in δ13C caused by oxidation of continental organic matter during glacial periods (1, 2, 3, 4, 5 and 6). LGM benthic foraminiferal Cd/Ca was about 15% lower in the same region; this lowering has been attributed to higher Cd in the LGM Atlantic oceana (assuming a constant oceanic Cd inventory) (3, 6, 7). Both interpretations agree that this region remained near the end of the “conveyor belt” with high ΣCO2 due to its accumulation as sinking particles decompose. Two recent reports have uncovered potential problems with this widely adopted consensus. First, laboratory culturing experiments on planktonic foraminifera suggest that foraminiferal δ13C may be sensitive to pH (8); if the same sensitivity holds for benthic foraminifera and if pH of the LGM ocean rose as much as indicated by early reports on LGM “paleo-pH” [based on foraminiferal δ11B (9, 10)], then a significant portion of the eastern tropical Pacific δ13C signal could be due to changes in pH rather than a whole-ocean δ13C shift due to continental organic matter oxidation (11). Second, it has been noted that the oceanic Cd inventory is sensitive to changes in the extent of reducing conditions, where CdS precipitates from sedimentary pore waters (12, 13). Depending on assumptions, a portion of lower Cd in the eastern tropical Pacific could be caused by enhanced removal of Cd from the ocean during glacial times rather than by a shift of Cd into other water masses. Neither argument can be considered established, however, so the current consensus may well survive. North and Tropical Atlantic. Despite some early substantial disagreements, for the past 5 years the consensus view of LGM Atlantic paleochemistry has been that there was 13C enrichment/Cd depletion of upper tropical and North Atlantic waters (1–2 km) and 13C depletion/Cd enrichment in deeper waters (3, 4, 5 and 6
Abbreviations: LGM, Last Glacial Maximum; kyr, thousand years; NADW, North Atlantic deep water; AABW, Antarctic bottom water; GNAI/DW, glacial North Atlantic intermediate/deep water; GRIP, Greenland Ice Project; GISP2, Greenland Ice Sheet Project 2.
CHARACTERISTICS OF THE DEEP OCEAN CARBON SYSTEM DURING THE PAST 150,000 YEARS: ΣCO2 DISTRIBUTIONS, DEEP WATER FLOW PATTERNS, AND ABRUPT CLIMATE CHANGE
8301
km) (3, 4 and 5, 14, 15 and 16). This distribution has been attributed to a shoaling of high 13C/low Cd North Atlantic source waters compensated by a greater influx of low 13C/high Cd Antarctic bottom waters. This replacement of some NADW by glacial Antarctic bottom water (AABW) notwithstanding, mid-depth northern Atlantic waters (3 km) have higher 13C and lower Cd than do waters of the eastern tropical Pacific. The structure and temporal variability are complex in detail (17). Whether the LGM NADW conveyor belt was global in scope as it is today has been debated considerably (see ref. 18 for my recent review of this debate). The matter may have been settled recently by studies of 231Pa/230Th by Yu et al. (19). 231Pa generated from decay of 235U in the Atlantic is “missing” from Atlantic sediments. Yu et al. make a convincing case that this deficiency is due to NADW-borne transport of 231Pa out of the Atlantic into the Antarctic Circumpolar Current, where 231Pa is trapped into sediments at levels exceeding its regional production rate. Because this deficiency persists during the LGM, the conveyor must have continued to move 231Pa out of the Atlantic. Northern Indian Ocean. Kallel et al. (20) reported that upper waters of the northern Indian Ocean had higher levels of 13C than those at present. Naqvi et al. (21) questioned this interpretation because it was based upon a Geochemical Ocean Sections/LGM comparison rather than a core top/LGM comparison (because there are significant differences between Geochemical Ocean Sections δ13C and core top C. wuellerstorfi δ13C). More recently, foraminiferal Cd evidence from several species of aragonitic and calcitic benthic foraminifera was presented (22); this evidence shows that upper northern Indian Ocean nutrient depletion (strongest in the Arabian Sea) could be reconciled with the evidence of Naqvi et al. (21). The major problem following this observation is the difficulty in accounting for the cause of 13C enrichment/Cd depletion. The Red Sea cannot be the source because its sill was too shallow during the LGM (due to sea level depression). Other sources (Indonesian basins, Antarctic sources) cannot be ruled out entirely but seem unlikely. In the “Global Picture” below, I suggest that upper glacial North Atlantic intermediate/deep water (GNAI/DW) is the source of this low ΣCO2 water.
FIG. 1. Schematic diagram of global deep water circulation during the LGM. Parallelograms represent hydrographic sections, divided into upper deep and lower deep circulations, as indicated by arrows. x, marks a possible site of sinking; ?, a certain regional source whose specific formationsite is unknown; ??, a questionable source. Antarctic. The Antarctic has been a major sticking point for deep water paleoceanography. Most published LGM δ13C data show values that are much lower than those of today (in some cases lower than anywhere else in the ocean) (4, 23, 24 and 25). Cd evidence is also self-consistent but contradictory to δ13C evidence: Cd is either the same as it is today or slightly lower (7, 23, 24). So δ13C data indicate that Southern Ocean deep water was high in ΣCO2, whereas Cd data indicate that it was moderate or lower in ΣCO2. The δ13C evidence also is difficult to reconcile with lowered glacial atmospheric pCO2 (25). Attempts have been made to resolve this conundrum. These explanations can account for part of the discrepancy but not all (26). I have reviewed this situation recently (24) and argue that the Southern Ocean “Mackensen Effect,” whereby δ13C of C. wuellerstorfi is observed to be low under waters of high productivity (27), is stronger during the LGM, hence 13C data are too low. However, this solution is not universally agreed to by stable isotope paleoceanographers. Northwest Pacific. The northwest Pacific has been another difficult area; however, in this case the problem includes internal inconsistency for each tracer as well as tracer-to-tracer discrepancies. Some evidence is consistent with formation of a low-Σ,CO2 LGM deep water [e.g., Cd is consistently lower in LGM benthic foraminifera at all depths compared with levels in the eastern tropical Pacific (7), but the few available core top Cd measurements in this region are inexplicably low (7, 28)]. Some northwest Pacific sites show δ13C similar to that found in the eastern tropical Pacific (29), but other sites show values that are enriched in 13C (4). Global Picture. On the broad scale considered here, LGM deep water paleochemical studies have three hits and two misses: Cd and δ13C data are consistent and informative on LGM chemical distributions in the eastern tropical Pacific, North and equatorial Atlantic, and northern Indian Ocean, but disagree substantially in the Southern Ocean and northwest Pacific. If the Southern Ocean disagreement is attributed to a productivity-related artifact in δ13C (and Cd evidence is accepted), and if the northwest Pacific is left as an open question, a synthesis of LGM paleochemical data can be offered as a testable hypothesis (Fig. 1). In Fig. 1, major flows are represented in “sections” along western boundaries of major ocean basins, with flow divided into upper deep and lower deep sections (approximately 1.5–2.5 and 2.5–5 km). Areas where bottom water is possibly formed, formation regions that are uncertain as to exact location but must have occurred within some broad region, and formation regions that are possible but
CHARACTERISTICS OF THE DEEP OCEAN CARBON SYSTEM DURING THE PAST 150,000 YEARS: ΣCO2 DISTRIBUTIONS, DEEP WATER FLOW PATTERNS, AND ABRUPT CLIMATE CHANGE
8302
not necessarily required are marked in Fig. 1. It is assumed that recirculation flows fill the basin interiors. The most important and well documented change in this LGM global circulation is the shift of CO2-depleted NADW to shallow depths, replaced in the deepest waters by a northward flow of high-CO2 Antarctic water. The exact source area of this NADW is not known but is certain to be found somewhere in northern reaches of the Atlantic. Because the 231Pa/230Th data of Yu et al. (19) require a net export of water from the Atlantic, this shallow GNAI/DW must penetrate into the Antarctic Circumpolar Current. As suggested by Toggweiler et al. (30), a shallower variety of North Atlantic deep water will have a more difficult time reaching high latitudes of the Southern Ocean where it can be recycled into the bottom water. . . . Instead, any high-13C water emerging beyond the tip of Africa might simply flow eastward into the Indian and Pacific basins, bypassing the Antarctic entirely. Although persistently low Cd in Antarctic deep waters is evidence against complete elimination of an Atlantic influence on Antarctic deep water chemistry, there nevertheless should be a tendency for less effective southward penetration of a shallower GNAI/DW. Extending Toggweiler's logic, I suggest here that much of the GNAI/DW that enters the northern portion of the Antarctic Circumpolar Current folds around the southern tip of Africa and moves northward as a western boundary current in the 1–2 km depth range. This flow can explain the low-ΣCO2 waters in the upper waters of the northern Indian Ocean. This circulation pattern also explains why the LGM Arabian Sea—proximate to the western boundary source—is lower in ΣCO2 than the Bay of Bengal, which has older recirculated waters. Part of the circulation scheme must remain vague because of data inconsistencies. However, northward penetration of high-ΣCO2 water from the Antarctic into the North Atlantic attests to formation of glacial AABW somewhere. There is little evidence on where in the Antarctic this glacial bottom water forms. Today, most primary AABW forms in on the continental shelf in the Weddell Sea with a lesser contributions from the Ross Sea and other sources (31). LGM sea level lowering of 120–130 m (32) and probable northward extension of year-round sea ice (33) may weaken this source. Other sources of AABW may replace the Weddell Sea; the Ross Sea and Southern Indian Ocean are possible replacements. Michel et al. (34) and Rosenthal (23) have suggested that bottom water formation may occur in the Indian Ocean sector, although they envision quite different initial properties. Michel et al. argue for formation of high-ΣCO2 bottom water (based on very light LGM foraminiferal δ13C observed in the Indian sector), whereas Rosenthal argues for nutrient-depleted deep water formation (based on Cd evidence). Wherever this Antarctic water forms, it must also enter the Pacific as it does today. The major question for LGM Pacific circulation is whether this AABW simply ages and returns southward as deep water (as it does today) or is supplemented by additional sources in the North Pacific. When possible sites of formation are mentioned, the Sea of Okhotsk and Bering Sea figure prominently. If deep water forms, it may be an upper level intermediate/deep water as seen in the glacial North Atlantic. Although there are loose ends to this story, a coherent picture of LGM Atlantic and Indian Ocean chemical characteristics can be fashioned from simple premises. For these regions, the scientific challenge is to fill in details of this picture and test it against other observations, such as 14C ventilation rates. In other regions, major advances are required to develop even a simplified picture of the circulation; however, some initial targets are obvious. (i) Why do δ13C and Cd disagree with one another in the Southern Ocean? (ii) Can a coherent picture of chemical distributions in the northwest Pacific be fashioned from either tracer? Rapid Climate Variability in the North Atlantic High Resolution Climate Change in the Ocean: A Brief Review. Just as we have a much more detailed picture of LGM chemical characteristics and inferred circulation of the Atlantic, we also have a better view of temporal variability of these characteristics. On longer time scales, paleochemical indicators show clear influence of Milankovitch 100–41-23 kyr orbital cycles; the Spectral Mapping Project (SPECMAP) investigated the relation of deep water cycles to orbital forcing and other responses in the climate system (30, 35). On shorter time scales, there is growing interest in decadal-millennial deep water variability. Greenland ice core evidence (36, 37, 38 and 39) shows that climate can shift abruptly between warmer and colder states. Broecker and colleagues (40, 41, 42, 43, 44, 45, 46, 47 and 48) have suggested that global deep water conveyor belt flow may be a prime driver of these changes, and hence there is significant effort directed at establishing deep water changes at high temporal resolution within an accurate time scale. The major challenge in this effort is that temporal resolution is limited by the low sedimentation rate of deep-sea sediments (typically only a few centimeters per thousand years) and biological reworking of the upper several centimeters of sediment. To study events shorter than millennia, it is necessary to work in regions of unusually high accumulation rate. Higher accumulation rates occur on continental margins, although slumps and other sedimentary disturbances occur almost as frequently, making it difficult to find undisturbed sections. High accumulation rates also occur in deep-sea “drift” deposits, where preferential scouring of fine-fraction material over a broad region combines with preferential deposition where transport weakens. Although these deposits are not common, in the North Atlantic there is at least one drift deposit at the depth of every major water mass and source. Hence, the search for short-term ocean climate variability has focused on these drift deposits as well as special regions of continental margins where undisturbed deposition can be documented [e.g., Cariaco Trench (49), Santa Barbara Basin (50), and others]. “Dansgaard-Oeschger” Interstadial events clearly recorded in Greenland ice cores are matched by events in foraminiferal species abundances North Atlantic cores (51, 52). Although the chicken–egg question of cause and effect is not resolved yet, a linkage between abrupt climate variability in Greenland and the surface North Atlantic is clearly demonstrated. The degree to which deep Atlantic circulation is involved is an open question. In the northern Atlantic where these events have their strongest expression in surface water properties, deep water paleochemical signals can be small even for extrema such as the LGM (18, 53), and benthic foraminifera are often scarce and sporadic, making it hard to construct a continuous high-resolution time series at these sites. Curry and Oppo (68) has evidence suggesting that at least the larger of these events influences the carbon isotope composition of benthic foraminifera from the tropical Atlantic. To answer the question more definitively, a benthic paleochemical record from a very high deposition site with a precise and accurate chronology is required. Bermuda Rise Paleoclimatology. For several years now, Lloyd Keigwin and I have worked on drift deposits of the Bermuda Rise, where accumulation rates vary from 10–20 cm/kyr (during warm climate periods) to 100–200 cm/kyr (during extreme glaciation). Located some 335 km northeast of Bermuda, sedimentation in this region is enhanced by fine-grained material eroded from the North American continental margin and transported to this site by energetic bottom currents (which weaken and allow fine-grained sediment to settle out). The bottom currents focus fine-grained sediments (clays, other detrital sediments, and coccoliths) onto this site (enhancing the accumulation rate several-fold), whereas coarse-grained materials such as planktonic and benthic foraminifera accumulate at a normal oligotrophic ocean rate (54, 55 and 56). The high accumulation rate increases temporal resolution and allows for an examination of
CHARACTERISTICS OF THE DEEP OCEAN CARBON SYSTEM DURING THE PAST 150,000 YEARS: ΣCO2 DISTRIBUTIONS, DEEP WATER FLOW PATTERNS, AND ABRUPT CLIMATE CHANGE
8303
century-scale (and shorter) climate variability. In addition to high temporal resolution, the Bermuda Rise has two other characteristics that make it appropriate for study of ocean climate changes. (i) It lies beneath oligotrophic subtropical waters, allowing it to serve as a representative of a large and typical ocean province; few other drift deposits are so located. (ii) The bottom depth is 4,400 m, giving benthic foraminifera a sensitive perspective on changes in lower North Atlantic deep water. Dilution of coarse-grained foraminiferal signal carriers has the disadvantage, however, that drift deposits require much larger sample sizes. Keigwin and Jones provided a precise accelerator mass spectrometry 14C-based chronology for the upper portion of the Bermuda Rise (57, 58); below the zone of radiocarbon dating, correlation of features in the benthic oxygen isotope curve to the orbitally tuned SPECMAP time scale extends the time scale. Analysis of the percentage of CaCO3 (%CaCO3) at this site (58) and at another drift deposit on the Bahamas Outer Ridge shows typical Atlantic warm high-carbonate/cold low-carbonate pattern. High temporal resolution reveals numerous millennial-scale events that are similar in frequency, timing, and pattern to “interstadial” events observed in the Renland (Greenland) ice core (58). Finally, in a study of the most recent deglaciation, Keigwin and I (15, 59) showed that surface and deep water properties are punctuated by a series of at least four millennialscale events. The most recent of these events is the “Younger Dryas” cooling, which occurred 13,000–11,500 calendar years ago, where both benthic Cd and δ13C indicate a strong and clear diminution of the percentage of low-ΣCO2 lower NADW. This evidence supports the idea of a linkage between thermohaline circulation and abrupt cooling during the Younger Dryas. The middle two Cd events coincide with pulses in freshwater fluxes into the Gulf of Mexico during deglaciation, and the earliest Cd event coincides with the most recent “Heinrich Event” pulse of detritus from icebergs into the northern North Atlantic. Hence this limited study of surface and deep water characteristics—combined with the extended record of %CaCO3—suggests that abrupt variability may be a major characteristic of subtropical surface and deep waters during glaciations and deglaciations. This case study of deglaciation provides evidence for millennial-scale climate variability during glaciations and deglaciations, but there is more significant public policy concern about the possibility of abrupt changes during warm climate periods. This concern was amplified by the report that extreme abrupt events characterized the last interglacial (“Eemian” or oxygen isotope stage “5e”) in the central Greenland Ice Project (GRIP) ice core (37). Subsequently, concerns were raised about the fidelity of deeper portions of the GRIP record because (i) the interval in question is not replicated in the nearby Greenland Ice Sheet Project 2 (GISP2) ice core, (ii) inclined beds (and probable folding disturbances) are seen in the GISP2 core, suggesting that the deepest portions of ice cores near bedrock are subject to stratigraphic disturbances, and (iii) there is no sensible correlation of air bubble δ18O2 record in deeper parts of both GISP2 (60) and GRIP (61) ice cores with that in the stratigraphically impeccable Antarctic Vostok ice core (62), suggesting that neither Greenland ice core contains a continuous undisturbed Eemian sequence. However, concern raised by the original report made it evident that we have very little detailed paleoclimatic information from previous warm interglacial periods, such as the Eemian. Although there are valid grounds for doubting GRIP Eemian fluctuations, little is known about the stability of stage 5e at high temporal resolution. Hence a surge of activity has been directed at the last interglacial period. Long Sediment Cores for Decadal- to Century-Scale Paleoclimatology. Because this question requires high temporal resolution, sediment drift deposits such as the Bermuda Rise are prime targets for studies of ocean climate characteristics of the last interglacial period. However, the desirable high sedimentation rates create a major problem; on the Bermuda Rise, the last interglacial period is beyond reach of most piston cores, which are rarely longer than 20 m, and more typically are considerably shorter. Deeper sediments can be obtained by the Ocean Drilling Project, but these come as narrow-diameter, discontinuous, 9-m sections. Recently, a superior alternative for studying sediment in the 20- to 50-m (below seafloor) depth zone has been developed. Working on the large French supply/research vessel Marion Dufresne, Yvon Balut has developed a capability for taking cores longer than 50 m with a giant large-diameter Kullenbergstyle piston corer. The capability draws upon vessel size and shape (the new Marion Dufresne is 120.5 m long with a displacement of 10,380 metric tons, and has a long straight starboard working rail), a highcapacity (18 tons) kevlar-cable winch, and a unique core handling system that allows for the deployment of 60-m-long core barrels. One such core, 51.7 m long, collected in the Indian Ocean in 1990 has already been described (63). Inspired by that success, a proposal was initiated in 1992 to bring the ship to the Bermuda Rise to obtain a high-resolution record of the last 15 kyr. The funding contribution from this U.S. National Science Foundation project was then combined with other international projects (with financial contributions from Canada, Germany, Britain, and France, as well as scientific personnel contributions from several other countries) to form the 1995 “IMAGES” coring expedition throughout the North Atlantic Ocean. During a 22-h period in the vicinity of the Bermuda Rise, three cores were collected during this program: MD95-2034 (47.2 m long, obtained in a 60-m core barrel), MD95-2035 (39.2 m long, obtained in a 40-m core barrel), and MD95-2036 (52.7 m long, recovered in a 60-m core barrel). MD95-2034 (33°41.46N, 57°34.54W, 4,461 m uncorrected) and MD95-2036 (33°41.44N, 57°34.55W, 4,461 m uncorrected) were collected at essentially the same site, and MD95-2035 (33°29.195, 57°53.157, 4,286 m uncorrected) was collected southeast of the other two cores in an area of somewhat lower deposition. Shipboard measurements of whole-core magnetic susceptibility, GRAPE (gamma ray absorption porosity estimation), and sound velocity were made. Cores MD95-2034 and MD95-2035 were split on board, photographed, and described, and their optical reflectance spectral properties were logged by a Minolta CM 2002 spectrophotometer with a 1-cm-diameter spot size which records 31 bands from the 400- to 700-nm wavelength. It was evident from shipboard magnetic susceptibility records that MD95-2034 and MD95-2036 were essentially identical apart from a few coring disturbances in upper sections. From reflectance spectrometry, a preliminary stratigraphy was developed for MD95-2034 as described below, the stage 5e section of MD95-2034 identified, and then the matching stage 5e section in MD95-2036 was identified by matching magnetic susceptibility profiles. This section of MD95-2036 was then split on board, optical reflectance were recorded, and then sampled were taken at 1-cm intervals; burrows and other disturbances that were visually evident were avoided. These samples were brought back to Massachusetts Institute of Technology within a week of coring, and they were immediately processed to begin foraminifera picking for isotope and geochemical analysis. Optical Spectrophotometry of Bermuda Rise Cores: Introduction. Because dominant contributors to optical characteristics of sediments in this region are calcium carbonate (white) and clays (dark), visible light reflectance is a function of sedimentary %CaCO3. As shown by Keigwin and Jones (58), %CaCO3 at this site records millennial-scale climate events with high resolution. Hence, optical reflectance spectrophotometry is of great utility in establishing stratigraphy of cores in this region. By matching chemically measured % CaCO3 records from the previous longest core from this site (KNR31-GPC5, 29 m) and optical reflectance records, the stratigraphy of these new cores can be quickly established. Beyond the interval covered by KNR31-GPC5, it is expected that the boundary between the last interglacial stage 5e and the previous glacial maximum (isotope stage 6) will be
CHARACTERISTICS OF THE DEEP OCEAN CARBON SYSTEM DURING THE PAST 150,000 YEARS: ΣCO2 DISTRIBUTIONS, DEEP WATER FLOW PATTERNS, AND ABRUPT CLIMATE CHANGE
8304
marked by a strong decrease in %CaCO3 (as on the transition between LGM and Holocene). By this means, an approximate stratigraphy could be built up quickly on board ship, allowing the crucial stage 5e section to be identified and sampled immediately. Although spectral reflectance data can be represented in many different ways, in this work data will be presented using the Commission Internationale de l'Eclairage “Lab” system, which represents spectral data as three parameters that approximate a subjective linear human physiological visual response to overall lightness (L, higher numbers are brighter), red-to-yellow balance (a, higher numbers are more red than yellow), and yellow-to-blue balance (b, higher numbers are more yellow than blue) (64). On the Bermuda Rise, as noted above, the dominant spectral signature is the overall lightness signal created by changes in relative proportion of white CaCO3 and dark clays. The color variables a and b are interesting because the Bermuda Rise receives variable fluxes of hematite originally eroded from Canadian Maritime Province Permo-Carboniferous red beds, which impart a reddish hue to the sediments (“brick-red lutite”) (65). Variability of this component differs significantly from that of CaCO3 and, hence, provides independent information on climate change. The comparison of spectral lightness (L) for cores MD95-2034 and MD95-2036 with the %CaCO3 record of KNR31-GPC5 shows close correspondence of spectral reflectance to features of the %CaCO3 record (Fig. 2). This correspondence is most apparent in the interval from isotope stage 3 to stage 5a (1,700–2,800 cm in KNR31-GPC5; 1,700–3,600 cm in MD95-2034; and 1,900–3,900 cm in MD95-2036). This rapid and simple shipboard measurement provides as much stratigraphic information as the slower %CaCO3 measurements. Examination of lightness records leaves little doubt that both MD95-2034 and MD95-2036 contain complete records of marine oxygen isotope stage 5 and a considerable portion of isotope stage 6. The upper portion of MD95-2034 was disturbed during core recovery, as was MD95-2036 to a lesser extent, so comparison between lightness and %CaCO3 is less helpful in this interval. The greatly extended isotope stage 2 (where accumulation rates are 100–200 cm/kyr) is relatively featureless as represented here, but it should be noted that when freshly split, this interval contained many black bands that are relic Zoophycos horizontal burrow structures containing sulfides which are unstable in the presence of oxygen (Fig. 3). These black bands fade within hours of exposure to the atmosphere. To eliminate these stratigraphically irrelevant structures, the spectrophotometer operator avoided these structures and attempted to find spots that were more representative of sediment typical of a 5-cm interval.
FIG. 2. Comparison of spectrophotometric reflectance (L) data from MD95-2034 and MD95-2036 with the %CaCO3 record from KNR31-GGC5 (58). Numbers on the MD95-2036 plot are marine oxygen isotope stages (centered in respective stages). Note the close correspondence of optical and chemical fluctuations between cores. The L lightness fluctuations of MD95-2034 and MD95-2036 are shown in Fig. 2; these records are highly similar. The apparent differences arise because (i) the average sample spacing for MD95-2034 was 5.1 cm, whereas that for MD95-2036 was 3.3 cm (this is the main factor causing visually apparent differences; by sampling more frequently, finer-scale variability and analytical noise become more evident); (ii) except for three sections containing and adjacent to stage 5e, MD95-2036 was measured using the 3 mm × 4 mm Colortron spectrophotometer, whereas all of MD95-2034 was measured with the 1-cm-diameter Minolta spectrophotometer (the smaller spot size contributes to higher variability); (iii) MD95-2034 was split on board, and spectra were measured immediately; MD95-2036 was transported back to Massachusetts Institute of Technology, and splitting and spectral measurements were made several months after sample collection. This difference between cores split on board and later on shore probably does not significantly affect color records, as can be seen by a detailed comparison of isotope stage 5 color records from MD95-2034 (shipboard spectrophotometry) and MD95-2036 (spectrophotometry months after coring for the 3,600- to 4,200-cm interval). The oxygen isotope stage 5 reflectance records of MD95-2034 and MD95-2036 are compared on simple linear depth scales in Fig. 4. Apart from slight scale differences, records are nearly identical down to the centimeter scale; slight adjustments of depth
CHARACTERISTICS OF THE DEEP OCEAN CARBON SYSTEM DURING THE PAST 150,000 YEARS: ΣCO2 DISTRIBUTIONS, DEEP WATER FLOW PATTERNS, AND ABRUPT CLIMATE CHANGE
8305
control points would make it difficult to discern that two different records were plotted. This evidence attests to the fidelity of both records in recording regional sedimentation characteristics unmarred by burrowing or local stratigraphic disturbances. This comparison also shows that there is no significant accuracy offset between the two spectrophotometers and that no artifactural color changes occurred during storage of and transport of the unsplit sections of MD95-2036.
FIG. 3. Video image of one section of MD95-2036 from isotope stage 2 containing Zoophycos relic structures. Several bands occur within this interval; only the most prominent one is marked by the arrow. Optical Spectrophotometry of Bermuda Rise Cores: Paleoclimatological Implications. Fig. 5 compares reflectance properties of all three of the new Bermuda Rise cores. Because MD95-2034 was collected at a site of slightly lower sedimentation rate, it contains a record of longer duration despite its shorter length, extending back to isotope stage 10. It can be seen that all of these cores display high resolution approaching century time scales and that patterns are similar for each core. This comparison also provides a dramatic example of the effect of bioturbation where some detail evident in higher accumulation rate cores (which average about 27 cm/kyr for their entire record lengths) is lost in a core with an average sedimentation rate of “only” 11 cm/kyr. Data from MD95-2034 were mapped onto a preliminary time scale using radiocarbon and oxygen isotope dates from KNR31-GPC5 (58) supplemented by the assumption that the grayscale transition at the beginning of isotope stage 5 corresponds in age to the SPECMAP 6.0 oxygen isotope marker (Termination II) and that to a first approximation—certainly not true in detail—sedimentation is uniform in stage 5. These records can then be compared with ice core data from Greenland and Antarctica using ice core time scales developed by Bender et al. (60) (Fig. 6). Although temporal correlations will require further critical examination and development of a benthic oxygen isotope record for the earlier portion of the Bermuda Rise, at a first glance one can see continuation of the rich structure and correlation with number and timing of Renland ice core interstadials that Keigwin and Jones noted for stages 3 through 5a. With a more detailed isotope record from GRIP and GISP2, these climate correlations can now be extended to specific tentative correlations between interstadial events and CaCO3 cycles and extended to the earlier interval of isotope stage 5. Clearly, the %CaCO3 climate record at this site (as seen through spectral reflectance) reveals more structure than the less structured marine isotope 5a–e partition based on lower accumulation rate cores (67). Isotope stage 5 shows at least 11 identifiable subdivisions. The most recent of these correspond to Greenland Interstadials 19, 20, and 21. The earliest of these %CaCO3 peaks correspond to the beginning of marine stage 5e (based on preliminary δ18O analyses by L. D. Keigwin, personal communication) and probably also to the early “hump” in the Vostok δD record. Based on the same preliminary δ18O data, the sudden %CaCO3 drop at the end of this peak also appears to lie within the warm portion of isotope stage 5e. Ice core data from Greenland and Antarctic and these new Bermuda Rise records argue for a more complex climate variability during marine oxygen isotope stage 5.
FIG. 4. Detailed comparison of spectrophotometric reflectance (Lab) data from MD95-2034 (thick line) and MD95-2036 (thin line) during isotope stage 5. Both cores are plotted on (different) linear depth scales, with no stretching of sections. This %CaCO3 (reflectance-based) variability in this region dominantly reflects a variable input of clay superimposed on a relatively constant calcium carbonate influx (56). Hence, these fluctuations are likely to be related to climate processes that affect clay flux into the western North Atlantic basin, including erosion on land and the continental rise, and transport by rivers and deep currents. Addition evidence for abrupt century-scale variability during stage 5 is seen in the a red-to-yellow color balance variable (Fig. 4). This parameter reflects transport of hematite from the Canadian Shield red beds, a transport that is clearly quite different from transport of clay minerals. For example, during the abrupt fall of
CHARACTERISTICS OF THE DEEP OCEAN CARBON SYSTEM DURING THE PAST 150,000 YEARS: ΣCO2 DISTRIBUTIONS, DEEP WATER FLOW PATTERNS, AND ABRUPT CLIMATE CHANGE
8306
%CaCO3 in mid-stage 5e, the hematite content rises abruptly for a few centuries and then falls back to its previous level, even though % CaCO3 remains low for much longer.
FIG. 5. Comparison of spectrophotometric reflectance (L) data from MD95-2034, MD95-2036, and MD95-2035. Note that MD95-2035 extends to isotope stage 10 because of its lower sedimentation rate. Numbers on MD95-2035 plot are marine oxygen isotope stages (centered in respective stages).
FIG. 6. Comparison of spectrophotometric reflectance (L) data from MD95-2034 with central GRIP δ18O data (37) and Antarctic Vostok δD data (66) (ice cores being plotted on time scales developed by the authors of refs. 60 and 62. Note that although the SPECMAP marine δ18O record is reference for all time scales (beyond the range of layer-counting and radiocarbon measurements), time scales are independently developed by correlation to major features that are not well developed during isotope stage 3. Hence, relative errors of several thousand years may occur for any one of these time scales, and it is not expected that the same interstadial events will be assigned exactly the same age for each time scale. Identification of interstadials is based upon subjective pattern recognition. Studies of the isotopic composition of planktonic and benthic foraminifera and Cd/Ca variability of benthic foraminifera are now underway. Early results suggest that the level of variability seen in these properties does not approach the extremes originally inferred from the GRIP ice core, significant climate variability beyond the traditional 5a–e classification exists throughout oxygen isotope stage 5 and significant variability exists even within stage 5e. Arguments based on the supposed climate stability of warm climate periods will need to be rethought based on these emerging results. Concluding Remarks The quest for uncovering the LGM ΣCO2 distribution and its link to deep ocean circulation changes has made significant progress in the Atlantic, northern Indian, and eastern tropical Pacific oceans, but the quest is still stalled regarding the Southern Ocean and the northwest Pacific. The temporal variability of ΣCO2 and circulation in the North Atlantic is becoming evident, with long-term Milankovitch orbital links as well as century- to millennial-scale variability that may be linked to events seen in Greenland ice cores. Warm climate periods are not as unstable as implied by the original GRIP ice core data, but significant fluctuations occur, and these need to be better understood. I thank conference organizers for inviting me to this stimulating meeting. I am deeply indebted to Yvon Balut and other key
CHARACTERISTICS OF THE DEEP OCEAN CARBON SYSTEM DURING THE PAST 150,000 YEARS: ΣCO2 DISTRIBUTIONS, DEEP WATER FLOW PATTERNS, AND ABRUPT CLIMATE CHANGE
8307
personnel of the Marion Dufresne coring effort: Laurent Labeyrie, chief scientist of IMAGES; Jean-Louie Turon, chief scientist of the ship during our coring efforts; and Yves Lancelot and especially Franck Bassinot for their efforts on our behalf. All of the crew and scientific shipboard personnel were crucial in this effort, especially the Magnetic Suseptibility Track efforts led by Larry Mayer, Kate Jarrett, Frank Rack, and Gavin Dunbar, and my optical team coworkers, especially Lynda Levesque. I thank Todd Sowers for putting the GRIP δ18O record on the GISP2 time scale. Graduate student Jess Adkins is thanked both for his shipboard work (along with Jeff Berry) and comments on this manuscript. My continuing collaboration with Lloyd Keigwin on Bermuda Rise cores is a key inspiration in this effort. This research was sponsored by National Science Foundation Grant OCE9402198 and National Oceanic and Atmospheric Administration Grant NA-3GGP0246. 1. Keigwin, L. D. & Boyle, E. A. (1985) in The Carbon Cycle and Atmospheric CO2: Natural Variations Archean to Present, Am. Geophys. Union Monograph 32, eds., Sundquist, E. A. & Broecker, W. S. (Am. Geophys. Union, Washington, DC), pp. 319–328. 2. Shackleton, N. J. (1977) in The Fate of Fossil Fuel CO2, eds., Andersen, N. & Mallahof, A. (Plenum, New York), pp. 401–427. 3. Boyle, E.A. & Keigwin, L.D. (1985/1986) Earth Planet. Sci. Lett. 76, 135–150. 4. Curry, W. B., Duplessy, J. C., Labeyrie, L. D. & Shackleton, N. J. (1988) Paleoceanography 3, 317–342. 5. Duplessy, J. C., Shackleton, N. J., Fairbanks, R. G., Labeyrie, L., Oppo, D. & Kallel, N. (1988) Paleoceanography 3, 343–360. 6. Boyle, E. A. (1988) Paleoceanography 3, 471–489. 7. Boyle, E. A. (1992) Annu. Rev. Earth Planet. Sci. 20, 245–287. 8. Spero, H. J., Bemis, B. E., Bijma, J. & Lea, D. W. (1995) Trans. Am. Geophys. Union 76, F292 (abstr.). 9. Sanyal, A., Hemming, N. G., Hanson, G. N. & Broecker, W. S. (1995) Nature (London) 373, 234–236. 10. Spivack, A. J., You, C.-F. & Smith, H. J. (1993) Nature (London) 363, 149–151. 11. Lea, D. W., Spero, H. J. & Bijma, J. (1996) Trans. Am. Geophys. Union 77, S159 (abstr.). 12. van Geen, A., McCorkle, D. C. & Klinkhammer, G. P. (1995) Paleoceanography 10, 159–169. 13. Rosenthal, Y., Boyle, E. A., Labeyrie, L. & Oppo, D. (1995) Paleoceanography 10, 395–413. 14. Boyle, E. A. & Keigwin, L. D. (1982) Science 218, 784–787. 15. Boyle, E. & Keigwin, L. D. (1987) Nature (London) 330, 35–40. 16. D. W.Oppo, D. W. & Lehman, S. J. (1993) Science 259, 1148–1150. 17. Sarnthein, M., Winn, K., Jung, S. J. A., Duplessy, J.-C., Labeyrie, L., Erlenkeuser, H. & Ganssen, G. (1994) Paleoceanography 9, 209–267. 18. Boyle, E. A. (1995) Phil. Trans. R. Soc. London B 348, 243–253. 19. Yu, E.-F., Francois, R. & Bacon, M. P. (1996) Nature (London) 379, 689–694. 20. Kallel, N., Labeyrie, L.D., Juillet-Laclerc, A. & Duplessy, J. C. (1988) Nature (London) 333, 651–655. 21. Naqvi, W., Charles, C. D. & Fairbanks, R. G. (1994) Earth Planet. Sci. Lett. 121, 99–110. 22. Boyle, E. A., Labeyrie, L. & Duplessy, J.-C. (1995) Paleoceanography 10, 881–900. 23. Rosenthal, Y. (1994) Ph.D. thesis (MIT/Woods Hole Oceanogr. Inst., Cambridge, MA). 24. Boyle, E. A. (1996) in The South Atlantic: Present and Past Circulation, eds., Wefer, G., Berger, W. H., Siedler, G. & Webb, D. (Springer, Berlin), pp. 423–443. 25. Lea, D. W. (1995) Paleoceanography 10, 733–748. 26. Broecker, W. S. (1993) Paleoceanography 8, 137–140. 27. Mackensen, A., Hubberten, H. W., Bickert, T., Fischer, G. & Fütterer, D. K. (1993) Paleoceanography 8, 587–610. 28. McCorkle, D. C., Martin, P. A., Lea, D. W. & Klinkhammer, G. P. (1995) Paleoceanography 10, 699–714. 29. Keigwin, L. D. (1987) Nature (London) 330, 362–364. 30. Imbrie, J., Boyle, E. A., Clemens, S. C., Duffy, A., Howard, W. R., Kukla, G., Kutzbach, J., Martinson, D. G., McIntyre, A., Mix, A. C., Molfino, B., Morley, J. J., Peterson, L. C., Pisias, N. G., Prell, W. L., Raymo, M. E., Shackleton, N. J. & Toggweiler, J. R. (1992) Paleoceanography 7, 701– 738. 31. Warren, B. A. (1981) in Evolution of Physical Oceanography, eds., Warren, B. A. & Wunsch, C. (MIT Press, Cambridge, MA), pp. 6–41. 32. Fairbanks, R. G. (1989) Nature (London) 342, 637–642. 33. Climate Long-Range Investigation, Mapping, and Prediction (1981) Seasonal Reconstructions of the Earth's Surface at the Last Glacial Maximum, Geol. Soc. Am. Map and Chart Series (Geol. Soc. Am., Boulder, CO). 34. Michel, E., Labeyrie, L., Duplessy, J.-C., Gorfti, N., Labracherie, M. & Turon, J.-L. (1995) Paleoceanography 10, 927–942. 35. Imbrie, J., Berger, A., Boyle, E. A., Clemens, S. C., Duffy, A., Howard, W. R., Kukla, G., Kutzbach, J., Martinson, D. G., McIntyre, A., Mix, A. C., Molfino, B., Morley, J. J., Peterson, L. C., Pisias, N. G., Prell, W. L., Raymo, M. E., Shackleton, N. J. & Toggweiler, J. R. (1993) Paleoceanography 8, 699–736. 36. Dansgaard, W., White, J. W. C. & Johnson, S. J. (1989) Nature (London) 339, 532–534. 37. Johnsen, S. J., Clausen, H. B., Dansgaard, W., Fuhrer, K., Gundestrup, N., Hammer, C. U., Iversen, P., Jouzel, J., Stauffer, B. & Steffensen, J. P. (1992) Nature (London) 359, 311–313. 38. GRIP Members (1993) Nature (London) 364, 203–207. 39. Grootes, P. M., Stuiver, M., White, J. W. C., Johnsen, S. & Jouzel, J. (1993) Nature (London) 366, 552–554. 40. Broecker, W. S., Peteet, D. M. & Rind, D. (1985) Nature (London) 315, 21–26. 41. Broecker, W.S. & Peng, T.-H. (1987) Global Biogeochem. Cycles 3, 251–259. 42. Broecker, W. S., Andree, M., Bonani, G., Wolfli, W., Oeschger, H. & Klas, M. (1988) Quat. Res. 30, 1–6. 43. Broecker, W. S. & Denton, G. H. (1989) Geochim. Cosmochim. Acta 53, 2465–2501. 44. Broecker, W. S., Bond, G., Klas, M., Bonani, G. & Wolfli, W. (1990) Paleoceanography 5, 469–478. 45. Broecker, W. S., Peng, T. H., Trumbore, S., Bonani, G. & Wolfli, W. (1990) Global Biogeochem. Cycles 4, 103–117. 46. Birchfield, G.E., Wang, H. & Wyant, M. (1990) Paleoceanography 5, 383–396. 47. Birchfield, G. E. & Broecker, W. S. (1990) Paleoceanography 5, 835–844. 48. Birchfield, E. G., Wang, H. & Rich, J. J. (1994) J. Geophys. Res. 99, 12459–12470. 49. Peterson, L. C., Overpeck, J. T., Kipp, N. G. & Imbrie, J. (1992) Paleoceanography 6, 99–120. 50. Behl, R. J. & Kennett, J. P. (1996) Nature (London) 379, 243–246. 51. Bond, G., Broecker, W., Johnsen, S., Mcmanus, J., Labeyrie, L., Jouzel, J. & Bonani, G. (1993) Nature (London) 365, 143–147. 52. Lehman, S. J. & Keigwin, L. D. (1992) Nature (London) 356, 757–762. 53. Jansen, E. & Veum, T. (1990) Nature (London) 343, 612–615. 54. Laine, E. P. & Hollister, C. D. (1981) Mar. Geol. 39, 277–300. 55. Bacon, M. P. & Rosholt, J. N. (1982) Geochim. Cosmochim. Acta 46, 651–666. 56. Suman, D. O. & Bacon, M. P. (1989) Deep-Sea Res. 36, 869–878. 57. Keigwin, L. D. & Jones, G. A. (1989) Deep-Sea Res. 36, 845–867. 58. Keigwin, L. D. & Jones, G. A. (1994) J. Geophys. Res. 99, 12397–12410. 59. Keigwin, L. D., Jones, G. A., Lehman, S. J. & Boyle, E. A. (1991) J. Geophys. Res. 96, 16811–16826. 60. Bender, M., Sowers, T., Dickson, M.-L., Orchardo, J., Grootes, P., Mayewski, P. & Meese, D. A. (1994) Nature (London) 372, 663–666. 61. Fuchs, A. & Leuenberger, M. C. (1996) Geophys. Res. Lett. 23, 1049–1052. 62. Sowers, T., Bender, M., Labeyrie, L., Martinson, D., Jouzel, J., Raynaud, D., Pichon, J. J. & Korotkevich, Y. S. (1993) Paleoceanography 8, 699– 736. 63. Bassinot, F. C., Labeyrie, L. D., Vincent, E., Quidelleur, X., Shackleton, N. J. & Lancelot, Y. (1994) Earth Planet. Sci. Lett. 126, 91–108. 64. Anonymous (1986) CIE Colorimetry (Commission Internationale de l'Eclairage, Paris), Publication 15.2. 65. Barranco, F. T., Jr., Balsalm, W. L. & Deaton, B. C. (1989) Mar. Geol. 89, 299–314. 66. Jouzel, J., Lorius, C., Petit, J. R., Genthon, C., Barkov, N. I., Kotlyakov, V. M. & Petrov, V. M. (1987) Nature (London) 329, 403–408. 67. Shackleton, N. J. (1977) Philos. Trans. R. Soc. London B 280, 169–182. 68. Curry, W. B. & Oppo, D. W. (1997) Paleoceanography 12, 1–14.
DIRECT OBSERVATION OF THE OCEANIC CO2 INCREASE REVISITED
8308
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8308–8313, August 1997 Colloquium Paper This paper was presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held November 13–15, 1995, at the National Academy of Sciences, Irvine, CA.
Direct observation of the oceanic CO2 increase revisited
PETER G. BREWER*, CATHERINE GOYET†, AND GERNOT FRIEDERICH*
© 1997 by The National Academy of Sciences 0027-8424/97/948308-6$2.00/0 PNAS is available online at http://www.pnas.org. ABSTRACT We show, from recent data obtained at specimen North Pacific stations, that the fossil fuel CO2 signal is strongly present in the upper 400 m, and that we may consider areal extrapolations from geochemical surveys to determine the magnitude of ocean fossil fuel CO2 uptake. The debate surrounding this topic is illustrated by contrasting reports which suggest, based upon atmospheric observations and models, that the oceanic CO2 sink is small at these latitudes; or that the oceanic CO2 sink, based upon oceanic data and models, is large. The difference between these two estimates is at least a factor of two. There are contradictions arising from estimates based on surface partial pressures of CO2 alone, where the signal sought is small compared with regional and seasonal variability; and estimates of the accumulated subsurface burden, which correlates well other oceanic tracers. Ocean surface waters today contain about 45 µmol·kg−1 excess CO2 compared with those of the preindustrial era, and the signal is rising rapidly. What limits should we place on such calculations? The answer lies in the scientific questions to be asked. Recovery of the fossil fuel CO2 contamination signal from analysis of ocean water masses is robust enough to permit reasonable budget estimates. However, because we do not have sufficient data from the preindustrial ocean, the estimation of the required Redfield oxidation ratio in the upper several hundred meters is already blurred by the very fossil fuel CO2 signal we seek to resolve. In a recent paper Ciais et al. (1) described the results of careful measurements of atmospheric CO2 distributions and isotopic ratios. While they noted that “there is still ample room for improvement in this technique,” they concluded that a large northern hemisphere terrestrial CO2 sink existed, and that the oceanic sink at temperate latitudes was small. This is in general agreement with the earlier work of Tans et al. (2) which estimated a minimal oceanic sink. In contrast, work on the other side of the Pacific Ocean by Tsunogai et al. (3) concludes that the Intermediate waters of the North Pacific form a very large contemporary CO2 sink, and provide oceanic data to support this conclusion. Siegenthaler and Sarmiento (4) have summarized a great deal of work and also conclude that the ocean is indeed an important sink for anthropogenic carbon, but that a “missing sink” is also probably located in the land biota. The differences are large: data presented by Ciais et al. (1) for 1993 indicate that the global oceanic CO2 sink is about 1.15 Gt-C·yr−1; Tans et al. (2) state that “the global ocean sink is at most 1 Gt-C·yr−1.” This is in contrast to oceanic tracer-based data and models which indicate an oceanic sink of 2 Gt-C·yr−1. How can such differing views hold? And why is there such confusion over what should be a reasonably straightforward oceanic geochemical signal? The importance of the large oceanic CO2 sink, and the fundamental principles of the chemistry and physics that drive it, have been cited in classic papers over many decades of this century (5, 6 and 7), and no flaw in these basic concepts is seen: CO2 remains an acidic gas, we still have a vast and alkaline ocean for it to react with, and the rate of transfer is determined by the large-scale circulation which brings the ocean and atmosphere into contact. Yet the current debate is confusing, and therefore in this paper we re-examine the elementary question of detection of the oceanic CO2 increase, seek to place some limits on the errors involved, and make suggestions for practical steps to reduce the uncertainty. We cannot here resolve the question of the ultimate size of the large-scale integrated ocean burden; but the detection of the fossil fuel CO2 signal by direct measurement, and the determination of its areal extent by careful expeditionary surveys offers the best way to achieve this and those measurements are being accomplished today. Background Brewer (8) first proposed and demonstrated a simple and direct estimate of detection of the oceanic anthropogenic CO2 increase by examination of Geochemical Ocean Sections Study (GEOSECS) data from the South Atlantic Ocean. He corrected contemporary deep observations for the effects of oxidative decomposition of marine organic matter, and the dissolution of carbonates, and derived a value for a corrected pCO2 that was linked to the invasion of the gaseous anthropogenic signal. An extended version of this analysis for North Atlantic waters, based upon Transient Tracers in the Ocean, North Atlantic Study (TTO NAS) data was recently given by Goyet and Brewer (9), who again calculated the invasion term for the fossil fuel CO2 component, and showed a strong correlation of this signal with the F-11 chlorofluorocarbon tracer along the isopycnal surface of Labrador Sea water. Other researchers (10) have adopted this technique with only small variations. Chen and Millero (11) independently published a somewhat different approach to this interesting problem, and Chen has since published widely on this topic (12, 13). Both the Brewer (8) and Chen and Millero (11) methods involve a simple arithmetic subtraction of the quantities of CO2 gas and alkalinity added to deep waters by respiration and carbonate dissolution once the surface waters are removed from their contact with the atmosphere and begin their century scale sojourn in the abyss. However the two approaches differ importantly in their normalization procedures and in the assumptions necessary to formalize the signal. This validity of these approximations to recognition of the fossil fuel signal has been the subject of much debate. Broecker et al. (14) attempted to greatly extend the question first posed, and asked whether not just the accumulated amount, but the time record of the CO2 chemistry of the past atmosphere,
*Monterey Bay Aquarium Research Institute, P.O. Box 628, Moss Landing, CA 95039; and †Woods Hole Oceanographic Institution, Woods Hole, MA 02543
DIRECT OBSERVATION OF THE OCEANIC CO2 INCREASE REVISITED
8309
could be recovered by inverting the oceanic water mass signals; not surprisingly, they found this to be impractical. Their closing comment, however, is worthy of note: “Once the preanthropogenic CO2 content of the atmosphere has been firmly established by ice core studies . . . then the oceanic distribution of TCO2 can be used to constrain some of the current uncertainties in models for the uptake of fossil fuel CO2 by the ocean.” Since that time the preanthropogenic atmospheric CO2 content has been established as close to 280 µatm, and the fossil fuel signal of ocean surface waters that we seek to identify has continued to rise, so that the surface ocean today contains some 45 µmol·kg−1 of CO2 in excess of that in the preindustrial era. Abbreviations: WOCE, World Ocean Circulation Study; TC, total carbon dioxide; TA, total alkalinity. Wallace (15) has carried out a survey of strategies for monitoring global ocean carbon inventories, and has attempted to assess uncertainties. He comments favorably on the need for providing reliable estimates of the build up of fossil fuel CO2 in the ocean, and notes that the anthropogenic CO2 content of the ocean now ranges from zero (deep waters) to 45 µmol·kg−1 (surface waters) thus providing a useful dynamic range; that individual measurement errors are between 1 and 5 µmol·kg−1; and that the manmade signal is superimposed on a natural background that has to be accurately recovered to use the signal. He notes that “a clear advantage of this approach is that it can provide model-independent estimates of the spatial variability of the excess CO2 distribution which in turn can be used to validate model predictions.” The mapping of sea surface pCO2 is relatively easy now that superior instrumentation has been developed, and it provides data on the distribution of natural sources and sinks. However, the driving signal for the fossil fuel term is not the natural pattern, but the forced disequilibrium between air and sea resulting from the rapid rate of atmospheric CO2 growth. If the oceanic uptake rate is at the high end of published estimates, then this signal must now have a globally averaged value of about +8 µatm to provide the needed driving force, but its observation on top of a natural background that varies by ±100 µatm is difficult indeed (2), and so far it has not been unequivocally detected. There is clearly an oceanic fossil fuel signal present, but it is the integrated amount and its time evolution that is hard to assess. Consider the simple large scale problem first: the surface area of the ocean is about 3.6 × 1014 m2. The globally averaged mixed layer depth has been estimated as about 75 m (7), and thus a volume of “fresh” sea water of about 2.7 × 1019 liters·yr−1 is exposed to the atmosphere. Recent work at the Joint Global Ocean Flux Study (JGOFS) Bermuda time series station (16) indicates that the CO2 content of mixed layer waters is increasing at a rate of about 1.7 µmol·kg−1·yr−1. Leaving aside for the moment the question of natural variability versus industrial atmospheric trends, and simply integrating this number from a northern hemisphere temperate gyre, on an ocean wide basis we would find a global uptake of 4.6 × 1013 mol·yr−1, or 0.55 Gt-C·yr−1. This is broadly consistent with the lower estimates of oceanic uptake of Tans et al. (2) and Ciais et al. (1), and it at once raises the question of how representative are single sites of a global balance, how to obtain a legitimate integrated signal, and how reliable the mean mixed layer (ventilation) depth might be. The average mixed layer depth estimate was derived so as to match the oceanic penetration of bomb radiocarbon over a decade or more; the mean equilibration time for CO2 is about 1 year, and that for 14CO2 is about 10 years, and so the two results are not entirely compatible. For instance, a greater effective mixed layer depth (a winter dominated signal) would increase the CO2 uptake rate significantly over the crude estimate above. It therefore seems timely to reconsider the problem of detection of the fossil fuel signal in the ocean by direct means, and to examine the concepts and assumptions involved in a more formal way. Concepts The relatively small, but rapidly growing, fossil fuel CO2 invasion signal in the ocean is written on top of a large and variable natural background; the problem is to normalize, or remove, or otherwise constrain the background signal so as to reveal the man-made component. The arithmetic turns out to be extraordinarily simple; but the problems that are thereby exposed lie at the root of our field and force us to ask some difficult questions. In the following discussion we use the notation TC to define the total quantity of CO2 in all its forms (CO2 + H2CO3 + HCO−3 + CO32−) in sea water, and TA to define the total alkalinity. Carbon dioxide gas is fixed in surface waters by photosynthesis, and returned as mineralized products at depth. This biogeochemical cycle is superimposed on the signal imposed by the physical effects of temperature and salinity distributions, and by any imbalances caused by the slow equilibration rate of CO2 with the atmosphere. The most commonly used equation to describe the biogeochemical cycle is that given by Redfield (17), and embellished and extended by Redfield et.al.(18):
The assumptions are that in living organic matter the oxidation state of carbon is that of carbohydrate, that nitrogen is present in the amino form, and that phosphorous may be represented as orthophosphate. This so called “Redfield Ratio” is critical to the problem; note that the addition or removal of CO2 gas during photosynthesis or decay does not change the total alkalinity, but that the companion removal or release of nitrate ion does. As indicated in the notation here, the uptake and release of nitrate ion is equivalent to removal and regeneration of nitric acid and must be accounted for in relating the observed alkalinity to the mass changes from calcium carbonate removal and addition. This was first described by Brewer et al. (19), and shown experimentally for the uptake side of the equation by Brewer and Goldman (20) and Goldman and Brewer (21). The effect of phosphate ion is more complex [see Bradshaw et al. (22) for a detailed account], for it appears in the acidimetric titration of sea water as a proton contributor in two steps; a correction for this of one H+ in the 18 Redfield protons is required here. There have been many attempts to revise the Redfield equation [e.g., Takahashi et al. (23) and Boulahid and Minster (24)], not normally through the inclusion of additional terms for trace constituents, but in an effort to increase the accuracy of the numerical coefficients for the principal reactants; it is remarkable that such a simple relationship should apparently hold over all the vast area of the earth's surface covered by the oceans. A recent and very thorough analysis is given by Anderson and Sarmiento (25). They examined the distribution of nutrients upon 20 neutral surfaces in the South Atlantic, Indian, and Pacific basins between 400 and 4,000 m depth and produced a revised set of values such that their preferred estimates are C/N/P/O = 117:16:1:–170. Thus in the deep ocean the addition of CO2 by respiration can be calculated by observing the oxygen deficit relative to
DIRECT OBSERVATION OF THE OCEANIC CO2 INCREASE REVISITED
8310
saturation with the atmosphere, and recourse to the preferred version of the Redfield ratio for the appropriate coefficients. We should note that the Anderson and Sarmiento (25) selection of a depth interval that excludes data from shallower than 400 m is in part due to the assumed presence in that depth range of the fossil fuel CO2 contamination signal described here. We are thus in a bind, for this is the very zone where the regeneration rates are largest (26), where local deviations from the mean may be expected to occur, and where we are most in need of information. There simply are no reliable data prior to the onset of penetration of fossil fuel CO2, and we must extrapolate through this zone in the best way possible. The oceanic cycles of phosphorus, nitrogen and oxygen have changed far less than that of CO2, and recovery of those ratios would be possible if multiple water mass mixing problems were not so severe. In surface sea water total alkalinity is observed to be a very well constrained function of salinity: the fossil fuel CO2 burden does not change the total alkalinity, and the seasonal variations are small for the increase in alkalinity from NO3 assimilation is partially offset by the decrease from calcium carbonate formation. Thus we can recognize subsurface addition of CO2 from carbonate dissolution by departures from the salinity to total alkalinity (TA) ratio established for winter time surface waters. For example Brewer et al. (27) give TA = 50.56S%0 + 547.0 µmol·kg−1, from an analysis of Transient Traces in the Ocean (TTO) data from North Atlantic surface waters, and the equivalent regional functions in other ocean basins will be increasingly well constrained in the near future from the Joint Global Ocean Flux Study/World Ocean Circulation Experiment (JGOFS/WOCE) Global Survey data, and from time series data such as that reported by Bates et al. (16). In this paper we will use the relationship TA = 45.785S%0 + 703.7 µmol·kg−1, (see Fig. 2), with a standard error of ± 6.2 µmol·kg−1 from our specimen analysis of recent Pacific data. To establish the CO2 correction for deep waters due to the dissolution of calcium carbonate, and the addition of acidic nitrate ion, we write (using Redfield's original coefficients for example) ∆TA = ∆TA (CaCO3) + (1 − 18)/138O2.
