AS scientific debate continues over the causes, mechanisms, and extent of global warming, the nations of the world have begun acting on the plausible assumption that human activities, particularly the release of significant amounts of greenhouse gases into the atmosphere, are leading to global warming. While not conclusive, evidence has been mounting that human-induced climate change is occurring.
At the Kyoto climate conference in December 1997, policymakers, climate experts, and