FIG. 1. Cruise tracks and station locations for WOCE Leg P17N. Note that we cannot use the observed nitrate ion concentration directly, for there is a significant nonmetabolized fraction (the preformed nitrate), and thus all corrections reduce to a function of the oxygen deficit of the water mass. To establish the deep water correction terms for total CO2, we therefore write ∆TC = 106/138∆O2 + 0.5(∆TA − [(1 − 18)/138 ∆O2]), which reduces to ∆TC = 0.8297∆O2 + 0.5∆TA [original Redfield ratios (17, 18)], or ∆TC = 0.738∆O2 + 0.5∆TA [from Anderson and Sarmiento (25)]. We now have, given the gas solubility data, the information we need to calculate a pair of values for TC and TA at some point at depth that in relates to the chemical properties acquired during ventilation at the surface many years ago. The term representing the initial TC is in effect a tracer of fossil fuel CO 2 input written on top of the natural background; it is a surprisingly robust signal and may be analyzed along ventilation surfaces of constant density much as any other tracer. The use of oxidative corrections to derive a quasi-conservative property is not a new concept, and
FIG. 2. Correlation between measured values of alkalinity and salinity for surface waters along the WOCE P17N transect. The strong linear correlation is to be expected, and the standard deviation is only ±6 µmol·kg−1.
DIRECT OBSERVATION OF THE OCEANIC CO2 INCREASE REVISITED
8311
properties such as “NO” and “PO” for nitrogen and phosphorus have been proposed and used for other purposes (28). From these two carbonate system properties we can also fully constrain the CO2 system (computing pH and pCO2 and thereby relating the signal to the atmosphere) if we know the thermodynamic constants. Errors and Sensitivity Behind this simple scheme lies complexity. Only a few specialists are aware of the details of CO2 chemistry. Let us give some simple examples. For clarity, we will take as a target a desire to constrain the system to ±10 µatm in pCO2; this represents about 7 years growth in today's atmospheric signal, and is about 5% of the accumulated fossil fuel change to date. We will see that we may not meet this goal, but it is a useful ready reference point. First, a signal of 10 µatm pCO2 is equivalent to 7 µmol·kg−1 in TCO2, and to 8 µmol·kg−1 in total alkalinity within the normal range of sea water values. Analytical errors of single samples today are 1 µmol·kg−1 in TCO2 and 3 µmol·kg−1 in total alkalinity, although improvements are possible. We need to establish the sensitivity to salinity, and the slope of the alkalinity/salinity line established for surface waters (see Fig. 2) indicates that to constrain the system to 10 µatm pCO2 we need to know salinity to 0.3%0, which is easily achievable by direct measurement, but a significant problem for remote sensing. We need to determine the oxygen deficit of subsurface waters; the direct dissolved O2 measurement of deep waters has an error of 1 µmol·kg−1 (roughly equivalent to 1 µmol TCO2); however the assumption of atmospheric equilibrium at the time of subduction may not be valid and is very hard to determine. We estimate the error from this source as perhaps 5 µmol·kg−1, equivalent to 5 µatm pCO2.
FIG. 3. Specimen calculation of a vertical profile of the partial pressure of CO2 at WOCE P17N Station 13, corrected for the effects of oxidative organic matter decomposition and carbonate dissolution so as to reveal an approximation to the original water mass surface conditions. The corrections resulting from the use of the original (1934) Redfield ratios (17, 18) and the more recent coefficients (1994) of Anderson and Sarmiento (25) are shown: both profiles are computed utilizing the thermodynamic constants of Goyet and Poisson (30). The deep water value of 280 ± 6 ppm given by the Anderson and Sarmiento coefficients is close to the established preindustrial atmospheric level. We need to know the Redfield ratio, and this problem was reviewed above. As an aside, we are fortunate that there is very little other chemical ambiguity; species representing the other oxidation states of carbon (e.g., CO and CH4) are present in insignificant quantities. We have shown that the principal correction to total alkalinity, other than carbonate uptake and dissolution, results from changes in the oxidation state of nitrogen and these are well understood; problems of denitrification do occur, but in the Anderson and Sarmiento (25) analysis these are significant only in the range between 1,000 and 3,000 m. The magnitude of other redox changes, such as those from sulfur, are trivial in the oxic ocean. Oceanic Example In earlier work (9) on North Atlantic data we have shown significant correlations between the computed invasion term for CO2, and the chlorofluorocarbon (F11) signal. Here we choose, in tribute to Roger Revelle and his love of his home in California, recent data from that region. The data are selected from WOCE Cruise P17N, and the station locations are shown in Fig. 1. We must first determine the correlation between total alkalinity and salinity for this region, and this is shown in Fig. 2. The correlation is remarkable, with a standard deviation of only 6 µmol·kg−1, and it is a tribute to the very careful experimental work accomplished. However, this represents the chemical state of the surface ocean at the time of the observations. The deep waters are formed in late winter at some
FIG. 4. Comparison of the calculated pCO2 profile from WOCE P17N stations utilizing the thermodynamic constants from several sources. The deep water values are in agreement to about ±10 ppm, but the warmer surface waters show a spread of about ±25 ppm. It is not necessary to use these constants to derive a ∆TCO2 value, but they are required to link calculated changes in oceanic chemical composition to the atmospheric driving signal.
DIRECT OBSERVATION OF THE OCEANIC CO2 INCREASE REVISITED
8312
other latitude, and we must find some way of either compensating for this (29) or of estimating the error. Changes due to CaCO3 fluxes are likely to be very small, and the NO3 cycle is likely the dominant concern. We then simply apply the equations above, and examine the signal. Fig. 3 shows a specimen calculation for WOCE P17N Station 13 off San Francisco. The calculation is run with both the original (1934) Redfield ratio (17, 18) and with the newer (1994) coefficient of Anderson and Sarmiento (25). While the ∆TC function may be calculated without recourse to the thermodynamic constants required for computing the partial pressure, we compute pCO2 here for clarity. The result shows clearly a consistent signal in the deep water below 500 m close to 280 µatm, and above that depth a rapidly rising signal that converges close to today's value near the surface. This picture is consistent with all other examinations of this signal and reveals a penetrating tracer related to the fossil fuel perturbation of the chemistry of the upper ocean. Interestingly the original Redfield ratio yields a deep water value about 30 µatm lower than the coefficient of Anderson and Sarmiento (25); we then have to make a value judgment as to the correct signal. In doing so we expose a weakness in our understanding, for although the preindustrial atmosphere and the ocean must have been in gross overall equilibrium, regional disequilibrium in the pCO2 of the deep water at the time of formation is clearly possible. Takahashi et al. (31) have compiled an extensive data set and analysis of seasonal surface ocean pCO2 conditions and find broadly that very large disequilibria do not occur; however, we do not know as well as we would like the initial conditions of water mass formation. Next we must make a choice of thermodynamic constants to carry out the calculation of pCO2; a recent review is provided by Millero (32). Although work on these constants has proceeded for decades, the choice is still controversial, and as we shall see the differences can be significant.
FIG. 5. Image of the calculated mixed layer depth for the North Pacific Ocean for the month of March 1995 from data supplied by the Fleet Numerical Meteorological and Oceanography Center model (37). In Fig. 4 we show results, not just for one station but for the entire WOCE P17N data set, for the initial pCO2 value, calculated with constants from Roy et al. (33), Hansson (34) and Mehrbach et al. (35) relative to the constants of Goyet and Poisson (30). While the deep water values are in agreement to within about ± 10 µatm (close to the target value for discussion in this paper), the surface values with a greater range of temperature and salinity show a spread of about ± 25 µatm; in fact the noise in the estimate resulting from the varying choice of thermodynamic constants for carbonic acid in sea water is about equal to that caused by uncertainty in the Redfield ratio. The selection of a set of thermodynamic constants for the ocean CO2 system appears in some way at many stages of the variety of experimental protocols and numerical calculations performed by all scientists in this area; there is no uniform international agreement on this, and experimentalists and modelers alike choose their personal preference. It is tempting to conclude from inspecting Fig. 4 that, since oceanic data treated with the constants of Goyet and Poisson (30) and Roy et al. (33) yields closely similar results, then these constants, determined in fully independent experiments, should be preferred. There is, however, not uniform agreement on this point, and it was pointed out in review of this manuscript that if the values for K1 and K2 are considered separately then the apparent agreement is not as good. That is that the impressive agreement shown in Fig. 4 may be the result of a fortunate compensation of positive and negative errors, rather than resulting from absolute accuracy of the values. The net result is that if we are to achieve an increase in our ability to observe
DIRECT OBSERVATION OF THE OCEANIC CO2 INCREASE REVISITED
8313
and model the oceanic carbon system, then a fundamental improvement in the accuracy of the thermodynamic constants is one of the required steps. This is an exceptionally challenging experimental task. The calculation of a signal which appears to be very closely related to a direct observation of the accumulated ocean fossil fuel CO2 invasion term, such as the profile given in Fig. 3, seems to be readily achievable. The principal debate is about the levels of accuracy acceptable for such a calculation, and the ultimate goal of integrating the signal on an ocean basin scale and relating the quantities to the global fossil fuel CO2 budget. The subsurface signals revealed are integrals mixed from several water mass sources, and divining the unique surface signature of any one oceanic region will be difficult. Moreover in practice there are strong regional and seasonal sources and sinks of CO2 in the surface ocean, and until we find strategies for dealing with these there will always be room for debate. Here there is room for optimism; the ability to gain seasonal CO2 data at a remote site has been greatly aided by the recent development of accurate and reliable new sensors that can be mated with buoy technology (36). The ability to predict in real time the mixed layer depth of the ocean has also undergone a revolution through advanced remote observational and computing techniques (37), and the results are now being made available to the civilian community. A specimen global map of calculated mixed layer depths produced by us from data supplied by the Fleet Numerical and Oceanography Center for the month of March 1995, close to the northern hemisphere vernal equinox and probably representing maximum values, is shown in Fig. 5. Conclusion The problem of detecting the fossil fuel CO2 signal in sea water is relatively easy. Modifications of the calculation of Brewer (8) or of Chen and Millero (11) all yield a robust signal of about 45 µmol·kg−1 in surface ocean waters today. In the North Pacific Ocean example given here the signal decays with depth, corresponding to the ventilation age of the water masses, and may be traced to about 400-m depth, consistent with deep winter time mixed layer formation in the northwest sector. The most consistent results are presently obtained by using the oxidative decomposition ratios of Anderson and Sarmiento (25) and the thermodynamic constants of either Goyet and Poisson (30) or Roy et al. (33). Refining the calculation will require increased knowledge of the Redfield ratio, and of surface winter time total alkalinity values. New sensing and sampling technologies now developed offer every prospect of yielding this information, and of improving estimates of long term ocean CO2 uptake. The very careful work of the teams of ocean scientists now producing data on the ocean carbon system as part of a coordinated global survey is gratefully acknowledged. This work was supported by a grant to the Monterey Bay Aquarium Research Institute from the David and Lucile Packard Foundation, and by National Aeronautics and Space Administration Grant Earth Observing System (EOS) NAG-232431 to the Woods Hole Oceanographic Institution. 1. Ciais, P., Tans, P. P., Trolier, M., White, J. W. C. & Francey, R. J. (1995) Science 269, 1098–1102. 2. Tans, P. P., Fung, I. Y. & Takahashi, T. (1990) Science 247, 1431–1438. 3. Tsunogai, S., Ono, T. & Watanabe, S. (1993) J. Oceanogr. 49, 305–315. 4. Siegenthaler, U. & Sarmiento, J. L. (1993) Nature (London) 365, 119–125. 5. Callendar, G. S. (1938) Q. J. R. Meteorol. Soc. 64, 223–240. 6. Revelle, R. & Suess, H. E. (1957) Tellus 9, 18–27. 7. Oeschger, H., Siegenthaler, U., Schatterer, U. & Gugelmann, A. (1975) Tellus 27, 168–191. 8. Brewer, P. G., (1978) Geophys. Res. Lett. 5, 997–1000. 9. Goyet, C. & Brewer, P. G. (1993) in Modeling Oceanic Climate Interactions, NATO Series I11, eds. Willebrand, J. & Anderson, D. L. T. (Springer, Berlin), pp. 271–297. 10. Jones, E. P. & Levy, E. M. (1981) J. Mar. Res. 39, 405–416. 11. Chen, C.-T. & Millero, F. J. (1979) Nature (London) 277, 205–206. 12. Chen, C.-T. (1982) Deep Sea Res. 29, 563–580. 13. Chen, C.-T. (1993) J. Oceanogr. 18, 257–270. 14. Broecker, W. S., Takahashi, T. & Peng, T.-H. (1985) Reconstruction of the Past Atmospheric CO2 Contents of the Contemporary Ocean: An Evaluation (U.S. Department of Energy, Washington, DC), Rep. DOE/OR 857. 15. Wallace D. W. R. (1995) Monitoring Global Ocean Carbon Inventories, Ocean Observing System Background Report 5 (Texas A&M Univ., College Station). 16. Bates, N. R., Michaels, A. F. & Knap, A. H. (1996) Deep Sea Res. 43, 347–383. 17. Redfield, A. C. (1934) James Johnstone Memorial Volume (Liverpool Univ. Press, Liverpool, U.K.), 176–192. 18. Redfield, A. C., Ketchum, B. H. & Richards, F. A. (1963) in The Seas, ed. Hill, M. N. (Wiley-Interscience, New York), Vol. 2, pp. 26–77. 19. Brewer, P. G., Wong, G. T. F., Bacon, M. P. & Spencer, D. W. (1975) Earth Planet. Sci. Lett. 26, 81–87. 20. Brewer, P. G. & Goldman, J. C. (1976) Limnol. Oceanogr. 21, 108–117. 21. Goldman, J. C. & Brewer, P. G. (1980) Limnol. Oceanogr. 25, 352–357. 22. Bradshaw, A. L., Brewer, P. G., Shafer, D. K. & Williams, R. T. (1981) Earth Planet. Sci. Lett. 55, 99–115. 23. Takahashi, T., Broecker, W. S. & Langer, S. (1985) J. Geophys. Res. 90, 6907–6924. 24. Boulahid, M. & Minster, J.-F. (1989) Mar. Chem. 26, 133–153. 25. Anderson, L. A. & Sarmiento, J. L. (1994) Global Biogeochem. Cycles 8, 65–80. 26. Martin, J. H., Knauer, G. A., Karl, D. M. & Broenkow, W. W. (1987) Deep Sea Res. 34, 267–285. 27. Brewer, P. G., Bradshaw, A. L., Shafer, D. K. & Williams, R. T. (1986) in The Changing Carbon Cycle: A Global Analysis, eds. Trabalka, J. R. & Reichle, D. E. (Springer, New York), pp. 348–370. 28. Broecker, W. S. (1974) Earth Planet. Sci. Lett. 23, 100–107. 29. Glover, D. M. & Brewer, P. G. (1988) Deep Sea Res. 35, 1525–1546. 30. Goyet, C. & Poisson, A. (1989) Deep Sea Res. 36, 1635–1654. 31. Takahashi, T., Olafsson, J., Goddard, J. G., Chipman, D. W. & Sutherland, S. C. (1993) Global Biogeochem. Cycles 7, 843–878. 32. Millero, F. J. (1995) Geochim. Cosmochim. Acta 59, 661–677. 33. Roy, R. N., Roy, L. N., Vogel, K. M., Moore, C. P., Pearson, T., Good, C. E., Millero, F. J. & Campbell, D. M. (1993) Mar. Chem. 44, 249–268. 34. Hansson, I. (1973) Deep Sea Res. 20, 461–478. 35. Mehrbach, C., Culberson, C. H., Hawley, J. E. & Pytkowicz, R. M. (1973) Limnol. Oceanogr. 18, 897–907. 36. Friederich, G. E., Brewer, P. G., Herlien, R. & Chavez, F. (1995) Deep Sea Res. 42, 1175–1186. 37. Clancy, R. M. & Sadler, W. D. (1992) Weather Forecasting 7, 307–327.
THE OBSERVED GLOBAL WARMING RECORD: WHAT DOES IT TELL US?
8314
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8314–8320, August 1997 Colloquium Paper This paper was presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held November 13–15, 1995, at the National Academy of Sciences, Irvine, CA.
The observed global warming record: What does it tell us?
T. M. L. WIGLEY*, P. D. JONES†, AND S. C. B. RAPER†
© 1997 by The National Academy of Sciences 0027-8424/97/948314-7$2.00/0 PNAS is available online at http://www.pnas.org. ABSTRACT Global, near-surface temperature data sets and their derivations are discussed, and differences between the Jones and Intergovernmental Panel on Climate Change data sets are explained. Global-mean temperature changes are then interpreted in terms of anthropogenic forcing influences and natural variability. The inclusion of aerosol forcing improves the fit between modeled and observed changes but does not improve the agreement between the implied climate sensitivity value and the standard model-based range of 1.5–4.5°C equilibrium warming for a CO2 doubling. The implied sensitivity goes from below the model-based range of estimates to substantially above this range. The addition of a solar forcing effect further improves the fit and brings the best-fit sensitivity into the middle of the model-based range. Consistency is further improved when internally generated changes are considered. This consistency, however, hides many uncertainties that surround observed data/model comparisons. These uncertainties make it impossible currently to use observed global-scale temperature changes to narrow the uncertainty range in the climate sensitivity below that estimated directly from climate models. Observations from land-based meteorological stations and ships at sea have been compiled, corrected for nonmeteorological biases, interpolated to a regular grid, and area-averaged to estimate changes in global-mean temperature over the past century or so. These data show an overall warming trend of about 0.5°C, with marked shorter time scale variability from year to year and decade to decade. It is suspected that part of the long-term warming trend is due to human activities, but determining just how much of the trend is human-induced is a difficult task. This paper describes the data and their development, and interprets the changes that have occurred in terms of anthropogenic and natural causal factors. Data Sets The primary data sets are those from land and marine areas. Extensive reviews of these data have been given by the Intergovernmental Panel on Climate Change (IPCC) (1, 2 and 3). Over land, the main data sets are those of Vinnikov et al. (4), Hansen and Lebedeff (5, 6), and Jones et al. (7, 8). The Hansen and Jones data sets are continually updated. Only the Jones data set has undergone rigorous quality control, and it is these data that are used in the standard IPCC global- and hemispheric-mean time series. The different land data sets have been compared in ref. 9. The Jones land data set has recently been updated through the addition of new historical data and a change of reference period from 1951–1970 to 1961–1990 (10). At the hemispheric- and global-mean levels, the effect of this update is small. The two main marine data sets are those of Jones et al. (ref. 9; see also ref. 11) and the U.K. Meteorological Office (UKMO) (12, 13). These two data sets have overlapping primary source material but differ in the way that they are corrected for instrumentation changes. The Jones et al. marine data set uses UKMO data from 1987 onward and COADS (Comprehensive Ocean-Atmosphere Data Set) data (14) before 1987. The data used in compiling area averages are sea surface temperature data. Marine air temperature data exist, but it is more difficult to correct these for instrumentation problems. Where reliable corrections can be made, the sea surface temperature and marine air temperature data agree almost perfectly (13). For global-mean (i.e., land plus marine) data, there are two data sets in common usage, the Jones data as listed in Trends ‘93 (15), which combine the older Jones land data (7, 8) with the Jones marine data, and the standard IPCC data set, which combines the newer Jones land data (10) with the UKMO marine data. Refs. 9 and 12 give details on how the land and marine data sets are merged. Because of the history of their development, these global data sets use different reference periods (Jones, 1950–1979; IPCC, 1961–1990). Numerous corrections have to be made to both the land data and the marine data to remove or correct for nonclimatological influences; extensive discussions have been given in refs. 1, 2, 9, 12, 13, 16, and elsewhere. Small residual biases may remain, but these are judged to cause errors in the overall global-mean trend of, at most, 0.1°C. Trend uncertainties also arise because of incomplete coverage; even now, there are no direct measurements for large areas of the southern oceans, although temperature data for these areas can be derived using satellite data (12, 13). Coverage changes add another element of uncertainty to the trend. IPCC (1, 2) judges the overall trend uncertainty to be ±0.15°C. Fig. 1 shows the standard IPCC global-mean temperature series compared with the Jones data. Differences arise for five main reasons: (i) because of the different reference periods used (corrected for approximately in Fig. 1); (ii) because of differences in the raw marine data sets (the UKMO data set is somewhat more comprehensive); (iii) because of differences in the method used to correct for marine instrumentation changes (discussed extensively in refs. 9, 12, 13, and 16); (iv) because of different ways in which the global means are calculated; and (v) because of small differences in the land data sets. The first reason, the effect of different reference periods, gives a difference of 0.046°C or 0.058°C, depending on how the data sets are compared. Over 1950–1979, the Jones data have a global mean of −0.005°C, whereas the IPCC data have a mean of −0.051°C; over 1961–1990, the means are 0.083°C and
*National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307-3000; and †Climatic Research Unit, University of East Anglia, Norwich NR4 7TJ, United Kingdom
THE OBSERVED GLOBAL WARMING RECORD: WHAT DOES IT TELL US?
8315
0.025°C, respectively. Note that neither data set has a global-mean value of precisely zero over its reference period. Primarily, this is because the reference period applies to the individual grid points and because there are coverage changes over the reference periods. In the Jones data set, the land and marine components have different reference periods. For compatibility in merging the data sets, the land data have to have their reference period adjusted from 1951–1970 to 1950–1979, an adjustment that can only be made approximately. This adjustment is not a problem with the IPCC data because both land and marine components use the same reference period (1961–1990). For land station data to be included in the analysis, however, full reference-period coverage was not required; only a minimum of 20 out of 30 years was needed. This can lead to biases in the anomalies relative to the reference period, which are reflected in the spatial mean anomaly averaged over the reference period. In Fig. 1, both data sets have been adjusted to have zero means over 1961–1990 by subtraction of 0.025°C from the IPCC data and 0.083°C from the Jones data. Abbreviations: IPCC, Intergovernmental Panel on Climate Change; UKMO, U.K. Meteorological Office; RMSE, root-mean-square error; GHG, greenhouse gas; ENSO, El Niño/Southern Oscillation; O/AGCM, ocean/atmosphere general circulation model; GFDL, Geophysical Fluid Dynamics Laboratory.
FIG. 1. Comparison of Jones and IPCC global-mean temperature data. Both data sets have been adjusted to have zero means over 1961–1990, and the annual data were then filtered with a 13-point Gaussian filter to highlight decadal and longer time scale changes. Data up to and including 1995 have been used. The second reason for differences (raw marine data differences) has not been specifically quantified, but it is likely to be relatively small. The third reason (different correction methods) is more important. Below, the combined influences of these two effects are calculated by differencing. To do this, we assume the effect of land data differences (the fifth reason) is negligible, calculate the effect of the fourth reason independently, and subtract this from the difference between the data sets after the effect of the first reason has been removed (see above and Fig. 1). The fourth reason (different hemispheric averaging methods) has not previously been discussed and is quantified here for the first time (to our knowledge). The two different methods are as follows. In calculating the global mean for the IPCC data set, the data for individual grid boxes are simply area-weighted and averaged. Because the fractional coverage in each hemisphere varies with time, with a relatively greater fraction covered in the Northern Hemisphere in the earlier years, this method may potentially bias the global mean toward the Northern Hemisphere in these years. The Jones global mean is calculated by area-averaging the hemispheres separately first and then averaging the hemispheric means. This method may put undue weight on the sparsely covered Southern Hemisphere in the early years. It is impossible to decide a priori which method is better. Fig. 2 shows the difference between the IPCC data recalculated using the Jones method and the standard IPCC values (recalculated values minus original values). This isolates the effect of the fourth reason. The differences are small and somewhat erratic, with no overall trend. Fig. 3 shows the residual difference, Jones data minus recalculated IPCC data, after adjustment of both data sets for the effect of reference-period differences. This plot has been calculated by subtracting the “error” shown in Fig. 2 from the annual data used to produce Fig. 1 This essentially isolates the influences of the different sea surface temperature data sets and the different ways these data sets have been corrected for instrumental biases. The low-frequency changes in this plot arise largely from the different instrumentation correction schemes, whereas the shorter time scale differences mainly reflect differences in the raw data. A clear overall trend (arising mainly over the period 1880–1910) is evident. This trend is reflected in the data differences shown in Fig. 1 and explains why the Jones data have a slightly greater overall warming trend than the IPCC data. Anthropogenic Causes of Global Warming Why has the globe warmed? Because we are confident that human activities have substantially changed the atmospheric composition in terms of greenhouse gases (GHGs; especially carbon dioxide) and aerosols, we are also confident that at least part of the observed warming is human-induced. The leading question is how much? To answer this, we first need to estimate the magnitude of the expected anthropogenic warming. To do this requires a knowledge of the anthropogenic forcing change, and a suitable model to convert this forcing to an estimated climate change. Fig. 4 shows the current central estimate of forcing changes as used in the latest IPCC calculations of global-mean temperature and sea-level change (ref. 17; further details are given in ref. 18). It is clear that CO2 is the main single factor, but
FIG. 2. Breakdown of differences between the Jones and IPCC global-mean temperature data over 1861–1994: the effect of different area-averaging methods. The differences shown are annual values of the difference IPCC data averaged according to the Jones method minus original IPCC data.
THE OBSERVED GLOBAL WARMING RECORD: WHAT DOES IT TELL US?
8316
aerosol influences are a close second in importance. Aerosols operate in two different ways to produce a negative forcing effect that is highly spatially heterogeneous compared with the forcing effect of CO2 (see refs. 19 and 20). In clear sky conditions, sulfate aerosols (derived mainly from SO2, which is produced, like CO2, by fossil-fuel combustion) reflect incoming solar radiation—this process is referred to as direct aerosol forcing (19, 20). In addition, there is an indirect cooling effect through the influence of aerosols, acting as cloud condensation nuclei, on cloud albedo (see refs. 21, 22, 23 and 24). Biomass burning also produces aerosols, and these are thought to have a net cooling effect too (ref. 25; this paper gives a much larger effect than more recent estimates). These three components give global-mean forcings with central estimates (to 1990) of −0.3 W/m2, −0.8 W/m2, and −0.2 W/m2, respectively (17, 18, 26), for a total aerosol forcing of −1.3 W/m2. These estimates are all highly uncertain. The most important uncertainty is that for the indirect effect, whose value is judged to lie somewhere between zero and −1.5 W/m2 (26).
FIG. 3. Breakdown of differences between the Jones and IPCC global-mean temperature data over 1861–1994: combined effect of different marine and land data sets, and different corrections to marine data. Land data set differences have a relatively minor effect. The differences shown are annual values of the difference Jones data minus IPCC data adjusted to use the Jones averaging method, with both data sets adjusted to have the same (zero) 1961–1990 means. Having determined a past forcing history, this may be used in an upwelling–diffusion energy–balance climate model to estimate the implied global-mean temperature changes. The model used here is that used in refs. 18, 27, and 28, the same model that has been used by IPCC (17). This model has a simplified ocean that accounts for oceanic lag effects, and differentiates the land and ocean areas in each hemisphere to model the spatially disparate forcing effects of sulfate aerosols and to account for the fact that the sensitivity of the climate system to external forcing differs over land and ocean areas. Land and ocean sensitivities are specified externally, allowing uncertainties in these parameters to be explored. The model ocean also has a variable upwelling rate, although this option is not used in the calculations presented here. Further details of the model are given in ref. 18, and its performance compared with coupled ocean/atmosphere general circulation models (O/AGCMs) is described in refs. 17 and 29. The main factor leading to uncertainties in the global-mean temperature response to any given forcing is the climate sensitivity, usually specified by the equilibrium global-mean warming for a CO2 doubling (∆T2×). Any calculations ignoring this uncertainty would be of very limited value. ∆T2× is thought to lie between 1.5°C and 4.5°C (30), with roughly 90% confidence. The best guess value for ∆T2× is 2.5°C (30). In the following analyses, the energy balance model has been used with a land/ocean sensitivity differential of 1.3 (explained in ref. 18), and constant upwelling rate, paralleling assumptions made in similar analyses carried out for IPCC (ref. 31, figure 8.4). The results of using the above IPCC forcing data with different global-mean climate sensitivities are given in Fig. 5, Fig. 6, and Fig. 8. For comparison purposes, we use the IPCC observed temperature data set. Because these data have a lower overall warming trend, the implied climate sensitivities are necessarily lower than they would be if we used the Jones data set.
FIG. 4. IPCC estimates of past anthropogenic forcing changes (17). Non-CO2 GHG forcing is the sum of the effects of CH4, N2O, halocarbons, and tropospheric and stratospheric ozone. Aerosol forcing combines direct and indirect sulfate effects and aerosols from biomass burning. In Fig. 5, only GHG forcing is used (a total of 2.6 W/m2 over 1765, the initial model simulation year, to 1990). The modeled and observed results shown have been adjusted to have a common 1861–1900 mean of zero. (The choice of reference period is somewhat arbitrary because only the temperature changes are modeled, not the absolute values.) It is clear from Fig. 5 that the best fit is obtained for a sensitivity below the 1.5°C lower bound of the 90% confidence range. The precise best-fit sensitivity depends on the method used for optimization and the time period over which the results are optimized, and also on the observed data set with which the model's results are compared. If the root-mean-square error (RMSE) is minimized over the full comparison period, 1861–1994, with no other constraints, and if raw annual IPCC global-mean temperatures are used in the comparison, the best-fit value for ∆T2× is 1.2°C (see Table 1). Table 1 gives results for comparisons with two other versions of the IPCC observed data set: IPCC annual data with the influence of the El Niño/Southern Oscillation (ENSO) phenomenon factored out (following ref. 32) and low-pass filtered IPCC data, as shown in Fig. 1. Similar implications arise no matter which data set is used. The ENSO component accounts for roughly 30% of the high-frequency variance and 9% of the total variance in the observed data. With the filtered data, the high-frequency component extracted accounts for 13% of the
THE OBSERVED GLOBAL WARMING RECORD: WHAT DOES IT TELL US?
8317
total variance. This latter value provides a useful bound on how much of the raw annual data variance can be explained by a deterministic fitting exercise that, like this one, considers only low-frequency forcing (namely 87%).
FIG. 5. Observed (OBS; IPCC) and modeled temperature changes for GHG forcing alone. Modeled changes are given for climate sensitivity (∆T2×) values of 1.5°C, 2.5°C, and 4.5°C. All series have been adjusted to have zero mean over 1861– 1900. 1994 is the last year shown. The same results have been presented in the 1995 IPCC report (31). The effect of including aerosol forcing is shown in Fig. 6. In this case, the best-fit value of ∆T2× is clearly higher than the upper 90% value of 4.5°C. If the RMSE against the raw IPCC data is minimized over 1861–1994, the best-fit sensitivity value is 6.3°C (see Table 1). This result is highly sensitive to the assumed amount of aerosol forcing largely because the transient temperature changes (i.e., the model output) become increasingly less sensitive to ∆T2× as ∆T2× increases (see ref. 31, and ref. 34, figure 8.1). Table 1 shows that, if the globalmean aerosol forcing is reduced from −1.3 W/m2 to −0.9 W/m2 (denoted the low case in Table 1), a change that is well within the aerosol forcing uncertainty range, then the best-fit sensitivity drops to 2.6°C. The addition of aerosol forcing improves the fit compared with the greenhouse-alone case (in terms of RMSE), but only marginally (see Table 1). The higher aerosol forcing case gives a slightly lower RMSE, but this result cannot be used to place more confidence on the higher forcing because RMSE results depend critically on assumptions made regarding other model parameters and the magnitudes of other forcings. Furthermore, as noted above, using the RMSE over the full analysis period is only one of many possible ways to judge a best fit; other methods have different best-fit implications. Table 1. Climate sensitivity estimates obtained by minimizing the RMSE between observed (IPCC) global-mean temperature changes and modeled changes over 1861–1994 Forcing Raw IPCC data IPCC data, ENSO removed Low-pass filtered IPCC data GHG Aerosol Solar ∆T2×, °C RMSE, °C R2 ∆T2×, °C RMSE, °C R2 ∆T2×, °C RMSE, °C R2 Yes — — 1.2 0.128 0.60 1.1 0.119 0.61 1.2 0.093 0.74 Yes Mid-IPCC — 6.3 0.121 0.64 6.0 0.114 0.65 6.4 0.083 0.79 Yes Low — 2.6 0.125 0.62 2.5 0.116 0.63 2.6 0.088 0.76 Yes Mid-IPCC Yes 3.0 0.117 0.67 2.8 0.107 0.69 3.0 0.077 0.82 Low Yes 1.8 0.118 0.66 1.7 0.109 0.68 1.8 0.079 0.81 Yes Results are given for five different forcing combinations and three different observed data series. The forcings used are IPCC GHG forcing; zero, central (Mid-IPCC), and “low” aerosol forcing (Mid-IPCC aerosol forcing is −1.3 W/m2 to 1990; low is −0.9 W/m2, a value that is substantially larger in magnitude than the IPCC lower bound); and solar forcing (33). The observed data cases are raw IPCC annual data; IPCC annual data with the ENSO influence factored out; and low-pass filtered IPCC data. The explained variance (R2) was calculated using R2 = 1 − (RMSE/SD)2, where SD is the standard deviation of the observed data (namely, 0.202°C for the raw IPCC data, 0.192°C for the data with ENSO removed, and 0.181°C for the low-pass filtered data).
Natural Causes of Global Warming Because the climate system varies naturally, one might expect at least part of the low-frequency, century time scale change in globalmean temperature to be due to natural factors. Overall, these factors could have produced a cooling or a warming. Natural variability can be split into externally forced changes and changes generated purely by internal processes [see also the paper by Keeling and Whorf (43), in this issue of the Proceedings]. The main external causes are solar irradiance changes and the effects of explosive volcanic activity (which produces a cooling aerosol layer in the stratosphere). Because the latter is a short-term effect, lasting at most a few years, and because the main focus of this paper is on the long-term observed trend, volcanic effects will not be considered here, but they can noticeably modify the results of the calculations. For solar influences, there is strong evidence of significant low-frequency changes in irradiance. These underlie irradiance changes related directly to the solar sunspot cycle (i.e., quasi-cyclic changes with a period of around 11 years) about which there is no dispute. To demonstrate the possible importance of solar forcing, the recent irradiance reconstruction of Hoyt and Schatten (33) is used (see Fig. 7). The inclusion of solar forcing raises two interesting issues: (i) the importance of the early forcing history (e.g., before 1800) in determining 20th century temperature changes and (ii) the choice of reference level (see ref. 35). Here, we begin our main calculations in 1765 (the initial year for GHG concentration changes) and use the estimated irradiance value in 1765 (1,371.55 W/m2) as the reference level. In other words, we assume that the climate system initially (1765) is in equilibrium with a solar irradiance of 1,371.55 W/m2; by default, therefore, we are assuming also that the irradiance was at this level for all time before 1765 (as illustrated in Fig. 7). An alternative would be to begin at the start of the Hoyt and Schatten solar record (1700) and use the 1700 irradiance value (1,367.52 W/m2) as a reference level. In this case, we are assuming, by default, a constant irradiance level of 1,367.52 W/m2 for all times before 1700 (see Fig. 7). In terms of top-of-the-atmosphere forcing, the difference in reference levels amounts to 0.71 W/m2 (cf. the IPCC central estimate of 1.3 W/m2 for anthropogenic forcing over 1765–1990), with the 1700-start case having a much larger overall (from 1700) forcing trend. It is pertinent to ask what affect does this difference have on temperature changes after 1861 (the start of our observed data analysis period)? We answer this question by comparing results for the two forcing cases. We note, however, that this comparison probably
THE OBSERVED GLOBAL WARMING RECORD: WHAT DOES IT TELL US?
8318
overestimates the magnitude of the history and reference-level effects. Using the 1700 value as the reference level may be too extreme because this date is within the anomalously low irradiance period corresponding to the Maunder sunspot minimum—a period often associated with the coldest part of the Little Ice Age. The 1765 value may be too high a reference level because most of the subsequent irradiance values are below this level (see Fig. 7). These two cases, therefore, probably bracket the range of possibilities. When the two simulations are compared, the history/reference-level effect is found to be very small; the difference in warming trend over the period 1861–1994 is only around 0.01°C, with the 1700-start case necessarily having the (very slightly) larger warming trend. This uncertainty is much less than that in the observed data (Fig. 1).
FIG. 6 Observed (OBS; IPCC) and modeled temperature changes for GHG and aerosol forcing. Modeled changes are given for climate sensitivity (∆T2×) values of 1.5°C, 2.5°C, and 4.5°C. All series have been adjusted to have zero means over 1861–1900. 1994 is the last year shown. The same results have been presented in the 1995 IPCC report (31).
FIG.7. Solar irradiance changes from Hoyt and Schatten (33). Their data span 1700–1992; estimated values are shown to 1994. To obtain top-of-the-atmosphere forcing values (to give changes that are then directly comparable with the anthropogenic forcings shown in Fig. 4), the values shown should be multiplied by 0.175. The main climate model calculations begin in 1765 and assume constant irradiance at the 1765 level before this date, as shown. Similar calculations using 1700 as a starting date with constant irradiance at the 1700 level for the time before 1700 lead to only negligible differences after the mid-19th century. This is an important result because it demonstrates that the warming since the mid- to late 19th century cannot be considered as a “recovery” from the cold period known as the Little Ice Age, a possibility alluded to in the 1990 IPCC report (1) and sometimes used by greenhouse skeptics as an alternative to anthropogenic forcing as an explanation of the 20th century warming trend. The 1700-start case includes an initial Little Ice Age period (indeed, it assumes that the globe was in a permanent Little Ice Age state before 1700!), yet results for this case differ only negligibly from the 1765-start case in terms of 20th century temperature changes; the memory of the climate system is simply not long enough to “remember” the low forcing interval before 1765. Full model results for the combined solar-plus-anthropogenic (greenhouse-plus-aerosol) forcing case are shown in Fig. 8.The addition of solar forcing gives a slight improvement in the best fit. The minimum RMSE based on the raw IPCC data drops to 0.117°C from 0.121°C (see Table 1), and the rapid observed warming over 1910–1940 and the subsequent leveling off of the temperature rise are matched more closely than for the pure anthropogenic forcing case. Similar RMSE improvements occur with the other observed data sets (Table 1). In addition, because the overall solar forcing trend is positive, the best-fit climate sensitivity is reduced, from 6.3°C to 3.0°C, based on the raw IPCC data. Using the low aerosol forcing gives an even lower sensitivity, 1.8°C (see Table 1). Part of the overall trend may also be due to internally generated natural variability; indeed, a considerable fraction of the annual to decadal time scale variability must be due to this factor. As noted above, roughly 13% of the total variance over 1861–1994 is at high frequencies (time scales on the order
FIG. 8. Observed (OBS; IPCC) and modeled temperature changes for GHG, aerosol, and solar forcing (i.e., as for Fig. 6 but with the solar forcing from Fig. 7 added). Modeled changes are given for climate sensitivity (∆T2×) values of 1.5°C, 2.5°C, and 4.5°C. All series have been adjusted to have zero means over 1861–1900. 1994 is the last year shown. The same results have been presented in the 1995 IPCC report (31).
THE OBSERVED GLOBAL WARMING RECORD: WHAT DOES IT TELL US?
8319
of 10 years or less), of which about 30% can be explained by ENSO. On longer time scales, the crucial issue here is whether or not part of the century time scale trend can be attributed to internally generated variability. The magnitude of this possible trend component has been estimated using the simple up-welling–diffusion energy–balance model employed here (see ref. 36) and using more complex coupled O/ AGCMs (37). An illustration of this unforced variability is given in Fig. 9, from the GFDL (Geophysical Fluid Dynamics Laboratory) O/ AGCM (37). If the O/AGCM data are searched for 134-year trends (i.e., the length of the observational record used here) the maximum global-mean trend is around 0.15°C. Somewhat larger century time scale, internally generated trend values have been given in ref. 36. Both models indicate, however, that only a relatively small (but nonetheless potentially important) fraction of the overall observed trend could be explained by internal processes. On shorter time scales, natural internally generated trends become more important relative to the anthropogenic signal, a fact that makes interpretation of shorter time scale changes increasingly difficult as the time scale decreases. For example, it has been noted that the observed warming trend over 1910–1940 is considerably larger than that expected from anthropogenic forcing (see Fig. 5 and Fig. 6). It is large, also, compared with the modeled effect of anthropogenic-plus-solar forcing (see Fig. 8). The discrepancy, 0.2–0.3°C over the 30year period, is, however, within what might be expected due to internally generated variability, which frequently shows trends of similar magnitude in model experiments. Note that the observed trend over this period is uncertain by at least 0.1°C (see Fig. 1). As an additional consistency check, we note that the residual variability obtained from the best-fit greenhouse/aerosol/solar forcing simulation (RMSE = 0.12°C if the raw IPCC data are used and 0.11°C if ENSO effects are removed) is similar to the average standard deviation over 134-year intervals in the GFDL internal variability simulation (namely, 0.09°C). Because the GFDL variability is a measure of the component that the present deterministic fitting exercise can never account for, this means that we are able to explain almost all of the variance that one could ever hope to explain. Given that the GFDL model underestimates the magnitude of interannual variability related to the ENSO phenomenon, and that the inclusion of volcanic forcing effects should reduce the best-fit RMSE value still further, this is an encouraging result. We note, however, that this result throws little light on which forcing assumption is best because all of the forcings considered here, when the model/observed data fit is optimized, yield similar RMSE values.
FIG. 9. Internally generated variability of global-mean temperature from the GFDL coupled O/AGCM control run. Only the first 200 years of the 1000-year run are shown. A long-term drift component of 0.023°C per century, probably a model artifact, has been removed. Although there is a quite striking overall consistency in the variance breakdown between the various anthropogenic and natural factors, this consistency does not preclude the possibility of a noticeable and practically significant century time scale trend in the observations due to internal factors. Such a trend could be either positive or negative. Accounting for this possibility introduces an additional element of uncertainty into any attempt to deduce the climate sensitivity from the observational data (38) even when the whole data record is considered. Table 1 shows that, because of external forcing uncertainties that lie well within the range of possibilities, the sensitivity could lie anywhere between 1.7°C and 6.4°C. If, however, part of the trend were internally generated, the range of possibilities may be substantially increased. For example, if 0.1°C of the overall observed warming trend of 0.5°C were internally generated (or, for that matter, the result of errors in the observational data), then the externally forced trend would only be 0.4°C. For any given forcing, the sensitivity required to explain a 0.4°C trend is much less than that required to explain a 0.5°C trend. This uncertainty factor, of course, may operate in either direction. Conclusions In attempts to explain the observed warming trend over the past century or so, inclusion of aerosol forcing improves the fit between observed and modeled data (in terms of the minimum RMSE value). However, the agreement between the observationally based (best-fit) sensitivity and the independent model-based range is not improved; the best-fit sensitivity changes from being too low (less than 1.5°C) if GHG forcing alone is used to too high (above 4.5°C) if the IPCC central estimate of aerosol forcing is included. These apparently anomalous sensitivities, however, can be easily reconciled with the model-based range, either through a small reduction in aerosol forcing or by including a solar forcing component. Neither addition changes the RMSE value significantly; including solar forcing leads to a slight improvement. Alternatively, as noted in ref. 38, a small, internally generated trend could help in reconciling these climate sensitivity differences, especially when considered in conjunction with the external forcing uncertainties. The potential importance of century time scale internal variability as a component of the observed warming has rarely been considered. Any such overall trend could be positive or negative. There is no way to determine whether such a trend component exists, let alone estimate its magnitude, but it is unlikely that it would be zero. In spite of the likelihood of an (unknowable) internally generated trend component, the model/observed data comparison presented here demonstrates that the observed century time scale global-mean temperature changes are consistent with a dominant anthropogenic influence and secondary influence from solar irradiance increases. The secondary nature of the solar role obtains because the overall solar forcing trend is substantially less than the central anthropogenic forcing estimate of 1.3 W/m2. The consistency of this explanation is reinforced by the magnitude of the residuals about the best fit, which is very similar to the magnitude of internally generated natural variability estimated independently from the GFDL coupled O/AGCM. This consistency, however, does not prove that there has been a large anthropogenic influence. Given uncertainties in the forcing (both anthropogenic and natural), it is still possible that part of the trend has been internally generated. Because
THE OBSERVED GLOBAL WARMING RECORD: WHAT DOES IT TELL US?
8320
of these various uncertainties, estimates of the value of the climate sensitivity cannot be improved by using observational data alone; the range of possible values for ∆T2× deduced in this way is substantially larger than the standard model-based range for ∆T2×, 1.5–4.5°C. It is, however, much more difficult to derive sensitivities below 1.5°C from the observational data than it is to obtain values above 4.5°C. Where do we go from here? In the most recent IPCC report (31) and in an earlier report (34), it was noted that studies of global-mean temperature alone are insufficient to show a compelling cause–effect relationship between anthropogenic forcing and climate change. Such studies, as shown above, can demonstrate that the observed warming is consistent with a substantial anthropogenic effect on climate but cannot accurately quantify this effect. To show a cause–effect linkage, more sophisticated techniques are required that make use of the patterns of observed climate change, either in the near-surface horizontal (latitude/longitude) plane (39, 40) or in the vertical (zonal mean/ height) plane (41, 42). Such pattern-based studies have shown increasing and statistically significant similarities between model predictions and observed temperature changes. These results, combined with the evidence from global-mean analyses, provide convincing evidence for a discernible human influence on global climate; but further work is required to better quantify the magnitude of the human influence and reduce uncertainties in the climate sensitivity. This work was supported by the U.S. Department of Energy (Grant DE-FG02-86ER60397) and the National Oceanic and Atmospheric Administration (Grant NA96AANAG0347). The National Center for Atmospheric Research is sponsored by the National Science Foundation. IPCC temperature data were provided by David Parker, U.K. Meteorological Office. 1. Folland, C. K., Karl, T. R. & Vinnikov, K. Y. (1990) in Climate Change: The IPCC Scientific Assessment, eds. Houghton, J. T., Jenkins, G. J. & Ephraums, J. J. (Cambridge Univ. Press, Cambridge, U.K.), pp. 195–238. 2. Folland, C.K., Karl, T. R., Nicholls, N., Nyenzi, B. S., Parker, D. E. & Vinnikov, K. Y. (1992) in Climate Change 1992: The Supplementary Report to the IPCC Scientific Assessment, eds. Houghton, J. T., Callander, B. A. & Varney, S. K. (Cambridge Univ. Press, Cambridge, U.K.), pp. 135– 170. 3. Nicholls, N., Gruza, G. V., Jouzel, J., Karl, T. R., Ogallo, L. A. & Parker, D. E. (1996) in Climate Change 1995: The Science of Climate Change, eds. Houghton, J. T., Meira Filho, L. G., Callander, B. A., Harris, N., Kattenberg, A. & Maskell, K. (Cambridge Univ. Press, Cambridge, U.K.), pp. 133–192. 4. Vinnikov, K. Y., Groisman, P. Y. & Lugina, K. M. (1990) J. Clim. 3, 662–677. 5. Hansen, J. & Lebedeff, S. (1987) J. Geophys. Res. 92, 13345–13372. 6. Hansen, J. & Lebedeff, S. (1988) Geophys. Res. Lett. 15, 323–326. 7. Jones, P. D., Raper, S. C. B., Bradley, R. S., Diaz, H. F., Kelly, P. M. & Wigley, T. M. L. (1986) J. Clim. Appl. Meteorol. 25, 161–179. 8. Jones, P. D., Raper, S. C. B. & Wigley, T. M. L. (1986) J. Clim. Appl. Meteorol. 25, 1213–1230. 9. Jones, P. D., Wigley, T. M. L. & Farmer, G. (1991) in Green-house-Gas-Induced Climatic Change: A Critical Appraisal of Simulations and Observations, ed. Schlesinger, M. E. (Elsevier Science Publishers, Amsterdam), pp. 153–172. 10. Jones, P. D. (1994) J. Clim. 7, 1794–1802. 11. Jones, P. D. & Briffa, K. R. (1992) Holocene 2, 174–188. 12. Parker, D. E., Jones, P. D., Folland, C. K. & Bevan, A. (1994) J. Geophys. Res. 99, 14373–14399. 13. Parker, D. E., Folland, C. K. & Jackson, M. (1995) Clim. Change 31, 559–600. 14. Woodruff, S. D., Slutz, R. J., Jenne, R. J. & Steurer, P. M. (1987) Bull. Am. Meteorol. Soc. 68, 1239–1250. 15. Jones, P. D., Wigley, T. M. L. & Briffa, K. R. (1994) in Trends ‘93: A Compendium of Data on Global Change, eds. Boden, T.A., Kaiser, D. P., Sepanski, R. J. & Stoss, F. W. (Carbon Dioxide Information Analysis Center, Oak Ridge, TN), pp. 603–608. 16. Folland, C. K. & Parker, D. E. (1995) Q. J. R. Meteorol. Soc. 121, 319–368. 17. Kattenberg, A., Giorgi, F., Grassl, H., Meehl, G. A., Mitchell, J. F. B., Stouffer, R. J., Tokioka, T., Weaver, A. J. & Wigley, T. M. L. (1996) in Climate Change 1995: The Science of Climate Change, eds. Houghton, J. T., Meira Filho, L. G., Callander, B. A., Harris, N., Kattenberg, A. & Maskell, K. (Cambridge Univ. Press, Cambridge, U.K.), pp. 285–357. 18. Raper, S. C. B., Wigley, T. M. L. & Warrick, R. A. (1996) in Sea-Level Rise and Coastal Subsidence: Causes, Consequences and Strategies, eds. Milliman, J. D. & Haq, B. U. (Kluwer, Dordrecht, The Netherlands), pp. 11–45. 19. Charlson, R. J., Langner, J., Rodhe, H., Leovy, C. B. & Warren, S. G. (1991) Tellus 43AB, 152–163. 20. Taylor, K. E. & Penner, J. E. (1994) Nature (London) 369, 734–737. 21. Twomey, S. A., Piepgrass, M. & Wolfe, T. L. (1984) Tellus 36B, 356–366. 22. Charlson, R. J., Lovelock, J. E., Andreae, M. O. & Warren, S. G. (1987) Nature (London) 326, 655–661. 23. Wigley, T. M. L. (1991) Nature (London) 349, 503–506. 24. Jones, A., Roberts, D. L. & Slingo, A. (1994) Nature (London) 370, 450–463. 25. Penner, J. E., Dickinson, R. E. & O'Neill, C. A. (1992) Science 256, 1432–1434. 26. Schimel, D. S., Alves, D., Enting, I., Heimann, M., Joos, F., et al. (1996) in Climate Change 1995: The Science of Climate Change, eds. Houghton, J. T., Meira Filho, L. G., Callander, B. A., Harris, N., Kattenberg, A. & Maskell, K. (Cambridge Univ. Press, Cambridge, U.K.), pp. 65–131. 27. Wigley, T. M. L. & Raper, S. C. B. (1987) Nature (London) 330, 127–131. 28. Wigley, T. M. L. & Raper, S. C. B. (1992) Nature (London) 357, 293–300. 29. Raper, S. C. B. & Cubasch, U. (1996) Geophys. Res. Lett. 23, 1107–1110. 30. Mitchell, J. F. B., Manabe, S., Meleshko, V. & Tokioka, T. (1990) in Climate Change: The IPCC Scientific Assessment, eds. Houghton, J. T., Jenkins, G. J. & Ephraums, J. J. (Cambridge Univ. Press, Cambridge, U.K.), pp. 131–172. 31. Santer, B. D., Wigley, T. M. L., Barnett, T. P. & Anyamba, E. (1996) in Climate Change 1995: The Science of Climate Change, eds. Houghton, J. T., Meira Filho, L. G., Callander, B. A., Harris, N., Kattenberg, A. & Maskell, K. (Cambridge Univ. Press, Cambridge, U.K.), pp. 407–443. 32. Jones, P. D. (1988) Clim. Monit. 17, 80–9. 33. Hoyt, D. V. & Schatten, K. H. (1993) J. Geophys. Res. 98, 18895–18906. 34. Wigley, T. M. L. & Barnett, T. P. (1990) in Climate Change: The IPCC Scientific Assessment, eds. Houghton, J. T., Jenkins, G. J. & Ephraums, J. J. (Cambridge Univ. Press, Cambridge, U.K.), pp. 239–255. 35. Wigley, T. M. L. & Raper, S. C. B. (1990) Geophys. Res. Lett. 17, 2169–2172. 36. Wigley, T. M. L. & Raper, S. C. B. (1990) Nature (London) 344, 324–327. 37. Stouffer, R.J., Manabe, S. & Vinnikov, K.Y. (1994) Nature (London) 367, 634–636. 38. Wigley, T. M. L. & Raper, S. C. B. (1991) in Climate Change: Science, Impacts and Policy, eds. Jäger, J. & Ferguson, H. L. (Cambridge Univ. Press, Cambridge, U.K.), pp. 231–242. 39. Santer, B. D., Taylor, K. E., Wigley, T. M. L., Penner, J. E., Jones, P. D. & Cubasch, U. (1995) Clim. Dyn. 12, 77–100. 40. Mitchell, J. F. B., Johns, T. C., Gregory, J. M. & Tett, S. F. B. (1995) Nature (London) 376, 501–504. 41. Santer, B. D., Taylor, K. E., Wigley, T. M. L., Johns, T. C., Jones, P. D., Karoly, D. J., Mitchell, J. F. B., Oort, A. H., Penner, J. E., Ramaswamy, V., Schwarzkopf, M. D., Stouffer, R. J. & Tett, S. F. B. (1996) Nature (London) 382, 39–46. 42. Tett, S. F. B., Mitchell, J. F. B., Parker, D. E. & Allen, M. R. (1996) Science 274, 1170–1173. 43. Keeling, C. D. & Whorf, T. P. (1997) Proc. Natl. Acad. Sci. USA 94, 8321–8328.
POSSIBLE FORCING OF GLOBAL TEMPERATURE BY THE OCEANIC TIDES
8321
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8321–8328, August 1997 Colloquium Paper This paper was presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held November 13–15, 1995, at the National Academy of Sciences, Irvine, CA.
Possible forcing of global temperature by the oceanic tides
CHARLES D. KEELING AND TIMOTHY P. WHORF Scripps Institution of Oceanography, University of California, San Diego, La Jolla, CA, 92093-0220
© 1997 by The National Academy of Sciences 0027-8424/97/948321-8$2.00/0 PNAS is available online at http://www.pnas.org. ABSTRACT An approximately decadal periodicity in surface air temperature is discernable in global observations from A.D. 1855 to 1900 and since A.D. 1945, but with a periodicity of only about 6 years during the intervening period. Changes in solar irradiance related to the sunspot cycle have been proposed to account for the former, but cannot account for the latter. To explain both by a single mechanism, we propose that extreme oceanic tides may produce changes in sea surface temperature at repeat periods, which alternate between approximately one-third and one-half of the lunar nodal cycle of 18.6 years. These alternations, recurring at nearly 90-year intervals, reflect varying slight degrees of misalignment and departures from the closest approach of the Earth with the Moon and Sun at times of extreme tide raising forces. Strong forcing, consistent with observed temperature periodicities, occurred at 9-year intervals close to perihelion (solar perigee) for several decades centered on A.D. 1881 and 1974, but at 6-year intervals for several decades centered on A.D. 1923. As a physical explanation for tidal forcing of temperature we propose that the dissipation of extreme tides increases vertical mixing of sea water, thereby causing episodic cooling near the sea surface. If this mechanism correctly explains near-decadal temperature periodicities, it may also apply to variability in temperature and climate on other times-scales, even millennial and longer. 1. Introduction The global record of atmospheric CO2 since 1958 shows evidence of quasi-10-year periodicity nearly synchronous with variability in global surface temperature (1). We have explored whether the latter is exceptional or also found in earlier instrumental temperature records dating back to 1855 (2). Our quest has led us into a controversial subject area, with no clear consensus for or against the existence of periodic quasi-decadal variations in temperature, or what might cause such variations. Claims of near 10-year periodicity in atmospheric temperature have often been linked with the sunspot cycle of approximately 11 years, but no consistent correlation has emerged (3). Changes in solar irradiance may explain solar forcing of temperature, but this mechanism seems negated by precise measurements of total irradiance that indicate only a very weak quasi-decadal variation (4). An association of temperature periodicity to the approximately 22-year solar magnetic “Hale” cycle has also been suggested (5), but without identification of a plausible temperature forcing mechanism. Damaging to claims that the Sun causes periodicities in temperature, or other aspects of weather, are reported interruptions in solar– weather correlations, most conspicuous during the 1920s (3). These failures involved not only temperature, but also regional precipitation, movement of pressure centers, winds, and storm frequency. They occurred without any obvious interruptions in the sunspot cycle. Following some of these failures, solar correlations seem to have reappeared, but with unexplained reversals of phase (3), notably an association of high sunspot numbers with cool periods before 1920, but with warm periods after 1950. Tidal action has also been suggested as a possible cause of periodicities in temperature and weather related either to the 18.6-year lunar nodical cycle or to its first harmonic. Many years ago (ref. 6, p. 222) it was proposed that abnormally great ocean tides modulate the amount of warm water entering the Arctic, Bering, and Baltic seas at 9-year intervals, perhaps by increasing the interchange of water across the sills separating these seas from the main oceans. Investigators more recently (7, 8) have proposed that weak but persistent north–south tidal currents, reversing over the 18.6-year nodical cycle, might periodically change sea surface temperature in the main oceans. Alternatively, it has been suggested (refs. 9 and 10, p. 41) that oceanic tides may affect sea surface temperature in shallow ocean waters by varying the intensity of vertical eddy diffusion there. Two other possible causes of quasi-decadal variations in global temperature, which we can only briefly mention here, are volcanic eruptions and internal oscillations of the coupled atmosphere–ocean system. Volcanic eruptions, which have produced great dust veils (11) and hence substantial blocking of sunlight, have tended to occur close to times of quasi-decadal cooling since 1855. The picture is complicated, however, because the timing of eruptions with respect to cooling has varied from one century to the next. Internal oscillations of the earth climate system may cause quasi-decadal variability according to results of recent state-of-the-art coupled general circulation models (e.g., ref. 12), reinforcing a school of thought in meteorology (ref. 13, p. 72) that the circulation of the oceans and atmosphere can change on decadal time scales with no external cause. Whether this tenet is true or not, the possibility of periodic forcing of temperature externally for example, by variable solar irradiance, oceanic tides, or volcanic activity, is not precluded, because an unstable coupled system can respond to periodic external forcing as well as oscillate freely. Here we present evidence that global temperature has fluctuated quasi-decadally since 1855, except for an interruption between about 1900 and 1945, thus supporting previous claims of failures of weather phenomena to maintain a correlation with the sunspot cycle near 1920. This interruption, although difficult to explain by a sunspot mechanism, does not rule out a tidal mechanism, because the astronomically driven tide raising forces since 1855 have exhibited strong 9-year periodicity only when quasi-decadal periodicity was evident in temperature data. Furthermore, unlike the perplexing shift in the phase of quasi-decadal temperature fluctuations with the sunspot cycle between the 19th and 20th centuries, there was no such shift in phase with respect to tidal forcing.
POSSIBLE FORCING OF GLOBAL TEMPERATURE BY THE OCEANIC TIDES
8322
To investigate further a possible relationship between tidal forcing and temperature appears to us to be worthwhile, because the oceanic tide raising forces vary in strength predictably over a wide range of time scales. If they should be shown to influence temperature on any time scale, they may explain periodic changes in weather and climate on other time scales as short as fortnightly and as long as the Milankovitch cycles. 2. Air Temperature Analysis and Sunspots To test the hypothesis of tidal forcing of temperature we have adopted a global compilation of temperature data over both land and in surface sea water (14), expressed as an anomaly beginning in 1855 and updated through mid-1995 (P. D. Jones, personal communication). We accept Jones' premise that sea surface and marine air temperatures follow each other closely on interannual time scales, so that combined land and marine temperature data portray global average variations in surface air temperature. The large scatter in monthly averaged global temperature data (dots plotted in Fig. 1) is not a strong encouragement to look for cyclic phenomena. Nevertheless, if the data are fit to a flexible nodal spline (15, 16) (solid curve in Fig. 1) to suppress the high frequency scatter, periods of persistently warmer and cooler conditions are indicated. Many of the warmer periods occurred at times of El Niño events (ref. 17, p. 623), suggesting coherent interannual variability. To detect possible fluctuations in global temperature on interannual time scales, we have fit monthly averages with nodal splines of successively greater stiffness (Fig. 2, Top), chosen to produce increasing degrees of low-pass filtering of the data. The choices were subjective, but are not critical to the outcome, because nearly the same distinct patterns are produced over a considerable range of stiffnesses. The stiffest spline (curve labeled 1) shows only a tendency for global air temperature to rise irregularly since 1855. The difference between the two looser splines produces an approximately decadal bandpass (Middle), whereas the difference between the stiffest and loosest curves shows a broad, low frequency bandpass (Bottom). In Fig. 3, a record of sunspot numbers (18) is compared with the near-decadal bandpass of temperature of Fig. 2. As shown by broken lines, four near-decadal peak temperatures immediately before 1905 were close to sunspot minima, while four immediately after 1960 were close to sunspot maxima. For several decades near 1920, peak temperatures in the bandpass were only about 6 years apart and did not correlate with sunspots at all. These results are thus consistent with the earlier findings reported by Herman and Goldberg (3), of an intermittent correlation of temperature with sunspots and a reversal of phase.
FIG. 1. Global surface temperature anomaly (combined land and marine) from 1855 through mid-1995 in degrees C (ref. 14 and P. D. Jones, personal communication). Monthly averages are shown as dots. The solid line is a spline fit (15) of these data with a standard error, σ, of 0.107°C. To examine further the oscillatory character of the global temperature record, we have computed its spectrum (Fig. 4) by the maximum entropy method (19), which is highly sensitive to spectral line detection. We have afterwards established the amplitudes and phases of 24 identified spectral peaks by least-squares fits, to avoid the problem that the maximum entropy method is not quantitatively reliable with respect to amplitudes and cannot establish phase relationships (20). Our method has been described in detail previously (21) and is applied here to an updated temperature record.
FIG. 2. Spline fits of the global temperature anomaly of Fig. 1 together with associated bandpasses. (Top) Global trends depicted by three superimposed spline fits (15), one consisting of a very stiff spline fit to yearly averages (dotted curve 1, σ of 0.106°C), and two consisting of splines of lesser stiffness fit to monthly averages (dashed curve 2 and solid curve 3, σs of 0.161 and 0.148°C, respectively). (Middle) Decadal bandpass (curve 3 minus curve 2). (Bottom) Low frequency bandpass (curve 3 minus curve 1). We reconstructed the time series of temperature by summing the resultant 24 computed sinusoidal spectral oscillations. Near the decadal time scale, a strong harmonic with a period of 9.3 years, when summed with a sideband at 10.3 years
FIG. 3. Comparison of mean sunspot number (ref. 18, 12-month running mean, upper curve) with the decadal bandpass of global surface temperature in degrees C, of Fig. 2 (Middle).
POSSIBLE FORCING OF GLOBAL TEMPERATURE BY THE OCEANIC TIDES
8323
(Fig. 4) produces oscillations with an average period of 9.8 years and a maximum amplitude, peak-to-peak, of 0.14°C (Fig. 5 Top). The two harmonics, of nearly equal amplitude in the reconstruction, beat with a period of 100 years and exhibit interference near 1925. Inclusion of additional harmonics to form a low-frequency bandpass (Fig. 5 Middle) does not cancel out the stronger, near-decadal oscillations of Fig. 5 Top, but additionally produces near 6-year oscillations between approximately 1900–1945. The pattern of oscillations is essentially as seen in the similarly broad bandpass derived from spline fits (Fig. 2 Bottom). Also, a full reconstruction (24 harmonics summed, Fig. 5 Bottom) reproduces much of the higher frequency variability seen in Fig. 1. Thus, the amplitudes and phases obtained by spectral analysis do not appear to be falsified by the restricted number of degrees of freedom of the spectral compositing, (cf. ref. 21 and Fig. 7).
FIG. 4. Maximum entropy spectrum of the global surface temperature anomaly of Fig. 1. The logarithm of power is plotted versus frequency in cycles per year. Above the 24 most prominent spectral peaks are shown their periods, in years. 3. A Proposed Mechanism for Lunisolar Tidal Forcing Our first objective in hypothesizing an oceanic tidal influence on global temperature is to propose a possibly plausible physical mechanism. We must bear in mind that the combined tide-raising forces of the Sun and Moon produce the same total tidal kinetic energy each year (D. E. Cartwright, personal communication). Nevertheless, the rate of production of kinetic energy varies for individual tidal events depending on the varying relative positions of the Earth, Moon, and Sun. We focus on oceanic tides, because the atmospheric tides are too small to be of practical importance in comparison (ref. 6, p. 218). The dominant constituents of oceanic tides are semi-diurnal and diurnal, but weaker, longer period elements also contribute to the tidal spectrum (ref. 22, p. 310). The strength of tides varies considerably with many cycles identified by Wood (ref. 23, pp. 201.42–201.60).
FIG. 5. Spectral bandpasses of the global temperature anomaly based on the maximum entropy spectrum of Fig. 4. (Top) Decadal (sum of two oscillations with periods of 9.31 and 10.27 years). (Middle) Low frequency (sum of nine oscillations with periods from 6.05 to 31.4 years, inclusive). (Bottom) Broad bandpass, including higher frequencies (sum of 24 spectral oscillations as described in text). Tides might influence temperature and weather by direct transport of heat (8), although the very low amplitude of long-period tides (24) makes this seem unlikely. A mechanism more likely to be dominant on the global scale would appear to be vertical mixing caused by a modulation of the distribution of the tidal potential between semi-diurnal, diurnal, and long-period elements, as proposed by Loder and Garrett (9). Even this mechanism is not obviously adequate to explain near-decadal temperature variations. A small set of strong tides, occurring over only a few days, would seem unlikely to produce cooler sea surface temperatures lasting for months or years. Also, since the total tidal energy dissipated each year is the same, a tidal cooling process, to be effective in causing interannual oscillations, must be strongly nonlinear. We propose that the dissipation of the strongest daily tides substantially increases vertical mixing in the oceans and thereby cools the overlying surface water, because the temperature of ocean water generally decreases with depth. This process is clearly nonlinear, because vertical eddy diffusivity depends quadratically on the tidal current velocity (9), and bottom friction dissipation on the third power of velocity (25). Also, prior mixing by only slightly weaker day-to-day tides in a series might reduce stratification and thereby promote greater impact of subsequent stronger events than otherwise. If tidally induced cooling should produce greater cloudiness or storminess, the attending lesser surface heating from the Sun or greater than average wind velocities should add even further to the nonlinearity. Finally, on longer time scales, greater ice formation and snow accumulation associated with cooling at high latitudes could provide additional positive feedback. These latter arguments are not dissimilar to those advanced to explain the ice ages, in which small changes in the latitudinal distribution of solar irradiance are believed to cause large climatic effects (26). Some direct evidence of tidal forcing of temperature exists. A recent study (27) reports fortnightly and monthly variability in sea surface temperature at tidal frequencies throughout the seas of Indonesia, a region important to world climate because of its proximity to the pool of warm water that affects the El Niño cycle. These temperature variations appear to be due to nonlinear dynamics, which cause a redistribution of tidal energy from semi-diurnal and diurnal to fortnightly and longer periods. The resulting modulation of the heat flux between the atmosphere and the ocean is shown to be large enough (many tens of watts m−2) to affect weather and climate. Some support of tidal forcing is also found in claims of increased storminess and rainfall in the first and third quarter of the lunar cycle, i.e., subsequent to the strong tidal forcing that accompanies new Moon and full Moon (28, 29, 30, 31 and 32). For a tidal mechanism of sea surface temperature variability to be plausible, it must be shown that tide raising forces generate, and therefore dissipate, enough energy in mixing ocean water to compete with other mixing mechanisms. The nontidal kinetic energy of the oceans, arising from the drag of the wind on the ocean surface, and augmented by the work of pressure forces at the the ocean surface and by the conversion of potential energy, generates about 6 × 10−3 W m−2 of kinetic
POSSIBLE FORCING OF GLOBAL TEMPERATURE BY THE OCEANIC TIDES
8324
energy (ref. 33, p. 7,683), therefore about 2 TW, integrated over the ice-free oceanic surface of 3.6 × 1014 m2. In comparison, the oceanic tides generate approximately 4 TW (34, 35), adequate to cause significant mixing. The mechanism of vertical mixing in the oceans is poorly understood, making it difficult to establish the importance of tidal mixing relative to other types of oceanic mixing. Nevertheless, recent studies suggest that the former may not be insignificant. Observations of sea water density (36) and a tracer experiment (37) indicate that mixing across density gradients (diapycnal mixing) is insufficient (about a factor 10 too small) to maintain the main thermocline of the open oceans. It is possible that vertical mixing occurs mainly at surface outcrops of density layers and in shallow areas where most tidal dissipation and consequent mixing takes place (27, 38, 39, 40, 41 and 42). Tidal forcing may thus significantly influence the open ocean thermocline (35). Tidal periodicities result from astronomical factors involving the rotation of the Moon and Earth in elliptical orbits about the Earth and Sun, respectively. The varying positions of the Earth, Moon, and Sun lead to global scale variations in the tide raising forces, but the distribution of tidal energy from these forces, and the fraction of the energy available to cause mixing, is unevenly distributed geographically. As a result, longer term tidal periodicities differ regionally (9). The quantitative prediction of possible tidal forcing of temperature therefore requires a detailed understanding of the actual tides not yet attained, but perhaps soon attainable (see e.g., ref. 35). Notwithstanding these complications, the possibility should not be dismissed that changes in temperature, globally averaged, may show the effects of tidal forcing on a scale in which astronomical periodicities survive in the data. Extremes in tide-raising forces, which are those tidal phenomena likely to affect temperature globally, occur at predictable times as a result of nearly optimal alignment and proximities of the Earth, Moon, and Sun (43), irrespective of local tidal behavior. We therefore feel justified here to begin an appraisal of possible tidal influence on temperature simply by comparing astronomical properties of tides with global temperature records. 4. Periodicities of the Oceanic Tides Our next objective is to describe the periodicities of extremely strong tidal events. In so doing we do not wish to imply that single tidal events are likely to be responsible for modulating sea surface temperature worldwide. They may be important, however, because they identify times of generally great tidal dissipation of kinetic energy, which could modulate temperature by means of an ensemble of events over days or even years.
FIG. 6. Timing of prominent lunisolar tide raising forces from A.D. 1850 to 2000 according to Wood (ref. 23, table 16). Each event is plotted as a vertical solid line whose length, above a threshold, is an approximate measure of the strength of forcing, expressed by the quantity γ, in degrees of arc per day, as described in the text. All events with γ greater than 17.02° d−1 are shown, resulting in approximately one displayed event per year. Selected dominant sequences of 18.03year events, labeled A-D, subdominant sequences, B* and C*, and an equinoctial sequence, EN, are identified by arcs connecting these tidal events and by dots or solid triangles (see text). Vertical hatched lines indicate times of events grouped 6 or 9 years apart, as described in the text. The motions of the Earth and Moon, although periodic, do not produce truly periodic strong tidal events, because these events require the near coincidence of four incommensurate recurring astronomical relationships, namely syzygy, perigee, eclipse, and perihelion. This circumstance, although adding complexity to the analysis, may, however, be an asset in proving a connection between tides and temperature, because interrupted or transient tidal periodicities should produce characteristic signatures of tidal forcing in temperature records. Eighteen-Year Repetitions of Strong Tides. The times of strong lunisolar tidal forcing since 1850 are shown in Fig. 6 by thin vertical solid lines of irregular height plotted versus time. As an approximate relative measure of the global tide raising forces of individual strong tidal events we have adopted the factor, γ, of Wood (ref. 23, p. 201), defined as the angular velocity of orbital motion of the Moon with respect to the perturbed motion of perigee, in degrees of arc per day at the moment of maximum forcing. This velocity is shown above a threshold of 17.02° d−1 by the length of each tidal line in Fig. 6. The tide raising forces define a hypothetical equilibrium tide (ref. 22, p. 316), which approximates the global average strength of the actual tides. Strong tide raising forces have occurred at very nearly 18-year intervals in staggered sequences, the most prominent of which are shown in Fig. 6 by arches of connecting lines and with black dots. The timing of several of these sequences is further indicated by vertical hatched lines. In the strongest dominant tidal sequence of the late 19th century, labeled B, forcing attained a maximum on December 31, 1880 (thick hatched line), and then declined. The subsequent dominant sequence of the 20th century, labeled C, reached a maximum 93.02 years later on January 8, 1974 (additional thick hatched line), and then similarly declined. Part of a dominant sequence before B, labeled A, and of one after C, labeled D, are also shown (without dots). All other tidal events depicted in Fig. 6 are also members of 18-year sequences. Also indicated in Fig. 6 by arches of connecting lines are two subdominant sequences, labeled B* and C*, which produced approximately 9-year events when combined, respectively, with Sequences B and C. Last, part of an equinoctial sequence, labeled EN, is shown (vertical broken lines). Between 1899 and 1947 this sequence produced 6-year events when combined with Sequences B and C* (vertical hatched lines and solid triangles).
POSSIBLE FORCING OF GLOBAL TEMPERATURE BY THE OCEANIC TIDES
8325
Astronomical Basis for 18-Year Events. Maximal tide raising forces occur only when the Sun and Moon are in direct mutual alignment. This occurs at syzygy (either full Moon or new Moon), provided also that the Moon or Sun be in eclipse with the Earth. The former two bodies must also be at the closest approach to the Earth, i.e., the Moon at perigee and the Sun at perihelion. Repetitions of syzygy, perigee, and eclipse are defined, respectively, by three lunar months (ref. 23, pp. 126–131): the synodic (29.5 days) representing every second recurrence of syzygy, the anomalistic (27.6 days) representing the recurrence of perigee, and the nodical (27.2 days) representing every second recurrence of the Moon positioned at its node, lying on the plane of the ecliptic, a requirement for an eclipse. The Earth and Sun attain closest approach (perihelion) once every anomalistic year. The anomalistic year is only slightly longer than the mean calendar year because perihelion advances very slowly, 1 day every 57 years (ref.22, p. 304). Perihelion presently occurs on January 2 in the Christian calendar. To assist in comprehending tidal periodicities we have listed selected periods of the near coincidence of two or more astronomical factors of tidal forcing in Table 1. As an exact, long-term measure of time we employ the “tropical year” of approximately 365.24 mean solar (calendar) days (ref. 22, p. 125), defined as the time between successive occurrences of the vernal equinox. (The tropical year, henceforth “year,” is essentially equal to the average calendar year. The anomalistic year is 1.00005 tropical years.) The shortest period in which syzygy, perigee, an eclipse, and perihelion nearly coincide is 18.030 tropical years (18.029 anomalistic years), consisting nominally of 223 synodic, 239 anomalistic, and 242 nodical months (see Table 1). Because even for these nearly commensurate periods there are offsets in timing (see Table 1), the maximal tidal-forcing event in any sequence of 18-year tides is preceded and succeeded by events having lesser tidal forcing. For Sequences A through D, shown in Fig. 6, the alignments of a given sequence are thus ever poorer, the greater the time interval from the climactic event of that sequence. Longer Term Patterns of Strong Tides. The interval between the climatic events of Sequences B and C is 93.02 years (compare with the thick hatched lines in Fig. 6), made up of five 18.03-year intervals and an additional interval of 2.87 years. The latter interval shows up as an offset of individual events of Sequence B from those of Sequence C. Because of this offset, there is no sustained periodicity related to dominant tidal events over more than about half a century (four 18-year cycles). Decadal Repetitions of Strong Tides. Subdominant tidal sequences, such as those labeled B* and C* in Fig. 6, have occurred almost midway between tidal events of dominant sequences. Each B* event occurred 8.97 years (111 synodic months) after a B event, the latter followed by another B* event 9.06 years (112 synodic months) later. The interval between individual B* and C* events, as between B and C events, was 2.87 years. In combination, the dominant and subdominant sequences, B, C, B*, and C*, have produced nearly decadal events, at times indicated in Fig. 6 by vertical hatched lines and large solid dots before 1900 and after 1945. Perigean Eclipse Cycle. The strength of strong tidal events, measured by the factor γ, is sensitive to their timing with respect to perihelion (ref.23, p. 218). The climactic events of Sequence B in December 1880 and Sequence C in January 1974 both occurred within a week of the date of perihelion. Weaker events of these sequences occurred ever further from perihelion because of the 0.029-year (11 day) deviation in timing of these events from exactly 18 anomalistic years. Near 1920, the timing with respect to perihelion was poor for both dominant Sequences B and C. As a consequence, another tidal cycle came into prominence in which the average timing of strong tidal events involved only perigee and eclipses. This cycle is characterized by two lunar orbital lines that rotate in opposite senses and at different rates. One, the line of apsides, is the major axis of the elliptical orbit of the Moon; rotating in the same sense as the Moon and Earth revolve around the Sun; it completes a revolution in 8.8475 tropical years on average (ref. 23, p. 179). The other is the line of nodes, defined as the intersection of the plane of the ecliptic with the plane of the Moon's orbit, the latter inclined to the former by 5.15°; it rotates in the opposite sense to the line of apsides, completing the well-known lunar nodical cycle in 18.6134 tropical years (ref. 23, p. 189). The line of apsides and line of nodes come into mutual alignment on average twice every 5.997 years. This time interval,
[1] the perigean eclipse cycle, is also the beat period between the anomalistic and nodical months. Because the timing of syzygy is not commensurate with Pe/2, one-half of the eclipse cycle expresses only the long-term average of the return period of perigean eclipses, the actual times varying from about 2.8 to 3.3 years (see Table 1). Six-Year Repetitions of Strong Tides. Prominent tidal events of the perigean eclipse cycle are shown in Fig. 6 between 1899 and 1947 as solid triangles. These events were associated with three sequences of 18-year tides, spaced approximately 6 years apart. Two of these sequences, B and C*, because they represent events occurring near the date of perihelion, also contributed to near-decadal periodicity, but the third sequence, EN, produced events near the autumnal equinox having little diurnal tidal contribution and relatively weak γ values (their semidiurnal tidal strength not well represented by γ values.) Of the three equinoctial tidal events, distinguished by broken lines, that of September 21, 1922, produced the greatest semidiurnal tide in an interval of 4 centuries (43). All three can be assumed to have caused remarkable tides in
*Period, in tropical years (365.2421988 days), for the number of synodic months indicated. Important average periods in boldface. †Synodic month, 29.53059 days; anomalistic month, 27.55455 days; nodical month, 27.21222 days. ‡In days. Parentheses indicate poor matching with the synodic month. §Must be a whole number. ¶Equal to (2.789396 + 2.870249 × 3 + 3.314935 × 2)/6. Equal to (5.659645 + 6.185184 × 2)/3. **Equal to (8.974580 + 9.055432)/2.
POSSIBLE FORCING OF GLOBAL TEMPERATURE BY THE OCEANIC TIDES
8326
regions where the semi-diurnal component predominates, such as in the north Atlantic Ocean (35). Sequence C events during this same period contributed strong forcing nearly midway between several of these 6-year events, but this did not lead to a sustained 3-year periodicity at Pe/2 because the other two associated 18-year sequences were too weak. Millennial Repetitions. The perigean eclipse cycle also influences tidal events on the millennial time scale because the return time for near coincidence of events of this cycle with perihelion is approximately 1,800 years. We propose that the repeat time of millennial extremes in tide raising forces, discussed below, relates to this return time, although the actual timing of such millennial events must be irregular, being sensitive to the exact time of syzygy (ref. 23, pp. 201–249). The near coincidence of perihelion with this cycle in the present millennium occurred near the time of a climactic tidal event in A.D. 1433 (ref. 6, p. 220). 5. Lunisolar Tides as a Forcing Agent of Temperature Our final objective is to demonstrate that the times of strong tidal forcing, based on astronomical factors, correlate with cool periods in the global temperature record at 6- to 10-year intervals. In Fig. 7 are shown four bandpasses of the global temperature anomaly obtained by either spectral analysis or spline fits. The upper three were shown previously; the lowest, including higher frequencies, is new. Also shown, by vertical hatched lines, are the times of selected tidal events as in Fig. 6. In referring to tidal events that occurred near the date of perihelion, we will cite the year of the event as though it occurred in January, unless we give the exact date. We first draw attention to the climactic tidal events of December 31, 1880, and January 8, 1974 (thick vertical hatched lines). Close to these dates (hereafter referred to as 1881 and 1974), spectrally derived oscillations in temperature, found by maximum entropy spectral analysis on the decadal time scale (curve 1, Decadal), show maximum rates of cooling, as indicated by the curve descending across the zero anomaly line. Thus, unlike the comparison of near-decadal temperature variations with the sunspot cycle (see Fig. 3), there is no shift in phase with respect to tidal forcing between centuries.
FIG. 7. Comparison of prominent 6- and 9-year tidal events, shown by vertical hatched lines, as in Fig. 6, with fluctuations in temperature shown by time plots derived spectrally or from spline fits. (Curve 1, Decadal) Decadal spectral bandpass as in Fig. 5 (Top). (Curve 2, Low Frequency/Spectral) Low frequency spectral bandpass as in Fig. 5 (Middle). (Curve 3, Low Frequency/Spline) Low frequency spline plot as in Fig. 2 (Bottom). (Curve 4, High Frequency) High frequency spline plot derived by subtracting curve 1 of Fig. 2 (Top) from the spline of Fig. 1. Next, still on the decadal time scale (curve 1), we examine the phasing of temperature variations with the timing of other tidal events (vertical hatched lines). The near coincidence of the climactic tidal events of 1881 and 1974 with maximum cooling rates did not extend to other decades, because the nearly 10-year intervals between cool periods exceeded the 9-year intervals between tidal events. The neardecadal temperature variations, nevertheless, have characteristics suggestive of tidal forcing. They are expressed as the sum of two harmonics (9.31 years and 10.23 years, see Fig. 4), which are close to the 9th and 10th harmonics of the 93-year tidal cycle. Also, they reinforced each other maximally close to the climactic tidal events of 1881 and 1974, and interfered maximally in the 1920s when the succession of 9-year tidal events was interrupted by an offset of 2.87 years. These two harmonics thus match the 93-year tidal cycle with its staggered sequences of 9-year events as well as can be expected for a single pair of spectral harmonics. A broader spectral bandpass including nine oscillations (Curve 2, Low Frequency/Spectral) shows additional relations of temperature to tidal events. This bandpass shows oscillations in phase with those of Curve 1, before 1900 and after 1945, when Curve 1 shows reinforcement, but 6-year oscillations between these dates, when Curve 1 shows interference. Except near the times of the climactic tidal events of 1881 and 1974, tidal events tended to coincide with temperature minima rather than maximum cooling rates. After the 9-year tidal event of 1863, as already noted above, these cool periods occurred approximately 1 year further apart than the tidal events. As a consequence, a decadal cool event, occurring in 1883, lagged the Sequence B tidal event of 1881, to the extent that the latter nearly coincided with the maximum decadal cooling rate. Moreover, the next such cool period, in 1893, lagged the Sequence B* tidal event by so much that it nearly coincided with a Sequence C* event, and was thus in phase with subsequent 6-year temperature oscillations. That these relationships are not an artifact of spectral analysis is shown by a lowpass spline fit (Curve 3, Low Frequency/Spline), which typically shows cool periods at the same times as by spectral analysis. This analysis, however, also shows the hint of a cool period near 1899, reinforcing the possibility that 6-year oscillations extended back to 1893. Thus, the 10-year spacing of cool events from 1863 to 1893 suggest an association with tidal forcing related to the perigean eclipse cycle; this forcing perhaps hastened (see Fig. 6) by the lesser strengths of the Sequence B* tidal events compared with Sequence C* events after 1900. (The tidal events of 1890 and 1893 indeed were of nearly equal strength.) The phasing of the temperature record with tidal events after 1956 is similar to that after 1863, leaving open the further possibility of similar tidal forcing 93 years after the events described for the late 19th century. Cool periods in a high-pass spline fit of temperature (Curve 4, High Frequency) also tend to coincide with 6- and 9-year tidal events. The cooling events of 1893 and 1899, discussed above, are seen to have been quite intense, although brief. In summary, since 1855, cool periods of global extent at near-decadal tidal intervals have typically occurred in episodes lasting about half a century, before, during, and after climactic tidal events spaced about a century apart. Between these episodes, also for about half a century, cool periods tended to occur synchronously with strong tidal events at 6-year inter
POSSIBLE FORCING OF GLOBAL TEMPERATURE BY THE OCEANIC TIDES
8327
vals. Almost all of the associated near-decadal and 6-year tidal events occurred at times of close mutual alignments of the Sun, Moon, and Earth with the Moon near perigee, but the former occurred closer to the date of perihelion than the latter. The changing importance of perihelion thus defined a near centennial tidal cycle, alternating between predominantly 6- and 9-year tidal forcing, which may be reflected in global temperature variations. 6. Further Implications of the Tidal Hypothesis Centennial Climatic Variability. To probe possible influences occurring on long time scales, we show in Fig. 8 the times and strengths of extreme tidal forcing back to A.D. 1600 and forward until A.D. 2140, according to Wood (23). As can be seen, the climactic tidal events that occurred 93.02 years apart in December 1880 and January 1974, were preceded, and will be succeeded, by other such climactic events. If our tidal hypothesis is valid these events should have been associated with times of unusually cool temperatures. A suggestion that this may have happened is indicated by the timing of cool episodes on a global scale identified by Jones and Bradley (ref. 44, pp. 658–659). These authors point out that the cool period that began near the end of the Middle Ages (the “Little Ice Age”) was not a monotonous period worldwide. Climactic evidence instead suggests several cool episodes, each lasting up to about 3 decades. These episodes, which appear to have been synchronous on the hemispheric and global scale, are shown as hatched rectangles in Fig. 8. As can be seen, they all occurred close to the times of near-centennial climactic tidal events, labeled Y, Z, A, and B. We have added an additional conforming cool period for the decades of the 1960s and 1970s, which would perhaps have been more pronounced in the absence of an anthropogenic warming component in recent years (45). As mentioned above in Section 4, we do not wish to imply that each of these cool periods could have been caused by only one or a few individual climactic tidal events. A more likely hypothesis is that they may have been caused by clusters of strong tidal events. A clustering tendency over several decades is suggested in Fig. 8 by a pyramidal pattern of more densely spaced quasi-annual tidal events in phase with climactic events of the dominant sequences. We postulate that a strongly nonlinear temperature response to an ensemble of strong tidal events may explain the global cooling that evidently occurred for several decades near the times of these climactic events.
FIG. 8. Timing of tidal forcing from A.D. 1600 to 2140, plotted with arcs connecting events of each prominent 18-year tidal sequence, as in Fig. 6. Also plotted are times of cool episodes seen in climate data identified by Jones and Bradley as of possibly global significance (ref. 44, pp. 658–659). A more recent global cool interval near 1970 is also shown, as discussed in the text. Dominant tidal Sequences A–D are labeled as in Fig. 6. Two earlier Sequences, Y and Z, and a later Sequence E are also labeled. Although climactic events of the dominant sequences from 1700 and 1974 are at 93-year intervals, there are irregularities in the time intervals between earlier and later climactic events, yielding an average near-centenniel interval of about 90 years. Longer-Term Climate Variability. The climactic tidal events shown in Fig. 8 alternated between stronger events (labeled Y, A, and C) and weaker events, with the two greatest occurring in 1610 [or in 1619 according to Cartwright (43)] and in 1787. Also, tidal events, as strong or nearly as strong, evidently occurred in 1247 and 1433 (11), thus forming, with the events in 1610 and 1787, a series with a repeat period of approximately 180 years. An even longer perspective of strong tidal forcing is gained from a study of the timing of astronomical alignments by Cartwright (43), who showed that the greatest tide raising forces in the past millennium occurred between A.D. 1340 and 1619, at 93-year intervals, when the perigean eclipse cycle was almost optimally timed with respect to perihelion. By his calculation, tides of such great magnitude will not occur again until A.D. 3182, an interval corresponding approximately to the near 1,800-year return period of optimal timing of perigean eclipses with perihelion discussed above in Section 4. According to the calculations of Otto Pettersson made many years ago, tides of great magnitude conforming roughly to this return period also occurred near 3500 B.C., 1900 B.C., and 200 B.C., as well as in A.D. 1433 (ref. 6, pp. 220–222). Although records of weather before the late Middle Ages appear to be too sketchy to test convincingly for a millennial tidal influence on climate, we point out that the 1,800-year tidal cycle, outlined above, implies that the climactic tides of the millennium before A.D. 1200 were weaker than those of recent centuries and should have promoted a warmer climate. The “medieval warm period,” between about A.D. 800 and 1000, followed by a decline in weather in the 1200s (ref. 13, pp. 177–187), conforms to this expectation, especially when considered together with the severe weather of the succeeding Little Ice Age. Moreover, for still earlier times, there is at least a hint of more unsettled weather (13) near the centuries of great tidal forcing as calculated by Pettersson. Over still longer times, the strength of tidal forcing must have changed in response to secular changes in the extreme distances of the Earth from the Moon and Sun, the inclination of the Earth's rotational axis to the ecliptic, and the position of perihelion with respect to the vernal equinox. Three of these quantities, involving the Earth and Sun, also determine the Milankovitch cycles, which have been shown with considerable statistical reliability to explain the timing of the ice ages over the past million years (26). It is widely held that solar irradiance reaching the Earth, because it is modified cyclically as a function of latitude by these three factors, has been the driving force for the ice ages, but the mechanism is less certain than the cyclic correlation. Also, the mechanism of the Milankovitch cycles does not appear to explain shorter period fluctuations seen in glacial data of the most recent ice age and the interglacial period preceding it (46, 47). Perhaps tidal forcing was also involved (48). 7. Discussion and Conclusions In our quest to understand quasi-decadal oscillations observed in atmospheric carbon dioxide, we have been led to investigate seemingly similar oscillations in global air temperature. Employing a much longer temperature record than that available for atmospheric CO2, we have found that near-decadal variations in global air temperature are characteristic of the past 141 years, except for a roughly 45year interruption centered near 1920. This pattern has also emerged using spectral analysis, specifically from the beating of two frequencies found to be close to the 9th and 10th harmonics of the lunisolar tidal cycle of 93 years. Furthermore, temperature oscillations with periods near 6 years were found in the temperature record by spectral analysis near the time of interference of the two
POSSIBLE FORCING OF GLOBAL TEMPERATURE BY THE OCEANIC TIDES
8328
near-decadal oscillations, and thus close in period to the 6-year repeat period of another prominent lunisolar tidal cycle. Is it possible that the oceanic tides influence global temperature? Perceptions of cyclic behavior in the climatic record have in the past involved so many exceptions and inconsistencies that the subject does not have a good reputation among scientists. Compelling evidence that a recognized periodicity is actually caused by an identifiable extraterrestrial forcing agent is difficult to find. Our task is not made any easier by the nonstationarity of the lunisolar tidal cycles resulting from the incommensurability of the various astronomical periodicities. Although patterns in the strength of tidal forcing often recur, they don't repeat identically even after hundreds of years. Evidence of a tidal connection therefore cannot rely solely on the usual past practice of looking for a correlation with temperature at a single tidal periodicity. Although we have delved into properties of the tides in some detail to test whether a correlation of tidal strength with temperature exists, much more might be accomplished by a closer attention to the possible physical basis for the correlations found. Until now, to mount such an effort has not seemed worthwhile, given the small perceived likelihood that any lunisolar tidal connection to climate exists. We have only touched upon a possible cause by proposing that strong tides increase vertical mixing in the oceans and thereby episodically cool the sea surface. Also, we have explored in detail only 6to 10-year periodicities seen in records of both temperature and tidal forcing. We propose, nevertheless, that the near synchronicities seen at these periodicities argue sufficiently in favor of a tidal-forcing hypothesis, to justify further investigation of a possible tidal mechanism of temperature and climate variability. We are grateful to many who gave generously of their time to discuss the subject of tides, solar phenomena, and climatic variation with us in the course of preparation of this article. We specifically thank Phillip Jones, David Cartwright, Amy Ffield, James Hansen, Reid Bryson, Harry van Loon, Thomas Wigley, Christopher Garrett, Thomas Royer, David Parker, Henry Diaz, Thomas Karl, and Fergus Wood. We are also grateful for discussions with Timothy Barnett, Robert Bacastow, Daniel Cayan, Myrl Henderschott, Ralph Keeling, and Walter Munk, at the University of California at San Diego. We further thank Dr. Jones and his coworkers for supplying us with their temperature data sets. Computer time was provided by the San Diego Supercomputer Center. Financial support was from the National Science Foundation (Grant ATM-91-21986) and from the U.S. Department of Energy (Grants FG03-90ER-60940 and FG03-95ER-62075). 1. Keeling, C. D., Whorf, T. P., Wahlen, M. & van der Plicht, J. (1995) Nature (London) 375, 666–670. 2. Jones, P. D. (1994) J. Clim. 7, 1794–1802. 3. Herman, J. R. & Goldberg, R. A. (1985) Sun, Weather, and Climate (Dover, New York). 4. Wigley, T. M. L. & Raper, S. C. B. (1990) Nature (London) 344, 324–327. 5. Newell, N. E., Newell, R. E., Hsiung, J. & Zhongxiang, W. (1989) Geophys. Res. Lett. 16, 311–314. 6. Lamb, H. H. (1972) Climate: Present, Past, and Future (Methuen, London). 7. Maxsimov, I. V. & Smirnov, N. P. (1965) Oceanology 5, 15–24. 8. Royer, T. C. (1993) J. Geophys. Res. 98, 4639–4644. 9. Loder, J. W. & Garrett, C. (1978) J. Geophys. Res. 83, 1967–1970. 10. Pittock, A. B. (1983) Q. J. R. Meteorol. Soc. 109, 23–55. 11. Lamb, H. H. (1970) R. Soc. London Philos. Trans. Ser. A 266, 425–533. 12. Latif, M. & Barnett, T. P. (1994) Science 266, 634–637. 13. Lamb, H. H. (1995) Climate, History and the Modern World (Routledge, New York), 2nd Ed. 14. Jones, P. D., Wigley, T. M. L. & Briffa, K. R. (1994) in Trends '93: A Compendium of Data on Global Change, eds. Boden, T. A., Kaiser, D. P., Sepanski, R. J. & Stoss, F. W. (CDIAC, Oak Ridge, TN), pp. 603–608. 15. Reinsch, C. H. (1967) Numer. Math. 10, 177–183. 16. Enting, I. G. (1987) J. Geophys. Res. 92, 10977–10984. 17. Quinn, W. H. & Neal, V. T. (1992) in Climate Since A.D. 1500, eds. Bradley, R. A. & Jones, P.D. (Routledge, London), pp. 623–648. 18. U.S. Department of Commerce, (1989) Solar Geophysical Data, Prompt Report No. 535 (GPO, Washington, DC), Part 1, p. 11. 19. Press, W. H., Flannery, B. P., Teukolsky, S. A. & Vetterling, W. T. (1992) Numerical Recipes in Fortran (Cambridge Univ. Press, New York). 20. Sonett, C. P. (1983) Weather and Climate Responses to Solar Variations, ed. McCormac, B. M. (Colorado Associated Univ. Press, Boulder), pp. 607–613. 21. Keeling, C. D. & Whorf, T. P. (1996) DEC-CEN Workshop on Climate Variability (Natl. Acad. Sci., Washington, DC), pp. 97–109. 22. Neumann, G. & Pierson, W. J., Jr. (1966) Principles of Physical Oceanography (Prentice–Hall, Englewood Cliffs, NJ). 23. Wood, F. J. (1986) Tidal Dynamics (Reidel, Dordrecht, The Netherlands). 24. Trupin, A. & Wahr, J. (1990) Geophys. J. Int. 100, 441–453. 25. Miller, G. R. (1966) J. Geophys. Res. 71, 2485–2489. 26. Imbrie, J. & Imbrie, K.P. (1979) Ice Ages: Solving the Mystery (Enslow, Short Hills, NJ). 27. Ffield, A. & Gordon, A. L. (1996) J. Phys. Oceanogr. 26, 1924–1937. 28. Rodés, L. (1937) Influye la Luna en el Tiempo, Memorias del Observatorio del Ebro No. 7 (Tortosa, Spain). 29. Adderley, E. E. & Bowen, E. G. (1962) Science 137, 749–750. 30. Brier, G. W. & Bradley, D. A. (1964) J. Atmos. Sci. 21, 386–395. 31. Bradley, D. A. (1964) Nature (London) 204, 136–138. 32. Hanson, K., Maul, G.A. & McLeish, W. (1987). J. Clim. Appl. Meteorol. 26, 1358–1362. 33. Oort, A. H., Anderson, L. A. & Peixoto, J. P. (1994) J. Geophys. Res. 99, 7665–7688. 34. Webb, D. J. (1982) Contemp. Phys. 23, 419–442. 35. Kantha, L. H., Tierney, C., Lopez, J. W., Desai, S. D., Parke, M. E. & Drexler, L. (1995) J. Geophys. Res. 100, 25309–25317. 36. Armi, L. (1978) J. Geophys. Res. 83, 1971–1979. 37. Ledwell, J. R., Watson, A. J. & Law, C. S. (1993) Nature (London) 364, 701–703. 38. Sandstrom, H. & Elliott, J. A. (1984) J. Geophys. Res. 89, 6415–6426. 39. Sherwin, T. J. (1988) J. Phys. Oceanogr. 18, 1035–1050. 40. Largier, J. L. (1994) J. Geophys. Res. 99, 10023–10034. 41. White, M. (1994) J. Geophys. Res. 99, 7851–7864. 42. Sjöberg, B. & Stigebrandt, A. (1992) Deep-Sea Res. 39 (2), 269–291. 43. Cartwright, D. E. (1974) Nature (London) 248, 656–657. 44. Jones, P. D. & Bradley, R. S. (1992) in Climate Since A.D. 1500, eds. Bradley, R. S. & Jones, P. D. (Routledge, London), pp. 649–665. 45. Wigley, T. M. L. (1997) Proc. Natl. Acad. Sci. USA 94, 8314–8320. 46. GRIP Members (1993) Nature (London) 364, 203–207. 47. Kotilainen, A. T. & Shackleton, N. J. (1995) Nature (London) 377, 323–326. 48. Ffield, A. (1994) Doctoral thesis, (Columbia University, New York).
SPECTRUM OF L00-KYR GLACIAL CYCLE: ORBITAL INCLINATION, NOT ECCENTRICITY
8329
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8329–8334, August 1997 Colloquium Paper This paper was presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held November 13–15, 1995, at the National Academy of Sciences, Ivine, CA.
Spectrum of l00-kyr glacial cycle: Orbital inclination, not eccentricity
RICHARD A. MULLER* AND GORDON J. MAC DONALD†
© 1997 by The National Academy of Sciences 0027-8424/97/948329-6$2.00/0 PNAS is available online at http://www.pnas.org. ABSTRACT Spectral analysis of climate data shows a strong narrow peak with period 100 kyr, attributed by the Milankovitch theory to changes in the eccentricity of the earth's orbit. The narrowness of the peak does suggest an astronomical origin; however the shape of the peak is incompatible with both linear and nonlinear models that attribute the cycle to eccentricity or (equivalently) to the envelope of the precession. In contrast, the orbital inclination parameter gives a good match to both the spectrum and bispectrum of the climate data. Extraterrestrial accretion from meteoroids or interplanetary dust is proposed as a mechanism that could link inclination to climate, and experimental tests are described that could prove or disprove this hypothesis. Using much improved dating techniques, Broecker and van Donk (1) in 1970 conclusively established that the dominant cycle in proxy climate records is 100 kyr. Broecker and van Donk did not commit themselves as to the origin of the 100-kyr cycle. In the years after 1970, it became customary to attribute the 100,000-year cycle to variations in the orbital eccentricity of the earth (2). Calculated variation of eccentricity shows a quasi-periodic behavior, with a period of about 100 kyr. Milankovitch (3, 4) proposed that eccentricity affected the climate through its effect on insolation: the average solar energy reaching the earth. In this paper we note five sets of observations which conflict with the suggestion that insolation variations associated with eccentricity are responsible for the dominant 100,000-year cycle. First, the eccentricity changes are small, between 0.01 and 0.05. The resulting changes in insolation are far too small to account for the dominant 100,000-year cycle observed in proxy climate records. Second, the orbital calculations which can be carried out with great accuracy back to several million years (5) show that the major cycle in eccentricity is 400,000 (400 kyr), rather than 100 kyr. A 400-kyr fluctuation is absent in most climate records, leading to specific disagreement between eccentricity and glacial data at both 400 ka and the present (the “stage 1” and “stage 11” problems). Many proposed explanations for the discrepancies have been advanced; in a recent review, Imbrie et al. (6) give a short list consisting of seven groups of models. Many of the models involve resonant or nonlinear behavior of the ice–ocean–atmosphere system; some derive the 100-kyr period from the envelope of the variation in the precession parameter. Well-dated climate proxy records show the 100,000-year cycle only over the last million years (7). Prior to this transition, the 100-kyr period is either absent or very weak. Calculated variation of eccentricity does not show any discontinuity a million years ago. If the eccentricity drove changes in insolation, it would be anticipated that variations in insolation due to changes in eccentricity would affect climate in earlier periods, as well as over the past million years. Since methods of dating have improved, a fourth possible problem with the Milankovitch insolation has developed: several recent observations suggest that the abrupt termination of the ice ages preceded warming from insolation (8), an effect we refer to as “causality problem.” The interpretation of these results is still controversial (9, 10, 11, 12 and 13). Furthermore, Imbrie et al. (9) argue that a true test of the Milankovitch theory must be performed in the frequency domain, not the time domain. The fifth problem with the Milankovitch insolation theory is found in the frequency domain. In this paper, we present a full resolution spectral analysis of δ18O proxy climate records. The analysis shows that the 100-kyr period is a single, narrow peak, a simple pattern that strongly confirms an astronomical origin, but which cannot be reconciled with any of the models presented in the review by Imbrie et al. (6) In contrast, an alternative model that we have proposed, which attributes the 100-kyr cycle to orbital inclination, passes all the spectral tests that the Milankovitch model fails (14). Climate Proxy Records The isotopic composition of the oxygen isotopes in sediment is believed to reflect the percentage of earth's water frozen in ice, and thus changes in the oxygen–isotope ratio δ18O are measures of the earth's climate. While we have examined a large number of records to test our conclusions, we use two primary records in this analysis: from ocean drilling project site 607 (15) and the Specmap (16) compilation. We chose these records because both had time scales that had not been tuned to match a presumed 100-kyr eccentricity cycle. Such tuning, had it been done, could have artificially narrowed the width of the 100 kyr spectral peak. The δ18O signals for these data for the past 600 kyr are shown in Fig. 1 a and Fig.1 b. The similarity between the two records is evident; the dominant feature is the 100-kyr cycle. The spectra for these data are shown in Fig. 1 c and Fig. 1 d. For site 607, which has unevenly spaced data, the spectrum is calculated using the methods of Lomb (17) and MacDonald (18); however, we obtained essentially identical results using interpolation and data taper followed by standard Fourier transform or by the Blackman–Tukey method (provided full lags were used). The Milankovitch model attributes the peak near 0.01 cycles per kyr (100-kyr period) to variations in the earth's eccentricity. The 0.024 cycles per kyr peak (41-kyr period) to changes in the obliquity (tilt of the earth's axis with respect to the ecliptic), and the 0.04 cycles per kyr peak (23-kyr period) to changes in the precession parameter (delay between perihelion and summer solstice). Note that the full width at half maximum (FWHM) of the 0.01 cycles per kyr peak (100-kyr period) is 0.0016 cycles per kyr, near the theoretic minimum width (0.0015 cycles per kyr) that can be obtained with a record of
*Department of Physics and Lawrence Berkeley Laboratory, University of California, Berkeley CA 94720; and †International Institute for Applied Systems Analysis, A-2361 Laxenburg, Austria
SPECTRUM OF L00-KYR GLACIAL CYCLE: ORBITAL INCLINATION, NOT ECCENTRICITY
8330
600-kyr length. A re-analysis of the original pacemaker core stack with full resolution also produces a single narrow peak (FWHM = 0.0019), which is the theoretically minimum width for a record of 464 kyr length. Likewise, the spectral analysis of data from site 806 (19) shows a single narrow peak.
FIG. 1. δ18O for past 800 kyr. (a) Data of site 607 from Ruddiman et al. (15). (b) Specmap stack of Imbrie et al. (16). (c) Spectral power of site 607. (d) Spectral power of Specmap. In the Milankovitch theory, the peak near 0.01 (100-kyr period) is attributed to eccentricity, the peak near 0.024 (41-kyr period) to obliquity, and the peak near 0.043 (23-kyr period) to precession. The narrow width of the 100-kyr peak strongly suggests a driven oscillation of astronomical origin. In contrast to dynamical astronomy, where dissipative processes are almost nonexistent, all known resonances within the earth–atmosphere system have energy transfer mechanisms that cause loss of phase stability. Narrowness of the 41-kyr and 23-kyr cycles is not necessarily significant, since the time scale of the data was tuned by adjusting the sedimentation rate to match the expected orbital cycles. The 100-kyr peak is incoherent with these other two cycles, there is no phase relationship. The fact that an unrelated peak is sharp can be considered as an a posteriori evidence that the tuning procedure yielded a basically correct time scale, although it could be incorrect by an overall stretch factor and delay. We did not anticipate the narrowness of the 100-kyr peak, assuming, as others have done, that it was due to forcing by variations in eccentricity. However, it is not easily reconciled with any published theory. The narrowness of the peak was missed in previous spectral analysis of isotopic data because of the common use of the Blackman–Tukey algorithm (20), which, as usually applied (lag parameter = 1/3), artificially broadens narrow peaks by a factor of 3. The Blackman–Tukey algorithm gained wide use in the 1950s because of Tukey's admonition that analysts could be misled by using classical periodograms in analyzing spectra having a continuous spectrum. For analysis of glacial cycles, these considerations did not arise, because the spectra are mixed spectra with very strong quasi-periodic peaks. Spectra of glacial cycles, as Tukey recognized, lend themselves to the use of conventional Fourier transforms. The region of the 100-kyr peak for the δ18O data is replotted in Fig. 2 a and Fig. 2 b with an expanded frequency scale. These plots can be compared with the spectral power of the eccentricity variations, shown in Fig. 2c, calculated from the detailed computations of Quinn et al. (5). Three strong peaks are present in the eccentricity spectrum: near 0.0025 cycles per kyr (400-kyr period), near 0.08 cycles per kyr (125-kyr period), and near 0.0105 cycles per kyr (95-kyr period). The disagreement between the spectrum of climate and that of eccentricity is evident. The absence of the 400-kyr peak in the climate data has long been recognized (for a review, see Imbrie et al. (6), and numerous models have been devised that attempt to suppress that peak. We note that the 100-kyr peak is split into 95- and 125-kyr components, in serious conflict with the single narrow line seen in the climate data. The splitting of this peak into a doublet is well known theoretically (22), and results from the phase-coherent modulation by the 400-kyr peak. But in comparisons with data, the two peaks in eccentricity were made into a single broad peak by the enforced poor resolution of the Blackman–Tukey algorithm. The single narrow peak in the climate data was likewise broadened and the resulting comparisons led to the belief that the theoretical eccentricity and the observed climate data were very much alike. The disagreement between the data (Fig. 2 a and Fig. 2 b) and the theory (Fig. 2 c and Fig. 2 d), cannot be accounted for by experimental error uncertainty. Tuning of the time scale to a specific peak (by adjusting the unknown sedimentation rates) can artificially narrow that peak as other peaks that are coherent with it [see, for example, Neeman (23)]. However, the data in Fig. 2 a and Fig 2 b were tuned only to peaks obliquity and precession that are incoherent with the 100-kyr eccentricity cycle, so that tuning cannot account for the narrow width. Likewise, chatter (errors in the time scale from mis-estimated sedimentation
SPECTRUM OF L00-KYR GLACIAL CYCLE: ORBITAL INCLINATION, NOT ECCENTRICITY
8331
rates) cannot reconcile the disagreement, since although chatter can smear a doublet into a single broad peak, it will not turn a doublet into a single narrow peak. Could a physical mechanism convert a 95- to 125-kyr doublet into a single narrow peak? Dissipative mechanisms could obscure the doublet, but (like chatter) they yield a single broad peak or a cluster of doublets. Resonances used to suppress the 400-kyr peak are not sharp enough to suppress one element of the narrow doublet. In principle, a strong nonlinear process could turn a doublet into a single peak, as it does (for example) in a laser; however, no such mechanism has been identified in the lossy, friction-filled environment of the earth and its atmosphere.
FIG. 2. Spectral fingerprints in the vicinity of the 100-kyr peak for data from site 607 (a); for data of the Specmap stack (b); for a model with linear response to eccentricity, calculated from the results of Quinn et al. (5) (c); for the nonlinear icesheet model of Imbrie and Imbrie (21) (d); and for a model with linear response to the inclination of the Earth's orbit (measured with respect to the invariable plane) (e). All calculations are for the period 0–600 ka. The 100-kyr peak in the data in a and b do not fit the fingerprints from the theories c and d, but are a good match to the prediction from inclination in e. Several nonlinear models reviewed by Imbrie et al. (6) derive the 100 kyr cycle from the envelope of the precession cycle. However, this envelope also has a split peak, since it derives ultimately from eccentricity (the envelope of the precession is the eccentricity). As an example, we show the spectrum of the ice sheet model of Imbrie and Imbrie (21) in Fig. 2d. As expected, it too shows the 95- to 125-kyr doublet, in disagreement with the data. None of the nonlinear models in the recent comprehensive review by Imbrie et al. (6) have the required laser-like mechanism, and they all predict a split peak. This is a fundamental disagreement, not fixed by adjusting parameters. Unlike the 400-kyr cycle, which is far enough mismatched from the 100 kyr to be suppressed (at least in principle) by the models, the lines in the 95–125 doublet are too close. We draw a remarkably strong conclusion that variations in the earth's eccentricity cannot be responsible for the 100-kyr cycle. Orbital Inclination: An Alternative 100-kyr Cycle We recently proposed that a different orbital parameter, the inclination of the earth's orbit to the invariable plane of the solar system, should be associated with the 100-kyr glacial cycle (14, 24). The invariable plane of the solar system is that plane perpendicular to the angular momentum vector of the solar system, and is approximately equal to the orbital plane of Jupiter. The dominant peak in the spectrum of the inclination is at 0.01 cycles per kyr (100-kyr period) in a remarkably close match to the 100-kyr peak observed in the climate spectra. According to theory, this 100-kyr peak is also split, but only by 10−3 cycles per kyr, and this cannot be resolved with the 600-kyr record length. The variation of inclination i with time is calculated using the long-term integrations of Quinn et al. (5) and projecting the variation of inclination to the invariable plane. The existence of the 100-kyr cycle of orbital inclination does not seem to have been previously noted by climatologists. It may have been missed for two reasons. Ever since the work of Milankovitch, the implicit assumption has been that insolation is the driving force for climate cycles, and the insolation is not directly affected by orbital inclination. In addition, the 100-kyr cycle is not evident until the orbital elements are transferred to the natural reference plane of the solar system, the invariable plane. The fit of orbital inclination to the δ18O data from Specmap is shown in Fig. 3. Only two parameters were adjusted in the fit: one to set the relative scale between inclination and δ18O and a lag representing the delayed ice response to inclination. The best fit had a lag of 33 ± 3 kyr, with inclination accounting for 43% of the variation in the δ18O signal (for a record extending back 900 kyr the fit is even better, with inclination accounting for 48% of the variation) (25). Note that the inclination cycle has no 400-kyr component: the 100-kyr cycle remains strong for the last 600 kyr. Thus attribution of the cycle to inclination provides a natural (no-parameter) solution to the stage 1 and stage 11 problems as well as to the causality problems. Bispectra Bispectral analysis can be used to give an independent test of the causal link between a theoretical driving mechanism and a response. A peak appears in the bispectrum only if three frequencies are present in the data, and the third is not only the sum or difference of the other two, but in phase lock with the sum or difference of their phases. The bispectrum can strongly suppress noise, and it can yield a completely independent test for proposed forcing mechanisms. In Fig. 4, we show the bispectrum of orbital inclination of δ18O (from
SPECTRUM OF L00-KYR GLACIAL CYCLE: ORBITAL INCLINATION, NOT ECCENTRICITY
8332
Specmap) and eccentricity [Elsewhere (26, 27) we give a detailed discussion of the calculations and their interpretations.] The strongest peak in inclination (Fig. 4a) is at (f1 = 0.009, f2 = 0.001), indicating that for the orbital variations, the signals near 0.001, 0.01, and 0.009 are frequency and phase locked. This same peak appears as the most significant signal in the δ18O bispectrum, confirming the hypothesis that the glacial cycles are driven by orbital inclination. In contrast, the bispectrum of eccentricity, Fig. 4c, shows little resemblance to the δ18O bispectrum. This bispectrum also supports that climate cycles are related to orbital inclination.
FIG. 3. Variations of the inclination vector of the Earth's orbit. The inclination i is the angle between this vector and the vector of the reference frame; Ω is the azimuthal angle = the angle of the ascending node (in astronomical jargon). In A–C, the measurements are made with respect to the zodiacal (or ecliptic) frame–i.e., the frame of the current orbit of the Earth. In D–F, the motion has been transformed to the invariable frame—i.e., the frame of the total angular momentum of the solar system. Note that the primary period of oscillation in the zodiacal frame (A) is 70 kyr, but in the invariable plane (D) it is 100 kyr. Linking Mechanisms Since orbital inclination does not affect insolation, we must search for another mechanism relating changes in orbital inclination to changes in global climate. The only plausible one we have found is accretion of interplanetary material: meteoroids and dust. As the orbit of the earth changes, it passes through different parts of the sun's zodiacal ring and encounters different regions of density of material. Changes in inclination will be reflected in changes of accretion. The meteoroids and dust will, through orbital processes, tend to concentrate in the invariable plane. As the earth passes through the invariable plane, accretion increases, and we speculate that glaciers grow, while recession of glaciers takes place during high inclinations when the earth's orbit tips out of the invariable plane. We emphasize that this mechanism is speculative, and that there is no known meteoroid or dust band that satisfies all the properties that we require, although it is possible that such a band could exist. We will offer some indirect evidence that accretion does vary with orbital inclination. Interplanetary dust accreting on the sun has previously been proposed as a driver of the ice ages (28, 29). Clube (30) discussed the possibility of accretion from a single large and unknown meteor stream affecting earth's climate, but he did not draw any conclusions with respect to the periodicity of glacial cycles. Hoyle and Wickramasinghe (31) calculated the effect that accreting dust in the atmosphere could have on the greenhouse effect through the seeding of ice crystals, and speculated that such accretion could have been responsible for the Little Ice Age. At a meeting of the Royal Astronomical Society, reported by G. Manley (32), Hoyle discussed the possibility that accretion could remove enough atmospheric water vapor to reduce the greenhouse effect and cause cooling. Stratospheric dust could also be an effective scavenger of other greenhouse gases, including ozone, and possibly could affect the concentration of components such as chlorine that are thought to be responsible for the destruction of ozone. The climatic effects of high-altitude dust and aerosols are known primarily from volcanic eruptions; global cooling of 0.5–1°C was estimated from the eruption of Krakatoa, and measurable climate changes have been attributed to El Chichon, Pinatubo, and other recent eruptions that injected several megatons of material into the stratosphere. Large explosive volcanic events occur typically once every century, so the average injection of volcanic material is approximately 100 kton/yr (33). Measurements by Kyte and Wasson (34) of iridium in oceanic sediments show that the long-term global
SPECTRUM OF L00-KYR GLACIAL CYCLE: ORBITAL INCLINATION, NOT ECCENTRICITY
8333
average flux from extraterrestrial materials for the period 35–70 Ma is 60–120 kton/yr, about the same as the long-term average from present-day volcanic eruptions.
FIG. 4. Bispectra of (a) inclination of the earth's orbit, (b) δ18O data from Specmap, and (c) the eccentricity of the earth's orbit. The inclination and eccentricity were taken from Quinn et al. (5) transformed to the invariable plane. Note the close match between the most significant peaks in the inclination bispectrum and δ18O bispectrum. The scale is linear, and the units arbitrary; for details of the bispectral method see MacDonald and Muller (26). Accretion could cause cooling (as volcanic eruption suggests) or warming (if cometary particles inject water). Large particles (10 µm) take a few hours to reach the ground: smaller particles (0.5 µm) take a few months. Gases can reside for much longer. Extraterrestrial accretion occurs at the top of the atmosphere, so the climate effects could be significantly different from those resulting from volcanic eruptions. In addition, the global distribution of dust from the two mechanisms is different; for example, stratospheric circulation patterns rarely carry volcanic material to the poles. Data on noctilucent clouds (mesospheric clouds strongly associated with the effects of high meteors and high altitude dust) supports the hypothesis that accretion increases significantly when the Earth passes through the invariable plane. A strong peak in the number of observed noctilucent clouds occurs on about July 9 in the northern hemisphere (35, 36) within about a day of the date when the Earth passes through the invariable plane. In the southern hemisphere the peak is approximately on January 9, also consistent with the invariable plane passage, but the data are sparse. This coincidence has not been previously noted, and it supports the contention that there is a peak in accretion at these times. On about the same date there is a similarly narrow peak in the number of polar mesospheric clouds (37) and there is a broad peak in total meteoric flux (38). It is therefore possible that it is a trail of meteors in the upper atmosphere, rather than dust, that is responsible for the climate effects.
FIG. 5. Spectrum of the accretion of extraterrestrial dust for the period 269–445 ka, determined from the helium-3 measurements of Farley and Patterson (39). Discussion The hypothesis that variations in inclination are responsible for the 100-kyr fluctuations cleanly solves three of the difficulties associated with the hypothesis that variations in eccentricity are responsible. First, the inclination shows a single narrow peak, in agreement with the spectrum of climate proxy records. Second, the variation in inclination does not show any peak at 400 kyr, again in agreement with observations. Third, the inclination hypothesis satisfactorily deals with the causality issue. The linkage of variations in inclination with climate suffers from the requirement that one must assume the dust concentrations to be sufficient to bring about significant changes in climate. Evidence that extraterrestrial accretion has varied with 100-kyr period is evident in the observations by Farley and Patterson (39). The spectrum of the observed accretion, as determined from fluctuations in helium-3, is plotted in Fig. 5. The only statistically significant peak is the predicted one with a period of 100 kyr. This association is suggestive, but we have not yet been able to calculate quantitatively the effect of various mechanisms of accretion on climate. The sudden onset of the 100-kyr peak about 1 million years ago can also be dealt with by the accretion hypothesis. We are required, however, to assume that the dustiness of the solar system underwent a discontinuous change at about a million years. This would require, for example, the breakup of a large comet. Again, Farley (40) has shown that indeed there appears to be discontinuity in the rate of accretion about 1 million years ago. We believe the inclination hypothesis is one that should be further investigated, both in terms of theory and in terms of observations of past rates of extraterrestrial accretion. Discussions with Walter Alvarez and the Renaissance Geology Group were very helpful. This work was supported in part by the Division of Environmental Sciences, U.S. Department of Energy, under contract DE-AC03-76SF00098.
SPECTRUM OF L00-KYR GLACIAL CYCLE: ORBITAL INCLINATION, NOT ECCENTRICITY
8334
1. Broecker, W. S. & van Donk, J. (1970) Rev. Geophys. Space Phys. 8, 169–197. 2. Imbrie, J. & Imbrie, K.P. (1979) Ice Ages, Solving the Mystery (Harvard Univ. Press, Cambridge, MA). 3. Milankovitch, M. (1941) Canon of Insolation and the Ice-Age Problem (Royal Serbian Academy, Belgrade, Yugoslavia). 4. Milankovitch, M. (1920) Théorie Mathématique des Phénomènes Produits par la Radiation Solaire (Gauthier-Villars, Paris). 5. Quinn, T. R., Tremaine, S. & Duncan, M. (1991) Astron. J. 101, 2287–2305. 6. Imbrie, J., Berger, A., Boyle, E., Clemens, S., Duffy, A., Howard, W., Kukla, G., Kutzbach, J., Martinson, D., McIntyre, A., Mir, A., Molfino, B., Morley, J., Peterson, L., Pisias, N., Prell, W., Raymo, M., Shackleton, N. & Toggweiller, J. (1993) Paleoceanography 8, 699–735. 7. Ruddiman, W.F., Raymo, M. & McIntyre, A. (1986) Earth Planet. Sci. Lett. 80, 117–129. 8. Winograd, I. J., Coplen, T. B., Landwehr, J. M., Riggs, A. C., Ludwig, K. R., Szabo, B. J., Kolesar, P. T. & Revesz, K. M. (1992) Science 258, 255– 260. 9. Imbrie, J., Mix, A. C. & Martinson, D. G. (1993) Nature (London) 363, 531–533. 10. Winograd, I. J. & Landwehr, J. M. (1993) U.S. Geological Survey Open-File (U.S. Geological Survey, Reston, VA), Rep. USGS 93–357. 11. Ludwig, K. R., Simmons, K. R., Winograd, I. J., Szabo, B. J. & Riggs, A. C. (1993) Science 259, 1626–1627. 12. Emiliani, C. (1993) Nature (London) 364, 583–584. 13. Landwehr, J. M., Winograd, I. J. & Coplen, T. B. (1994) Nature (London) 368, 594. 14. Muller, R. A. & MacDonald, G. J. (1995) Nature (London) 377, 107–108. 15. Ruddiman, W. F., Raymo, M. E., Martijnson, D. G., Clement, B. M. & Backman, J. (1989) Plaeoceanography 4, 353–412. 16. Imbrie, J., Hays, J., Martinson, D., McIntyre, A., Mix, A., Morley, J., Pisias, N., Prell, W. & Shackleton, N. (1984) in Milankovitch and Climate Part 1, eds. Berger, A., Imbrie, J. & Shackleton, N. (Riedel, Dordrecht, The Netherlands), pp. 269–305. 17. Lomb, N. (1976) Astrophys. Space Sci. 39, 447–462. 18. MacDonald, G. J. (1989) Rev. Geophys. 27, 449–469. 19. Berger, W. H., Yasuda, M. K., Bickert, T., Wefer, G. & Takayama, T. (1994) Geology 22, 463–467. 20. Blackman, R. B. & Tukey, J. W. (1958) The Measurement of Power Spectra (Dover, New York). 21. Imbrie, J. & Imbrie, J. Z. (1980) Science 207, 943–952. 22. Berger, A. & Loutre, M. (1991) Earth Planet. Sci. Lett. 111, 369–382. 23. Neeman, B. U. (1993) Orbital Tuning of Paleoclimatic Records: A Reassessment (Lawrence Berkeley National Laboratory, Berkeley, CA), Rep. LBNL-39572. 24. Muller, R. A. (1994) Glacial Cycles and Extraterrestrial Accretion (Lawrence Berkeley Laboratory, Berkeley, CA), Rep. LBL-35665. 25. Muller, R. A. & MacDonald, G. J. (1994) Glacial Cycles: Orbital Mutation Dominated for the Last 900,000 Years (Lawrence Berkeley Laboratory, Berkeley, CA), Rep. LBL-35667. 26. MacDonald, G. J. & Muller, R. A. (1994) Bispectral Fingerprint Identifies 100 ky Climate Cycle: Orbital Inclination (Lawrence Berkeley Laboratory, Berkeley, CA), Rep. LBL-36214. 27. Muller, R. A. & MacDonald, G. J. (1997) Geology 25, 3–6. 28. Hoyle, F. & Lyttleton, R. A. (1939) Proc. Cambridge Philos. Soc. 35, 405–415. 29. McCrea, W. H. (1975) Nature (London) 255, 607–609. 30. Clube, S. (1987) Philos. Trans. R. Soc. London 323, 421–436. 31. Hoyle, F. & Wickramasinghe, N. C. (1991) Nature (London) 350, 467. 32. Manley, G. (1953) Nature (London) 173, 1206–1208. 33. Kondratyev, K. Y. (1988) Climate Shocks: Natural and Anthropogenic (Wiley, New York). 34. Kyte, F. T. & Wasson, J. T. (1986) Science 232, 1225–1229. 35. Thomas, G. E. (1991) Rev. Geophys. 29, 553–575. 36. Foyle, B. & Haurwitz, B. (1966) Space Sci. Rev. 6, 279–340. 37. Thomas, G. E. & Olivero, J. J. (1989) J. Geophys. Res. 94, 14673–14681. 38. Whipple, F. L. & Hawkins, G. S. (1956) J. Meteorol. 13, 236–240. 39. Farley, K. & Patterson, D. B. (1995) Nature (London) 378, 600–603. 40. Farley, K. (1995) Nature (London) 378, 153–156.
CAN INCREASING CARBON DIOXIDE CAUSE CLIMATE CHANGE?
8335
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8335–8342, August 1997 Colloquium Paper This paper was presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held November 13–15, 1995, at the National Academy of Sciences, Irvine, CA.
Can increasing carbon dioxide cause climate change?
(global warming/climate sensitivity/climate modeling/greenhouse effect) RICHARD S. LINDZEN Building 54, Room 1720, Massachusetts Institute of Technology, Cambridge, MA 02139
© 1997 by The National Academy of Sciences 0027-8424/97/948335-8$2.00/0 PNAS is available online at http://www.pnas.org. ABSTRACT The realistic physical functioning of the greenhouse effect is reviewed, and the role of dynamic transport and water vapor is identified. Model errors and uncertainties are quantitatively compared with the forcing due to doubling CO2, and they are shown to be too large for reliable model evaluations of climate sensitivities. The possibility of directly measuring climate sensitivity is reviewed. A direct approach using satellite data to relate changes in globally averaged radiative flux changes at the top of the atmosphere to naturally occurring changes in global mean temperature is described. Indirect approaches to evaluating climate sensitivity involving the response to volcanic eruptions and Eocene climate change are also described. Finally, it is explained how, in principle, a climate that is insensitive to gross radiative forcing as produced by doubling CO2 might still be able to undergo major changes of the sort associated with ice ages and equable climates. The title suggested for this paper (by Dave Keeling) is tantalizing for its ambiguity. At some level, the answer is philosophically trivial. After all, our knowledge is rarely so perfect that we can say anything is absolutely impossible. In connection with this question we can go a bit further, and state that increasing CO2 is likely to cause some climate change, and that the resulting change will involve average warming of the earth. However, this answer is almost as trivial as the first. The climate is always undergoing change, and if the changes due to increasing CO2 are smaller than the natural variability, then these changes will be of only modest concern except as an exercise in weak signal detection. The more serious question then is do we expect increasing CO2 to produce sufficiently large changes in climate so as to be clearly discernible and of consequence for the affairs of humans and the ecosystem of which we are part. This is the question I propose to approach in this paper. I will first consider the question of whether current model predictions are likely to be credible. We will see why this is unlikely at best. I will then show how we might estimate and bound climate sensitivity both directly and indirectly from existing data. Finally, I will consider the relationship of changes in mean temperature to changes in the structure of climate. It has been suggested that small changes in mean temperature are important because major changes in past climate were associated with major changes in the equator-to-pole temperature difference, but only small changes in the mean temperature. I will argue that the changes in mean temperature may be only residuals of the changes in the meridional temperature distribution rather than the cause. Current Forecasts Present projections of the climatic effects of increasing CO2 are based on models of varying degrees of complexity. The relative similarity of all these predictions for the increase in global mean temperature has lent a degree of plausibility to the resulting predictions. We shall, in this section, analyze the nature of these “traditional” results to understand what the physical basis is for the common prediction. In the following section we will examine some of the processes crucial to these predictions to see whether they are known to sufficient accuracy for the purpose of climate predictions. Before doing this, it will be necessary to briefly review the physics of the “greenhouse effect.” Although this process is usually summarized by the assertion that infrared-absorbing gases inhibit the ability of the earth's surface to emit thermal radiation, and thus force the surface to get warmer, the reality is substantially more complex. Möller and Manabe (1) made an early start toward understanding this matter. In this one-dimensional study, both radiative and radiative–convective equilibria were calculated using assumed distributions for humidity and cloudiness. The simplistic picture corresponds essentially to radiative equilibrium, for which Möller and Manabe calculated the equilibrium temperature of the surface to be about 350 K, which is 95 K warmer than the black-body temperature of 255 K. When convection is included by introducing a simple convective adjustment, the surface temperature comes down to the observed value of 288 K. Convective adjustment reduced the greenhouse effect by about 75%, by allowing for the fact that radiation is not the only form of heat transfer in the atmosphere. When infrared opacity is high, evaporation and mechanical transport are more efficient ways for the surface to cool. Lindzen (2) offered a more complete schematic of the realistic operation of the natural greenhouse effect. One begins by recognizing that water vapor, the atmosphere's main greenhouse gas, decreases in density rapidly with both height and latitude. Surface radiative cooling in the tropics, which has the highest concentration of water vapor, is negligible. Heat from the tropical surface is carried upward by cumulus convection and poleward by the Hadley circulation and planetary-scale eddies to points where radiation can more efficiently transport the heat to space. Where radiation can more efficiently carry the heat depends on the radiative opacity and the motions themselves. In point of fact, without knowing the dynamical heat fluxes, it is clear that one cannot even calculate the mean temperature of the earth. It is interesting, in this regard, to look at model intercomparisons of meridional heat flux, and their comparison with observationally based estimates. An extensive study (3) shows that such differences reach 2 PW (petawatts). As shall be noted later, this is roughly equivalent to differences in vertical fluxes of about 25 W·m−2—much larger than the 4 W·m−2 change that a doubling of CO2 is expected to produce. A particularly acute example of the problem with dynamic fluxes is revealed when one couples
Abbreviations: OLR, outgoing long-wave radiation; GCM, general circulation model; GFDL, Geophysical Fluid Dynamics Laboratory; GISS, Goddard Institute for Space Studies; TOA, top of atmosphere.
CAN INCREASING CARBON DIOXIDE CAUSE CLIMATE CHANGE?
8336
models for the atmosphere with ocean models. Here, theclimate tends to drift unless one applies so-called flux corrections. Examples of such corrections are given for all the leading models (4). The corrections have to be applied on a latitude-by-latitude basis, and the magnitude of the correction can be as large as 100 W·m−2. As can be understood from our discussion of the realistic nature of the greenhouse effect, these dynamic fluxes do not represent systematic biases independent of the CO2 forcing; rather they are essential to calculating the response to increased CO2. The issue is not that the forcing due to CO2 is buried within these larger uncertainties, but rather whether we can reckon the response reliable. The role of water vapor is nonlinear. Assuming 80% relative humidity in a 2-km boundary layer, and a fixed relative humidity above the boundary layer, Fig. 1 shows how, for a given temperature distribution, outgoing long-wave radiative flux varies as one perturbs the relative humidity above the boundary layer [Fig. 1 was calculated using the radiative transfer code (5); a similar calculation appears in ref.6 ]. One sees that the effect of a 5% change in relative humidity depends on the base humidity being perturbed. For low base humidities, a 5% change is associated with about 5 W·m−2. For high base humidities, the change is about half of this. For purposes of comparison, the 4 W·m−2, which a doubling of CO2 is expected to produce, is roughly equivalent to a 4–8% change in relative humidity. Note that uncertainties in measurements of humidity are on the order of 20% or more, though things appear to have improved over the past 2 years. We shall look at the improved data soon. However, it is again clear that we are dealing with uncertainties and errors that are large compared with the climatic impact of CO2. Here too, these errors occur in a field that is crucial to calculating the response to CO2, since the water vapor feedback is essentially responsible for the model predictions of large warming due to increasing CO2. Clearly, even superficial agreement between observations and model-derived water vapor would be inadequate to establish the model feedback. This potentially important positive feedback was first identified by Manabe and Wetherald (7). Using a simple one-dimensional radiative–convective model, they found that assuming constant relative humidity led to a significantly enhanced response to increased CO2 over what would have been obtained with fixed specific humidity. The point, simply, is that with fixed relative humidity, specific humidity must increase with warming. Upper-level water vapor (above 2–3 km in the tropics) dominates the radiative role of water vapor, despite the fact that most of the atmosphere's water vapor is found below 800 millibars [1 millibar (mb) = 100 Pa] (8). Of course, given the nonlinearity of the radiative effect of water vapor, the average radiative response to water vapor is not equal to the response to an average water vapor, and, therefore, one-dimensional studies are inappropriate. However, the results of the above one-dimensional studies remain indicative of general properties.
FIG. 1. The changes in outgoing long-wave radiation (OLR) for increase and decrease of 5% in relative humidity above 2 km as a function of the unperturbed relative humidity. The most useful way of viewing feedbacks is by means of the formula
[1] where fi is the ith feedback factor. For fixed relative humidity, the water vapor feedback factor is about 0.4. This turns out to be much larger than the factors due to clouds and snow in present models. However, as may be seen from the formula, the addition of smaller factors on top of the 0.4 due to water vapor rapidly increase the response. Without the water vapor feedback the impact of model cloud and snow feedbacks would be small (2). It is worth reviewing the basis for the assumption of constant relative humidity in (7). It is based on the crudely observed picture from ref. 9 reproduced in Fig. 2. It was argued in (7) that the overall relative humidity varied only between about 30% and 50%, and that the pattern was similar for both winter and summer, suggesting that the atmosphere was attempting to maintain a given relative humidity regardless of temperature. There were, of course, very few measurements available for ref. 9. However, subsequent analyses of radiosonde data showed a fairly similar picture (10). Unfortunately, the radiosonde data have proven extremely unreliable (11). In particular, radiosonde data tended to replace readings of very low humidity with relative humidities of 20%. Nevertheless, these primitive observations received a certain amount of credibility insofar as they were consistent with humidities predicted in general circulation models (GCMs). However, recently, the 183-GHz channel on the SSM/T-2 satellite has provided detailed data on the global distribution of relative humidity. Fig. 3 shows a global daily map for relative humidity between 500 and 300 mb for May 5, 1995. We see hugely more variability
FIG. 2. Latitude (in deg)–height (in km) distribution of relative humidity for both summer and winter (taken from ref. 9).
CAN INCREASING CARBON DIOXIDE CAUSE CLIMATE CHANGE?
8337
than was suggested in ref. 9. We also see confined moist regions associated with active convection and comparatively rapid transitions to extremely dry regions away from the convection. Incidentally, this observation resolves a problem in ref. 12, where, in studying the moisture budget of the tropical troposphere, it was found that it was impossible to account for the humidities observed by radiosondes in clear subsiding regions. The satellite data show that these regions are, indeed, dry. Some indication of why current models misrepresent the vertical distribution of humidity is given in ref. 13, where the authors calculated the correlations of interannual variability of humidity at various levels with the variability at the surface for both traditional radiosonde data and for the output of a Geophysical Fluid Dynamics Laboratory (GFDL) GCM. The results are shown in Fig. 4. In the data, upper-level humidity rapidly decorrelates from surface humidity, while in the model all levels are highly correlated. This strongly points to the likelihood of strong artificial coupling of levels in this model. Judging from the results in ref. 14, similar problems are likely in the Goddard Institute for Space Studies (GISS) model, and they appear to be independent of the choice of convective parameterization. To be sure, there are the already-mentioned problems with radiosonde data which are discussed in detail in ref. 15. Current radiosonde data are much better (though they still display a moist bias in dry regions) and show the decoupling of water vapor behavior above and below the trade inversion much better. Fig. 5 illustrates this. Models also show a tendency to underestimate humidity in moist regions and to overestimate it in dry regions (16).
FIG.3. Relative humidities in the layer 500–300 mb derived from 183-GHz soundings from SSM/T-2 for May 5, 1995. The color scale for relative humidity is shown below the panel. Figure provided by R. Spencer, National Aeronautics and Space Administration/Huntsville (personal communication). There are potential problems with the vertical distribution of temperature as well. For simple radiative convective models, the vertical profile of temperature in the troposphere is essentially fixed. Thus, the response to tropopause level forcing from doubled CO2 must consist in warming throughout the troposphere, including the surface. In principle, warming at the top of the troposphere (without warming at the surface) would be sufficient to balance the forcing. Data indicate a significant degree of independence for temperature changes at different levels (17). Of course, GCMs do not explicitly assume rigid vertical coupling of temperature in the troposphere; however, it is possible that coupling is stronger than in nature.
FIG. 4. Correlation of interannual variations in specific humidity with interannual variations in specific humidity at the surface as observed in radiosonde data, and as calculated in a GFDL climate model forced by observed variations in sea surface temperature. [Reproduced with permission from ref. 13(Copyright 1996, American Meteorological Society).] The above-described problems with heat fluxes and humidity, as well as the potential problems with vertical structure of temperature, all serve to render model feedbacks extremely uncertain. In view of these problems, it is important to
CAN INCREASING CARBON DIOXIDE CAUSE CLIMATE CHANGE?
8338
consider whether there are alternative approaches to determining climate sensitivity. Observational Determinations of Climate Sensitivity The purpose of the present section is to assess various approaches to using data to infer climate sensitivity, given that current GCMs are unlikely to be adequate for this task. Direct Approach. As already noted, a doubling of CO2 is generally taken to imply a forcing at the tropopause of about 4 W·m−2. The question of climate sensitivity amounts to asking how much must the earth's surface warm to compensate for this forcing. A simplistic approach to the question of climate sensitivity would be to study the temporal variation of globally integrated OLR with varying globally averaged temperature. The ratio of the temperature variations to the variations in OLR would represent the climate sensitivity. However, a priori, naturally occurring changes in global mean temperature on time scales of from weeks to years may not form proper surrogates for warming due to increased CO2 (18). Another problem with this approach is that OLR is not the sole contributor to the radiative response. In principle, we should look at the change in total radiative flux at tropopause levels. For the tropics, however, OLR in clear sky regions appears to be the dominant contributor to the total flux change (19). Still another part of the problem is that naturally occurring changes in mean temperature on these time scales are significantly associated with changing regional patterns of warming rather than global warming (20). Insofar as the water vapor feedback is involved in climate sensitivity, Fig. 1 shows that moisture changes in dry regions are much more important than changes in moist regions. A global change involving an intensification or reduction of existing differences between moist and dry regions can lead to a change in OLR even in the absence of change in mean temperature. It will clearly be necessary to distinguish such changes from those specifically associated with changes in the mean temperature. It should be noted that changing patterns can be associated with changes in circulation and changes in temperature, both of which play a role in the moisture budget (12). Fig. 3, in fact, suggests the interesting possibility that the primary feedback process might consist in the change in areal coverage of the very dry regions. Presumably, natural variations include a full range of such possibilities so that observed ratios of average temperature variations to variations in total OLR would show a significant scatter. A primary problem associated with the direct measurement of climate sensitivity will be to distinguish changes in flux associated with changes in mean temperature from those associated with pattern changes not associated with changes in the mean temperature. There is, moreover, no assurance that all changes in mean temperature will be appropriate surrogates for global warming.
FIG. 5. Histograms of relative humidity at different pressure levels obtained from recent radiosondes at near equatorial stations. Figure provided by R. Spencer (personal communication). The question of the sensitivity of tropical temperature is an important matter in its own right. In particular, the tropics have distinctly different basic physical balances from those in the extratropics (21). Tropical sensitivity is also an important factor in global sensitivity. GCM results characteristically indicate no special difference between tropical and extratropical sensitivity to a doubling of CO2. This is seen in Fig. 6. Although there is enhanced response at polar latitudes, the response is relatively flat from the equator to about 40°. While there are reasons to suppose that the model response in the tropics is excessive, this has been widely argued about and is not essential to the present discussion; it will, however, be important for the discussion in the next section. What is relevant here is that the changing tropical sensitivity has a profound effect on the response to increased CO2 globally. In calculations performed with the Center for Oceans, Land, Atmosphere GCM, it was found that constraining surface temperature in small regions of the tropics was sufficient to substantially reduce the globally averaged response (23). Moreover, recent work (24) leads one to conclude that models underestimate the degree of mixing from the tropics to the extratropics, and hence may underestimate the effect of the tropics on the extratropics. In general, therefore, reduced tropical sensitivities will imply reduced global sensitivity, though the extent of the reduction may well be greater than indicated in ref. 23, since models appear to understate the degree of coupling between the tropics and extratropics. In dealing with the climate sensitivity of the tropics, we are dealing with the sensitivity of a system open not only to changes in top of atmosphere (TOA) radiative forcing but also to changes in meridional flux. This is illustrated in Fig. 7. Sensitivity would essentially be the ratio of average tropical temperature change to change in total flux. However, for the meridional flux, we can be reasonably confident that for increasing temperature, ∆Fmeridional ≥ 0, provided that we are considering zonal averages over the whole tropics. In this case, if we include only TOA flux changes, we will get an upper bound for sensitivity, as is readily seen from Eq. 2:
[2]
FIG. 6. Meridional distributions of zonally averaged change in surface air temperature due to a doubling of CO2 for December, January, February (A, Community Climate Model; B, GFDL; C, GISS; D, Oregon State University). From ref. 22.
CAN INCREASING CARBON DIOXIDE CAUSE CLIMATE CHANGE?
8339
FIG. 7. Schematic illustration of the fluxes relevant to the heat budget of the tropics. If the average is over only a sector of the tropics, we have no such assurance. This is one of the main problems in ref. 19. The results there are nonetheless both suggestive and instructive. Chou used Earth Radiation Budget Experiment data to assess the changes in net flux over the region (−30° ≤ latitude ≤ +30°, −100° ≤ longitude ≤ +100°) between April 1987 (a warm El Niño year) and April 1985 (a colder La Niña year). The average surface temperature for the region he examined was about 0.3°C warmer in 1987 than in 1985. The change in net flux was about 4 W·m−2. This flux change was almost entirely due to OLR over clear sky regions (as opposed to clear sky OLR, which commonly refers to the calculated OLR that would occur in the absence of clouds). The changes over cloudy regions were dominated by clouds whose infrared and visible effects tended to cancel to a large extent locally, but to an important extent there is also cancellation between different regions. As we see from Fig. 3, clear sky regions tend to be very dry. The radiative response must be due to drying over these regions, and this is consistent with Chou's results, which show that the flux changes are almost exclusively restricted to the subsiding region of the Pacific. The explicit drying of already dry regions appears to be more important than net drying. Indeed, there can be moistening of already moist regions and even net moistening while still having a negative water vapor feedback. Taken at face value, this suggests a very low sensitivity for the tropics compared with most GCMs, where a change of about 2°C is associated with 4 W·m−2. Using Eq. 1, this would imply a water vapor feedback factor of about −2 rather than + 0.4, which is typical of current models. However, given that Chou considered only a sector, we do not know if this is an over- or underestimate of the actual tropical sensitivity, since feedbacks require a consideration of complete systems including both convective and subsiding regions, and limited sectors include unknown proportions of each. Indeed, consideration of other months and years can even lead to apparent feedbacks of opposite sign for such limited regions. We also are unable to distinguish pattern changes not directly related to mean temperature from changes that are. It is nonetheless useful to examine any differences between GCM-generated TOA flux changes and those found by Chou for runs using the same sea surface temperatures used by Chou. A preliminary attempt has been made in ref. 25, focusing only on OLR and using the Atmospheric Model Intercomparison Program's data for various GCMs. This is by no means a test of sensitivities. However, it does provide some information on how a very important component of sensitivity is replicated in models. The results demonstrate that most models overestimated the observed “sensitivity” appreciably (though one, in fact, underestimated it). However, this work did not check in detail as to how much of the difference was due to errors in pattern or to water vapor feedbacks directly, nor did it focus on the OLR in clear regions which dominated Chou's results. In particular, it would appear from Fig. 4 that a very important consideration ought to be how dry and how large the areal coverage is of the very dry subsiding regions. One important methodological point which emerged from Covey's study (25) was that model results for regional “sensitivity” varied pronouncedly from the models' global sensitivity, indicating rather clearly that what Chou was observing was not global sensitivity. Such a situation is not remedied by considering the statistics of many month pairs as opposed to the use of a single pair by Chou. Despite the problems in ref. 19, it does point the way toward a proper observational determination of the sensitivity to global forcing. It would consist, at best, in the measurement of the complete TOA flux integrated over the whole earth (averaged over, say, a month) for several years, and the measurement of surface temperature over the same period. One would form the pattern correlation of temperature for each pair of months in the record, as well as the average rms difference of the temperatures and the difference of the globally averaged temperatures. By sorting according to these three quantities, one might, hopefully, be able to disentangle the dependence of flux on both patterns and mean temperature. Presumably the latter would be indicative of the climate sensitivity we are seeking. Comparing pairs of months separately would avoid the problems in averaging associated with the nonlinearity of the effect of water vapor. Performing such a study for tropical latitudes separately would allow some insight into the physical origins of the sensitivity. Of, course, there would remain the problem of whether states differing only in mean temperature formed proper surrogates for global climate change. There is also the problem that total insolation varies with an annual cycle due to the varying distance of the earth from the sun. This may require that comparisons be restricted to the interannual variability of each month. However, none of the quantities needed for such a study require any truly new instruments. Indeed, it appears that the needed data may be marginally available from existing satellites and surface data. The necessary length of record will likely be determined by the need to obtain sufficiently large numbers of month pairs to populate all relevant possibilities. This number of appropriate month pairs is greatly reduced if one cannot find a suitable correction for the varying solar distance. A caveat that requires some consideration is the obvious fact that Chou's results (19) suggest how it is possible for OLR to change in response to changes in circulation without accompanying changes in mean temperature. If our question is how much must the earth's temperature change to compensate for 4 W·m−2 forcing, then Chou's results show that it is at least physically possible for such compensation to occur without net warming. This should alert us to the possibility that simple definitions of climate sensitivity are by no means guaranteed to be relevant. It should be mentioned that there have been attempts other than Chou's to directly measure climate sensitivity. Unfortunately, these generally assumed that local or seasonal changes in temperature could be considered as surrogates for climate change. However, as noted in ref. 12, the warmest regions are associated with convection and high upper level humidities, while dry subsiding regions are associated with cooler temperatures regardless of feedbacks. Thus, a study like that of Raval and Ramanathan (26) inevitably shows a positive cor
CAN INCREASING CARBON DIOXIDE CAUSE CLIMATE CHANGE?
8340
relation of surface temperature and “enhanced” greenhouse effect. Indeed, should there be a strong negative feedback associated with enhanced drying in subsiding regions (and/or expanded dry regions), such an approach would indicate a spuriously increased water vapor feedback. (This, itself, might lead to a useful test.) A similar problem pertains to the study by Rind et al. (27). They compared the summer tropics with the winter tropics. However, the summer tropics are associated with ascending moist air, while the the winter tropics are associated with dry subsiding air—again independent of the actual feedback. Indirect Approach: Volcanic Sequences. It has long been noted that volcanic veils provide a short-term perturber of global temperature. Whether, the climatic response to such perturbations provides a test of climate sensitivity is less clear. The problem was addressed crudely in ref. 28. In that paper, a simple energy balance climate model with a box-diffusion ocean was used. The ocean was taken to have an insulated boundary at 300 m to simulate the effect of upwelling and avoid the problems associated with unbounded oceans. Climate sensitivity was specified. Volcanic veils were assumed to set up within 3 months of eruption and decay with an exponential decay time of 13 months. Diffusion, in such models, is a surrogate for all the processes in real oceans that couple the mixed layer with the thermocline. The coefficient is chosen to match chemical tracer data. This is, of course, extremely crude, but might be adequate for global response to global forcing. Using such a model, it was noted that the response to a volcano during the first 2 years following eruption was, given the uncertainties in both temperature measurements and aerosol optical properties, unable to distinguish between sensitivities ranging (in terms of the equilibrium response to double CO2) from 0.15°C to 6°C. In this connection, it should be noted that a study of the response of the GISS GCM to Pinatubo did mention that it was only a test of the short-term physics in their model (29). Recently, C. Giannitsis and I have recalculated the response to volcanos with a model that, at least, distinguishes land and sea, tuning the coupling between the two by using the seasonal cycle (R.S.L. and C. Giannitsis, unpublished work). The results, for the purposes of this discussion, are similar to those reported in ref. 28 in that a reasonable correspondence between calculated response and observed response is obtained for a wide range of sensitivities, at least for the first 2 years following eruption. For longer periods, there is an interesting dependence on sensitivity. For low sensitivities, the response rapidly decays to essentially zero. However, for higher sensitivities, there is a rapid decay of the response to about 30% of the maximum response, with the remainder decaying on the ocean response scale, which is very long. The reason for this difference is that climate sensitivity is also a measure of how tightly air and sea temperatures are coupled. High sensitivity is associated with weak coupling, allowing the establishment of significant disequilibration of the sea surface temperature. This was noted in detail in ref. 31. As a practical matter, 30% of the peak response is too small relative to natural variability to be detected. However, it was suggested in ref. 28 that a sequence of strong volcanos occurring over several decades would produce a measurably different response for different sensitivities. Such a sequence did occur between Krakatoa in 1883 and Katmai in 1912, with a noticeable absence of large eruptions until the 1950s. Of course, there is a great deal of uncertainty over the exact strength of the forcing due to these volcanos. Our results were based on what we believe to be the conservative assumption that Krakatoa was no stronger than Pinatubo. The results show that for sensitive climates (0.6°C for a doubling of CO2), each volcano builds on the residual base of earlier volcanos leading to a substantial long-term cooling (0.5°C between 1883 and 1912). For low sensitivity, the response consists in a sequence of essentially independent “blips.” The observed temperature record certainly shows nothing more than isolated “blips.” Given the uncertainties in the volcanic forcing, it would be inappropriate to place undue confidence in this result. However, it is consistent with low sensitivity. The results stem from the long response time associated with large sensitivity, and argue for short response times. It is also possible to reduce response times by assuming lower ocean heat diffusivity. However, this gives rise to larger discrepancies between predictions and observations of temperature change over the past century. The commonly claimed “broad consistency” depends on long ocean delays. Indirect Approach: Eocene. Fig. 6 suggests another possibility for the indirect estimate of tropical sensitivity. Pastclimates involved marked changes in the equator-to-pole temperature difference. In the case of ice ages, this difference may have been due in part to the increased meridional gradient in radiative forcing due to the increased high-latitude albedo associated with the ice itself. This renders difficult the specification of the forcing that was acting on the tropics. However, for warmer climates, like that of the Eocene, the change in albedo from the present may not have been large, and the reduced equator-to-pole temperature difference almost certainly called for an increased heat flux out of the tropics. At present, this flux is about 5 PW, of which the ocean contributes about 1 PW (10). It may be estimated that a reduction of the equator-to-pole temperature difference from about 40°C to 20°C will require that the present flux be increased to about 6 PW. Although it is currently popular to seek such changes as arising from shifts in ocean circulation, they can also arise quite readily from changes in atmospheric heat flux. The strength of forcing of atmospheric eddies depends not only on the meridional gradient of radiative forcing but also on the intensity of the tropical Hadley circulation, which supplies the momentum to the unstable subtropical jet. The latter is strongly influenced by both orbital parameters and the distribution of land and sea (2, 32), both of which were almost certainly different during the Eocene. Such forcings are potentially much larger than one expects from the net external radiative forcing (especially from orbital variations), and do not, in fact, call for net average external forcing. In any event, a positive ∆Fmeridional of about 1 PW is equivalent to a ∆FTOA of about 12 W·m−2 for the tropics. This ought to have cooled the tropics, and, indeed, early estimates of Eocene equatorial temperatures indicated that the tropics may have been as much as 5°C cooler than they are today. This is only modestly less than current model sensitivity. However, recent corrections to these early estimates have reduced the equatorial cooling to less than 1°C (33), which is more in line with the sensitivity estimates based on the sequence of volcanos around the turn of the past century. The response in the extratropics is consistent with meridional temperature structure being significantly determined by dynamic processes rather than detailed radiative responses at each latitude (34, 35). Again, there are legitimate questions about this procedure, not the least of which concern the reliability and representativeness of the paleoclimatic data. The role of potentially higher levels of CO2 during the Eocene could have contributed to reduced equatorial cooling, though current assessments (36) suggest that CO2 levels during the Eocene were only double present values, and such changes would cancel only 4 W·m−2. The Nature of Past Climate Change The primary variable in most global warming discussions is global mean temperature. The suggested values for change on the order of 2°C do not, on the face of it, seem catastrophic. However, it is commonly noted that major changes in past climate were, in fact, associated with relatively small changes in global mean temperature, but that the global changes were well correlated with changes in the mean. The basis for these
CAN INCREASING CARBON DIOXIDE CAUSE CLIMATE CHANGE?
8341
claims is essentially Fig. 8, which is based on ref. 37 but is generally attributed to ref. 38. What is shown in this figure is the meridional distribution of surface temperature for various past climates scaled by the change in global mean temperature. The fact that temperatures so scaled seem to lie on a single curve has led to the conclusion that mean temperature determines the meridional distribution uniquely. However, the “universality” of the relation is almost certainly an artifact of the fact that equatorial temperature changes are, according to this curve, very small. Thus, climate changes involving primarily changes in the equator-to-pole temperature difference will inevitably scale approximately with the mean temperature, since changes in the mean temperature are simply a residual of the changes in the equatorto-pole temperature when equatorial temperatures are approximately fixed. If, as suggested by GCM results in Fig. 6, equatorial temperature changes are not much smaller than extratropical changes, then the “universality” of the curve in Fig. 8 would disappear. In either case, we no longer have any direct relation between global mean temperature and overall climate. This is consistent with the fact that we have no convincing mechanism whereby changes in mean temperature automatically determine the changes in meridional heat flux. At the same time, we do have mechanisms for changing the meridional heat flux even in the absence of changes in the mean external radiative forcing (32). In view of the above, there appears to be little reason to assume the modest changes in mean temperature that are claimed for increased CO2 will automatically be associated with major global climate change. Similarly, there is no reason to suppose that a climate insensitive to changing CO2 cannot, nonetheless, undergo profound climate change.
FIG. 8. Universal latitude variation of climate change. Based on ref. 37. Conclusion The brief conclusion of this paper is that current GCMs are inadequate for the purpose of convincingly determining whether the small changes in TOA flux associated with an increase in CO2 are capable of producing significant climate change. However, we may not be dependent on uncertain models to ascertain climate sensitivity. Observations can potentially directly and indirectly be used to evaluate climate sensitivity to forcing of the sort produced by increasing CO2 even without improved GCMs. The observations needed for direct assessment are, indeed, observations that we are currently capable of making, and it is possible that the necessary observations may already be in hand, though the accuracy requirements may be greater than current data provide. Still, the importance of the question suggests that such avenues be adequately explored. Since the feedbacks involved in climate sensitivity are atmospheric, they are associated with short time scales. Oceanic delays are irrelevant, since observed surface temperatures are forcing the flux changes we are concerned with. The needed length of record must be determined empirically. Indirect estimates, based on response to volcanos, suggest sensitivity may be as small as 0.3–0.5°C for a doubling of CO2, which is well within the range of natural variability. This is not to suggest that such change cannot be detected; rather, it is a statement that the anticipated change is well within the range of what the earth regularly deals with. It is further noted that the common assertion that even small changes in mean temperature can lead to major changes in climate distribution is ill-founded and, likely, wrong. Work reported here was done cooperatively with E. Schneider, C. Giannitsis, and D. Kirk-Davidoff. This work was supported by Grant 914441-ATM from the National Science Foundation and Grant NAGW 525 from the National Aeronautics and Space Administration. Ten percent of this research was funded by the U.S. Department of Energy's National Institute of Global Environmental Change (NIGEC) through the NIGEC Northeast Regional Center at Harvard University (Department of Energy Cooperative Agreement DEFC03–90ER61010) and through the Computer Hardware, Advanced Mathematics and Model Physics program. Financial support does not constitute an endorsement by the Department of Energy of the views expressed in this article. 1. Möller, F. & Manabe, S. (1961) Z. Meteorol. 15, 3–8. 2. Lindzen, R. S. (1993) Annu. Rev. Fl. Mech. 26, 353–378. 3. Gleckler, P. J., Randall, D. A., Boer, R., Colman, G., Dix, M., Galin, V., Helfand, M., Kiehl, J., Kichl, A., Kitch, A., Lau, W., Liang, X.-Z., Lykossov, V., McAvaney, B., Miyakoda, K. & Planton, S. (1994) Cloud-Radiative Effects on Implied Oceanic Energy Transports as Simulated by Atmospheric General Circulation Models, Report No. 15 of the Program for Climate Model Diagnostics and Intercomparisons (Lawrence Livermore Radiation Laboratory, Livermore, CA). 4. Gates, W. L., Cubasch, U., Meehl, G., Mitchell, J. & Stouffer, R. (1993) An Intercomparison of Selected Features of the Control Climates Simulated by Coupled Ocean-Atmosphere General Circulation Models (World Meteorological Organization, Geneva), Publ. WMO/TD-No. 574. 5. Chou, M. D., Krats, D. P. & Ridgway, W. (1991) J. Climate 4, 424–437. 6. Thompson, S. L. & Warren, S. G. (1982) J. Atmos. Sci. 39, 2667–2680. 7. Manabe, S. & Wetherald, R. T. (1967) J. Atmos. Sci. 24, 241–259. 8. Shine, K. P. & Sinha, A. (1991) Nature (London) 354, 382–384. 9. Telegadas, K. & London, J. (1956) A Physical Model of Northern Hemisphere Troposphere for Winter and Summer (Research Div. College of Engineering, New York Univ., New York), Scientific Report No. 1, Contract AF19(122)-165. 10. Peixoto, J. P. & Oort, A. H. (1992) Physics of Climate (Am. Inst. Phys., New York). 11. Elliot, W. P. & Gaffen, D. J. (1991) Bull. Am. Meteorol. Soc. 72, 1507–1520. 12. Sun, D.-Z. & Lindzen, R. S. (1993) J. Atmos. Sci. 50, 1643–1660. 13. Sun, D.-Z. & Held, I. M. (1996) J. Clim. 9, 665–675. 14. DelGenio, A. D., Lacis, A. A. & Ruedy, R. A. (1991) Nature (London) 351, 382–385. 15. Wade, C. G. (1994) J. Atmos. Ocean. Tech. 11, 687–700. 16. Schmetz, J. & van de Berg, L. (1994) Geophys. Res. Lett. 21, 573–576. 17. Lee, W.-J. & Mak, M. (1994) J. Atmos. Sci. 51, 2137–2144. 18. Lindzen, R. S., Kirtman, B., Kirk-Davidoff, D. & Schneider, E. (1994) J. Climate 8, 1681–1684. 19. Chou, M.-D. (1994) J. Climate 7, 1684–1692. 20. Wallace, J. M., Zhang, Y. & Renwick, J. A. (1995) Science 270, 780–783. 21. Lindzen, R. S. (1991) Prospects for Tropical Modeling, Proceedings of the European Centre for Medium-range Weather Forecast Conference on Tropical Meteorology, Reading (internal report, available from ECMWF, Shinfield Park, Reading RG2 9AX, U.K.). 22. MacCracken, M. C. & Luther, F. M., eds. (1991) Projecting the Climatic Effects of Increasing Atmospheric Carbon Dioxide (U.S. Dept. of Energy, Washington, DC), pp. 280–319 (available as NTIS, DOE ER-0237 from Natl. Tech. Inf. Service, Springfield, VA).
CAN INCREASING CARBON DIOXIDE CAUSE CLIMATE CHANGE?
8342
23. Schneider, E. K., Lindzen, R. S. & Kirtman, B. P. (1997) J. Atmos. Sci., in press. 24. Stone, P. H. & Nemet, B. (1996) J. Atmos. Sci. 53, 1663–1674. 25. Covey, C. (1995) Correlation Between Outgoing Longwave Radiation and Surface Temperature as a Measure of Climate Sensitivity: A Model Intercomparison, Report No. 30 of the Program for Climate Diagnostics and Model Intercomparisons (Lawrence Livermore Radiation Laboratory, Livermore, CA). 26. Raval, A. & Ramanathan, V. (1989) Nature (London) 342, 758–761. 27. Rind, D., Chiou, E. W., Chu, W., Larsen, J., Oltmans, S., Lerner,J., McCormack, M. P. & McMaster, L. (1991) Nature (London) 349, 500–503. 28. Lindzen, R. S., (1995) in Natural Climate Variability on Decade-to-Century Time Scales, ed. Martinson, D. G. (National Acad. Press, Washington, DC), 182–186. 29. Hansen, J., Lacis, A., Ruedy, R. & Sato, M. (1992) Geophys. Res. Lett. 19, 215–218. 30. Budyko, M. I. & Izrael, Y.A. (1991) in Anthropogenic Climate Change, eds. Budyko, M. I. & Izrael, Y. A. (Univ. Arizona Press, Tucson), pp. 277– 318. 31. Hansen, J., Russell, G., Lacis, A., Fung, I. & Rind, D. (1985) Science 28, 857–859. 32. Lindzen, R. S. & Pan, W. (1994) Clim. Dyn. 10, 49–57. 33. Zachos, J. C., Stott, L. D. & Lohmann, K. C. (1994) Paleoceanography 9, 353–387. 34. Stone, P. H. (1978) J. Atmos. Sci. 35, 561–571. 35. Sun, D.-Z. & Lindzen, R. S. (1994) J. Atmos. Sci. 51, 757–772. 36. Sinha, A. & Stott, L. D. (1994) Global and Planetary Change 9, 297–307. 37. Hoffert, M. I. & Covey, C. (1992) Nature (London) 360, 573–576.
GASES IN ICE CORES
8343
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8343–8349, August 1997 Colloquium Paper This paper was presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held November 13–15, 1995, at the National Academy of Sciences, Irvine, CA.
Gases in ice cores
MICHAEL BENDER*, TODD SOWERS†, AND EDWARD BROOK*‡
© 1997 by The National Academy of Sciences 0027-8424/97/948343-7$2.00/0 PNAS is available online at http://www.pnas.org. ABSTRACT Air trapped in glacial ice offers a means of reconstructing variations in the concentrations of atmospheric gases over time scales ranging from anthropogenic (last 200 yr) to glacial/interglacial (hundreds of thousands of years). In this paper, we review the glaciological processes by which air is trapped in the ice and discuss processes that fractionate gases in ice cores relative to the contemporaneous atmosphere. We then summarize concentration–time records for CO2 and CH4 over the last 200 yr. Finally, we summarize concentration–time records for CO2 and CH4 during the last two glacial–interglacial cycles, and their relation to records of global climate change. Ice crystals near the surface of a glacier are compressed by the continual addition of snow at the surface. As the ice crystals travel down into the glacier, they grow and reorient themselves into a closer packing. Density rises, open porosity decreases, and by some depth between 40 and 120 m, the crystals are sintered together into an impermeable mass (“glacial ice”) in which about 10% of the volume is composed of isolated bubbles. The gas trapped in these bubbles is close in composition to contemporaneous air and allows us to reconstruct changes in atmospheric chemistry on three important time scales: the last 200 years (relating to anthropogenic change), the Holocene (last 10,000 years), and the last several hundred thousand years (relating to glacial–interglacial cycles). In this paper we discuss the gas trapping process and summarize data on the changes in the greenhouse gas concentrations of air over the three time scales of interest. Physics of Gases in Glaciers The density of ice at the surface of an ice sheet is typically 0.3–0.35 g cm−3; the corresponding porosity is 62–67%. Settling and packing cause the density to rise rapidly to about 0.55 g cm−3 by a depth of 10–30 m. Below, recrystallization and other processes drive a somewhat slower increase in density, which continues until individual crystals are fused together into an impermeable mass of glacial ice (1). At the “bubble closeoff depth,” about 10–15% of the volume is air, and the density is about 0.81–0.84 g cm−3 (2, 3). The “firn” is the zone of porous snow and ice above the closeoff depth, and the depth interval in which bubbles close is termed the “firn–ice transition.” Below the transition, densification continues by the compression of bubbles due to hydrostatic pressure. When the snow accumulation rate at the surface of an ice sheet is greater than about 4 cm yr−1 (expressed as the ice-equivalent thickness of annual layers), discrete seasonal layers of snow are preserved that have characteristic physical and chemical properties. Wintertime layers initially have higher densities than summertime layers. This density contrast is maintained during the densification process (Fig. 1). The firn-ice transition occurs at the same density for wintertime and summertime layers, but wintertime layers attain the closeoff density at a shallower depth. Consequently, there is an interval of about 10 m in which wintertime layers are more extensively sealed than summertime layers. In this interval, permeable and impermeable layers alternate in the ice sheet (3). There are three regimes of gas transport in the firn (4, 5.) The uppermost layer, which appears to extend down to about 10-m depth at some sites where firn air has been sampled and analyzed, is affected by convective mixing driven by surface wind stress. Underlying the convective zone is the “stagnant air column,” in which transport is by molecular and atomic diffusion only. Diffusivities of gases are typically about 1 m2 day−1 at 10-m depth. Below, they decrease with increasing density (5), due to a combination of lower porosity and higher tortuosity (the latter factor accounts for the extra distances gas atoms and molecules must travel as they wind their way through the ice crystals to move from one depth to another). The diffusivity of the firn is such that air at the base of the stagnant column today has a “CO2 age” ranging from about 6 yr for the GISP2 core (central Greenland) to about 40 years at Vostok (East Antarctica). In point of fact, however, air in firn at a given depth is not of a single age. The composition of firn air is convoluted by a number of processes. Air in the convective zone responds instantaneously to changes in atmospheric chemistry. These changes then propagate down into the stagnant column by molecular or atomic diffusion. Schwander et al. (6) calculated that the age of maximum abundance at the base of the firn is about 0.65·zt2/D, where zt is the height of the stagnant column and D is the free air diffusivity of the gas at the ambient temperature and pressure of the site. A small fraction of the gas is younger, and there is a long tail to older ages. The diffusivity of an element or compound decreases with increasing mass and increasing atomic or molecular diameter. Thus each element or compound diffuses at a different rate, and each isotope of a compound diffuses at a different rate. In consequence, the covariation between the composition of one gas and another (e.g., CO2 and CH4) in firn is different from their historical covariation in air. The isotopic composition of a gas (e.g., CO2) in firn air also varies with the concentration of that gas in a way that is different from the historical relationship. The concentrations of gases and isotopes that diffuse most rapidly will be closest to their current atmospheric concentrations. Because light isotopes diffuse more rapidly, the concentration of a gas in firn air will be more depleted in heavy isotopes than was the atmosphere at the time it had the same concentration as a firn air sample. Differential diffusivity is a first-order effect that must be taken into account when interpreting data on the concentration and isotopic composition of gases in firn air and ice cores (7). Abbreviation: kyr, thousands of years.
*Graduate School of Oceanography, University of Rhode Island, Kingston, RI 02881; †Department of Geosciences, 447 Deike Building, Pennsylvania State University, University Park, PA 16802; and ‡Departments of Geology and Environmental Science, Washington State University, Vancouver, WA 98686
GASES IN ICE CORES
8344
FIG. 1. (From Martinerie et al., ref. 3). (Upper) Closed porosity vs. density in the firn at Summit, Greenland. Note the monotonic relationship between the two properties. When density = 0.83, open porosity = 0. The decrease in closed porosity at higher densities is due to compression of bubbles. (Lower) Closed porosity vs. depth for maximum-density layers and minimum-density layers. Note how maximum-density layers are completely closed by 75-m depth, and minimum-density layers by 83-m depth. In the intervening depth interval, parts of the firn remain permeable but gases cannot migrate vertically. Several additional factors influence the composition of air in the firn. Gravitational fractionation is one (8, 9, 10). The pressure of gases increases with depth below the surface of the firn according to the barometric equation: P/Pz = exp(mgz/RT). m = mass in units of g mol−1, g is the gravitational acceleration constant, z is depth, R is the ideal gas constant, and T is Kelvin temperature.
FIG. 2. δ15N (Left) and CH4 (Right) vs. depth in firn air from the GISP2 site at Summit, Greenland. The subsurface maximum in δ15N is due to thermal fractionation (15N is enriched at 5- and 10-m depth because the firn at these depths, which remains at the mean annual temperature, is colder than air at the surface during summertime, when sampling was done). The increase below 20-m depth is due to gravitational fractionation. CH4 decreases very slowly to the top of the bubble closeoff zone at 70-m depth. Below it decreases very rapidly because gases cannot migrate vertically and the age of the gas in the firn increases as rapidly as the age of the ice (about 4 yr/m). As Craig et al. (8) and Schwander (10) recognized, this equation applies not only to bulk air but to each individual constituent of air in that (dominant) depth interval of the firn where transport is essentially entirely by diffusion (the stagnant air column). The rate at which the enrichment-per-mass unit increases with depth, expressed in the δ notation, is (∆mg/RT −1)·1,000, or about 0.005‰/amu per meter at typical firn air temperatures. The relative enrichment with depth for different species is directly proportional to the mass difference. The firn air data for the GISP2 site, central Greenland, demonstrate the expected enrichment for the δ15N of N2 (Fig. 2). The enrichment or depletion is significant for nearly all species, corresponding to 3 ppmv of CO2 at the base of deep firn profiles, for example. Seasonal changes in the concentrations of gases in air cause seasonal variations in firn air chemistry. The magnitude of these variations relative to their secular trends depends on location and property. The effect is perhaps largest for O2, CO2, and δ13C of CO2 in Greenland. Seasonal variations are damped out with depth and become very small below 30–50 m. Thermal fractionation also affects the isotopic and elemental composition of firn air. Severinghaus (11) and Severinghaus et al. (12) first recognized the importance of thermal fractionation in porous environmental media in their studies of the composition of air in sand dunes. Temperature gradients cause fractionation, with heavier gases or isotopes being enriched in colder regions. For 15N, the fractionation is about 0.025‰/°C. Thermal fractionation is large in firn because gases diffuse faster than heat. In consequence, steep seasonal temperature gradients occur in the upper 5 m of the firn and gases nearly equilibrate with these temperature gradients. This effect produces large seasonal variations in isotopic compositions and in the O2/N2 ratio in the top few meters of the firn. The seasonal anomalies decrease with depth, and for most species are insignificant below 30 m. O2 is an exception; the concentration of this gas in air is changing so slowly (on a percentage basis) that seasonal thermal gradients are significant down to 60-m depth. These processes combine to influence the composition of gas throughout the firn, and at its base where gases are trapped as bubbles in impermeable ice. Here, two modes of trapping are possible. First, seasonal layering may be absent and air may be trapped throughout the bubble closeoff zone. In this case, the composition of the bulk trapped gases in ice cores will be further convoluted because of the finite closeoff interval. At Vostok, for example, the bubble closeoff zone is about 8 m
GASES IN ICE CORES
8345
thick. A single layer of ice traps bubbles throughout the 300 yr it moves through this zone. This process accounts for the largest share of the dispersion of gas ages in a single sample of ice. At Summit, Greenland, on the other hand, high-density layers are completely sealed as they pass through the top of the bubble closeoff zone. Sealing forms vertically impermeable layers that prevent additional diffusive mixing and “locks in” the composition of gas present in the open, intervening, low-density layers. In such a case, individual ice samples can resolve time periods as short as a decade. The influence of lock-in at Summit can be seen from the firn air data at GISP2 (Fig. 2). δ15N ceases to rise below 70 m depth. CH4 and CO2 concentrations fall rapidly below this depth; CO2 and CH4 “ages” of firn air increase about as rapidly as ice ages below 70 m (4 yr/m). These firn air results are similar to those of Schwander et al. (4), who developed the method for firn air sampling and studied the composition of air in the firn at the nearby GRIP site. The final process influencing the composition of gas in polar firn and ice cores is effusion. Craig et al. (8) suggested this process to account for the depletion of O2 and Ar relative to N2 in polar ice samples, as measured first by Raynaud and Delmas (13) and later by Craig et al. (8) and Sowers et al. (9). Craig et al. (8) pointed out that O2 and Ar, with diameters of about 3 Å, were smaller than N2 (diameter 3.3 Å). They suggested that cracks and imperfections in the ice of about 3 Å spacing would allow the effusive loss of O2 and Ar while selectively retaining nitrogen. It is not clear whether effusion takes place in situ during bubble closeoff or after ice cores are retrieved. Some indications for the former possibility are that deep firn air samples are enriched in O2 and Ar relative to N2, and that the O2 and Ar depletions of ice core samples do not increase after cores have been on the surface for a few days. The processes affecting gases in ice cores need to be taken into account in reconstructions of the composition of the past atmosphere. First, measured concentrations of gases in ice cores and firn air need to be corrected for effects of gravitational fractionation and, where appropriate, thermal fractionation. Second, gas records are useful only when dated absolutely or on a time scale common to other records. Because bubbles close at depths of 40–120 m, gases are younger than the ice enclosing them. The gas age–ice age difference (∆age) is as great as 7 kyr in glacial ice from Vostok; it is as low as 30 yr in the rapidly accumulating Antarctic core DE 08. There are substantial uncertainties associated with ∆age, limiting our ability to interpret some records. This is not a problem when reconstructing the anthropogenic transient from ice core studies, because one can align the recent part of ice core records with direct observations and assume that ∆age is constant below the interval of overlap. Once trapped in bubbles, air in ice cores is subject to two additional processes. First, bubbles are compressed under hydrostatic pressure (14). Second, gases eventually begin to dissolve in the ice as air hydrates (14, 15). Nucleation is kinetically limited, and at ambient temperatures and pressures occurs over order 104 yr (15). As a result, air hydrates form at depths of about 400–1,500 m; cold temperatures and slow accumulation favor formation at shallower depths. Different gases form clathrates at different pressures (16). In the long zone over which air hydrates and bubbles coexist in ice cores, the composition of gases in bubbles must be different from the composition in bulk ice. When gas samples are extracted after crushing of ice for a sufficiently long period of time, as is done for CO2 analysis, individual compounds are apparently not fractionated despite the fact that overall extraction efficiency is <100%. The evidence for this statement is that coherent records of CO2 are obtained by analyzing different ice cores, despite the fact that the dissolution of gases occurs in samples of different ages. Reconstructions of the Anthropogenic Transient from Ice Core and Firn Air Chemistry CO2. The most extensive study of the preindustrial CO2 concentration of air and its anthropogenic rise is that of Etheridge et al. (17). Their results are based largely on studies of the DE 08 ice core, from Law Dome, Antarctica (66° 43 S, 113° 12 E; elevation 1,250 m). The high accumulation rate, about 1.2 m/yr, and warm annual temperature (−19°C) at the site of this core (which causes the closeoff depth to be relatively shallow) allow time to be resolved exceptionally well. Etheridge et al. (17) estimate the gas age–ice age difference to be only 30 yr and the duration of the bubble closeoff process to be 8 yr. They supplement their results from DE-08 with the nearby DSS core and compare their CO2–time curve with that from Siple Dome, analyzed by Neftel et al. (18). Results are summarized in Fig. 3. The CO2 concentration of air varies in the range 275 ± 5 ppm between 1350 and 1800. The anthropogenic rise begins between 1780 and 1870, the later age corresponding to the time when CO2 clearly begins rising above the highest level of the previous 500 yr. As inferred from the study of the DE 08 core, CO2 generally rises at a rate that tends to increase with time. The most dramatic exception is the interval of constant concentration inferred for the period between 1935 and 1945. The DE 08 record is in good agreement with the DSS and Siple data. The exception is the period between 1935 and 1945, when CO2 concentrations from the latter two cores fall below those of DE 08 by up to 5 ppmv. Box-diffusion model deconvolutions, which assume CO2 concentrations to be in good agreement with the Siple results, invoke a net biosphere release of about 0.1–0.5 Gt C/yr during this period (20, 21). Etheridge et al. (17) suggest that one can balance the carbon cycle between 1935 and 1945 with a combination of
FIG. 3. (Upper) CO2 vs. time before present, as inferred by Etheridge et al. (17) from ice core studies. (Lower) CH4 vs. time before present, as inferred by Etheridge et al. (19) from ice core studies.
GASES IN ICE CORES
8346
lower respiration, increased growth due to CO2 fertilization and changes in temperature and precipitation, and suppressed CO2 input to the atmosphere as a result of exceptional El Nino events. Otherwise the new data of Etheridge et al. (17) are consistent with the Siple CO2 data and estimates of carbon fluxes derived therefrom. These invoke rising CO2 beginning around 1780, with the biosphere as the major source. Fossil fuel becomes the dominant source around 1910. The biosphere continues to be a source until 1940 in the deconvolution of Siegenthaler and Oeschger (20), when it becomes neutral, and 1970 in the deconvolution of Keeling et al. (21), when it becomes a sink.
FIG. 4. Selected ice core records for the last 350 kyr. Data from refs. 34, 35, 36, 37. CH4. Two recent records (19, 22) supplement earlier work of Craig and Chou (23), Etheridge et al. (24), Khalil and Rasmussen (25), and Stauffer et al. (26) to provide a detailed curve of the anthropogenic increase in atmospheric CH4. The results of Blunier et al. (22) suggest a stable preindustrial value of about 720 ppbv over Antarctica between about 1000 A.D. and 1750 A.D. The Antarctic concentration rise after 1840, as measured in the DE 08 core by Etheridge et al. (24), is shown in Fig. 3. The factors maintaining the preanthropogenic CH4 background, and those responsible for the anthropogenic increase,
GASES IN ICE CORES
8347
have been discussed in a number of papers (refs.27 and 28 and citations therein). According to Chappellaz et al. (29, 30), preindustrial CH4 production was dominated by tropical wetlands, with grassland and temperate to boreal forests also making significant contributions. They argue that the natural source has diminished by 20% since preindustrial times. Anthropogenic sources are, of course, far greater. Chappellaz et al. (29) estimate that an anthropogenic decrease of about 20% in the atmospheric OH concentration has caused the atmospheric lifetime of CH4 to rise and thereby contributed to the anthropogenic increase.
FIG. 5. Selected ice core records for the last 100 kyr. Data from refs. 30, 34, 35, and 38. Other Gases. Studies of trapped gases in ice cores have revealed preanthropogenic levels of two other properties. The preanthropogenic N2O concentration is estimated to have been about 275–285 ppbv, compared with a contemporary value of about 311 ppbv (24, 31, 32). Because of its high solubility, the precise measurement of this gas in ice cores is very difficult. The δ13C of CO2 over Antarctica decreased from a preindustrial level of about −6.5‰ to approximately −8.0‰ today (33). The decrease roughly mirrors the CO2 increase. Firn Air Studies. Sampling and analysis of gas in the firn is an emerging approach for improving existing records of an
GASES IN ICE CORES
8348
thropogenic transients (4, 7). Firn air sampling allows one to collect 1,000 liters or more of isotopically unfractionated air as old as 90 yr. These samples can be used to measure concentrations of trace organic compounds, O2/N2 ratios, and the isotopic composition of greenhouse gases, all measurements that are far more difficult to make on ice core samples. Glacial–Interglacial Changes in Atmospheric Chemistry Recorded in Ice Cores Selected climate records are summarized in Fig. 4 and Fig. 5, covering the periods from 0 to 350 kyr and from 0 to 100 kyr before the present (B.P.), respectively. The δ18O of calcitic foraminifera from deep sea sediments is a proxy indicator for ice volume. The δD or δ18O of ice from ice cores is a proxy indicator of temperature in the area of the ice core. Recent inversions of borehole temperature data conclude that the glacial–interglacial temperature change in Greenland was about 20°C (39, 40). The dust content of ice cores is a proxy for dryness of the source areas. High dust contents are generally attributed to dry source areas from which dust is readily suspended into the atmosphere. CO2 and CH4 concentrations are measured in bubbles of polar ice as described above. June solar insolation at 65° N latitude is plotted at the bottom of the figure. On a broad scale, the atmospheric CO2 concentration is clearly linked to climate change. Lorius et al. (41) argued that the glacial– interglacial CO2 change (with a small contribution from CH4) was responsible for a 2°C temperature change when all feedbacks are considered. Obviously this number is highly uncertain but it is very likely that increasing CO2 concentrations in air contributed to global warming during glacial terminations 1 and 2. The link between CO2 and higher-frequency climate changes is weak. For example, there is no minimum in CO2 corresponding to the cold period (Glacial Stage 5d) at about 110 kyr B.P. Similarly, CO2 does not rise during the interstadial events of the past 35 kyr (ref. 42; Fig. 5). Many factors have been invoked to explain the glacialinterglacial change in the CO2 concentration of air. The cause of this change was vigorously debated for a decade beginning in 1982. Recently attention to this subject has diminished, more because the protagonists are exhausted than because the issue is resolved. It is widely accepted that the pCO2 of the atmosphere is regulated by the pCO2 of surface seawater, because about 99% of the CO2 in the ocean/atmosphere system resides in the oceans. In his classic paper initiating the debate, Broecker (43) noted that lower glacial temperatures would cause the CO2 concentration of air to fall, whereas higher salinities would cause CO2 to rise. A glacial ocean temperature decrease of 1.5°C would cause CO2 to fall by 20 ppmv. This change would be roughly offset by a salinity increase of 1‰. Recent controversial results showing that equatorial temperatures may have increased by up to 5°C suggest that the glacial–interglacial change in sea surface temperatures may have been much larger than Broecker initially estimated. If this is correct, the temperature change may have contributed 20–30 ppm of the glacial–interglacial change in atmospheric pCO2 after correcting for the compensating effect of salinity (the temperature dependence of pCO2 on sea surface temperature is about 13 ppmv/°C; ref. 43). Clearly other factors must contribute to the glacialinterglacial difference in atmospheric pCO2. These factors must act either by decreasing the total CO2 concentration of surface seawater or by increasing the surface alkalinity. Changes in ocean circulation, the nutrient/carbon ratio of organic matter, and varying rates of nutrient utilization in Antarctic waters have been invoked to account for the changing surface water total CO2 concentrations. Removal of CaCO3 during sea level rise, changing ratios of CaCO3 to organic carbon in biogenic debris falling out of the surface ocean, and changes in the calcium carbonate compensation depth in the deep ocean have been invoked to change the alkalinity of surface waters. There have been large changes in the CH4 concentration of air during the last 200 kyr (Fig. 4 and Fig. 5). CH4 changes are very closely linked with changes in climate (29, 44). CH4 differs from CO2 in two important ways. First, CO2 changes are about five times more important than those of CH4 in driving glacial–interglacial temperature changes (41). Second, CH4 changes, which are believed to be caused mainly by changes in emission rates, are somewhat better understood and serve much better as a climate proxy. Chappellaz et al. (27) generally interpreted changes in atmospheric CH4 as reflecting increases in the extent of low-latitude wetlands due to increased precipitation and perhaps temperatures. Recent records of atmospheric CH4 variations over the last 200 kyr (Fig. 4 and Fig. 5) reveal several very interesting features. As noted above, there is a large glacial–interglacial change, with Antarctic CH4 concentrations varying between glacial levels of 350 ppbv and interglacial levels of 700 ppbv. There are changes of somewhat smaller magnitude associated with climate changes linked to the 20-kyr climate cycles associated with precession. Examples are CH4 maxima associated with warm periods at 80 kyr and 100 kyr B.P. Third, there are large CH4 variations associated with interstadial events of the last 40 kyr. These events, recorded most dramatically in Greenland ice cores (e.g., ref. 30; Fig. 5), reflect rapid warming over Greenland, slow cooling, and rapid cooling back to baseline glacial values. During the last 40 kyr at least, interstadial warmings are accompanied by increases of about 150 ppbv in the CH4 concentration of air. Chappellaz et al. (30) invoked wetter and warmer conditions in the tropics to explain the CH4 increases. Support for this idea comes from the fact that the longer interstadial events, at least, are recorded in Antarctica, illustrating their global nature (45), and are sometimes coincident with deposition of cave deposits in Botswanaland, which Holmgren et al. (46) linked to increased tropical precipitation. Finally, the CH4 concentration of the atmosphere was surprisingly variable during the Holocene. Methane was at a maximum early in the Holocene, fell to a broad minimum about 5 kyr B.P., and has risen during the last 3 kyr (22). Only the abrupt minimum at about 8.2 kyr B.P. has been directly linked to a climate event, namely an abrupt cooling that occurred at that time. The causes of the mid-Holocene minimum and subsequent slow rise are topics of current investigation. This research was supported by grants from the New England Regional Center of the National Institute of Global Environmental Change, Department of Energy (Subagreement 901214-HAR), and the Office of Polar Programs, National Science Foundation. 1. Herron, M. M. & Langway, C. C., Jr. (1980) J. Glaciol. 25, 373–385. 2. Raynaud, D. & Lorius, C. (1977) in Proceedings of the Grenoble Symposium, August/September 1975 (International Association of Hydrological Sciences, United Kingdom), Vol. 118, pp. 326–335. 3. Martinerie, P., Raynaud, D. Etheridge, D. M., Barnola, J.-M. & Mazaudier, D. (1992) Earth Planet. Sci. Lett. 112, 1–13. 4. Schwander, J., Barnola, J.-M., Andrie, C., Leuenberger, M., Ludin, A., Raynaud, D. & Stauffer, B. (1993) J. Geophys. Res. 98, 2831–2838. 5. Sowers, T., Bender, M., Raynaud, D. & Korotkevich, Y. S. (1992) J. Geophys. Res. 97, 15683–15697. 6. Schwander, J., Stauffer, B. & Sigg, A. (1988) Annu. Glaciol. 10, 141–145. 7. Battle, M., Bender, M., Sowers, T., Tans, P., Butler, J., Elkins, J., Ellis, J., Conway, T., Zhang, N., Lang, P. & Clarke, A. (1996) Nature (London) 383, 231–235. 8. Craig, H., Horibe, Y. & Sowers, T. (1988) Science 242, 1675–1678. 9. Sowers, T., Bender, M. & Raynaud, D. (1989) J. Geophys. Res. 94, 5137–5150.
GASES IN ICE CORES
8349
10. Schwander, J. (1989) in The Environmental Record in Glaciers and Ice Sheets, eds. Oeschger, H. & Langway, C. (Wiley, New York). 11. Severinghaus, J. P. (1995) Ph.D. dissertation (Columbia University, New York). 12. Severinghaus, J. P., Bender, M. L., Keeling, R. F. & Broecker, W. A. (1996) Geochim. Cosmochim. Acta 60, 1005–1018. 13. Raynaud, D. & Delmas, R. (1977) in Proceedings of the Grenoble Symposium, August/September 1975 (International Association of Hydrological Sciences, United Kingdom), Vol. 118, pp. 377–381. 14. Gow, A. J. & Williamson, T. (1975) J. Geophys. Res. 80, 5101–5108. 15. Price, P. B. (1975) Science 267, 1802–1804. 16. Miller, S. L. (1969) Science 165, 489–490. 17. Etheridge, D. M., Steele, L. P., Langenfelds, R. L., Francey, R. J., Barnola, J.-M. & Morgan, V. I. (1995) J. Geophys. Res. 101, 4115–4128. 18. Neftel, A., Moor, E., Oeschger, H. & Stauffer, B. (1985) Nature (London) 315, 45–47. 19. Etheridge, D. M., Pearman, G. I. & Fraser, P. J. (1992) Tellus 44B, 282–294. 20. Siegenthaler, U. & Oeschger, H. (1987) Tellus 39B, 140–154. 21. Keeling, C. D., Bacastow, R. B., Carter, A. F., Piper, S. C., Whorf, T. P., Heimann, M., Mook, W. G. & Roeloffzen, H. (1989) in Aspects of Climate Variability in the Pacific and Western Americas, Geophysical Monograph, ed. Peterson, J. H. (Amer. Geophysical Union, Washington, DC), Vol. 55, pp. 165–236. 22. Blunier, T., Chappellaz, J., Schwander, J., Stauffer, B. & Raynaud, D. (1995) Nature (London) 374, 46–49. 23. Craig, H. & Chou, C. C. (1982) Geophys. Res. Lett. 9, 1221–1224. 24. Etheridge, D. M., Pearman, G. I. & de Silva, F. (1988) Annu. Glaciol. 10, 28–33. 25. Khalil, M. A. K. & Rasmussen, R. A. (1987) Atmos. Environ. 21, 2445–2452. 26. Stauffer, B., Fischer, G., Neftel, A. & Oeschger, H. (1985) Science 229, 1386–1388. 27. Aselmann, I. & Crutzen, P. J. (1989) J. Atmos. Chem. 8, 307–358. 28. Fung, I., John, J., Lerner, J., Matthews, E., Prather, M., Steele, L. P. & Fraser, P. J. (1991) J. Geophys. Res. 96, 13033–13065. 29. Chappellaz, J., Fung, I. & Thompson, A. (1993) Tellus 45B, 228–241. 30. Chappellaz, J., Blunler, T., Raynaud, D., Barnola, J. M., Schwander, J. & Stauffer, B. (1993) Nature (London) 366, 443–445. 31. Khalil, M. A. K. & Rasmussen, R. (1988) Annu. Glaciol. 10, 73–80. 32. Machida, T., Nakazawa, T., Fujii, Y., Aoki, S. & Watanabe, O. (1995) Geophys. Res. Lett. 22, 2921–2924. 33. Friedli, H., Lotscher, H., Oeschger, H., Siegenthaler, U. & Stauffer, B. (1986) Nature (London) 324, 237–238. 34. Shackleton, J. H. & Pisias, N. G. (1985) in The Carbon Cycle and Atmospheric CO2: Natural Variations Archean to Present, Geophysical Monograph Series, eds. Sundquist, E. T. & Broecker, W. S. (Amer. Geophysical Union, Washington, DC), Vol. 32, pp. 303–317. 35. Jouzel, J., Barkov, N. I., Barnola, J. M., Bender, M. L., Chappellaz, J., Genthon, C., Kotlyakov, V. M., Lipenkov, V., Lorius, C., Petit, J. R., Raynaud, D., Raisbeck, G., Ritz, C., Sowers, T., Stievenard, M., Yiou, F. & Yiou, P. (1993) Nature (London) 364, 407–412. 36. Barnola, J. M., Pimienta, P., Raynaud, D. & Korotkevich, Y. S. (1991) Tellus 43B, 83–90. 37. Petit, J. R., Mounier, L., Jouzel, J., Korotkevich, Y., Kotlyakov, V. & Lorius, C. (1990) Nature (London) 343, 56–58. 38. Grootes, P. M., Stuvier, M., White, J. W. C., Johnsen, S. & Jouzel, J. (1993) Nature (London) 366, 552–555. 39. Cuffey, K., Alley, R., Grootes. P., Bolzan, J. & Anandakrishnan, S. (1994) J. Glaciol. 40, 341–349. 40. Johnsen, S., Dahl-Jensen, D., Dansgaard, W. & Gundestrup, N. (1995) Tellus 47B, 624–629. 41. Lorius, C., Jouzel, J., Raynaud, D., Hansen, J. & Le Treut, H. (1990) Nature (London) 347, 139–145. 42. Sowers, T. & Bender, M. (1995) Science 269, 210–214. 43. Broecker, W. (1982) Geochim. Cosmochim. Acta 46, 1689–1705. 44. Chappellaz, J., Barnola, J. M., Raynaud, D., Korotkevich, Y. S. & Lorius, C. (1990) Nature (London) 345, 127–131. 45. Bender, M., Sowers, T., Dickson, M. L., Orchardo, J., Grootes, P., Mayewski, P. A. & Meese, D. A. (1994) Nature (London) 372, 663–666. 46. Holmgren, K., Karlen, W. & Shaw, P. A. (1995) Quat. Res. 43, 320–328.
TREE RINGS, CARBON DIOXIDE, AND CLIMATIC CHANGE
8350
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8350–8353, August 1997 Colloquium Paper This paper was presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held November 13–15, 1995, at the National Academy of Sciences, Irvine, CA.
Tree rings, carbon dioxide, and climatic change
GORDON C. JACOBY* AND ROSANNE D. D'ARRIGO Tree-Ring Laboratory, Lamont-Doherty Earth Observatory, Palisades, NY 10964
© 1997 by The National Academy of Sciences 0027-8424/97/948350-4$2.00/0 PNAS is available online at http://www.pnas.org. ABSTRACT Tree rings have been used in various applications to reconstruct past climates as well as to assess the effects of recent climatic and environmental change on tree growth. In this paper we briefly review two ways that tree rings provide information about climate change and CO2: (i) in determining whether recent warming during the period of instrumental observations is unusual relative to prior centuries to millennia, and thus might be related to increasing greenhouse gases; and (ii) in evaluating whether enhanced radial growth has taken place in recent decades that appears to be unexplained by climate and might instead be due to increasing atmospheric CO2 or other nutrient fertilization. It is found that a number of tree-ring studies from temperature-sensitive settings indicate unusual recent warming, although there are also exceptions at certain sites. The present tree-ring evidence for a possible CO2 fertilization effect under natural environmental conditions appears to be very limited. Longer time series than those presently available from instrumental records are needed to evaluate whether recent climatic shifts are unusual and might be evidence of anthropogenic change due to increasing CO2 and other greenhouse gases. Longer records of natural climate variability and forest growth information can also help validate climate and carbon budget models used for prediction of future climate (e.g., see ref. 1). Large-scale changes in sources and sinks of carbon in the terrestrial biosphere (due to climatic change, direct CO2 fertilization, forest regrowth, increased decay rates, or other factors) can act as either negative or positive feedbacks to the earth's climate system (e.g., see refs. 2 and 3). Recent studies based on isotopic measurements of atmospheric CO2 suggest that there may in fact be a large CO2 sink in the land biosphere of northern temperate latitudes (30–60°N) (4, 5). Below we outline some of the tree-ring evidence for recent climate and forest growth changes and their relevance for studies of the global carbon cycle. We focus on two issues: (i) whether recent climatic changes during the period of instrumental observations appear to be unusual relative to the past, and (ii) whether enhanced radial growth has taken place that appears to be unexplained by climate and might be due to increasing atmospheric CO2 or other nutrient fertilization. Do Temperature-Sensitive Tree-Ring Records Indicate that Recent Warming is Unusual? Tree-ring measurements can help to distinguish anthropogenic from natural environmental change. These data can be used to determine whether recent climatic changes are unusual and possibly due to anthropogenic effects (specifically, increasing CO2 and other trace gases) (e.g., see ref. 6) or are still within the range of natural climate variability. Several recent studies, outlined briefly below, have evaluated tree-ring and other proxy data with this goal in mind. Cook (7) reviewed high-resolution temperature histories from tree ring and coral proxies to evaluate to what degree the 20th century warming has been anomalous relative to prior centuries to millennia. At northern latitudes, these histories include temperature-sensitive tree-ring series for northern Alaska (8), the north Polar Urals (9), and the Arctic as a whole (10). All three of these series indicate unusual 20th century warming. Recent tree-ring data from Mongolia indicate that there is unusual warming in that region (11), in agreement with the Arctic reconstruction (10). Jacoby and D'Arrigo (8) describe recent warming in Alaska relative to past tree growth variations (see Fig. 1). This study describes a summer temperature reconstruction based on maximum latewood density which shows evidence of recent warming of 0.5° to 1°C over the past century. By contrast, the ring-width data, which appear to integrate temperature conditions throughout the year (8), indicate more pronounced recent warming of annual temperatures of 2° to 3°C. Briffa et al. (12) used a 1,000-year long tree-ring temperature record from Siberia to infer that the twentieth century (1901–1990) summer warmth has been unusual relative to the past millennium. Bradley and Jones (13) reconstructed Northern Hemisphere summer temperatures back to A.D. 1400 by using a combination of historical, tree-ring, and ice-core data and found recent conditions to be very warm relative to the past. In contrast, a millennium-long record from Fennoscandia indicates that it was warmer in Fennoscandia during the so-called Medieval Warm Period (14) than it is today, possibly due to cooling of the North Atlantic (12). In the Southern Hemisphere, a multimillennial summer temperature reconstruction from southern South America shows no evidence of unusual recent warming, in agreement with instrumental records (15, 16). However, tree-ring and other records from Tasmania and New Zealand do indicate anomalous warming in recent decades relative to the past (7, 17, 18). The Tasmanian huon pine record, which is multimillennial in length, indicates that the warming of recent decades is highly unusual, with only one marginally warmer interval over the past several thousand years. This warm-season temperature reconstruction suggests that the recent warming in Tasmania is anomalous although not entirely unprecedented (7). In summary, a number of temperature-sensitive middle to higher latitude tree-ring records from both hemispheres show evidence that recent warming in these regions may be anomalous. A few of these series are millennial in length. These few very long records allow evaluation of temperature variations prior to the Little Ice Age cold interval, which could bias interpretation of recent warm conditions (13). Other sources of proxy data in several areas also support these indications of unusual recent warming relative to the past [e.g., ice cores (19)].
*To whom reprint requests should be addressed, e-mail:
[email protected].
TREE RINGS, CARBON DIOXIDE, AND CLIMATIC CHANGE
8351
FIG. 1. Reconstructions of central Alaska temperatures. (Upper) Five-year averaged annual (October–September) temperature reconstructed by using ring widths. (Lower) Summer (May–August) temperatures reconstructed by using maximum latewood density. Note the increase in reconstructed summer temperature over the past 100 years is only about 0.5° to 1.0°C, whereas the reconstructed annual temperature has increased about 1.5° to 2°C. The cooler period in annual temperatures prior to 1900 was broken by several warm intervals. Dashed line in Upper is 5-year recorded temperatures for central Alaska. Note that the reconstruction underestimates temperatures since about 1970. This is attributed to the effects of moisture stress (8). The tree-ring data used in these studies are from sites selected to amplify the climatic signal due to temperature, and are not necessarily representative of large components of the land biosphere nor indicators of large-scale enhanced carbon sequestration. Changes in radial growth in these trees do not provide information about possible shifts in respiration or allocation of carbon below-ground. Warming may also be causing negative feedbacks to forest productivity, which can counteract enhanced growth in other areas. For example, some temperature-limited sites may now be showing the negative effects of moisture stress (partially due to increased evapotranspiration caused by warmer temperatures) or insect infestation related to recent warmer conditions (ref. 8 and see Fig. 1). Is There a CO2 Fertilization Effect in Tree Rings? Another means by which tree rings are being used to test for anthropogenic effects is by evaluating whether direct CO2 fertilization due to increasing atmospheric CO2 (ordinarily limiting to plant growth) is presently enhancing the growth of natural vegetation. The response of plant growth to a direct CO2 fertilization effect has been demonstrated in numerous laboratory experiments, usually using seedlings (e.g., ref. 20). Modeling suggests that this enhanced growth should result in greater carbon sequestering of land ecosystems, provided that this “beta factor” is sufficiently large (e.g., refs. 21 and 22). Little is known, however, about whether such an effect is occurring on a large scale in natural vegetation, where environmental conditions are exceedingly complex. Here we review several tree-ring studies which evaluate the possible effects of direct CO2 fertilization on radial growth of trees growing in natural environmental settings. LaMarche et al. (23) presented one of the first studies which purported to find evidence for a possible CO2 fertilization effect in tree rings. Their study was based on ring-width chronologies of high-elevation bristlecone and limber pines growing in the southwestern United States, which show unusual enhanced growth over the past century. One reason for their conclusion that this enhanced growth is due to CO2 fertilization is that high-elevation plants may be more CO2-limited than those at lower elevations (24). Yet no quantitative modeling was presented by LaMarche et al. (23) to rule out the possible contribution of favorable climatic change to account for the growth increases. Graumlich (25) found no such evidence for CO2 fertilization in high-elevation foxtail pine and other species in the Sierra Nevada. She based her conclusions on the observations that (i) recent trends were not unusual relative to those in the pre
TREE RINGS, CARBON DIOXIDE, AND CLIMATIC CHANGE
8352
anthropogenic period, and (ii) recent growth variations were largely explainable by climate–growth relationships. A contrasting view is presented by Graybill (26) and Graybill and Idso (27), who argued that CO2 fertilization is detectable in certain pine species growing at high elevations of the southwestern United States, but only if they show a strip-bark growth form. In trees with a strip-bark morphology, any added CO2 should be allocated primarily to the active cambial region, resulting in a greater response (27). Graumlich (25) speculated that the disparate conclusions for trees in the southwest might be reconciled, since the LaMarche et al. (23) trees were also of a strip-bark morphology. By contrast, the trees in her study did not show this feature. Other studies include a paper by Kienast and Luxmoore (28), who showed negative results for a CO2 fertilization effect in trees in the Rocky Mountains of Colorado. D'Arrigo and Jacoby (29) did not find evidence for a CO2 fertilization effect at the northern treeline of North America, based on evaluation of residual trends following modeling of climate–growth relationships. One possible explanation is that a threshold level of CO2 increase is needed before an effect can be detected. Another is that other factors, including cold temperatures, a short season of cambial cell division, and nitrogen deficiency could preclude a direct CO2 response in the extreme boreal forests. The unexplained increase in growth of lodgepole pine at a high-elevation site in the San Jacinto Mountains of California (30) did not occur in limber pine near the same site and, as noted in the study, could be related to changes in winter precipitation. One of the most thorough analyses of representative boreal forest growth involved the measurement of ring widths and density of trees in mature, closed canopy, white spruce stands at 11 locations in western Canada (31). A limited number of trees were felled, and seven disks were cut from each to obtain data on cross-sectional area and taper to enable calculations of volumetric and biomass growth rate change. Jozsa and Powell concluded that biomass productivity and annual growth layer weights are related to long-term and yearly climatic variability with possible response to spruce budworm activity (31). They do not present any indication that there is a systematic growth trend that could be related to CO2 fertilization. This is an extremely important study of mature trees in natural forest stands. Thus the results are widely relevant to real-world situations. Discussion and Conclusions We have briefly described some of the tree-ring evidence presently being used to assess whether recent growth changes are unusual relative to the past, and might be evidence for warming due to greenhouse gases and/or direct CO2 fertilization. A number of temperaturesensitive records, some of which date back for several millennia, do indicate unusual recent warming. Yet this is by no means taking place at all sites. Another caveat is that the trees studied here are from particular sites selected to amplify climatic signals, and are not necessarily representative of large components of the land biosphere nor enhanced sequestering of carbon (2). In addition, above-ground radial growth changes do not provide information about respiration or below-ground effects. Other changes (e.g., drought stress) could lead to negative feedback effects (e.g., refs. 2 and 8). The evidence for CO2 fertilization is inconclusive at present for trees growing in natural settings, where there can be many other limiting and interacting factors. Controlled experiments simulating natural conditions underway at the Biosphere 2 facility will attempt to evaluate the combined effects of different environmental factors, and compare plant responses in different simulated ecosystems and between species (32). Such controlled studies may provide additional insights which can help resolve the uncertainties of the CO2 fertilization issue. Even if trees with a strip-bark growth form are most likely to show this effect, these types of trees are only a small component of the land biosphere. The evidence described here provides only partial information regarding the behavior of the land biosphere. There are still many uncertainties, and it is unlikely that these issues will be resolved in the very near future. Additional studies and improved spatial and temporal coverage of tree-ring data are needed to decrease uncertainties about whether anthropogenic effects are presently taking place. Research for this paper was supported by the Climate Dynamics Program of the National Science Foundation, ATM 94-06732. This is Lamont–Doherty Earth Observatory contribution no. 5594. 1. Mitchell, J. F. B., Johns, T. C., Gregory, J. M. & Tett, S. F. B. (1995) Nature (London) 376, 501–504. 2. Jacoby, G. C. & D'Arrigo, R. D. (1995) in Biotic Feedbacks in the Global Climatic System: Will the Warming Feed the Warming?, eds. Woodwell, G. M. & Mackenzie, F. T. (Oxford Univ. Press, New York), pp. 108–118. 3. Woodwell, G. M. & Mackenzie, F. T., eds. (1995) Biotic Feedbacks in the Global Climatic System: Will the Warming Feed the Warming? (Oxford Univ. Press, New York). 4. Ciais, P., Tans, P., Torlier, M., White, J. W. C. & Francey R. J. (1995) Science 269, 1098–1102. 5. Denning, A. S., Fung, I. Y. & Randall, D. (1995) Nature (London) 376, 240–243. 6. Wigley, T. M. L., Jones, P. D. & Raper, S. C. B. (1997) Proc. Natl. Acad. Sci. USA 94, 8314–8320. 7. Cook, E. R. (1995) Climate Dynamics 11, 211–222. 8. Jacoby, G. C. & D'Arrigo, R. D. (1995) Global Biogeochemical Cycles 9, 227–234. 9. Graybill, D. A. & Shiyatov, S. G. (1992) in Climate Since A. D. 1500, eds. Bradley, R. S & Jones, P. D. (Routledge, London), pp. 393–414. 10. D'Arrigo, R.D. & Jacoby, G.C. (1993) Climatic Change 25, 163–177. 11. Jacoby, G. C., D'Arrigo, R. D. & Davaajamts, T. (1996) Science 273, 771–773. 12. Briffa, K. R., Jones, P. D., Schweingruber, F. H., Shiyatov, S. G. & Cook., E. R. (1995) Nature (London) 375, 156–159. 13. Bradley, R. S. & Jones, P. D. (1993) The Holocene 3, 367–376. 14. Hughes, M. K. & Diaz, H. F. (1994) Climatic Change 26, 109–142. 15. Lara, A. & Villalba, R. (1993) Science 260, 1104–1106. 16. Villalba, R., Boninsegna, J. A., Lara, A., Veblen, T. T., Roig, F. A., Aravena, J. C. & Ripalta, A. (1996) in Climatic Variations and Forcing Mechanisms of the Last 2000 Years, eds. Jones, P. D., Bradley, R. S. & Jouzel, J. (Springer, New York), pp. 161–189. 17. Cook, E. R., Bird, T., Peterson, M., Buckley, B., D'Arrigo, R. & Francey, R. (1992) The Holocene 2, 205–217. 18. D'Arrigo, R. D., Buckley, B., Cook, E. R. & Wagner, W. S. (1995) Palaeogeography, Palaeoclimatology, Palaeoecology 119, 293–300. 19. Thompson, L. G. (1992) in Climate Since A. D. 1500, eds. Bradley, R. S. & Jones, P. D. (Routledge, London), pp. 517–548. 20. Allen, L. H. & Amthor, J. S. (1995) in Biotic Feedbacks in the Global Climatic System: Will the Warming Feed the Warming?, eds. Woodwell, G. M. & Mackenzie, F. T. (Oxford Univ. Press, New York), pp. 51–84. 21. Reynolds, J. F., Harley, P. C., Thomas, R. B. & Strain, B. R. (1992) Plant, Cell and Environment 15, 271–282. 22. Wullschleger, S. D., Post, W. M. & King, A. W. (1995) in Biotic Feedbacks in the Global Climatic System: Will the Warming Feed the Warming?, eds. Woodwell, G. M. & Mackenzie, F. T. (Oxford Univ. Press, New York), pp. 85–107. 23. LaMarche, V. C., Jr., Graybill, D. A., Fritts, H. C. & Rose, M. R. (1984) Science 225, 1019–1021. 24. Cooper, C. F. (1986) Science 231, 859–860. 25. Graumlich, L. J. (1991) Ecology 72, 1–11. 26. Graybill, D. A. (1987) in Proceedings of the International Symposium on Ecological Aspects of Tree-Ring Analysis, eds. Jacoby, G. C. & Hornbeck, J. W., (U.S. Dept. of Commerce, Springfield, VA), pp. 463–474. 27. Graybill, D. A. & Idso, S. B. (1993) Global Biogeochem. Cycles 7, 81–95.
TREE RINGS, CARBON DIOXIDE, AND CLIMATIC CHANGE
8353
28. Kienast, F. & Luxmoore, R. J. (1988) Ecology 76, 487–495. 29. D'Arrigo, R. D. & Jacoby, G. C. (1993) Global Biogeochem. Cycles 7, 525–535. 30. Jacoby, G. C. (1986) in Climate-Vegetation Interactions, NASA Conference Publication 2440, eds. Rosenzweig, C. & Dickinson, R. (NASA, Greenbelt, MD), pp. 114–118. 31. Jozsa, L. A. & Powell, J. M. (1987) Can. J. For. Res. 17, 1075–1079. 32. Nelson, M., Burgess, T., Alling, A., Alvarez-Romo, N., Dempster, W., Walford, R. & Allen, J. (1993) Bioscience 43, 225–236.
GEOCHEMISTRY OF CORALS: PROXIES OF PAST OCEAN CHEMISTRY, OCEAN CIRCULATION, AND CLIMATE
8354
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8354–8361, August 1997 Colloquium Paper This paper was presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held November 13–15, 1995, at the National Academy of Sciences, Irvine, CA.
Geochemistry of corals: Proxies of past ocean chemistry, ocean circulation, and climate ELLEN R. M. DRUFFEL Department of Earth System Science, University of California, Irvine, CA 92697
© 1997 by The National Academy of Sciences 0027-8424/97/948354-8$2.00/0 PNAS is available online at http://www.pnas.org. ABSTRACT This paper presents a discussion of the status of the field of coral geochemistry as it relates to the recovery of past records of ocean chemistry, ocean circulation, and climate. The first part is a brief review of coral biology, density banding, and other important factors involved in understanding corals as proxies of environmental variables. The second part is a synthesis of the information available to date on extracting records of the carbon cycle and climate change. It is clear from these proxy records that decade time-scale variability of mixing processes in the oceans is a dominant signal. That Western and Eastern tropical Pacific El Niño-Southern Oscillation (ENSO) records differ is an important piece of the puzzle for understanding regional and global climate change. Input of anthropogenic CO2 to the oceans as observed by 13C and 14C isotopes in corals is partially obscured by natural variability. Nonetheless, the general trend over time toward lower δ18O values at numerous sites in the world's tropical oceans suggests a gradual warming and/or freshening of the surface ocean over the past century. The biogeochemical cycling of carbon on Earth has undergone marked changes over the past glacial-to-interglacial period, and nearly all of these changes have occurred prior to scientific observation. The key to understanding present and future changes in climate and the cycling of biogeochemically important elements (i.e., C, N, S, P, O, Cd, Ba, Si, etc.) is to adequately account for the past changes in their cycling. To enable retrospective studies, a number of substrates on Earth have acted as integrators of these changes. Within the bands of tree rings lie records of relative seasonal rainfall and 14C/12C ratios (1) of the atmospheric CO2 used during photosynthesis. Another recorder of climate is oceanic sediment, which contains a layered time history of faunal shells that lived in overlying surface and deep waters. Ice cores have been used to reveal past air temperatures (2) and atmospheric CO2 concentrations (3) during the past 100,000 years. The skeletons of corals provide an unaltered record of the chemical and physical conditions that existed in the surrounding seawater at the time of accretion of its calcium carbonate skeleton. The advantages of corals as an oceanic recorder are the enhanced time resolution (biweekly to seasonal) available from the high growth rate, and the absence of mixing processes that are present in all oxic sediments (i.e., bioturbation). There are two aspects of this synthesis paper. The first is a brief review of the field of coral geochemistry. A brief discussion of coral biology, taxonomy, and microstructure is presented. Factors affecting coral skeletal growth and isotopic and chemical records within the coralline aragonite are reviewed, and a discussion of the annual and subannual density banding within the skeletal matrices is included. Second, a synthesis is presented of the available proxy records from corals as they relate to past climate, ocean circulation changes, and anthropogenic input of excess CO2. For some databases, the anthropogenic CO2 appears clear and globally distributed. For other databases it appears that variability on interannual to decadal time scales complicates a straightforward attempt to separate natural perturbations from the anthropogenic CO2 signal on the Earth's environment. Shen (4) presents a review of the types of coral archives available from the perspective of historical El Niño-Southern Oscillation (ENSO) influences on the tropical Pacific Ocean. Also, Druffel et al. (5) present a review of coral growth and the factors that affect it. The present paper attempts to compliment the information contained in these two publications. Most of the records available to date are from surface corals that are restricted to the upper 50–75 m of the ocean's surface. However, there is a growing effort devoted to the study of deep-sea corals and gorgonians found at all depths of the world's oceans (≤10,000 m). Most of the discussion in this review centers on surface corals, though new and exciting research involving the use of deep species as paleoclimatologic tools are mentioned. Review of Corals as Geochemical Proxies Biology and Microstructure. Corals are from the order Scleractinia, a group in the subclass Zoantharia. Scleractinians include solitary and colonial species of corals, many of which secrete external skeletons of aragonite. The oldest known scleractinians are shallow water corals from the Middle Triassic (6). The polyp portion of the coral secretes calcium carbonate (CaCO3) as the mineral aragonite. The cylindrical outer portion of the polyp is called the polyp wall and is terminated above by the horizontal oral disc and below by the basal disc (see Fig. 1). The oral disc is bound by one or more ringlets of tentacles that are covered with stinging nematocysts, which aid in the collection of food and in defense. The mouth is located in the center of the oral disc and leads to the interior gastrovascular cavity where the organs of digestion, absorption, and excretion lie. Within the polyp are three layers of tissue: ectoderm, mesogloea, and endoderm. Parts of the ectoderm that are in direct contact with the skeleton are known as the calcioblast and are responsible for secretion of the aragonite. Within the endoderm of most hermatypic (reef-building) corals are found symbiotic dinoflagellatealgae, called zooxanthellae. Scleractinians reproduce both sexually and asexually. In most species, ova and sperm are located in the same polyp. Fertilization takes place inside the gastrovascular cavity, and the larvae are ejected through the mouth. The young larvae or planulae settle on a hard, solid surface to form a coral colony. A few days after fixation, the first complete set of mesenteries
Abbreviations: ENSO, El Niño-Southern Oscillation; SST, sea surface temperature(s); DIC, dissolved inorganic carbon.
GEOCHEMISTRY OF CORALS: PROXIES OF PAST OCEAN CHEMISTRY, OCEAN CIRCULATION, AND CLIMATE
8355
and septa are formed. Also, asexual reproduction takes place in an existing coral colony, thereby rendering the distance between coral polyps roughly constant as the surface area of the colony increases.
FIG. 1. Anatomical structure of a coral polyp and its association with the skeleton (redrawn from ref. 7). The microstructure of a coral skeleton reveals an intricate matrix that is reflective of the coral's habits. The skeleton consists of minute, orthorhombic crystalline needles of aragonite, about 2 µ in diameter. When a planulae first settles to form a coral colony, it accretes a horizontal layer of aragonite called the basal disk (Fig. 1). As the polyp grows upward, the margin of the base turns upward, forming a cup, called the epitheca. The epitheca is believed to contain daily growth bands (8) that are formed as a result of daily changes in the shape of the calcioblastic layer (9). Also extending from this basal disk are vertical sheets of septa that support and separate the mesenteries in the polyps (Fig. 1). Each septum consists of numerous crystal clusters called sclerodermites. A sclerodermite is made from fibrous aragonitic needles radiating from a single calcification center believed to be an amino acid. Dissepiments are horizontal layers of aragonite that form both inside and outside the corallite. They form skeletal boundaries found at the bottom of the coral polyp and are left behind as the result of upward growth of the polyp. Massive hermatypic corals are more desirable than the branching varieties as the morphology to use for paleoreconstructions. First, massive corals form round, wave-resistant structures that can include hundreds of years of uninterrupted growth. Second, the accretion rate of calcium carbonate is much higher for hermatypic corals that contain symbiotic zooxanthellea than for deep species. There is a considerable amount of evidence that links zooxanthellar activity with calcification rate (10). The fundamental significance of the association between zooxanthellea and coral host has long been controversial as there are several mechanisms through which scleractinians might benefit from the presence of the algae, including translocation of photosynthate-derived carbon to the coral polyp (11, 12). It is important to note that some authors have observed high-Mg calcite in coral skeletons. MacIntyre and Towe (13) examined Porites lobata from an Oahu reef and concluded that the calcite was contained within the tubular organic network formed by microborers (i.e., fungus or algae). Thus, it appears that the presence of calcite in coral skeletons indicates accretion from noncoral sources and reflects contamination. Methods available to detect calcite are x-ray diffractometry (≥1% calcite can be detected) and visual inspection under sunlight (calcite twinkles, whereas coralline aragonite is a dull off-white). Density Banding in Corals. Ma (14) suggested that groupings of constricted epithecae reflected seasonal variations in the growth rate of coral skeletons. Later, Wells (8) studied middle Devonian fossil corals and cautiously interpreted fine ridges on the surface of the coral epitheca (see Fig. 1) to be daily growth bands. He found about 400 ridges (days) per annum, which agreed with astronomical expectations of the deceleration of the Earth's period of rotation since the Devonian. The most significant record contained in most surface coral skeletons are annual density bands. They are primary skeletal characteristics that consist of a high and low density portion per year discernible by x-ray of a thin slab cut along the axis of upward corallite growth (15, 16). Annual variations in density represent changes in both the rate of linear skeletal extension and calcification. These growth bands were first conclusively demonstrated as annual by Knutson et al. (17). They found excellent agreement between autoradiographs of Enewetak coral, which depict nuclear bomb products incorporated during nuclear blasts (mostly 90Sr), and xradiographs, which reveal density structure. At a few locations, multiple banding within an annual growth period has been identified, though this is not common (18, 19). Also, Buddemeier and Kinzie (18) have identified lunar bands in P. lobata from Kahe Point, Oahu, Hawaii. Using x-radiography and scanning densitometry of Montastrea annularis from St. Croix, U.S. Virgin Islands, Dodge and Brass (20) found that skeletal extension (linear growth) is negatively correlated with density and positively correlated with calcification (mass addition). Hudson et al. (21) used alizarin-staining techniques and x-radiography to observe annual density bands in M. annularis from the Florida Straits and the Gulf of Honduras. They noticed that dense aragonitic skeleton accreted during the warm summer months of July through September and that thicker, less dense bands accreted during the cooler months of October through June. At some locations, such as the Gulf of Eilat in the Red Sea, corals (Porites lobata) accrete high density bands during the season of cold sea surface temperatures (SST). In Florida, thicker, dense layers, known as stress bands, also formed during the unusually severe winters of late 1969, 1963, 1957, 1941, 1898, 1894, 1885, and 1856. These cold winters or “cold fronts” are the result of weather phenomena that originate in the Northwest and pass over the reef in a southeastern direction. This correlation between stress bands and recorded cold fronts is proof that the density bands in these corals are indeed annual in nature. Glynn and Wellington (22) studied Pavona clavus colonies from the Galapagos Islands and revealed that dense bands form during the warm water months of January through March and thicker, less dense bands form during the cold, upwelling season, usually April through December. These workers attributed increased growth rates during certain years as the result of ENSO events. These events are linked to the intensity of the Southern Oscillation Index, which is the difference in air pressure between Easter Island and Darwin, Australia (23). When this index is low, the Peru Current decreases its flow and upwelling along the western coast of South America is restricted to below the sea surface for a period of 4 to 10 months. Simultaneously, a thin skin of warm, low-nutrient tropical surface water moves into the Galapagos region from the West and North. Just as stress bands in M. annularis from the Florida Straits depict recorded cold fronts, unusually high growth rates during El Niño years provide markers of known years for the Galapagos corals. Severe ENSO events, such as that during 1982–83, caused interruption of normal growth of Galapagos corals due to the partial or complete loss of their zooxanthellae or bleaching (22). In areas with smaller temperature changes during this ENSO event, such as Costa Rica, the corals were not bleached and hence contained a complete record of the 1982–83 ENSO (24). Just as dendrochronologists cross date wood samples to extend tree-ring chronologies (25), so can sclerochronology be used to obtain lengthened records of coral growth. Stress bands, like those observed in the Florida Straits coral, could be used as markers of known years in dead colonies. Though annual growth rates of some massive hermatypic corals show
GEOCHEMISTRY OF CORALS: PROXIES OF PAST OCEAN CHEMISTRY, OCEAN CIRCULATION, AND CLIMATE
8356
no systematic change for a given colony during its lifetime (26, 27), intercolony variations in annual growth rate have been observed (22, 27). Therefore, stress band matching would prove more useful for sclerochronologists than direct growth rate comparisons among coral skeletons. A very fruitful method of long time-series reconstruction of coral records over the past glacial cycle involves measurement of the ingrowth of 230Th (28) (see section titled “Dating of Corals”). Factors Affecting Growth. Optimal coral growth occurs at SST that range from 20° to 26°C (29, 30). The limits of coral tolerance range from a low of about 16°–18°C to a high of 30°–32°C, depending on the species under consideration. Thus, coral growth is limited to the temperate and tropical latitudes between about 35°N and 32°S. For massive corals, skeletal growth rate is dependent on SST (31, 32). The influence of light as an environmental factor has been extensively studied but is still not well understood. The rate of growth of M. annularis was positively correlated with the number of sun hours per day (33) and the water depth of the corals (31). Calcification rates during daylight periods are 3- to 15-times higher than those during dark periods (34). Goreau and Goreau (35) showed that coral with zooxanthellae calcified about 19-times faster than those that had lost their algae. These zooxanthellae-free corals calcified even less in the light than did the corals with zooxanthellae in the dark. This illustrates that photosynthesis is not the only role that the zooxanthellae play as a symbiont. Buddemeier (27) suggested a negative correlation between skeletal density and available light. Many other environmental factors affect the growth of corals. For instance, salinity must be between 25‰ and 40‰ to support healthy corals (36). The depth in which corals live also affects their growth. For example, M. annularis shows an optimal growth rate at 10 m depth, but is only 10% of this rate at 30 m depth (37). Sedimentation has long been known as a detriment to the well-being of corals (38). Large amounts of sediment covering the colony may interfere with feeding processes and zooxanthellar photosynthesis. There is also an energy loss by the polyps due to increased mucous production used for shedding the sediment. In summary, optimal growth conditions for hermatypic corals include temperatures within the range of 20°–26°C, low sediment load, growth at 5–15 m depth, normal ocean salinities, and relatively protected water. These conditions yield an optimum vertical growth rate of 10–15 mm/yr for massive species (39). Dating of Corals. Dating of fossil corals has been done by measuring concentrations of radioisotopes incorporated within the skeletal matrix at the time of accretion (i.e., 14C, 228Ra, 210Pb) and those that have grown in since accretion (i.e., 230Th, 231Pa). For example, 14C dating of coralline aragonite from Enewetak and Oahu has revealed that subtle changes in eustatic sea level have occurred during the last 6,000 years in the North Pacific Ocean (18, 40). Moore and Krishnaswami (41) revealed that 228Ra (t1/2 = 6.7 yr), which is incorporated in corals at about the same ratio to calcium as that in seawater, can be used to date corals less than 30 years of age. Lead-210 (t1/2 = 22.7 yr) dating in corals has been shown to agree within 10% with annual banding chronologies of corals from the North Atlantic over the past 100 years of growth (26, 41, 42). Several bomb-produced radionuclides have also aided in the determination of coral growth rates. Bomb 14C is detected above natural levels in all surface coral bands younger than about 1957 (19, 43, 44, 45 and 46). Other products of thermonuclear weapons testing (e.g., 90Sr and 239,240Pu) are found in post-1950 corals and reveal past levels of these isotopes in the surrounding seawater (47, 48, 49 and 50). Measurements of 234U and its in-growth daughter, 230Th, in corals from the Florida Keys and the Bahamas revealed that sea level was close to its present interglacial level at about 85,000, 130,000, and 190,000 years ago (51). A second in-growth method was presented by Ku (52), who measured 231Pa in corals from Barbados; he found good agreement between the ages of corals from the last interglacial determined from the in-growth of 231Pa and those determined using the 230Th in-growth method. A most exciting discovery in the field of coral geochemistry was made by Edwards et al. (28). They developed techniques for measuring 230Th in corals by using isotope dilution thermal ionization mass spectrometry (TIMS). This new method greatly reduced both the sample size (from tens of grams to 250 mg of coral) and the uncertainty in the age determination (by a factor of 10–20) compared with the previously used alphaspectrometry methods. It is now possible to date corals with ages of 30, 12,000, and 123,000 years with uncertainties of ±3, ±60, and ±1,000 years [two sigma errors, respectively (28, 53)]. This discovery made it possible to more accurately date the last interglacial period (28), the Younger Dryas, and the levels of 14C in the surface ocean dissolved inorganic carbon during the last glacial maximum (53, 54). Carbon and Oxygen Isotopes in Coral Skeletons. Most of the records obtained from corals have been of stable oxygen (δ18O) and carbon (δ13C) isotopes in annual and seasonal bands. The majority of the carbon found in coral skeletons originates as dissolved inorganic carbon (DIC) in sea water. Goreau (34) proposed the following as the dominant reaction that occurs during the accretion of calcium carbonate by corals: Ca2+ + 2 HCO−3 → H2O + CaCO3 + CO2. McConnaughey (55) refined this by suggesting intermediate reactions involving dehydration of the HCO−3 (2HCO−3 → 2H2O + 2CO2), then the subsequent rehydration and disassociation of the hydrogen ions to form CO =3, which is used in the precipitation of CaCO3. It is believed that zooxanthellae increase CaCO 3 accretion by using the CO2 produced by this reaction during photosynthesis. This is one of the reasons corals with symbiotic algae accrete calcium carbonate at such rapid rates. Weber (56) found that the δ13C (per thousand deviation of 13C/12C ratio with respect to that of PDB-1 standard) of skeletal carbonate ranged from 1‰ to −1‰, whereas the δ13C of the DIC in the surrounding seawater was 3‰ to 1‰. Earlier, the lower δ13C values were attributed in part to the incorporation of a small amount of metabolic carbon from the coral's diet of zooplankton (δ13C = −18‰; ref. 56) to the carbonate matrix (58). Recent interpretation shows that the reasons for carbon and oxygen isotope disequilibria in corals with respect to those in seawater are twofold (55, 59). First, kinetic isotopic effects during the hydration and hydroxylation of CO2 cause the simultaneous depletions of 18O and 13C with respect to their lighter isotopes (16O and 12C) by as much as 4–6‰ and 10–15‰, respectively. There is an approximately linear correlation between δ18O and δ13C signals in corals (59, 60 and 61) and in a pure calcite gorgonian (62), due to the degree of kinetic disequilibria that exists in the calcioblastic layer. Second, alteration of the skeletal δ13C signal is caused mainly by changes of the δ13C in the internal DIC pool within the calcioblastic layer, which are caused by varying rates of photosynthesis and respiration by the coral polyp and its symbiotic algae. During photosynthesis, selective removal of 12CO2 by the zooxanthellae during photosynthesis causes an increase of δ13C in the DIC left behind to form the accreted aragonite. Likewise, during respiration of organic matter, addition of 12C-enriched CO2 decreases the δ13C of the skeletal carbonate (63). These metabolic effects are responsible for large shifts of skeletal δ13C resulting primarily from the seasonally variant ambient light incident on the coral surface (64). McConnaughey (59) showed that nonseasonal changes of δ13C can be minimized by sampling along major growth axes (i.e., tops of coral heads). The periodicity of subannual δ13C and high density bands in three Atlantic corals varied as a function of seasonal cloud cover (65). The δ18O patterns in the same
GEOCHEMISTRY OF CORALS: PROXIES OF PAST OCEAN CHEMISTRY, OCEAN CIRCULATION, AND CLIMATE
8357
corals, however, were a function of both SST and salinity in the seawater measured at each site. δ13C of carbonate is only slightly affected by ambient temperature (0.035‰/°C). Epstein and Mayeda (66) first showed that δ18O in mollusk aragonite varied inversely with SST. It was subsequently demonstrated that the δ18O of coralline aragonite also varied inversely with SST though different genera were offset by constant amounts from this δ18O vs. SST relationship (67, 68). Offsets demonstrate that precipitation occurs faster than the establishment of oxygen isotope equilibrium between CaCO3 and H2O in the calcioblastic layer. Several authors have successfully used δ18O as a monitor of past SST (5, 19, 65, 69, 70). The slope of SST vs. δ18O is similar to the mollusk results (66), a 0.22‰ rise in δ18O per 1°C decrease (Fig. 2). The other important influence in coral δ18O is the isotopic signature of seawater, which is related to the salinity in the surface water through evaporation and precipitation cycles. Several investigators have demonstrated that subannual δ18O measurements from surface corals are a predictable function of SST and salinity (65, 70, 71 and 72). Radiocarbon in modern corals was measured first by Knutson and Buddemeier (73), who detected bomb levels of 14C in the outer 10 cm of a massive coral from Fanning Island. They hypothesized that corals hold records of the input of bomb-produced radionuclides to the surface waters of the ocean. Independently, two groups reconstructed time histories of ∆14C in two Atlantic corals. Nozaki et al. (46) reported variations in ∆14C and in stable isotopes (δ13C and δ18O) from a 200-year Dipoloria strigosa from North Rock, Bermuda. Druffel and Linick (43) observed changes of ∆14C from a 360-year core of M. annularis from the Florida Keys. These and other papers have shown that ∆14C of corals closely matches the ∆14C of DIC in surrounding seawater (44, 45, 74, 75). This isotope record is not affected by isotope fractionation, as all ∆14C measurements are normalized to a δ13C value of −25‰ by first measuring the δ13C value (76) in the coral. Radiocarbon in corals has been used to reconstruct time histories of climate and circulation (see first two sections of part 2 of this paper).
FIG. 2. Correlation between annual range in historical sea surface temperature (SST) records (∆T°C) and annual range in δ18O values (∆δ18O) in coral bands from the Galapagos Islands, Fanning Island, and Canton Island. The line shown represents a slope of 0.22‰ per 1°C decrease in SST predicted by the temperature-dependent aragonite-water fractionation factor. This slope is the same as that defined for the equilibrium precipitation of calcite (66). (Data are from ref. 69.) Minor and Trace Element Composition of Corals. Divalent cations that are abundant in seawater (i.e., Mg, Sr, and Ba) appear as minor elements in corals (1 µmol/mol Ca), and other less-abundant divalent cations (M) are in trace abundances (<1 nmol/mol Ca–1 µmol/mol Ca). Most minor and trace element concentrations in corals are incorporated independent of bulk tissue concentrations (77) and are proportional to ambient levels in the surrounding seawater. Thus the M/Ca ratio in the skeleton can be used to reconstruct past concentrations of M in seawater, assuming Ca levels have remained unchanged in seawater. Shen (4) reports that the distribution coefficients based on the measured divalent M/Ca ratios in corals and estimated seawater concentrations of the respective metals vary only within an order of magnitude. This indicates kinetic control of the M2+ incorporation, as a wider range is expected from a thermodynamic control for this process (78). Using flame atomic absorption spectrophotometry (AAS), Schneider and Smith (79) measured Sr/Ca ratios in Hawaiian corals and found that this ratio varied as a function of SST, though there were problems with corals that had undergone diagenesis. Subsequent work using TIMS methods has shown that there are corals whose Sr/Ca records track SST with an apparent accuracy of better than ±0.5°C (80). DeVilliars et al. (81) report that both variable extension rate and biogenic cycling of Sr vs. Ca can complicate some Sr/Ca records in serving as simple sclerothermometers. The U/Ca ratio in corals is not a simple function of the U/Ca ratio in seawater. Flor and Moore (82) suggested that decaying organic matter in skeletons of dead corals may release U that exchanges with the carbonate matrix. Shen and Dunbar (83) report a correlation between U/Ca and δ18O in corals from the Galapagos Islands, although they cannot rule out additional control of U/Ca by salinity and carbonate ion content of surface waters. Work by Shen et al. (84) using graphite furnace AAS revealed that Cd/Ca measurements in Bermudian corals uncovered records of historical upwelling and industrial fallout. Only after exhaustive oxidative and reductive cleaning procedures was it possible to obtain the lattice-bound Cd concentrations in the corals, separated from the organic and detrital phases that were much more enriched. These phases contributed to the much higher concentrations of trace metals reported in corals by previous investigators (85, 86). Likewise, Ba/Ca ratios in a Galapagos coral revealed historical changes of the nutrient content in surface waters surrounding the Galapagos Islands (87). Other investigators have reported records in corals for a variety of other minor and trace elements, including Pb, V, Zn (42), Cu (88), Mn (88, 89 and 90), Mg (91, 92), and rare earth elements (93). An in-depth review of minor and trace element records from corals is presented by Shen (4). The use of these records as climate recorders and pollution monitors appears later in the paper (see first and third sections of part 2). Phosphate has been shown to act as a crystal poison of calcification (94). Dodge et al. (95) showed that concentrations of P in corals from three sites in the North Atlantic were consistent with the history of sewage and other pollution in the surrounding reef. Rasmussen (96) demonstrated that high levels of phosphate in seawater were inversely proportional to concentrations of Sr in Porites from the Great Barrier Reef. She further showed that morphological alteration of the coral skeleton is associated with high phosphate levels in the surrounding seawater that came from agricultural fertilizers. Corals as Proxies of the Carbon Cycle and Climate Change Synthesis of the information obtained from coral proxy records is done in the second part of this paper. Three questions are
GEOCHEMISTRY OF CORALS: PROXIES OF PAST OCEAN CHEMISTRY, OCEAN CIRCULATION, AND CLIMATE
8358
addressed about the cycling of excess carbon in the oceans and changes in climate over the past several hundred years. (i) What are the regional changes in climate within the Atlantic and Pacific Ocean basins? (ii) How have coral parameters revealed variability of climate and ventilation in the North Atlantic and South Pacific? and (iii) What evidence exists of anthropogenic CO2 (fossil fuel and biogenic CO2) in the ocean? Regional and Global Changes in Climate. Changes of climate on Earth have been recorded in corals for millions of years, though relatively few of the clues have been decoded. These climate records are revealed through physical, chemical, and isotopic imprints in corals. There are several examples in coral records of how changes in regional wind patterns, rainfall, and occurrences of ENSO demonstrate the existence of variability in climate within a given ocean basin. Early on, Dodge and Vaisnys (97) established correlations between the linear growth rate of Bermuda corals and recorded air temperature and air pressure variations, demonstrating that paleoclimatic records were recoverable from corals. Also, from growth band analysis of recent and fossil corals from Barbados, Dodge et al. (98) observed both average band width and its variability were lower in corals from 105,000 years ago (kya) than those from recent times or from 85 kya and 125 kya. Despite this link, they were unable to identify a specific climatic variable as the dominant cause of the differences. Later, Shen et al. (99) used multiple tracers for extracting climate records from corals. In a 47-year record (quarter-annual sampled) of Pavona clavus from the Galapagos Islands, highly significant linear correlations were found between four of the tracers examined (δ18O, δ13C, Ba/Ca, Cd/Ca) and Peruvian coast SST. These correlations result from seasonal upwelling cycles and periodic interruptions of these cycles caused by changes in climate (i.e., ENSO). In a striking example of past climate reconstruction in the tropical western Pacific, Cole et al. (71) measured δ18O in monthly samples from a 96-year Tarawa Atoll coral and found that δ18O reflected seawater salinity changes induced by regional variations in rainfall. They determined that ENSO extremes in the central and western tropical Pacific were occasionally decoupled from those observed in coastal South America. From spectral analysis of the δ18O data, the distribution of variance had shifted among annual and interannual periods between about 1930 and 1950, in concert with an observed shift in the strength of the Southern Oscillation. This is an example of regional changes in climate within an ocean basin that can be reconstructed from coral skeletons. Periodic wind bursts associated with ENSO are also recorded in western Pacific coral. Historic Mn/Ca records from a Tarawa Atoll coral revealed high ratios during the ENSO years of 1965, 1972, and 1976 (100). They concluded that these high Mn periods were due to remobilized Mn from lagoon sediments caused by vigorous westerly wind bursts that happened along the equator during ENSO events. A new tracer of past ENSO events, controlled by the climatic forcing of the tropical winds, was uncovered by this novel study. Large decadal changes in ventilation rate that were induced by changes in climatic parameters were uncovered by geochemical tracer records (101, 102). From post-bomb records of high precision ∆14C in Bermuda and Florida corals, Druffel (102) used an inverse model to calculate the water mass renewal rate (WMRR) with respect to CO2 ventilation in the Sargasso Sea. Results showed that the WMRR in the Sargasso Sea was high in the early 1960s, decreased in the late 1960s, and remained low throughout the 1970s. This factor-of-three change in the ventilation of the upper North Atlantic reflects the long-term change of climate within an ocean basin. Climate change has been demonstrated at other locations of the world's oceans. Chakraborty and Ramesh (103) demonstrated that Indian summer monsoon rainfall records can be reconstructed from δ18O measured in corals from the Lakshadweep archipelago in the Arabian Sea. Isdale and others (104, 105) reported that fluorescent bands in coral skeletons are coincident with periods of high runoff from local rivers feeding into the Great Barrier Reef. Humic materials from terrestrial plants are a major contributor to this fluorescence. This demonstrates a significant correlation between ancient rainfall in Australia and a tracer easily quantified in coastal reefs. Dunbar et al. (106) reported the variability of δ18O measured for the past four centuries from Pavona clavus from Urvina Bay in the Galapagos Islands. They found cooler temperatures during the early 1600s and early 1800s, which is in agreement with many North American tree ring records. They suggest that major shifts in the dominant ENSO modes within their data throughout time indicate reorganizations of the tropical Pacific climate system. A compilation of the δ18O data from several coral series in the tropical Pacific region reveals a striking trend (107). Although it is not clear-cut, there is a general decrease of δ18O values from the latter part of the last century to the present in nearly all of the coral records available. This appears to be evidence of large-scale warming in the upper layers of the ocean, or a general freshening of these waters, or a combination of these. Is this evidence of global warming as related to anthropogenic inputs of CO2? It is a tantalizing set of records, nonetheless, and decoupling of the SST and water mass components of this trend has been set as a high priority for future research. Temporal Changes in Ocean Circulation. How have coral parameters revealed variability of climate and ventilation in the North Atlantic and South Pacific? First, there are several modes of circulation in the ocean. Vertical mixing processes include upwelling and winter mixing, and horizontal processes encompass upper ocean currents (i.e., Gulf Stream) and eddies. Some of the best tracers of vertical and horizontal mixing processes are those that record nutrient levels (i.e., Cd/Ca ratios), SST (i.e., δ18O and Sr/Ca), and relative degrees of upwelling (i.e., ∆14C). The available data sets show that temporal variability within the oceans on interannual-to-decade time scales is prevalent during the past few hundred years. A useful paleocirculation indicator is δ18O in corals, as it records past SST changes in regions that have little or no changes in salinity. From seasonal δ18O measurements in numerous tropical Pacific corals, records of past ENSO events were reconstructed and compared with historical records from the South American coast (69, 106). Century time-scale changes in mixing have been observed in the North Atlantic. From an 800-year-old Bermuda coral, Patzold and Wefer (108) revealed a decrease in the summer SST values during the Little Ice Age. They also found lower overall growth rate during this period, suggesting a relation between coral growth rate and nutrient availability during periods of cooler climate in the North Atlantic. Radiocarbon in oceanic DIC is controlled mainly by changes in ocean circulation rather than by atmospheric exchange of CO2. There is a large reservoir of DIC in the surface ocean, and it takes a long time (10 years) for the 14CO2 in the atmosphere to turn over with respect to exchange with the surface ocean. An excellent example of the use of 14C as a circulation tracer is the study by Toggweiler et al. (45). They reconstructed the pre-bomb ∆14C values at numerous surface locations in the tropical Pacific using corals and concluded that low ∆14C waters off Peru had their origin as 8°C water in the subantarctic region of the southwestern Pacific. Using comparisons between a model simulation of radiocarbon (109) and coral ∆14C distributions, they suggested that diapycnal alteration of successively less dense features of the South Pacific
GEOCHEMISTRY OF CORALS: PROXIES OF PAST OCEAN CHEMISTRY, OCEAN CIRCULATION, AND CLIMATE
8359
thermocline is responsible for upper ocean ventilation in the southern and tropical Pacific. Thus, Toggweiler et al. (45) showed that the circulation of subsurface waters in a major ocean basin was not obvious when using traditional conservative properties (temperature and salinity) and that the lagrangian tracer 14C was needed for the patterns and sources to become clear. Temporal variations in circulation were revealed using high precision (±2–3‰) ∆14C measurements in a 356-year-old coral core from the southern Great Barrier Reef (110). This study revealed increased decade time-scale variability of ∆14C between A.D. 1680 and 1730 (Fig. 3), coincident with the Maunder minimum in solar activity. Changes in vertical mixing and changes in the transport of the low-14C, South Equatorial Current source waters to the western Coral Sea region are likely processes that could account for this variability. These data also suggest century time-scale changes in the periodicity of ENSO as it is manifest in the southwestern Pacific, and this change correlates with an apparent reduction in ventilation of upper waters by fossil fuel CO2 during the first half of the 20th century (110). The main conclusion from these records is that the amplitude of interannual and decadal variability complicates the task of assessing long-term anthropogenic trends. This makes it a challenge, though still a possibility, to decouple the natural variability from the anthropogenic CO2 signal in the oceans. Evidence of Anthropogenic CO2 in the Global Ocean. Corals are diaries that record within their pages many types of environmental information, including nuclear fallout products and excess CO2 input to the ocean. What evidence exists of anthropogenic CO2 (excess CO2 from fossil fuel and biomass burning) in the ocean today?
FIG. 3. Biannual Australian coral n∆14C values (in ‰, detrended with respect to a third-order polynomial fit) plotted with incidences of strong and very strong ENSO events before A.D. 1800 (a) and including medium ENSO events after A.D. 1800 (b) as reported by Quinn et al. (111). Solid peaks represent correlation between ENSO events and coincident (±2–4 years) n∆14C lows, and open areas indicate the absence of this correlation. (Data are from ref. 110). Records of aeolian input of heavy metal pollution to the oceans was presented by Shen et al. (84). They measured Cd/Ca in a Bermuda coral and found that it revealed a record of aeolian fluxes of industrial Cd to the western North Atlantic from the continental United States via the westerlies. Peaks of high Pb/Ca were also observed in the same coral and resulted from the use of tetraethyl lead in gasoline for cars (42). The initial rise in the Cd from about 1900 to 1925 was coincident with a rise in zinc production (from which Cd is a byproduct) that occurred at a time when flue dust recovery of Cd was negligible. The second rise of Cd from 1942 to the present was due both to Cd flue dust that still escaped during Zn ore production, as well as to other smelting exhausts (84). Records of both global and localized inputs of nuclear fallout to the oceans are available from corals. Toggweiler and Trumbore (49) measured 90Sr in corals from the Pacific and Indian Oceans and found that localized inputs of 90Sr to the North Pacific atolls between 1952 and 1958 were higher than previously had been believed. They also concluded that major surface currents through the Indonesian archipelago were responsible for the transport of large amounts of 90Sr, and thus water, from the Pacific to the Indian. Finally, this study revealed that there was a major transport of water from temperate to tropical regions of the Pacific on an interannual time scale. Records of 239,240Pu have also been extracted from Caribbean island corals to show a close correlation with the fallout history of 90Sr (50). Fossil fuel CO2 has been produced since the middle of the 19th century, and its invasion from the atmosphere into the ocean has been recorded by corals and sclerosponges. This excess CO2 is unique in its carbon isotopic signatures, owing to the absence of 14C (∆14C = −1,000‰) and low δ13C signature (−28‰), in comparison with those in atmospheric CO2 (0‰ and <−7‰, respectively). Efforts to extract the fossil fuel CO2 signal from the δ13C signature in corals has yielded mixed results owing to the alteration of the DIC δ13C signature by kinetic isotopic and metabolic effects. This δ13C record is preserved without fractionation in sclerosponge skeletons, as these animals accrete aragonite in equilibrium with the surrounding seawater DIC. Using the sclerosponge Ceratoporella nicholsoni from Jamaica, Druffel and Benavides (112) measured a δ13C decrease of 0.5‰ from 1850 to 1972 and attributed this to fossil fuel CO2 invasion into surface waters. They deconvolved this time history to reveal that the input of an additional source of excess CO2, the reduction in the size of the terrestrial biosphere due to biomass burning, peaked at the turn of the 20th century. Reduction of the 14C/12C ratio in atmospheric CO2 since 1850, known as the Suess effect, is due mainly to the input of 14C-free CO2 from fossil fuel burning into the atmosphere and ocean (113). The Suess effect in the surface ocean has been measured in corals from the Atlantic and Pacific Oceans and varies regionally from −3‰ to −12‰ (43, 74, 110, 114). However, decade time-scale changes in mixing of the upper oceans have caused the Suess effect to vary temporally and spatially, much more so than if ocean mixing had remained in steady state. Large changes in ∆14C of 10–20‰ on decade time scales mask the Suess effect at a number of locations in the southwest Pacific Ocean and Sargasso Sea (110, 114). More work needs to be done at other locations in the surface world's oceans to decouple the Suess effect from decade time-scale changes in circulation. At present, the magnitude of the anthropogenic CO2 signal in surface waters is difficult to ascertain in surface corals. Changes in mixing parameters that occur at some sites on interannual-to-decade time scales compete with the anthropogenic signal, making it difficult to deconvolve the two. This
GEOCHEMISTRY OF CORALS: PROXIES OF PAST OCEAN CHEMISTRY, OCEAN CIRCULATION, AND CLIMATE
8360
problem also arises in studies of the concentration of DIC in the upper ocean. The δ13C signal in seawater DIC (115) and as recorded in sclerosponges remains as a promising method for reconstructing the time history of the anthropogenic CO2 signal into the surface ocean. To decouple the natural and anthropogenic effects on the carbon cycle, a more intensive effort has to be made to define natural variability in the upper ocean. We must also identify other accurate integrators of the anthropogenic CO2 signal. Future Research Directions. Reconstruction of past SST changes in the world's oceans during the period of rising atmospheric pCO2 is a top priority. A lowering of δ18O in numerous coral records over the past 150 years may be evidence of a general rise of SST throughout the temperate and tropical surface oceans (107). Measurements of high precision Sr/Ca ratios will help resolve this important issue. Also, intense sampling and analyses of coral along a transect known for its sensitive response to climate change will add significantly to our knowledge of the global ocean's response to climate change. Such an area is a meridional section through the warm pool in the western tropical Pacific. The reconstruction of multiple tracer records from the same corals are essential to fully glean all of the available information regarding past climate and circulation change in the oceans. Measurements of nutrient-like elements (i.e., Ba/Ca), 14C, and stable isotopes on monthly time scales will enhance pursuits of the timing of past changes in water transport and climate. This approach can be extended to the last deglaciation, where the question of the timing of glacial melting and the onset of deep water formation is still not well known. A recorder of salinity changes in ocean water contained within the skeletons of corals should be pursued with vigor. One of the most exciting and potentially fruitful areas of coral geochemical research is the study of deep-sea species. The primary carbon source to several species of deep calcareous corals is from the DIC in the surrounding seawater (116). The life spans of single specimens have been shown to be several hundred years (62, 117) to more than 1,000 years (118). It is now possible to reconstruct time histories of tracers in deep water masses, and thus extend our view to three dimensions within the world's oceans. Records of deep water stable isotopes (δ18O and δ13C) (117), trace metals (119), and bomb fallout products (62) are starting to emerge in the literature. We need to understand the processes controlling isotopic fractionation of the CaCO3, as well as high-resolution age dating of the fine structure within the skeleton of deep species using the highly precise chronometer 230Th. Subsequently, we can reconstruct time histories of nutrient distributions and ventilation of the deep ocean basins (120). Reconstruction of past dissolved organic carbon (DOC) concentrations in seawater is very important for understanding the changes in the carbon cycle over the past 100 years, as well as the past glacial–interglacial cycle. Whether DOC concentrations are recorded within corals remains to be demonstrated. Tritium, a bomb product used as a tracer of ocean circulation, could be reconstructed from the organic matter within corals. Accelerator mass spectrometry measurements of a variety of radionuclides in corals (e.g., 10Be and 7Be) could help to understand particle cycling in seawater. Individual organic compounds would be of interest as indicators of primary production or pollution in coastal locales. It is likely that these records, and many more, are hidden within the pages of the coral diary. I dedicate this article to Roger Revelle. I fondly remember his comments and encouragement in that melodic, deep voice. I thank Jess Adkins and an anonymous reviewer for their helpful reviews of the manuscript, Sheila Griffin for assisting with drawings and preparation, Andrea Grottoli-Everett for advice, and Ed Urban and Karl Turekian for editorial treatment. This work was funded by the Chemical Oceanography Division of the National Science Foundation through Grant OCE-9314691. 1. Suess, H. E. (1980) Radiocarbon 22(2), 200–209. 2. Dansgaard, W., Clausen, H. B., Gundestrup, N., Hammer, C. U., Johnsen, S. F., Kristinsdottir, P. M. & Reeh, N. (1982) Science 218, 1273–1277. 3. Neftel, A., Moor, E. & Stauffer, B. (1985) Nature (London) 315, 45–47. 4. Shen, G. T. (1993) Bull. Inst. Fr. Etud. Andines 22(1), 125–158. 5. Druffel, E. R. M., Dunbar, R., Wellington, G. & Minnis, S. (1990) in Global Ecological Consequences of the 1982/83 El Niño Southern-Oscillation, ed. Glynn, P. (Elsevier, New York), pp. 233–253. 6. Stanley, G. (1981) Geology 9, 507–511. 7. Wells, J. (1956) in Treatise on Invertebrate Paleontology, ed. Moore, R. (Geological Society of America and Univ. of Kansas Press, Lawrence), pp. 328–444. 8. Wells, J. W. (1963) Nature (London) 197, 948–950. 9. Barnes, D. (1972) Proc. R. Soc. Lond. B 182, 331–351. 10. Goreau, T. F. (1959) Ecology 40, 67–90. 11. Muscatine, L. & Cernichiari, E. (1969) Biol. Bull. (Woods Hole, Mass.) 137, 506–523. 12. Smith, D., Muscatine, L. & Lewis, D. (1969) Biol. Rev. 44, 17–90. 13. MacIntyre, I. & Towe, K. (1976) Science 193, 701–702. 14. Ma, T. (1934) Proc. Imp. Acad. (Tokyo) 10, 353–356. 15. Buddemeier, R., Maragos, J. & Knutson, D. (1974) J. Exp. Mar. Biol. Ecol. 14, 179–200. 16. MacIntyre, I. & Smith, S. (1974) Proc. Int. Symp. Coral Reefs, 2nd. (Great Barrier Reef Committee, Brisbane, Australia), pp. 277–287. 17. Knutson, D. W., Buddemeier, R. W. & Smith, S. V. (1972) Science 177, 270–272. 18. Buddemeier, R. & Kinzie, R. (1975) in Growth Rhythms and the History of the Earth's Rotation, eds. Rosenberg, G. D. & Runcorn, S. K. (Wiley, London), pp. 135–147. 19. Druffel, E. R. M. (1987) J. Mar. Chem. 45, 667–698. 20. Dodge, R. & Brass, G. (1984) Bull. Mar. Sci. 34(2), 288–307. 21. Hudson, J., Shinn, E. A., Halley, R. B. & Lidz, B. (1976) Geology 4, 361–364. 22. Glynn, P. & Wellington, J. (1990) in Global Ecological Consequences of the 1982/83 El Niño Southern-Oscillation, ed. Glynn, P. (Elsevier, New York). 23. Wyrtki, K., Stroup, E., Patzert, W., Williams, R. & Quinn, W. (1976) Science 191, 343–346. 24. Carriquiry, J., Risk, M. & Schwarcz, H. (1988) Palaios 3, 359–364. 25. Fergusson, C. (1968) Science 159, 839–846. 26. Dodge, R. & Thomson, J. (1974) Earth Planet. Sci. Lett. 23, 313–322. 27. Buddemeier, R. (1974) Proc. Int. Symp. Coral Reef 2nd. (Great Barrier Reef Committee, Brisbane, Australia), pp. 259–267. 28. Edwards, R. L., Chen, J. H. & Wasserburg, G. J. (1986/7) Earth Planet. Sci. Lett. 81, 175–192. 29. Stoddart, D. (1969) Biol. Rev. 44, 433–498. 30. Jokiel, P. L. & Coles, S. L. (1977) Mar. Biol. 43, 201–208. 31. Highsmith, R. (1979) J. Exp. Mar. Biol. Ecol. 37, 105–125. 32. Glynn, P. & Wellington, G. (1983) Corals and Coral Reefs of the Galaagos Islands (Univ. of California Press, Berkeley). 33. Bak, R. P. M. (1974) Proc. Int. Symp. Coral Reefs 2nd. (Great Barrier Reef Committee, Brisbane, Australia), pp. 229–233. 34. Goreau, T. F. (1963) Ann. N.Y. Acad. Sci. 109, 127–167. 35. Goreau, T. & Goreau, N. (1959) Biol. Bull. Mar. Biol. Lab. Woods Hole, Mass. 117, 239–250. 36. Coles, S. L. & Jokiel, P. L. (1978) Mar. Biol. 49, 187–195. 37. Baker, P. & Weber, J. (1975) Earth Planet. Sci. Lett. 27, 57–61. 38. Dodge, R., Aller, R. & Thompson, J. (1974) Nature (London) 247, 574–577. 39. Buddemeier, R. & Kinzie, R. (1976) Oceanogr. Mar. Biol. 14, 183–225. 40. Easton, W. & Olson, E. (1976) Geol. Soc. Am. Bull. 87, 711–719. 41. Moore, W. & Krishnaswami, S. (1972) Earth Planet. Sci. Lett. 15, 187–190. 42. Shen, G. T. & Boyle, E. A. (1988) Chem. Geol. 67, 47–62. 43. Druffel, E. M. & Linick, T. W. (1978) Geophys. Res. Lett. 5, 913–916. 44. Konishi, K., Tanaka, T. & Sakanoue, M. (1982) Proc. Int. Symp. Coral Reef 4th, ed. Gomez, E. D. (Marine Science Center, Univ. of the Philippines, Manila). 45. Toggweiler, J. R., Dixon, K. & Broecker, W. S. (1991) J. Geophys. Res. 96, 20, 467–20, 497.
GEOCHEMISTRY OF CORALS: PROXIES OF PAST OCEAN CHEMISTRY, OCEAN CIRCULATION, AND CLIMATE
8361
46. Nozaki, Y., Rye, D. M., Turekian, K. K. & Dodge, R. E. (1978) Geophys. Res. Lett. 5, 825–828. 47. Purdy, C., Druffel, E. R. M. & Livingston, H. (1989) Geochim. Cosmochim. Acta 53, 1401–1410. 48. Moore, W., Krishnaswami, S. & Bhat, S. (1973) Bull. Mar. Sci. 23(2), 157–176. 49. Toggweiler, J. R. & Trumbore, S. E. (1985) Earth Planet. Sci. Lett. 74, 306–314. 50. Benninger, L. & Dodge, R. (1986) Geochim. Cosmochim. Acta 50, 2785–2797. 51. Broecker, W. & Thurber, D. (1965) Science 149, 59–61. 52. Ku, T.-L. (1968) J. Geophys. Res. 73(6), 2271–2276. 53. Edwards, R. L., Beck, J. W., Burr, G., Donahue, D., Chappell, J. M. A., Bloom, A. L., Druffel, E. R. M. & Taylor, F. M. (1993) Science 260, 962– 967. 54. Bard, E., Hamelin, B., Fairbanks, R. G., Zindler, A. & Mathieu, G. (1990) Nucl. Instrum. Methods Phys. Res. Sect. B 52, 461–468. 55. McConnaughey, T. (1989) Geochim. Cosmochem. Acta 53, 163–171. 56. Weber, J. (1974) Proc. Int. Symp. Coral Reef 2nd. (Great Barrier Reef Committee, Brisbane, Australia), pp. 289–298. 57. Land, L. S., Lang, J. C. & Smith, B. N. (1975) Limnol. Oceanogr. 20, 283–287. 58. Weil, S., Buddemeier, R. W., Smith, S. V. & Kroopnick, P. M. (1981) Geochim. Cosmochim. Acta 45, 1147–1153. 59. McConnaughey, T. (1989) Geochim. Cosmochem. Acta 53, 151–162. 60. Emiliani, C., Hudson, J. H., Shinn, E. A. & George, R. Y. (1978) Science 202, 627–629. 61. Swart, P. K. (1983) Earth Sci. Rev. 19, 51–80. 62. Druffel, E. R. M., King, L. L., Belastock, B. A. & Buesseler, K. O. (1990) Geochim. Cosmochim. Acta 54, 1493–1500. 63. Smith, S. & Kroopnick, P. (1981) Nature (London) 294, 252–253. 64. Cole, J. (1992) Interannual-Decadal Variability in Tropical Climate Systems: Stable Isotope Records and General Circulation Model Experiments (Columbia University, New York). 65. Fairbanks, R. G. & Dodge, R. E. (1979) Geochim. Cosmochim. Acta 43, 1009–1020. 66. Epstein, S. & Mayeda, I. (1953) Geochim. Cosmochim. Acta 4, 213–224. 67. Weber, J. & Woodhead, P. (1970) Chem. Geol. 6, 93–117. 68. Weber, J. & Woodhead, P. (1972) J. Geophys. Res. 77, 463–473. 69. Druffel, E. R. M. (1985) Geophys. Monogr. 32, 111–122. 70. Dunbar, R. B. & Wellington, G. M. (1981) Nature (London) 293, 453–455. 71. Cole, J., R. Fairbanks, & Shen, G. (1993) Science 260, 1790–1793. 72. Swart, P. & Coleman, M. (1980) Nature (London) 283, 557–559. 73. Knutson, D. & Buddemeier, R. (1973) in Radiocarbon Contamination of the Marine Environment (International Atomic Energy Agency, Vienna), p. 735. 74. Druffel, E. R. M. & Suess, H. E. (1983) J. Geophys. Res. 88(C2), 1271–1280. 75. Druffel, E. R. M. & Griffin, S. (1995) Radiocarbon 37(2), 517–524. 76. Stuiver, M. & Polach, H. A. (1977) Radiocarbon 19(3), 355–363. 77. Buddemeier, R., Schneider, R. & Smith, S. (1981) Proc. Int. Symp. Coral Reef 4th 2, 81–85. 78. Shen, G. & Stanford, C. (1990) in Global Ecological Consequences of the 1982–83 El Niño-Southern Oscillation, ed. Glynn, P. (Elsevier, New York), pp. 255–283. 79. Schneider, R. & Smith, S. (1982) Mar. Biol. 66, 121–131. 80. Beck, J., Edwards, R. L., Ho, E., Taylor, F. W., Recy, J., Rougeri, F., Joannot, P. & Henin, C. (1992) Science 257, 644–647. 81. DeVilliars, S., Shen, G. T. & Nelson, B. T. (1995) Geochim. Cosmochim. Acta 58, 197–208. 82. Flor, T. & Moore, W. (1977) in Proc. Int. Symp. Coral Reef 3rd (Rosenstiel School of Marine and Atmospheric Sciences, University of Miami, Miami), pp. 555–561. 83. Shen, G. & Dunbar, R. (1994) Geochim. Cosmochim. Acta 59, 2009–2024. 84. Shen, G., Boyle, E. & Lea, D. (1987) Nature (London) 328, 794–796. 85. St. John, B. (1974) Proc. Int. Symp. Coral Reef 2nd. 2, 461–469. 86. Livingston, H. & Thompson, G. (1971) Limnol. Oceanogr. 16, 786–796. 87. Lea, D., Shen, G. & Bole, E. (1989) Nature (London) 340, 373–376. 88. Linn, L. J., Delaney, M. L. & Druffel, E. R. M. (1990) Geochim. Cosmochim. Acta 54, 387–394. 89. Shen, G., Campbell, T. M., Dunbar, R. B., Wellington, G. M., Colgan, M. W. & Glynn, P. W. (1991) Coral Reefs 10, 91–101. 90. Delaney, M. L., Druffel, E. R. M. & Linn, L. (1993) Geochim. Cosmochim. Acta 57, 347–354. 90. Swart, P. (1981) Palaeogeogr. Palaeoclimatol. Palaeoecol. 34, 115–136. 92. Oomori, T., Kaneshima, T. K., Nakamura, Y. & Kitano, Y. (1983) Galaxea 1, 77–86. 93. Sholkovitz, E. & Shen, G. (1995) Geochim. Cosmochim. Acta 59(1), 2749–2756. 94. Simkiss, K. (1964) Biol. Rev. 39, 487–505. 95. Dodge, R., Jickells, T. D., Knap, A. H., Boyd, S. & Bak, R. P. M. (1984) Mar. Poll. Bull. 15(5), 178–187. 96. Rasmussen, C. Proc. Int. Symp. Coral Reef 6th (1988) (Coral Reef Symposium Executive Committee, Townsville, Queensland, Australia), pp. 325– 330. 97. Dodge, R. & Vaisnys, J. (1975) Nature (London) 258, 706–708. 98. Dodge, R., Turekian, K. & Vainys, J. (1977) Proc. Int. Symp. Coral Reef 3rd (Rosenstiel School of Marine Atmospheric Sciences, University of Miami, Miami), pp. 361–365. 99. Shen, G., Cole, J. E., Lea, D. W., Linn, L. J., McConnaughey, T. A. & Fairbanks, R. G. (1992) Paleoceanography 7(5), 563–588. 100. Shen, G., Linn, L. J., Campbell, T. M., Cole, J. E. & Fairbanks, R. G. (1992) J. Geophys. Res. 97(C8), 12689–12697. 101. Jenkins, W. J. (1982) J. Mar. Res. 40, (Suppl.) 265–290. 102. Druffel, E. R. M. (1989) J. Geophys. Res. 94, 3271–3285. 103. Chakraborty, S. & Ramesh, R. (1993) Terra Nova 5, 545–551. 104. Isdale, P. (1984) Nature (London) 310, 578–579. 105. Susic, M., Boto, K. & Isdale, P. (1991) Mar. Chem. 33, 91–104. 106. Dunbar, R. B., Wellington, G. M., Colgan, M. W. & Glynn, P. W. in Proceedings of Seventh Annual Pacific Climate (PACLIM) Workshop, Int. Ecol. Stud. Prid. Technol., eds. Betancourt, J. L. & Tharp, V. L. (California Dept. of Water Resources Press, Los Angeles), pp. 183–198. 107. Cole, J. E. (1996) in Climatic Fluctuations and Forcing: Mechanisms of the Last 2000 Years, eds. Jones, P. D., Bradley, R. S. & Jouzel, J. (Springer, Berlin), pp. 333–355. 108. Patzold, J. & Wefer, G. (1992) Fourth International Conference on Paleoceanography (Kiel, Germany), pp. 224–225. 109. Toggweiler, J., Dixon, K. & Bryan, K. (1989) J. Geophys. Res. 94(C6), 8217–8242. 110. Druffel, E. R. M. & Griffin, S. (1993) J. Geophys. Res. 98, 20249–20259. 111. Quinn, W. H., Neal, V. T. & DeMayolo, S. E. A. (1987) J. Geophys. Res. 92, 14449–14461. 112. Druffel, E. R. M. & Benavides, L. M. (1986) Nature (London) 321, 58–61. 113. Suess, H. E. (1953) in Proceedings Conference Nuclear Processes in Geological Settings, (University of Chicago Press, Chicago), pp. 52–56. 114. Druffel, E. (1997) Science 275, 1454–1457. 115. Quay, P. D., Tilbrook, B. & Wong, C. S. (1992) Science 256, 74–79. 116. Griffin, S. & Druffel, E. R. M. (1989) Radiocarbon 31, 533–542. 117. Smith, J. (1993) Isotopes in a Deep-Sea Coral from Orphan Knoll (McMaster University, Hamilton, Ontario, Canada). Dissertation. 118. Druffel, E., Griffin, S., Witter, A., Nelson, E., Southon, J., Kashgarian, M. & Vogel, J. (1995) Geochim. Cosmochim. Acta 59, 5031–5036. 119. Adkins, J. & Boyle, E. (1994) Eos 75(44), 347. 120. Adkins, J., Boyle, E. A., Cheng, H., Edwards, R. L. & Druffel, E. R. M. (1995) Eos 76(46), F290.
A LONG MARINE HISTORY OF CARBON CYCLE MODULATION BY ORBITAL-CLIMATIC CHANGES
8362
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8362–8369, August 1997 Colloquium Paper This paper was presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held November 13–15, 1995, at the National Academy of Sciences, Irvine, CA.
A long marine history of carbon cycle modulation by orbital-climatic changes TIMOTHY D. HERBERT Department of Geological Sciences, Brown University, Providence, RI 02912
© 1997 by The National Academy of Sciences 0027-8424/97/948362-8$2.00/0 PNAS is available online at http://www.pnas.org. ABSTRACT Pacing of the marine carbon cycle by orbital forcing during the Pliocene and Pleistocene Ice Ages [past 2.5 million years (Myr)] is well known. As older deep-sea sediment records are being studied at greater temporal resolution, it is becoming clear that similar fluctuations in the marine carbon system have occurred throughout the late Mesozoic and Tertiary, despite the absence of large continental ice sheets over much of this time. Variations in both the organic and the calcium carbonate components of the marine carbon system seem to have varied cyclically in response to climate forcing, and carbon and carbonate time series appear to accurately characterize the frequency spectrum of ancient climatic change. For the past 35 Myr, much of the variance in carbonate content carries the “polar” signal of obliquity [41,000 years (41 kyr)] forcing. Over the past 125 Myr, there is evidence from marine sediments of the continued role of precessional ( 21 kyr) climatic cycles. Repeat patterns of sedimentation at about 100, 400, and 2,400 kyr, the modulation periods of precession, persistently enter into marine carbon cycle records as well. These patterns suggest a non-linear response of climate and/or the sedimentation of organic carbon and carbonates to precessional orbital perturbations. Nonlinear responses of the carbon system may help to amplify relatively weak orbital insolation anomalies into more significant climatic perturbations through positive feedback effects. Nonlinearities in the carbon cycle may have transformed orbital-climatic cycles into long-wavelength features on time scales comparable to the residence times of carbon and nutrient elements in the ocean. We tend to think that we live in the most exciting time. To a geologist, the most exciting time is the late Pleistocene world (the past 800,000 years), a period of unusually sensitive climate. Certainly, there are many surprises that have come from gathering paleoclimatic records over the past few million years (Myr) or so. These surprises include the demonstration that ice ages come and go with a regularity that can only come from pacing by changes in the Earth's orbit (1, 2); that these ice ages involve changes in sea level on the order of 125 m (3); that the deep ocean circulation rearranges itself with climatic state (4, 5 and 6); and that the climatic changes, at least for the last two major glacial cycles, go hand in hand with variations in methane and carbon dioxide contents of the atmosphere (7, 8). The observed response of late Pleistocene climates to orbital perturbations raises a number of puzzles, many of which will reappear as we examine older climatic records. From a variety of proxy records, we have learned that seasonal anomalies enter strongly into the mean state of the climate system over the past few million years. This is seen clearly in the resemblance of the famous “Milankovitch” radiation curve of July temperatures at 65° north to the oxygen isotope record of ice volume and sea level. Were the insolation anomalies to be calculated on an annual basis, the 21,000-year (21-kyr) precessional component would vanish. In addition, a strong, 100-kyr cycle not present in the insolation curve runs through the oxygen isotopic record of sea level changes and other late Pleistocene climate records. The 100-kyr cycle dominates the 160-kyr-long greenhouse gas records of polar ice cores (7, 8), although the Vostok methane record (8) seems to show particular sensitivity to the 21-kyr precessional forcing as well. It is difficult to explain the 100-kyr cycle of glaciation from the meager 100-kyr insolation anomaly directly due to eccentricity variations, although it may arise as a rectified response to the precessional cycle within a narrow band of the equator (9). Rather, most climate modelers invoke a large nonlinearity or linearities of the climate system to transfer energy from the precessional cycle, which is amplitude-modulated at the eccentricity periods of about 100, 400, and 2,400 kyr. Amplitude-modulated signals vary about a mean value according to a lower frequency function, which in the case of precession is the orbital eccentricity. The low frequency does not appear directly in the output of a linear response to such a forcing. Nonlinear systems tend to transform amplitude-modulated forcing into an outcome with significant energy at the modulating periods and, hence, could produce long-wavelength climatic cycles from the much shorter precessional perturbations (10, 11). In this contribution, I shall follow the variability of past climates, as it manifests itself in carbon records in deep-sea sediments over the past 125 Myr. This time span has seen major changes in the state of the Earth's climate: changes from the warm, equable climates of the middle Cretaceous Period to the progressive cooling and glaciation of high latitudes during the Tertiary Period (12, 13, 14 and 15). The dynamics of the carbon records will be viewed with a particular lens: that of the spectral window tuned to the Milankovitch frequency band. By comparing variations on the same time scale, and with similar forcing, with those that alter late Pleistocene climate, I wish to make three points. First, the large-scale modulations of the marine carbon system, well characterized for the last 2.5 Myr, persist at similar frequencies into the more remote marine record. The prime recorder that will be used to measure these modulations is the calcium carbonate content of marine sediments. Second, in some cases, we may be able to assess whether the ancient marine climate–carbon connection was synchronized between hemispheres in ancient climates, as it has been in the more recent glacial cycles. And third, I will argue that marine carbon records suggest that pervasive nonlinearities in the climate system can translate relatively high frequency (20kyr), seasonal, insolational variance into low frequency patterns that might be mistaken for tectonically driven changes in climate and ocean geochemistry.
Abbreviations: kyr, thousand years; Myr, million years; Ma, million years ago; K/T, Cretaceous–Tertiary.
A LONG MARINE HISTORY OF CARBON CYCLE MODULATION BY ORBITAL-CLIMATIC CHANGES
8363
The Pleistocene Precedent Geologically measurable aspects of the marine carbon cycle trace important perturbations in the Earth's climate; these measurements also suggest that the marine carbon cycle may have feedbacks that amplify climatic variance. One of the first convincing links between marine sediments and global climate change of the Pleistocene came from Arrhenius' (16) study of the calcium carbonate content of sediment cores from the equatorial Pacific Ocean. Arrhenius correctly connected the variations in deep-sea sediment composition with ice age cycles known from the land and also noted the cyclical nature of the changes over time. It is now clear that over the past 2.5 My, nearly all aspects of the Earth's climate have been paced by variations in the Earth's orbital elements (17, 18 and 19). In the deep sea, these changes include rearrangement of the deep circulation such that present differences in carbon chemistry between the abyssal Atlantic and Pacific nearly vanished during glacial stages (6, 20). This restructuring is evidenced by changing lateral gradients in carbon isotopes recorded by benthic foraminifera (4, 6), and by compensating changes in calcium carbonate content in Pacific and Atlantic sediments (21). During ice ages, carbonate microfossils tend to be better preserved in the Pacific and to show poorer preservation in the Atlantic. The carbonate variations trace important modulations in deep-water sources over time and, therefore, in the heat and salinity budgets of the global ocean circulation. Carbon isotopic records measured from the skeletal remains of foraminifera indicate major changes in the carbon budget of the glacial world as compared with the present. A global shift to lighter isotopic compositions is best explained as a transfer of terrestrial organic matter (with mean isotopic composition of −25‰ to the ocean during ice ages after the destruction of forests and soils in the colder and drier glacial world (22). The mean oceanic shift in δ13C of −0.4‰ (22, 23) corresponds to the addition of some 450 × 1015 g of carbon to the ocean during glacial times. By itself, such a transfer would have tended to promote a higher CO2 content of the atmosphere. Ice core records, however, indicate the reverse; the paradox must be resolved by changes in the carbon cycle in the ocean that favor increasing carbon storage in the ocean during glacial periods. To date, hypotheses involving changes in the production of marine organic carbon in the glacial world (24, 25 and 26) and in the alkalinity structure of the glacial ocean (27) have been proposed but have not produced a satisfying resolution. As we look to time scales of millions of years, it becomes clear that the spectral content of Pleistocene carbon records mirrors that of other paleoclimatic indices. For example, the transition in ice volume frequency at 1 Myr ago from an earlier regime of 41 kyr (obliquity) to the late Pleistocene rhythm of large-amplitude, 100-kyr cycles, shows clearly in carbonate time series as well as in changes in oxygen isotope ratios (28, 29). In both regimes, carbonate variations record in large measure the dance of deep-water sources over time. The faithfulness of marine carbonate records in replicating the mid-Pleistocene switch observed in oxygen isotopic data from 41- to 100-kyrdominated spectra is significant for it indicates that nonlinearities in carbonate preservation and production are not so inherently large as to inevitably produce 100-kyr and longer cycles from precessional climate forcing alone (we will return to this point later). If changes in ocean alkalinity and nutrient structure are to be inferred for late Pleistocene, 100-kyr ice age cycles (27), then the early Pleistocene regime saw largely 41-kyr repeat cycles of high latitude origin in the oceanic carbon pump. Organic carbon contents of marine sediments also vary at Milankovitch frequencies in the late Pleistocene (30, 31, 32 and 33). If these can be read as paleo-productivity records, then there is evidence for substantial modifications in the export of carbon from surface waters in soft form. It appears, however, that the patterns of organic carbon accumulation obtained from sediments are highly regional, and it is difficult at present to reconstruct a global mass balance for the export of organic matter during glacial periods. Signatures of Orbital Cycles A general understanding of the “fingerprint” of orbitally driven cycles is necessary if we are to follow the sensitivity of the climate system in general, and the carbon cycle in particular, to orbital perturbations in the more distant past. Direct insolation forcing comes from the obliquity cycle, with a current mean period of 41 kyr, and from precessional variations, with periods of 23 and 19 kyr. Considerations of tidal friction indicate that the periods of these orbital cycles should be shorter in the past; Berger and colleagues (34) estimate that the obliquity repeat time would be 38.75 kyr by 100 Ma (mid-Cretaceous) and that the precessional periods would lie at about 22 and 18 kyr, respectively. Eccentricity forcing is likely to remain at similar frequencies to the present day, that is, at a short cycle of 109 kyr actually comprised of components near 95 and 125 kyr, and nearly “line” frequencies of 413 and 2,400 kyr (35, 36). These frequencies should primarily enter the climate system as modulations (the variance around the mean) of precession, as the direct radiative forcing due to eccentricity variations is minuscule. However, the importance of 100-kyr energy in the late Pleistocene record of ice volume tells us that Milankovitch climate theory may be missing a major component of climate variance. Geologists have a number of tools to use as they examine older marine records for continuing evidence of orbitalclimatic forcing. By using multiple constraints, a convincing case can often be made that orbital frequencies are correctly identified. Criteria include repeat periods estimated by paleomagnetic and paleontological data consistent with orbital cycles, continuity of cyclic sedimentation, statistical evidence for periodicity from spectral analyses, successful correlation of patterns between different study locations, and internal consistency of the data viewed with the orbital model of sedimentation (for example, evidence of the expected amplitude modulating frequencies of precession when a precessional signal is identified). In most cases, the repeat times of carbon cycles can be estimated to a relative error of 10–20% over the past 84 Myr, the period of frequent, well-dated magnetic reversals. Good matches over long time periods between the repeat times of sedimentary cycles and expected orbital periods are powerful indications that a cause and effect connection has been found. Difficulties in dating older marine sediments accurately mean that other consistency arguments become more important in fingerprinting an orbital-climatic connection. One can also compare the character of sedimentary cycles to the expected forcing. For example, obliquity cycles are nearly sinusoidal in nature, whereas the precessional rhythm is strongly modulated by eccentricity cycles at about 100, 400, and 2,400 kyr. Such diagnostic features can often be observed in sedimentary time series. As we perform time series analyses of pre-Pleistocene deep-sea sediments, some of the questions we ask are as follows. (i) Is there evidence of continuing paleoceanographic sensitivity to orbital forcing since the Mesozoic, as seen through carbon records? (ii) If so, how does the climatic response to the possible mix of orbital components evolve over time? (iii) Is there evidence that Northern and Southern hemispheres continue to be synchronized to orbital-climatic cycles in the distant past? (iv) Do long-wavelength, eccentricity-driven cycles appear in pre-Pleistocene carbon records? (v) Can we deduce some features common to orbitally driven changes in the carbon cycle over the long late Mesozoic and Tertiary marine record?
A LONG MARINE HISTORY OF CARBON CYCLE MODULATION BY ORBITAL-CLIMATIC CHANGES
8364
Spectral Record of Tertiary Age Carbonate Sediments Bloemendal and deMenocal (37) have documented a spectral shift in carbonate records at 2.6 million years ago (Ma) that seems coincident with the onset of significant Northern Hemisphere glaciations. They show that carbonate proxy records from the Oman margin record a growth in 41-kyr spectral power at the expense of 23-kyr precessional power across this transition. Further work has demonstrated that a similar transition occurs in the equatorial Atlantic. The change is believed to reflect the penetration of high latitude climate variations, which are most sensitive to obliquity forcing, into low latitudes. Although deMenocal (38) ascribes most of the variation in sediment carbonate content to dilution by atmospheric dust, the fact that a similar spectral transition occurs in nearly dust-free equatorial Pacific sediments (39) argues again for a coordinated shift in the frequency with which deep-water sources and carbonate alkalinity oscillated as climate evolved to cooler and more glaciated conditions. As well sampled records of pre-Pliocene deep-sea sediments accumulate, it is becoming clear that the marine carbon cycle sensed orbital forcing in warmer climates of the middle and earlier Tertiary. Ocean drilling has provided glimpses of this sensitivity at a number of locations; these patterns have been supplemented by studies of pelagic sediments now exposed on land (40, 41, 42 and 43). One of the longest continuous records of orbitally mediated carbonate cycles comes from Ocean Drilling Program Leg 154 (44). Variations in carbonate content at Milankovitch frequencies were traced to middle Paleocene (55 Ma) times, with no suggestion of a halt in the Milankovitch response (44). Records from Oligocene age onward show evidence of obliquity (41 kyr) forcing, confirming a suggestion by Mead and colleagues (45) that this cycle has been an important paleoclimatic frequency since the inception of glaciation in the Southern Hemisphere at the end of the Eocene Epoch, 33.5 Ma. Low Frequency Components in the Tertiary Carbon Cycle The importance of 100-kyr cyclicity in the late Pleistocene is a puzzle in the Milankovitch theory of climate. Its disappearance in the earlier Pleistocene and latest Pliocene suggests a simpler world in which paleoclimatic patterns are more easily explained by a linear response to orbital forcing. However, there are a number of observations to suggest that eccentricity frequencies haunt earlier times as well. In some cases, prePleistocene time series show variance centered on the 100-kyr cycle in a manner similar to the last 1 Ma, but in others, there is good evidence for the 400-kyr component, the single largest term in the astronomical eccentricity series. Pliocene (3–4 Ma) carbonate records from the equatorial Pacific show a frequency structure almost identical to that of the late Pleistocene (Fig. 1). An 100kyr cycle of dissolution has about twice the amplitude attributable to obliquity and precessional periods. Deep-ocean carbon chemistry was clearly modulated by orbital forcing in a climate without large Northern Hemisphere ice sheets. Carbonate content of equatorial Pacific sediments fluctuated at a dominant period of 400 kyr in late Miocene times (46), another indication of a long history of eccentricity pacing of deep-ocean carbonate dissolution. This 400-kyr period is also seen clearly in carbon isotopic records of benthic foraminifera of midMiocene age (47) that imply considerable changes in the marine carbon budget. The authors attribute these changes to periods of enhanced burial of marine organic matter during climatic coolings. What is intriguing is the large amplitude of the long-wavelength variations (1‰ cycles in δ13C), in view of the minimal direct radiative forcing at eccentricity frequencies.
FIG. 1. Spectra of early Pliocene (3.5–4.5 Ma) carbonate variations at Deep Sea Drilling Site 573, equatorial Pacific, and at Ocean Drilling Program Site 661, tropical Atlantic. Frequency is plotted as a function of stratigraphic position; no “tuning” has been performed to achieve a match to an orbital model. Spectral peaks interpreted to correspond to the short cycle of eccentricity (109 kyr), the obliquity cycle (41 kyr), and the two components of precession (23 and 19 kyr) are indicated by vertical lines. Interhemispheric Synchrony of Precessional Carbonate Cycles, Late Cretaceous Deep-sea coring has given us a number of overlapping records in the South Atlantic and Indian oceans (48, 49) of late Cretaceous and early Tertiary age that provide paleoceanographers with an unusual opportunity to develop a global picture of the dynamics of carbon cycling in the older ocean. In addition, intense interest in the Cretaceous–Tertiary (K/T) extinction level has generated detailed studies of land sections where former deep-sea sediments outcrop. Good independent dating of sections from magnetic reversal stratigraphy and from micropaleontological studies add confidence to the correlation of high frequency sedimentary variations between sites. These records allow us not only to document the pervasive orbital influence on marine carbon records, but also, using the unusually well documented time line at the K/T boundary, to assess whether carbon cycle variations may have been coordinated between hemispheres, as they have been in the Pleistocene. Cores recovered by the Deep Sea Drilling project from the South Atlantic display continuous carbonate cyclicity from Santonian (85 Ma) to mid-Paleocene (60 Ma) time. Drill sites come from two subsiding rises on opposite sides of the South Atlantic that lay at about 30° south at the time of deposition. Variations in carbonate content can be logged by reflectance spectroscopy calibrated to discrete measurements of calcium carbonate content (Fig. 2). Carbonate cycles are present at sites that ranged in paleodepth from 0.5 to 3.5 km. Carbonate variations are substantial (lows of 35%, highs of 85%) and highly structured. The mean repeat time of the cycles estimated from paleomagnetic stratigraphy is 23.3 ± 4 kyr, a good match to the expected precessional period (48). Just as importantly, the cycles display the typical hierarchy of modulation expected of precessional anomalies. Nodes of low amplitude occur at 4–5 cycles, 20–25 cycles (Fig. 2A), and 100–120 cycles (Fig. 2B). These almost certainly match eccentricity modulation at about 100-, 400-, and 2,300-kyr repeat times. The cycles appear to be directly correlative between sites, indicating a basin-wide climatic control on sedimentation. Along with the amplitude modulation expected of a direct response to precessional seasonal anomalies in insolation, the South Atlantic carbonate cycles contain evidence for cycles in accumulation that cannot be so readily explained. Changes in sediment accumulation rate can be estimated by following the spacing of the carbonate cycles over time. Intervals of thicker
A LONG MARINE HISTORY OF CARBON CYCLE MODULATION BY ORBITAL-CLIMATIC CHANGES
8365
cycles, an indication of higher sediment fluxes, coincide with intervals of high amplitude oscillations in carbonate content. The most evident modulations in sedimentation rate occur at intervals of about 400 and 2,500 kyr. A linear response of carbonate sedimentation to precessional forcing would not generate long-wavelength anomalies in deposition because the increase in deposition rate during one phase of a precessional cycle would be precisely canceled by the succeeding decrease, i.e., the precessional signal itself varies at any given latitude about a constant mean. Therefore, the low frequency energy envelope of precessional variations must have been transferred in this portion of the geological record into low frequency changes in oceanic carbonate deposition through an unidentified nonlinear mechanism.
FIG. 2. Time series of late Cretaceous (Campanian through Maastrichtian) deep-sea carbonate variations from South Atlantic Deep Sea Drilling Project Site 516F (DSDP 516F). Reflectance measurements at 700 nm logged every 4 cm by a visible light spectophotometer have been used as a proxy for carbonate content. Time control comes from magnetic reversal stratigraphy correlated to the Berggren et al. (50) time scale. Carbonate variations are modulated at a number of scales, the longest of which is close to 2.5 Myr, the long nodal cycle of eccentricity (B). The detail above shows the organization of individual precessional carbonate cycles into bundles of 100- and 400-kyr spacing (A). The late Cretaceous carbonate variations show a South Atlantic Ocean dominated almost entirely by low latitude (precessional) climate variations. Because the cycles appear in both shallow (paleodepth ≤ 1 km) and deeper locations, it seems likely that they reflect in large part changes in the productivity of calcareous organisms rather than in the alkalinity of deep waters. Simple models of sediment accumulation suggest that the carbonate cycles reflect increases both in carbonate accumulation (high carbonate beds) and in dilution by terrigenous material during the antiphase (dark, low carbonate concentration layers) of the 20-kyr cycle. Late Cretaceous deep-sea sediments now exposed in Spain give us the unusual chance to test whether precessional variations in the marine carbon cycle may have been synchronized between hemispheres. As in the South Atlantic, high frequency oscillations in carbonate have been detected in a number of outcrops of Maastrichtian age sampled by ten Kate and Sprenger (51). Although magnetic reversal stratigraphy has not been successful in each study site, the lithological patterns are similar enough between sites that time lines can be drawn quite confidently between the Spanish sections. I believe that it is possible to correlate the distinctively modulated patterns of carbonate variations in the 500 kyr before the K/T boundary between the North and South Atlantic with high confidence (Fig. 3). Tie lines between the sites include the K/T boundary, the position of the magnetic reversal boundary separating polarity zone C29R from the normal polarity zone C30N, and the node of low amplitude carbonate variations that I attribute to the 400-kyr eccentricity amplitude envelope. Surprisingly, the carbonate oscillations seem to be in phase between hemispheres. In every South Atlantic section, the K/T boundary occurs at the top of the high carbonate phase of the precessional cycle. Precisely the same pattern occurs in Spain. The studied sections were situated in subtropical locations (about 30° south and north) on opposite sides of the late Cretaceous equator. It appears likely, then, that carbonate variations in both hemispheres, at least in the Atlantic basins, were somehow synchronized in a manner similar to the late Pleistocene. (Recall that insolation anomalies due to precession are out of phase between hemispheres; synchrony of ocean chemical changes on opposite sides of the equator implies the dominance of one hemisphere.) The most plausible means of linking the carbonate records between the North and South Atlantic is through a deep-ocean circulation climatically weighted to
A LONG MARINE HISTORY OF CARBON CYCLE MODULATION BY ORBITAL-CLIMATIC CHANGES
8366
one hemisphere. Changes in carbonate content could have been effected by some combination of varying nutrient inventories and varying corrosiveness of deep waters to carbonate minerals.
FIG. 3. Comparison of latest Cretaceous carbonate cycles from the Southern Hemisphere [Deep Sea Drilling Project Site 516F (DSDP 516F)] and from the Northern Hemisphere (Zumaya, Spain). Similar modulations are seen in both records; note the node of low variability that is correlated to the 400-kyr eccentricity modulation of precessional amplitude. The K/T extinction boundary occurs in both hemispheres at the peak of a high carbonate cycle. Spanish data are from ten Kate and Sprenger (51). The 100- and 400-kyr Cycles in Carbonate and Organic Carbon Deposition, Middle Cretaceous The middle Cretaceous (Aptian through Cenomanian stages) is widely believed to have had the warmest global climate of the last 125 Myr (12, 52). One common feature of deep-sea deposits of this age are organic-carbon enriched “black shales,” which have been found throughout the Mediterranean–Atlantic basins, and into the high latitude Indian Ocean (39, 53). Modern deep-sea sediments generally have organic carbon contents of 0.5% or less; therefore, the frequent occurrence of layers with 1–25% organic carbon in the middle Cretaceous pelagic record may indicate unusual oceanic circulation during the thermal maximum. The data presented below indicate that the biological productivity of warm Cretaceous oceans was highly variable; warmer temperatures apparently did not make for a less dynamic environment. Detailed studies from one location, a continuous core drilled through Aptian and Albian deep-sea sediments of central Italy (54) can be applied to other mid-Cretaceous ocean basins because studies of pelagic sediments outcropping in the Mediterranean region and recovered by deep-sea drilling show similar patterns (40). Time series sampling demonstrates that organic carbon-enriched beds fall in a consistent position in a hierarchy of sedimentary oscillations linked to orbital forcing. Black shales occupy 5-cm-thick intervals of low carbonate in the curves presented in Fig. 4. Typical deposition rates for the intervals studied are 0.5 cm/kyr, so the dark–light, low–high carbonate pair has a period of 20 kyr. However, there are longer term oscillations of 50 cm (100-kyr) long, in which intervals of two to three oscillations of high carbonate content alternate with two to three pairs of lower carbonate cycles. It is in the troughs of this 50-cm cycle that black shales occur. A further hierarchy of cyclicity occurs at 2–2.5 m (400–500 kyr). This wavelength makes an “envelope” on the amplitude of the 50-cm, 100-kyr cycle. At nodal points, carbonate variations are small, and black shales are usually absent. Spectral analysis of the middle Cretaceous lithological records clearly demonstrates the presence of an important periodicity at 100 kyr (55) with much greater amplitude than at the precessional periods. Simple “tuning” assumptions used to sharpen spectra in fact detect two line components whose frequency ratio precisely matches the beat frequencies of the 100-kyr cycle (95 and 125 kyr) in the Cretaceous carbonate records (55).
FIG. 4. Middle Cretaceous series of carbonate variations driven by a combination of precession and eccentricity cycles. The series in A comes from strata at 102 Ma; the series in B, from strata at 110 Ma; and the series in C, from strata at 115 Ma. The average sedimentation rate is just under 0.5 cm/kyr. Organic carbon-rich beds occur in 5-cm-thick bands of low carbonate content.
A LONG MARINE HISTORY OF CARBON CYCLE MODULATION BY ORBITAL-CLIMATIC CHANGES
8367
Geochemical work demonstrates that the basic alternation of lithologies is a simple mixing of biogenic skeletal components (both siliceous and calcareous) and terrigenous material (54). The organic carbon-enriched layers are actually impoverished in the remains of hard parts of the marine plankton. The covariance of both siliceous and carbonate deposition makes it unlikely, therefore, that the carbonate oscillations reflect changes in the saturation state of the deep basin waters. Instead, the evidence favors several-fold variations in the productivity of the planktonic ecosystem at periods of 20, 100, and 400 kyr. Large variations in productivity could in turn generate large changes in deposition rate, which could skew the interpretation of frequency spectra collected as a function of depth. This time–depth distortion could explain the observation that the individual 20-kyr carbonate oscillations are poorly resolved in frequency analysis (55). If we use the bedding cycles as time markers, it is possible to develop a reasonable approximation to the dynamics of accumulation of both biogenic and terrigenous materials in the middle Cretaceous ocean. This reconstruction does indeed show the fingerprint of a precessional cycle highly modulated by 100- and 400-kyr cycles of eccentricity. The orbital chronometer works as follows. If we assume that each 10-cm cycle in carbonate content has approximately the same time value (in fact, the precessional period varies by 12% around its mean), and we pick cycle maxima and minima objectively, then it is possible to total the accumulation of carbonate, biogenic silica, organic matter, and terrigenous matter per unit time (one 20-kyr cycle). It is then possible to assess the relative variances of the different components and to detect modulations in the accumulation rate that might be otherwise masked. Fig. 5 shows bed-by-bed (10-kyr increment) totals of carbonate and aluminum (a proxy for clay content) for an upper Albian (102 Ma) time series. The variance of the carbonate accumulation is about twice that of the terrigenous component, reinforcing the earlier hypothesis of a strong control on deposition rate by fluctuations in carbonate production (a conclusion further strengthened by micropaleontological data from ref. 56). Carbonate accumulation patterns also show significant periodicities centered at about 100 and 400 kyr (dashed curves indicated in Fig. 5). These variations could only arise if eccentricity modulations of precession enter in some way directly into deposition rate. A weak oscillation in aluminum flux is present at the 100-kyr period, essentially out of phase with carbonate (and silica) deposition. But a quite regular, large amplitude cycle occurs at the 400-kyr period. The flux of clays and other terrigenous materials, therefore, covaried with carbonate with about equal amplitude and phase at the 400-kyr cycle. This covariance mostly masks the 400-kyr signal when only weight percent measures of the sediment are inspected.
FIG. 5. Comparison of accumulation rates of carbonate and aluminum over an estimated 1.6 Myr of middle Cretaceous history from central Italy. A 20-kyr repeat time for carbonate cycles has been assumed, consistent with independent estimates of sedimentation rate, and with spectral analysis of geochemical time series. Geochemical fluxes were estimated by binning components in sequential 10-kyr half-cycles defined by maxima and minima in carbonate content. Sinusoids of about 100 and 400 kyr (dashed lines) have been drawn to highlight the dominant patterns of carbonate accumulation. Note that the variability of carbonate fluxes is larger than that for aluminum at the 100-kyr period but that the two components covary at the 400-kyr period. The 3-fold changes in carbonate accumulation on 100-kyr time scales are believed to result from changes in the productivity of calcareous plankton. We are left then, with a picture of large variations in the flux of both biogenic and terrigenous materials to many Cretaceous ocean basins. The shortest recurrences were paced by precessional insolation anomalies. It appears that at the 20- and 100-kyr wavelengths, the high marine productivity episodes alternated with low productivity intervals in which the clay accumulation was enhanced. As the dominant mode of clay flux to the sheltered basins of the Italian region was presumably by eolian deposition, the evidence favors a synchronization between wetter continental conditions, with low clay influx and higher marine producitivity, and drier conditions, with enhanced transport of clays to the deep sea and concomitantly lower productivity in the study area. The changes in productivity could be caused by large changes in the supply of nutrients to the ocean basins with changes in climate, or, more probably, by changing patterns of exchange between ocean basins. Switches from “estuarine” to “lagoonal” configurations (57) could have altered the carbon dynamics on the Milankovitch time scale. The Cretaceous organic, carbon-rich intervals correspond to extremes in precessional forcing, in a manner strikingly similar to analogous deposits from the Eastern Mediterranean over the past 0.5 Myr (58). Models that use Cretaceous geography and orbital end member geometries do in fact suggest monsoonal sensitivity of the Cretaceous climate that could explain the sedimentary variations (52, 59). The largest depositional anomalies in the Italian sequence, however, coincide with eccentricity periodicities that would be absent from a direct analysis of radiative forcing. As observed in late Pleistocene records, geologic data
A LONG MARINE HISTORY OF CARBON CYCLE MODULATION BY ORBITAL-CLIMATIC CHANGES
8368
suggest that climatic nonlinearities can translate high frequency variations into longer period geochemical flux anomalies by mechanisms still unknown. One consequence of the climate–sediment nonlinearities may be to transfer depositional anomalies of carbon and nutrient storage from time scales shorter than the residence times of these elements to 100-kyr, 400-kyr, and even longer scales, which could well affect their mean oceanic inventories. We also suspect, although we do not know yet, that the terrigenous input of carbon to the Cretaceous oceans may have also varied in phase with the clay accumulation cycles observed in the deep sea. Are There Common Threads? It would be simplistic to assume that all variations in carbonate and organic carbon time series from the deep sea represent precisely the same climatic response to orbital forcing. The spectral mix of frequencies observed in sediment records in fact varies on time scales of millions of years. It does appear, however, that variations in the deposition of both carbon phases have always been climatically sensitive at the orbital frequencies. If the Pleistocene–Pliocene precedent holds for earlier times, spectra of carbonate time series in marine records give a good representation of the frequency mix of paleoclimatic energy in the entire climatic system. Precessional forcing appears to be a constant in the accumulation of organic and inorganic carbon in the deep sea, and in the influx of terrigenous materials. Obliquity signals appear strongly only after the cooling of the high latitude Southern Hemisphere began. The persistence of periodicities associated in some way with eccentricity forcing is one of the major surprises that come from looking at older records. Energy at about 100-, 400-, and in some cases, 2,400-kyr wavelengths is especially clear before the onset of large-scale Southern Hemisphere glaciation in the earliest Oligocene, but it also appears in many later carbon records. The dominance of the 41-kyr obliquity cycle in long Oligocene, early Miocene, and early Pleistocene carbon records make it clear that 100-kyr cyclicity is not inherent in the carbonate recorder itself. Rather, eccentricity-related cycles in carbonate content and in carbon isotopic composition most likely reflect real changes in ocean and atmospheric circulation at these frequencies. One of the unresolved puzzles of late Pleistocene climate, the dominance of an orbital term with little direct insolation forcing, therefore grows in importance with a longer time perspective. What we do not know yet is the extent to which anomalies in marine carbon deposition related to climate change represent not only passive tracers of changes in ocean circulation but also active feedback responses that could modify climate sensitivity. As tracers, the aspects of the carbon system that can be monitored geologically include changes in the path and sources of deep-water formation, changes in wind stress and nutrient inventories to the surface biota, and transfers of carbon between terrigenous and marine reservoirs. Feedbacks implied by ancient carbon cycles range from variable uptake by the ocean of atmospheric carbon dioxide due to changes in temperature and circulation, to anomalies in the global rates of organic and carbonate carbon deposition rate that could affect the long-term greenhouse gas content of the atmosphere. It is particularly important to understand the origin of long-wavelength marine carbon cycles as these could potentially alter factors such as sedimentary storage of nutrient elements and the carbon dioxide content of the atmosphere for long enough periods of time to affect weathering cycles on land and evolution and biogeochemical cycling in the ocean. If the data presented above are representative of ocean history for the past 125 Myr, the gap between the relatively short duration (20 and 41 kyr) orbital insolation and tectonic perturbations to climate and biogeochemical cycles may be bridged by nonlinear climate–carbon connections. These may cause changes in the oceanic output of carbon and associated elements on time scales normally considered to fall within the tectonic window (cf. ref. 60). If we are to look for common threads over time, it is necessary to look for robust mechanisms that do not depend, for example, on the existence of extensive glaciation (for example, see ref. 24), or on other boundary conditions not appropriate for the span of late Mesozoic and Tertiary time. A simple model should also explain why different variables might be coupled in a response to orbital forcing. I suggest that the interaction between radiative forcing, marine sea surface temperatures, and the hydrological cycle may give a rich array of climatic responses at Milankovitch frequencies over long periods of geological time. For the carbon cycle, small changes in the density structure of deep waters can modulate the basin–basin fractionation of alkalinity and nutrients (cf. ref. 57) and explain dissolution patterns imposed on many sediment carbonate records. The existence of carbonate dissolution cycles in basinal deposits during most of the Tertiary and late Mesozoic suggests that there have been at least two regions of deep-water formation, with different initial dissolved inorganic carbon characteristics, throughout this time. The density differences between potential deep-water-forming areas may be extremely small and would be sensitive to changes in both sea surface temperature and salinity. The Mediterranean may be a small-scale example of this phenomenon. It seems likely that the present evaporative water balance shifted at a number of time in the past to fresher conditions in response to enhanced African monsoons, those regulated by the precessional cycle (58). During such periods, the exchange of waters between the Atlantic and Mediterranean may have been reversed, leading to a nutrient-trap configuration and the development of anoxic bottom waters and organic-carbon-rich sediment layers (61). Because water vapor is such a potent greenhouse gas, changes in sea surface temperature could amplify global temperature anomalies. On land, changes in the water balance could promote or disfavor the storage of organic carbon in soils and biomass. The signs of terrestrial climate change may be imprinted in cycles in the carbon isotopic composition of sea water, and in the variations in deposition rate of terrigenous dust in deep-sea sediments. Geological records hint that all of these responses happen commonly, and that they have important nonlinear aspects that enhance modest initial forcings. I thank S. D'Hondt, E. Erba, A. G. Fischer, C. D. Keeling, J. Park, and I. Premoli Silva for many helpful discussions, and two anonymous reviewers who helped improve the final manuscript. Portions of this work were funded by the Petroleum Research Foundation of the American Chemical Society and by grants from the National Science Foundation. 1. Hays, J. D., Imbrie, J. & Shackleton, N. J. (1976) Science 194, 1121–1132. 2. Imbrie, J., Hays, J. D., Martinson, D. G., McIntyre, A., Mix, A. C., Morley, J. J., Pisias, N. G. & Shackleton, N. J. (1984) in Milankovitch and Climate, eds. Berger, A., Imbrie, J., Hays, J., Kukla, G. & Saltzman, B. (Reidel, Hingham), pp. 269–305. 3. Fairbanks, R. G. (1989) Nature (London) 342, 637–642. 4. Shackleton, N. J., Imbrie, J. & Hall, M. A. (1983) Earth Planet. Sci. Lett. 65, 233–244. 5. Duplessy, J. C., Shackleton, N. J., Fairbanks, R. G., Labeyrie, L., Oppo, D. & Kallel, N. (1988) Paleoceanography 3, 343–360. 6. Raymo, M. E., Ruddiman, W. F., Shackleton, N. J. & Oppo, D. W. (1990) Earth Planet. Sci. Lett. 97, 353–368. 7. Barnola, J. M., Raynaud, D., Korotkevich, V. S. & Lorius, C. (1987) Nature (London) 329, 408–414. 8. Chappellaz, J., Barnola, J. M., Raynaud, D., Korotkevich, Y. S. & Lorius, C. (1990) Nature (London) 345, 127–131. 9. Short, D., Mengel, J. G., Crowley, T. J., Hyde, W. T. & North, G. R. (1991) Quat. Res. 35, 157–173.
A LONG MARINE HISTORY OF CARBON CYCLE MODULATION BY ORBITAL-CLIMATIC CHANGES
8369
10. Wigley, T. M. L. (1976) Nature (London) 264, 629–631. 11. Imbrie, J. & Imbrie, J. Z. (1980) Science 207, 943–953. 12. Savin, S. M. (1977) Annu. Rev. Earth Planet. Sci. 5, 319–355. 13. Wolfe, J. A. (1978) Am. Sci. 66, 694–703. 14. Spicer, R. A. & Parrish, J. T. (1990) J. Geol. Soc. (London) 147, 329–341. 15. Barron, E. J. (1983) Earth Sci. Rev. 19, 305–338. 16. Arrhenius, G. (1952) Repts. Swed. Deep Sea Exped. 5, 1–227. 17. Hovan, S. A., Rea, D. K., Pisias, N. G. & Shackleton, N. J. (1989) Nature (London) 340, 294–296. 18. Shackleton, N. J., Berger, A. & Peltier, W. R. (1990) Trans. R. Soc. Edinburgh Earth Sci. 81, 251–261. 19. Tiedemann, R., Sarnthein, M. &. Shackleton, N. J. (1994) Paleoceanography 9, 619–638. 20. Oppo, D., Fairbanks, R. G., Gordon, A. L. & Shackleton, N. J. (1990) Paleoceanography 5, 43–54. 21. Berger, W. H. (1973) J. Foraminiferal. Res. 3, 187–195. 22. Shackleton, N. J. (1977) in The Fate of Fossil Fuel CO2 in the Oceans, eds. Andersen, N. R. & Malahoff, A. (Plenum, New York), pp. 401–427. 23. Curry, W. B. & Crowley, T. J. (1987) Paleoceanography 2, 489–517. 24. Broecker, W. S. (1982) Prog. Oceanogr. 11, 151–197. 25. Sarmiento, J. L. & Toggweiler, J. R. (1984) Nature (London) 308, 621–624. 26. Martin, J. H. (1990) Paleoceanography 5, 1–14. 27. Boyle, E. A. (1988) J. Geophys. Res. 93, 15701–15715. 28. Ruddiman, W. F., Raymo, M. & McIntyre, A. (1986) Earth Planet. Sci. Lett. 80, 117–129. 29. Park, J. & Maasch, K. (1993) J. Geophys. Res. 98, 447–461. 30. Muller, P. J. & Suess, E. (1979) Deep Sea Res. 26A, 1347–1362. 31. Pedersen, T. F. (1983) Geology 11, 16–19. 32. Lyle, M. (1988) Nature (London) 335, 529–532. 33. Sarnthein, M., Winn, K., Duplessy, J.-C. & Fontugne, M. R. (1988) Paleoceanography 3, 361–399. 34. Berger, A., Loutre, M. F. & Dehant, V. (1989) Paleoceanography 4, 555–564. 35. Berger, A. & Loutre, M. F. (1991) Quat. Sci. Rev. 10, 297–317. 36. Berger, A., Loutre, M. F. & Laskar, J. (1992) Science 255, 560–566. 37. Bloemendal, J. & deMenocal, P. (1989) Nature (London) 342, 897–900. 38. deMenocal, P. B. (1995) Science 270, 53–59. 39. Herbert, T. D. & Mayer, L. A. (1991) J. Sediment. Petrol. 61, 1089–1108. 40. Cotillon, P. & Rio, M. (1984) Init. Rep. Deep Sea Drill. Proj. 77, 339–378. 41. Schwarzacher, W. (1987) Earth Planet. Sci. Lett. 84, 22–26. 42. Hilgen, F. J. (1991) Earth Planet. Sci. Lett. 107, 349–368. 43. Gale, A. S. (1995) in Orbital Forcing Timescales and Cyclostratigraphy, Geol. Soc. Special Publication 85, eds. House, M. R. & Gale, A. S. (Geol. Soc., London), pp. 177–197. 44. Shackleton, N. J., Curry, W. B. & Richter, C. (1995) Proc. Ocean Drill. Prog. Init. Repts. 154, 1–1111. 45. Mead, G. A., Tauxe, L. & LaBreque, J. L. (1986) Paleoceanography 1, 273–284. 46. Moore, T. C., Jr., Pisias, N. G. & Dunn, D. A. (1982) Mar. Geol. 46, 217–233. 47. Woodruff, F. & Savin, S. M. (1991) Paleoceanography 6, 755–806. 48. Herbert, T. D. & D'Hondt, S. L. (1990) Earth Planet. Sci. Lett. 99, 263–275. 49. Huang, Z., Boyd, R. & O'Connell, S. (1992) Proc. Ocean Drill. Prog. Sci. Res. 122, 259–277. 50. Berggren, W. A., Kent, D. V., Swisher, C. C., III & Aubry, M.-P. (1995) in Geochronology, Time Scales and Global Stratigraphic Correlation, Soc. Econ. Paleontol. Mineral. Spec. Publ. 54, eds. Berggren, W. A., Kent, D. V., Aubry, M.-P. & Hardenbol, J. (Soc. Econ. Paleontol. Mineral., Tulsa, OK), pp. 129–212. 51. ten Kate, W. G. H. & Sprenger, A. (1993) Sediment. Geol. 87, 69–101. 52. Barron, E. J., Arthur, M. A. & Kauffman, E. G. (1985) Earth Planet. Sci. Lett. 72, 327–340. 53. Barker, P. F. & Kennett, J. P. (1988) Proc. Ocean Drill. Prog. Init. Rep. 113, 293–328. 54. Herbert, T. D., Stallard, R. F. & Fischer, A. G. (1986) Paleoceanography 1, 495–506. 55. Park, J. & Herbert, T. D. (1987) J. Geophys. Res. 92, 14027–14040. 56. Premoli Silva, I., Ripepe, M. & Tornhaghi, M. E. (1989) Terra Nova 1, 443–448. 57. Berger, W. H. (1970) Geol. Soc. Am. Bull. 81, 1385–1402. 58. Rossignol-Strick, M. (1985) Palaeogeogr. Palaeoclimatol. Palaeoecol. 49, 237–263. 59. Oglesby, R. & Park, J. (1989) J. Geophys. Res. 94, 14793–14816. 60. Lasaga, A. C., Berner, R. A. & Garrels, R. M. (1985) in The Carbon Cycle and Atmospheric CO2: Natural Variations Archean to Present, eds. Sundquist, E. T. & Broecker, W. S. (American Geophysical Union, Washington, DC), Monogr. 32, pp. 397–411. 61. Sarmiento, J. L., Herbert, T. D. & Toggweiler, J. R. (1988) Global Biogeochem. Cycles 2, 427–444.
DEPENDENCE OF GLOBAL TEMPERATURES ON ATMOSPHERIC CO2 AND SOLAR IRRADIANCE
8370
Proc. Natl. Acad. Sci. USA Vol. 94, pp. 8370–8377, August 1997 Colloquium Paper This paper was presented at a colloquium entitled “Carbon Dioxide and Climate Change,” organized by Charles D. Keeling, held Nov. 13–15, 1995, at the National Academy of Sciences, Irvine, CA.
Dependence of global temperatures on atmospheric CO2 and solar irradiance DAVID J. THOMSON Mathematics of Communications Research Department, Bell Laboratories, Murray Hill, NJ 07974
© 1997 by The National Academy of Sciences 0027-8424/97/948370-8$2.00/0 PNAS is available online at http://www.pnas.org. ABSTRACT Changes in global average temperatures and of the seasonal cycle are strongly coupled to the concentration of atmospheric CO2. I estimate transfer functions from changes in atmospheric CO2 and from changes in solar irradiance to hemispheric temperatures that have been corrected for the effects of precession. They show that changes from CO2 over the last century are about three times larger than those from changes in solar irradiance. The increase in global average temperature during the last century is at least 20 times the SD of the residual temperature series left when the effects of CO2 and changes in solar irradiance are subtracted. Although it is generally conceded that the average surface temperature of the Earth has increased by about 0.6°C during the last century, there is little agreement on the cause of this warming. The primary cause of this disagreement is uncertainty about the relative contribution to this warming of atmospheric CO2 and changes in solar irradiance. The purpose of this paper is to describe some data analysis that may help to discriminate between solar and CO2 effects, and to give estimates of the relative magnitudes of these two effects. The difference between analysis such as those described in ref. 1 and those here is that this data analysis is based on deseasonalized temperature time series where the effects of precession were included.* The detection of precession in instrumental temperature series and the necessity of including it when removing the annual cycle from temperature data was demonstrated in ref. 2. I also describe some of the statistical peculiarities and limitations of these data series and suggest where better data are needed. The paper begins with a discussion of the data being analyzed and, to delineate the issues, presents some ordinary least-squares fits of the temperature data with atmospheric CO2 concentration and changes in solar irradiance. I next discuss the mathematical methods used and describe some statistical properties of the various data series. This is followed by some simple estimates of the transfer functions between fossil fuel consumption and atmospheric CO2 levels and from CO2 levels and changes in solar irradiance to temperature. The penultimate section summarizes recent findings on destabilization of the annual cycle, followed by conclusions. In these analyses I do not directly take into account the effects of stratospheric aerosols nor various internal feedback mechanisms such as cloud cover. Stratospheric aerosols are generally believed (3, 4) to result in cooling, so their omission makes the estimates for sensitivity conservative. Similarly, while a detailed understanding of internal feedback mechanisms, such as water vapor, is necessary to predict temperature changes from first principles, one may use measurements to assess the general climate response to forcing without having to consider the internal feedbacks explicitly, much as one can design a filter using operational amplifiers without detailed consideration of the quantum mechanics, or even current flow, in the individual transistors in the amplifiers. The estimates given here depend neither on general circulation models nor on the assumptions that underlie such models. The transfer functions are estimated directly from observations of temperature and CO2 and, for solar irradiance, a physically based proxy data series. Data Sources and Preparation For measurements of surface air temperature I use the low-pass filtered Jones–Wigley Land plus Marine data (5) shown in figures 9 and 10 of ref. 2. The bandwidth of the low-pass filter was 0.5 cycle/year so the Nyquist rate is one sample per year. These series differ from the ones usually seen in two important aspects: First, I replaced the standard “deseasonalizing” procedure used to produce temperature anomaly series with a projection filter separation into low-pass, annual, and high-frequency components so, implicitly, the usual “box-car” running-mean smoother has been replaced with a low-pass filter. Second, instead of assuming a constant amplitude climatology with a period of 1 calendar year, I allowed the phase of the annual components to track the observed phase. Thus, the significant changes in the annual cycle caused by the changing balance between direct insolation, periodic at one cycle per tropical year, and transported heat, periodic at one cycle per anomalistic year, has been removed from the data, eliminating the spurious monthly trends associated with temperature anomaly series (2, 6). Note that, although the time-resolution of these series is one year, the series is as smooth as that given by the usual “boxcar” procedure at decade-scale resolution. The “Global” temperature series used here is the arithmetic average of the Northern and Southern Hemisphere series. The Northern Hemisphere, Southern Hemisphere, and Global series are denoted by Tn(t), Ts(t), and Tg(t) respectively, with t the Gregorian calendar date. In this paper I use the average temperature over the 65 years from 1854 to 1918 as a base reference. There are several reasons to prefer this period to the usual 1951–1980 reference period. First, based on the sunspot record, solar activity in the 1854–1918 period appears to be representative of the 245-year available record and 65 years covers most of the 88-year Gleisberg cycle (7). Second, median fossil fuel consumption in this period was only about 6% of the current rate and the
*Briefly, the annual temperature cycle at a given latitude consists of direct insolation plus transport effects. The direct insolation components vary with the tropical year, the time from equinox to equinox, 365.2422 days, while the net radiation received by earth and hence the mean transport vary as the anomalistic year, the time from perihelion to perihelion, or 365.2596 days. Precession, loosely defined, is the change in the longitude of perihelion measured from the vernal equinox.
DEPENDENCE OF GLOBAL TEMPERATURES ON ATMOSPHERIC CO2 AND SOLAR IRRADIANCE
8371
concentration of atmospheric CO2 increased by 4.7% between 1854 and 1918 (8), only one-quarter of the 1951–1980 rate. Third, chloroflourocarbons and similar ozone-depleting chemicals were not yet in use, and as there is considerable evidence (9, 10) for stratospheric control of climate, it is desirable to use a reference period before changes in stratospheric chemistry by chloroflourocarbons began. Fourth, the major erratic changes in the timing of the seasons described in ref. 2 appear to have begun about 1920, so the early reference period is largely free of their effects. Opposed to these considerations are poorer spatial coverage of the temperature series and the presence of several major volcanic events (11) during the earlier period. There are no solar irradiance measurements without the confounding influence of the atmosphere in the early period and only a few near the end of the 1951–1980 period. Keeling's CO2 measurements began in 1958, so the 1951–1980 period has better CO2 data than the earlier period. The average and median values of the 1854–1918 data are 171.9 mK and 180.0 mK (Northern Hemisphere) and 150.7 mK and 159.0 mK (Southern Hemisphere) below the 1951–1980 reference. Because the SD (σ) of the raw Tg(t) series during the 65-year reference period is 57.7 mK, the 1990 temperature of 635 mK above the base temperature is, at a minimum, an 11 σ increase. The CO2 data are as listed in table A.6 of ref. 8 up to 1955, and the Mauna Loa averages through 1994 from the Oak Ridge National Laboratory data set ndp001r5 (12) since 1955. The early data has been interpolated from irregular and inhomogeneous observations and, statistically, is too smooth. I denote this series by C(t), and log2CO2(t) by CL(t). [Radiation theory predicts that temperature is a logarithmic function of atmospheric CO2 concentration (13). Use of the base 2 logarithm of the CO2 data gives coefficients that directly describe the effect of doubling CO2.] I have also used Marland's fossil-fuel production series (14), denoted F(t), as an adjunct to the CO2 measurements. This series starts in 1860, later than the temperature data, but the early data appear to be better than the corresponding CO2 data. The combination of these data with the CO2 data is described later. For solar irradiance I used the Foucal–Lean (15) reconstruction. This series, L(t), is independent of the temperature data, matches direct solar irradiance measurements since they have been available, and is a reconstruction from other solar measurements before then. I emphasize, however, that this series is a proxy, not direct measurements, so that inferences drawn from it may not have the reliability of inferences from direct observations. Changes in the period of the sunspot cycle were suggested (16) as a solar irradiance proxy. Although there is a high apparent correlation between this proxy and a heavily smoothed version of the Hansen temperature series (17), this correlation is not reliable. The jackknife variance (see below) of the low-frequency coherence between the temperature and the sunspot period is large, more than four times that expected under Gaussian theory. Because of this, lack of a physical basis for the proxy, and the failure of other statistical tests described in ref. 2, this proxy is not used here. Least-Squares Fits Both CO2 and changes in solar irradiance have been invoked to explain the observed increase in global temperature. Fig. 1 shows the filtered Jones–Wigley global temperature series, Tg(t) together with a least-squares fit to it using C(t) plus a constant, and a second leastsquares fit to Tg(t) using L(t) plus a constant. Each of these fits explains more than 75% of the variance over the full 1854–1990 interval (the residual SDs are 74.2 mK and 83.8 mK, respectively). Including both C(t) and L(t) simultaneously as explanatory variables further reduces the residual SD to 62.5 mK. Nearly 87% of the variance is explained, and both partial F statistics (18) are highly significant, so neither variable can be dropped. Examining these residuals further, one finds that their autocorrelation at a 1-year lag is 0.914 so that the conditions required for the Gauss–Markov theorem (the basis of least-squares) to be valid (19) are not satisfied. This is not simply a technical mathematical quibble, but indicates serious fitting problems whose existence may be verified by repeating the fitting process over different time intervals and observing the change in the estimated coefficients. For example, a least-squares fit to Tg(t) with just the proxy solar irradiance L(t) and a constant the interval 1854–1918 gives a negative temperature response to increasing proxy solar irradiance. Consequently, one must use statistical time-series methods that are reliable when serially correlated residuals are present.
FIG. 1. The Jones–Wigley global temperature series, low-pass filtered to a time resolution of 1 year, with independent least-squares fits to solar irradiance changes and to atmospheric CO2. Mathematical Preliminaries Many of the problems in the analysis of climate data require new methods for time-series analysis. Most of the commonly used timeseries methods were derived under the assumption of stationarity, that is, their derivations assume that the statistics of the observed process are independent of the choice of time origin. The climate seems to be nonstationary; over a few years the annual cycle of the seasons makes cyclostationary (or periodically correlated) processes a better model than stationary processes, implying that Fourier transforms of frequencies offset by multiples of one cycle per year will be correlated. On longer time scales, evolution of the Earth's orbit obviously results in vast shifts in the climate, and I recently have shown (2) that these changes in the orbit must be considered in the analysis of instrumental data series as well. Here these effects are accounted for in the way the data were filtered. Because both solar and CO2 effects alter the seasonal cycle as well as low frequencies, I have not removed common modulation terms at 0 and 1 cycle/year. Finally, anthropogenic changes are altering the composition of the atmosphere and the climate system. Thus, analysis methods that contain an implicit assumption of stationarity should be used with caution.
DEPENDENCE OF GLOBAL TEMPERATURES ON ATMOSPHERIC CO2 AND SOLAR IRRADIANCE
8372
Similarly, confidence intervals for parameters derived from time series are often based on an assumed Gaussian distribution. It appears that this assumption is invalid for climate data: there are outliers, predominantly from volcanic effects, and, surprisingly, some quantities appear to be more stable than one would expect from a Gaussian distribution. Even discounting the intrinsic complications of the climate and solar data, the available data is far from homogeneous. There are, for example, no direct measurements of solar irradiance, and only a few direct CO2 measurements before 1958. In addition, the spatial coverage of the temperature averages has changed markedly. Fluctuations in the fossil-fuel production series (14) appear to be roughly proportional to the level of production, so the absolute errors in the 19th century were lower than they are at present. The methods used in this paper are a mixture of time-domain and frequency-domain techniques. The multiple-window (20, 21, 22) method of spectrum estimation is used as an indirect basis for most estimates. This basic method is formally extended in several ways: Repeating the analysis with each window omitted in turn, or jackknifing over windows (23), gives estimates of the variances of spectra and coherences that do not depend on a Gaussian assumption. Quadratic-inverse theory (24) tests for correlations between closely spaced frequencies, and a variant (25) allows checking for nonstationary effects. The singular value decomposition of a multiple-window spectrogram (21) provides insight into some of the stationarity issues. In addition, a variety of prewhitening and robust estimation methods typified by those in (26, 27) have been used throughout. The Jones–Wigley data, however, has been so carefully prepared that robust procedures detect little more than volcanic activity and do not significantly change the results. Space limitations preclude including many of these checks and comparisons, and I have omitted those that give expected results. The stochastic structure of both climate and solar data is amazingly complicated when compared to that commonly encountered in textbook examples or even most engineering problems. Because analysis methods for nonstationary time series do not appear to be well enough developed to cope with these series, nor the stochastic structure of the data well enough understood, I use methods that allow simple models for the trends and forcings while leaving nearly stationary residuals. Stochastic Properties of the Data Series Because one cannot reliably estimate the relationships between random series without understanding their internal structure, I began this study with a brief examination of the properties of the individual data series. Power spectra of the various series have the usual “red” characteristics of much geophysical data and are not shown. Most of the high-energy, low-frequency part of the spectrum is a result of the trends under consideration, and the stochastic structure is more obvious if the data are prewhitened by filtering. Because the spectra appear similar, I computed the geometric mean of the spectra of Tg(t), L(t), and CL(t), Fourier transformed this average spectrum to give an autocorrelation sequence, then computed a fourth-order autoregressive prewhitening filter from the autocorrelations. I then use this AR-4 filter on all the data series and recompute the spectra. The spectra have been computed with a multiple-window (20) estimate typically using a time-bandwidth product of 6.0 and the 10 lowest-order windows so the spectrum estimates have a nominal x-square distribution with 20 degrees of freedom. Their 5% and 95% confidence intervals have been determined by jackknifing over windows (23). The jackknife method is known analytically to give good performance over a range of distributions, to be reasonably distribution free, and, empirically, to be sensitive both to unresolved structure in the spectrum and to nonstationarity. The jackknife variances of the spectrum of the prewhitened temperature series are, on average, about 87% of that expected under Gaussian theory (28). In contrast, the jackknife variance of the spectrum of L(t), Fig. 2, is nearly 10 times that expected near the 22-year Hale cycle and considerably lower than expected around the 11-year sunspot cycle. In a stationary Gaussian series such excursions would be extremely rare. Examining the cause for the low jackknife variance, the quadratic-inverse test for unresolved spectral details (24), Fig. 3, for the Northern Hemisphere is lower than expected except at low frequencies. (The geometric mean, across frequencies, is 3.1, compared to an expected value of 4, and the estimate is below the expected value over much of the frequency range.) This implies that the sampling variations within the analysis bandwidth in the estimated spectrum are smaller than one would expect in a Gaussian random process with a constant spectrum. From this, I infer that much of the apparently random structure in the temperature data is not random but, more probably, a consequence of either deterministic forcing or internal oscillations. Because of the similarity between the statistics of the solar and temperature data, and the stationary temperature residuals (described below), solar forcing is a more likely explanation than internal oscillations. The unresolved structure test in both hemispheres have local maxima near the 22-year Hale cycle and significant maxima at low frequencies. Fig. 4 is a plot of an estimate of the frequency derivative d/df In{S(f)} [strictly, the relative quadratic-inverse coefficients, b1(f)/S(f) in the notation of ref. 24] for the two hemispheric temperature series and L(t). The higher-order coefficients have similar characteristics, but lower amplitudes. The similarity of the low-frequency features in the temperature and irradiance data lend credence to many of the suggested solar-climate relations and also imply that extracting the statistical details of these relationships will be complicated and difficult. Turning to the low-frequency maximum in Fig. 3, narrower bandwidth spectrum estimates hint at unusual activity at low frequencies in both Tn and Ts. The harmonic F-test shows moderate evidence for a periodic component with an estimated period of 180 years. The estimated period is longer than the length of the data series, and while the nominal significance level is 99.7%, this is not much higher than 1 – 1/N = 0.993, the level where one expects at least one false detection of a line in noise. However, because the ±1 – σ confidence interval on the period of 153 to 220 years includes the 208-year Suess period, the signal is detected in both hemispheres, and the magnitude-squared coherence between them is about 0.6, the possibility that there is a complicated natural signal in the data near a period of 200 years should not be dismissed. The
FIG. 2. The estimated jackknife variance of the natural log of the spectrum of solar irradiance changes. The dashed curve is the expected value and varies slowly as the adaptive weighting process changes the degrees-of-freedom of the spectrum estimate. Note the high peak near the 22-year Hale cycle and the low variance near the 11-year cycle, 0.09 cycle/year.
DEPENDENCE OF GLOBAL TEMPERATURES ON ATMOSPHERIC CO2 AND SOLAR IRRADIANCE
8373
similarity of the filtered amplitude estimates, 4.4 mK and 4.0 mK, and phases, −135.0° and −138.5° (referenced to 1900.0), in the Northern and Southern Hemispheres, respectively, further support this conclusion. Analysis of long series of 14C data (21) shows significant components at periods of 230.6, 208.4, and 199.4 years, so because the frequency difference between the latter two periods corresponds to 4,600 years, one cannot expect these components to be resolved in the 137-year series, nor the single line F-test to work reliably.
FIG. 3. Quadratic-inverse tests for unresolved spectral details for the prewhitened Northern and Southern Hemisphere temperature series. The distribution of the test statistics is approximately X24] and the dotted lines show the expected value and 95% levels. Note that the Northern Hemisphere estimate is lower than expected over much of the frequency range. A possible alternative explanation to deterministic forcing as the cause of the low variances of the temperature spectra is nonstationarity, and this was tested for using the methods in ref. 25. This test shows less variability in time than expected at periods longer than 8 years. Thus, near the 11-year solar cycle, both the tests for unresolved structure and stationarity show more stability than one would expect in a random series. Examination of the nonstationary quadratic-inverse coefficients, am(f), shows that the a2(f)s, approximately, the second time-derivatives of the spectra† are reasonably significant and similar in both hemispheres. Looking at details of the spectrum of the solar irradiance series, L(t), the predominant feature is the character of the jackknife variance (Fig. 2). There is significant excess variability near 22-year periods and a paucity around 11-year periods. The unresolved structure test is uniformly higher than expected out to a frequency of about 0.13 cycle/year, that is, until the high-frequency edge of the 11-year band, followed by a low region. The nonstationary quadratic-inverse coefficient a1(f), roughly the time-derivative of the spectrum, or S(f) in both hemispheres shows decreasing power near 22-year periods and increasing power in the 7- to 11-year period band. The singular value decomposition of the log-spectrogram described in ref. 21 suggests mild nonstationarity in Tn(t), Ts(t), and L(t) concentrated about the 104-year Suess period, and possibly related frequencies. The phases of the annual cycle of the temperature data shows similar characteristics. Note that this 104-year period is not a simple additive term, but a periodic modulation of the stochastic structure of the process.
FIG. 4. Stationary quadratic-inverse coefficients, b1(f)/S(f), for the Northern and Southern Hemispheres and solar irradiance. The dashed lines near ±0.4 are one SD from the mean for a Gaussian process with a white spectrum. Note the close tracking between all three series at low frequencies and of the temperature series generally. I emphasize that the statistical peculiarities just described are those of the individual data series; they contrast markedly with the statistics of the temperature residuals to be described. To the extent that the climate has a linear response to forcing, nonstationarity in the forcing should appear as a similar nonstationarity in the temperature series so, internal oscillations aside, removing the effects of such forcing should leave stationary residuals. The estimate of the spectrum of the log2 CO2(t) data, shown in Fig. 5. shows two important features: first, the range of the spectrum, seven decades, is larger than that of either the temperature or irradiance data, and second, the jackknife variance estimate of ln ` ( f) has a maximum of 12.06 at periods near 62 years, compared to an expected value of about 0.13, or 90 times expected. Thus the 5% to 95% confidence region for the CO2 spectrum estimated by the jackknife method covers a range of about 104, instead of the factor of 2 that would be expected from Gaussian theory. This uncertainty in the spectrum of CO2 variations carries through to the coherences and transfer function estimates and, consequently, standard fre
FIG. 5. An estimate of the spectrum of the CO2 data from 1854–1990 with 5% and 95% confidence intervals determined by jackknifing. The large uncertainty in this estimate precludes making reliable frequency-domain estimates of the transfer functions.
†In
a nonstationary process the spectrum S(f1, f2) is a function of two frequencies and several representations are useful. (Stationary processes are much simpler because the spectrum is concentrated on the line f1 = f2 and, consequently, a function of a single frequency.) A one-dimensional Fourier transform of S(f1, f2) taking the variable f1 – f2 into to gives a dynamic spectrum D(f, to) where the frequency f = (f1 + f2)/2. The “time derivatives” of the spectrum referred to are expansion coefficients of lnD(f, t) and, approximately, of the form ∂/∂t lnD(f, t).
DEPENDENCE OF GLOBAL TEMPERATURES ON ATMOSPHERIC CO2 AND SOLAR IRRADIANCE
8374
quency-domain estimates of transfer functions are not presented here. Temperature Response Models and Transfer Function Estimates I attempted to assess the simultaneous effects of solar variability and greenhouse gases by estimating transfer functions. Because both effects have been advanced as plausible explanations for the increasing temperature, one should expect that changes in solar irradiance and greenhouse gas concentrations should appear coherent as well. It has been shown (29) that, considered as a pair, changes in temperature and CO2 were coherent, with changes in temperature leading those in CO2, instead of vice-versa, as popularly supposed. Marland's fossilfuel production series, in contrast, leads both the temperature and CO2 data. Changes in solar irradiance also lead those in both CO2 and temperature. In addition to these complicated relationships, there are other difficulties to consider; the physically distributed and multicomponent nature of the problem, combined with spatially and temporally discrete data, makes the appropriate form of the transfer functions obscure. The roughly periodic character of the solar cycle makes lead-lag relations ambiguous. In an attempt to untangle these relationships, I tried several different forms of estimates; the most successful of these is described below. Here I define “successful” as low prediction error, few free parameters, and physically realistic implications. For example, I consider a negative low-frequency response to either increases in solar irradiance or CO2 unacceptable. Transfer functions implying noncausal response to changes in solar irradiance imply either that the transfer function is unacceptable, or that there is a problem with the assumed irradiance. I also require that the estimates be reasonably insensitive to linear filtering operations applied to the raw data, which eliminates many of the possible models. The rationale for this requirement is twofold: first, if there is a linear relationship between the input and output of a system, estimates of the transfer function should not change significantly when the same linear filter is applied to both input and output, and second, residuals in the series are autocorrelated, so if the filtering reduces the residual autocorrelation, the requirements of the Gauss–Markov theorem are more nearly satisfied. The estimated transfer functions are clearly not unique, but other acceptable estimates with similar prediction errors have similar implications. A simple model for the atmospheric CO2 concentration at time t is [1] C(t) = ζC(t − 1) + βF(t) + γTg(t − 1), where the coefficient ζ on the previous concentration allows for exchange and sequestration by the oceans and biosphere, β is the part of the carbon present in fossil fuels that contributes to atmospheric CO2, and γ allows for the temperature forcing described in ref. 29. Doing a robust fit with the more accurate Keeling data weighted a factor of 6 higher than the pre-1958 data gives ζˆ = 0.999974, implying that the atmosphere acts, as expected, as an integrator. The estimated coefficients are βˆ = 0.524, somewhat lower than the estimate in ref. 8 or direct estimates between CO2 and the integrated fossil fuel record. This is offset by the temperature feedback, γˆ = 1.196 GT/K. Replacing the Tg(t) term with L(t) gives a slightly poorer fit with a coefficient 0.38 GT/(W/m2). In the following paragraphs, I describe some simple timedomain estimates of the transfer functions. If one takes the simplest conceptual model [2] T(t) s0L(t) + c0CL(t) and substitutes Eq. 1 for C(t) one obtains, approximately, the equation [3] T(t) T + αT(t − 1) + sL(t) + cCL(t) with suitable definitions of the constants, and F(t) incorporated into CL(t). This can be generalized to the class of models
with suitable choices of A, P, and Q. When A = 0 the autoregressive feedback terms are absent, and the model that of a direct finite impulse response filter from L(t) and C(t) to temperature. In particular the models A, P, Q = 0, 1, 0 are the direct least-squares fits, mentioned earlier, to L(t) only; 0, 0, 1 to C(t) only; and 0, 1, 1 to both simultaneously. I choose the interval 1854–1965 to estimate the coefficients for the particular model and use the 25-year interval, 1966–1990, for validation. Reconsider the direct least-squares estimate model 0, 1, 1. The coefficients obtained by fitting from 1854–1965 are ŝ0 = 0.1115, ĉ0 = 1.984. The mean-square error in the fitting period is 6,695 (mK)2, or an rms error of 81.8 mK. This fit is not particularly impressive but, nonetheless, it suggests that changes in solar irradiance lead those in temperature by a few years. Using the data for the 25 years 1966–1990, with the coefficients found for the earlier period gives a prediction error of 39,820 (mK)2, for an rms error of 199.5 mK. Because the autocorrelations of the residuals in all these cases are large, the assumptions of the Gauss–Markov theorem are violated, and ordinary least-squares does not give reliable results. To see how badly, replace Eq. 2 with [4] ∆rT(t) d + s0∆rL(t) + c∆rCL(t), where ∆rT(t) = T(t) − rT(t − 1) and similarly with L and C. Using the 1-year autocorrelation of the temperature data, r = 0.881 for all three series gives s0 = −0.0018, c = 3.121 with corresponding fitting and prediction errors of 1,369 (mK)2 and 2,436 (mK)2 respectively. Because the innovations variance of the temperature data
where B(v) 1.07 is a bias correction depending on the degrees-of-freedom (30) of the estimate and ` ( f) the estimated spectrum at frequency f, is about 1,303 (mK)2 and the variance of ∆rT is 1,835 (mK)2 the errors in Eq. 4 are far less correlated than are those of Eq. 2, so the conditions of the Gauss–Markov theorem are closer to being satisfied. Note that using this simplest of filters on the data increases the estimated sensitivity to CL by 57% and changes the sign of ŝ0. Because a decrease in average temperature as a response to an increase in solar irradiance is implausible, I reject such simple models. Direct frequency-domain estimates of the transfer functions appear to vary too rapidly to be reliably estimated from the available data as their Fourier transforms are significantly noncausal. For example, the impulse response estimated between Tn(t) and L(t) is noncausal with a width about the length of a solar cycle. This is an artifact of the near periodicity of the solar cycle as, without other knowledge or assumptions, one cannot assign the response to a periodic process to a given time without ambiguities of N periods. Similar problems are common in transfer function estimation problems ranging from magnetotellurics (31) to the design of telephone equalizers. The estimated impulse response between CL(t) and T(t) is causal and oscillatory with about a 20-year duration. If this
DEPENDENCE OF GLOBAL TEMPERATURES ON ATMOSPHERIC CO2 AND SOLAR IRRADIANCE
8375
characteristic persists with improved early CO2 data, it implies that large transient responses may be possible. In addition to thermal inertia, the observed feedback from temperature to CO2 also generates persistence. To model this together with the effects of CO2 and solar irradiance changes on the temperature record, I fit the filtered Jones–Wigley series as a 1, 1, 1 model, explicitly Eq. 3. Here the feedback term αT(t − 1) results in a rational (as opposed to polynomial) approximation of the transfer functions allowing more rapid low-frequency changes in them than possible with simple models, it includes some persistence, and generally permits simple feedback effects. Direct least-squares estimates of the coefficients and some performance measures are given in Table 1. To assess the model's predictive power I start at 1966 and iterate forward to t = 1990 using the observed solar irradiance and CO2 data. The predictions were run in two ways: in the first, the temperature data from the previous year was used; in the second, the model was started with the 1965 temperatures and the prediction of the previous year's temperature used from there on. The 1966– errors in Table 1 refer to the first case. The second is not as accurate, but is not ridiculously inaccurate either; for example, the 25-year prediction of the 1990 Northern Hemisphere temperature is 622 mK, not too dissimilar to the 648 mK observed. The data, fit, and extrapolations obtained with these models are shown in Fig. 6 (Northern Hemisphere), and Fig. 7 (Southern Hemisphere). The general 25-year prediction agrees reasonably well with the observed temperature in level, if not in exact details. It might be noted that the autoregressive nature of the feedback term in these models causes them to be biased slightly toward zero. Examination of the residuals from this model shows them to be mundane and the exotic characteristics of the individual series missing. First, the residuals show no significant departures from normality; their range is ±2.35 σ for the global series, there is no detectable skewness, and the standardized fourth moment is 2.65 for the 1854–1965 data, and similar in the validation interval. The Southern Hemisphere residuals are slightly long-tailed. The extreme residual, −107 mK, is a result of the Krakatau eruption in 1883. Thus the errors of 30.4 mK and 26.9 mK obtained for the global temperature with a model having only four parameters imply that the 1990 temperature of 636 mK is about 21 SD above the 1854–1918 base level measured on the scale of the residual SD. The feedback term, αT(t − 1) in Eq. 3, reduces this to about a 9.3σ increase above what persistence would normally generate. The probability of a Gaussian random variable being 9σ from the mean is about 10−19. Second, the range of the spectrum of the residuals from this model is only 18, just slightly larger than expected for a white-noise process. This spectrum hints at a weak 4.6-year echo-like term. Table 1. The estimated parameters for Eq. 3 for the three temperature series considered Temperature series α s c Northern Hemisphere 0.8965 0.0106 0.200 Southern Hemisphere 0.8122 0.0033 0.426 0.8817 0.0069 0.255 Global
–1965, mK 37.1 32.4 30.4
1966–, mK 34.7 25.1 26.9
The column headed –1965 gives the rms error in the modeling period, 1854–1965. The column headed 1966– gives the rms errors for a 1-year prediction using the parameters estimated from the 1854–1965 and averaged over the 25-year validation interval, 1966–1990.
FIG. 6. The Northern Hemisphere temperature data, solid line, and fits. The model coefficients were determined using data before 1965 with the fit shown by the dashed line. After 1965 the fit includes CL(t), L(t) and the previous year's temperature. The smoother line, short dashes, shows the prediction starting with the 1965 temperature using only CL(t) and L(t). Third, the residuals were tested for stationarity using the quadratic-inverse methods of ref. 25 and, generally, see Fig. 8, the test statistics are slightly less than their expected value, suggesting that the residuals due not contain any serious nonstationary terms. There are, nonetheless, some features of interest. The “time derivative” of the spectrum a1(f)/a0(f) is mostly negative, implying that the residual spectrum is decreasing as a function of time. This probably reflects nothing more than the improvements in spatial coverage and data quality that have occurred over the course of the record. The tests for both hemispheres show significant nonstationarity at periods of 4 years, perhaps an artifact of leap years in the original monthly averaging process. In addition, the Southern Hemisphere test has a significant peak in the El-Nino band, and the Northern Hemisphere data are quite nonstationary at the quasibiennial oscillation frequency. At low frequencies, the test statistics from both hemispheres are well below their expected value. Thus the tests for stationarity reveal no evidence for long-term unexplained variability.
FIG. 7. The Southern Hemisphere temperature and fits. The curves are as in Fig. 6.
DEPENDENCE OF GLOBAL TEMPERATURES ON ATMOSPHERIC CO2 AND SOLAR IRRADIANCE
8376
FIG. 8. Quadratic-inverse tests for stationarity of the Northern Hemisphere (solid) and Southern Hemisphere (dashed) temperature residuals from model Eq. 3. The test statistics have an approximate distribution and have been slightly smoothed for plotting. In both hemispheres the test statistic at a 4-year period exceeds the 98% level. The transfer functions for this model are
for solar irradiance, and similarly for CL. The long-term response is given by the low-frequency response; evaluated at f = 0 with the coefficients estimated above one obtains 0.109K/(W/m2) for solar irradiance changes and 2.075 K/(2 × CO2) To simplify the comparison = 22,536 (mK)2 and of the different units one may examine the low-frequency variance explained for the two components and = 6,067 (mK)2 in the Northern Hemisphere, with corresponding figures of 28,320 (mK)2 and 317 (mK)2 in the Southern Hemisphere. Thus CO2 explains over 3 times as much variance in changes in solar irradiance in the Northern Hemisphere, over 100 times as much in the Southern. Using simply the change in CO2 concentration from 287.7 in 1854 to 352.7 parts per million in 1990 the lowfrequency transfer function predicts an increase of 0.61 K for CO2 while an irradiance change of 2.4 W/m2 (from 1365.7 to 1368.1 W/m2 results in an estimated temperature change of 0.26 K. Given the great improvement that the addition of the feedback term makes on the direct least-squares estimates, it seems reasonable to investigate related models. A suite of such models, incorporating a range of time delays and prewhitening filters was tested. None of these simple models significantly improve on Eq. 3. Changes in the Timing of the Annual Cycle I presented evidence (2) that the frequency of the annual temperature cycle at many locations is closer to the anomalistic year than it is to the tropical year, and, in addition, the character of the phase of the annual cycle began to change rapidly near the middle of this century with the average change in phase coherent with the changes in atmospheric concentration of CO2. I also suggested that the cause of the rapid phase shift was the changing balance between direct insolation at a given latitude and transported energy. Recall that, at temperate latitudes, the periodicity of insolation is the tropical year and peaks at the summer solstice. The total radiation received by the earth, on the other hand, has the periodicity of the anomalistic year and is maximum at perihelion, currently in early January. In the Northern Hemisphere the phases of the two components are thus nearly opposed and, consequently, the phase of the resultant is easily perturbed. The Sable Island example shown in ref. 6 demonstrates that the proposed mechanism can explain the rapidly changing phase seen at individual stations. The geographic distribution of these phase changes is available in (32, 33), and confirmatory changes are directly observable in the CO2 record (34). The observed phase changes must be taken seriously as an indicator of climate change caused by increasing concentrations of atmospheric CO2. First, the increasing variability of the phases and the change in distribution (Fig. 4 of ref. 2) is so large that the probability of such an occurrence in a stationary process is nearly zero. Second, because the effects of observational errors and similar noise-like effects on the data will be uncorrelated between the low frequencies where the increase in temperature is observed and the annual cycle, one must consider them as independent evidence for a changing climate. Discussion and Conclusions More effort must be made to obtain improved CO2 data for the 19th and early 20th centuries if possible. Similarly, extensive effort to develop better, or independent, solar irradiance proxies is desirable. Analysis of the low-frequency and annual parts of the temperature records yield at least three largely independent indicators of climate change; the change in distribution of individual station phase trends about the mean, the change in average phase, and the increase in average temperature. All are unprecedented in the instrumental record. The probability of the observed changes occurring through natural, as opposed to anthropogenic, causes appears to be exceedingly small. First, although a major ice age causes a larger temperature change than has happened so far in response to CO2, the temperature increase that occurred between 1920 and 1990 would have taken more than 2,000 years even at the historically rapid rate of the last deglaciation. During deglaciation, transient warming in the North Atlantic after a Heinrich event was faster (35), but there is no evidence for a Heinrich event during the last few centuries. Second, internal climate oscillations, shifts in storm tracks, and the like obviously can change local and regional climate by much larger amounts than observed in the hemispheric averages. These regional fluctuations appear to be superimposed on the general global trend. Additionally, such internal oscillations pro
DEPENDENCE OF GLOBAL TEMPERATURES ON ATMOSPHERIC CO2 AND SOLAR IRRADIANCE
8377
duce warm and cool regions that interchange over decade to century time scales (32, 36), but whose effects largely cancel in hemispheric averages. Third, while there is reasonable evidence for greater climate variability during the Holocene than has been observed during the period where instrumental data are available (37, 38), there is no evidence in the statistics that a major unidentified source of natural variation is present during the instrumental record. Such a source would have to mimic, perversely, either solar irradiance changes or the changes in atmospheric CO2 to cause the observed temperature changes and to be mistaken for them. Similarly, while mindful of the many caveats on data quality, spatial coverage, etc. given in ref. 1, the appearance of possible leap-year artifacts at a level below 10 mK in the residuals suggests that the data cannot be as untrustworthy as is occasionally implied. The residual temperature variation remaining once the known effects of precession, solar irradiance changes, and atmospheric CO2 concentration are removed bound unknown effects to about 200 mK peak-to peak in the hemispheric average series during the last century. Consider the null hypothesis that the observed temperature fluctuations and atmospheric CO2 levels are independent: The probability that the hemispheric temperatures would fluctuate purely by chance in such a way to produce the observed coherences with CO2 is exceedingly low. Given that the records encompass more than a century, the probability is so low that one would not expect to see such an event by chance during the age of the earth. The probability of the observed coherence between atmospheric CO2 and changes in the timing of the seasons shown in figure 13 of ref. 2 without a causal connection is similarly low. Consequently one must strongly reject the hypothesis of independence between atmospheric CO2 and temperature. The alternative hypothesis, that increasing levels of atmospheric CO2 plus a slight change in solar irradiance are causally responsible for the observed changes in temperature, in contrast, results in test statistics that are ordinary in every way. Because major changes in climate as a response to human use of fossil fuels have been predicted for more than a century (39, 40), their detection can hardly be considered surprising. From examining the data records I conclude: Changes in solar irradiance explain perhaps one-quarter of the increase in temperature during the last century. The changes in atmospheric CO2 concentration resulting from human consumption of fossil fuels cause most of both the temperature increase and the changes in the seasonal cycle. 1. Houghton, J. T., Meira Fihlo, L. G., Callender, B. A., Harris, N., Kattenberg, A. & Maskell, K. (1996) Climate Change 1995 (Cambridge Univ. Press, Cambridge, U.K.). 2. Thomson, D. J. (1995) Science 268, 59–68. 3. Santer, B. D., Taylor, K. E., Wigley, T. M. L., Penner, J. E., Jones, P. D. & Cubasch, U. (1995) PCMDI Rep. 21 (Lawrence Livermore Nat. Lab., Livermore, CA). 4. Schwartz, S. E. & Andrae, M. O. (1996) Science 272, 1121–1122. 5. Jones, P. D. (1994) J. Climate 7, 1794–1802. 6. Karl, T. R., Jones, P. D., Knight, R. W., White, O. R., Mende, W., Beer, J. & Thomson, D. J. (1996) Science 271, 1879–1883. 7. Sonett, C. P., Giampapa, M. S. & Matthews, M. S., eds (1991) The Sun in Time (Univ. of Arizona Press, Tucson). 8. Keeling, C. D., Bacastow, R. B., Carter, A. F., Piper, S. C., Whorf, T. P., Heimann, M., Mook, W. G. & Roeloffzen, H. (1989) in Aspects of Climate Variability in the Pacific and the Western Americas, ed. Peterson, D.H. (American Geophysical Union, Washington, DC), pp. 165–363. 9. Robock, A. (1996) Science 272, 972–973. 10. Haigh, J. D. (1996) Science 272, 981–986. 11. Bradley, R. S. & Jones, P. D. (1992) Climate Since AD 1500 (Routledge, London). 12. Boden, T. A., Kaiser, D. P., Sepanski, R. J. & Stoss, F. W. (1994) Trends 93: A Compendium of Data on Global Change Pub. ORNL/CDIAC-65, (Oak Ridge National Laboratory, Oak Ridge, TN). 13. Manabe, S. & Wetherald, R. T. (1967) J. Atmos. Sci. 24, 241–259. 14. Marland, G., Anders, R. J. & Boden, T. A. (1994) Trends '93: A Compendium of Data on Global Change Pub. ORNL/CDIAC-65, (Oak Ridge National Laboratory, Oak Ridge, TN), pp. 505–584. 15. Foukal, P. & Lean, J. (1990) Science 247, 347–357. 16. Friis-Christensen, E. & Lassen, K. (1991) Science 254, 698–700. 17. Hansen, J. & Lebedeff, S. (1987) J. Geophys. Res. 92, 13345–13372. 18. Draper, N. R. & Smith, H. (1981) Applied Regression Analysis (Wiley, New York). 19. Seber, G. A. F. (1977) Linear Regression Analysis (Wiley, New York). 20. Thomson, D. J. (1982) Proc. IEEE 70, 1055–1096. 21. Thomson, D. J. (1990) Philos. Trans. R. Soc. London A 330, 601–616. 22. Percival, D. B. & Walden, A. T. (1993) Spectral Analysis for Physical Applications: Multitaper and Conventional Univariate Techniques (Cambridge Univ. Press, Cambridge, U.K.). 23. Thomson, D. J. & Chave, A. D. (1991) in Advances in Spectrum Analysis, ed. Haykin, S. (Prentice–Hall, Englewood Cliffs, NJ), Chapter 2. 24. Thomson, D. J. (1990) Philos. Trans. R. Soc. London A 332, 539–597. 25. Thomson, D. J. (1993) Proc. SPIE 2027, 236–244. 26. Thomson, D. J. (1977) Bell System Tech. J. 56, 1769–1815 and 1983–2005. 27. Kleiner, B., Martin, R. D. & Thomson, D. J. (1979) J. R. Stat. Soc. B-41, 313–351. 28. Thomson, D. J. (1994) Proc. IEEE Int. Conf. on Acoustics, Speech and Signal Processing, Adelaide, Australia 6, 73–76. 29. Kuo, C., Lindberg, C. R. & Thomson, D. J. (1990) Nature (London) 343, 709–714. 30. Jones, R. H. (1976) J. Am. Stat. Assoc. 71, 386–388. 31. Larsen, J. C. (1992) Philos. Trans. R. Soc. London A 338, 169–236. 32. Mann, M. E. & Park, J. (1994) J. Geophys. Res. 99, 25819–25833. 33. Mann, M. E., Park, J. & Bradley, R. S. (1996) Geophys. Res. Let. 23, 1111–1114. 34. Keeling, C. D., Chin, J. F. S. & Whorf, T. P. (1996) Nature (London) 382, 146–149. 35. Bond, G., Broecker, W., Johnsen, S., McManus, J., Labeyrie, L., Jouzel, J. & Bonani, G. (1993) Nature (London) 365, 143–147. 36. Mann, M. E. & Park, J. (1995) Nature (London) 378, 266–270. 37. Overpeck, J. T. (1996) Science 271, 1820–1821. 38. O' Brien, S. R., Mayewski, P. A., Meeker, L. D., Meese, D. A., Twickler, M. S. & Whillow, S. I. (1995) Science 270, 1962–1964. 39. Arrhenius, S. (1996) Philos. Mag. 41, 237–276. 40. Uppenbrink, J. (1996) Science 272, 1122.
DEPENDENCE OF GLOBAL TEMPERATURES ON ATMOSPHERIC CO2 AND SOLAR IRRADIANCE
8378
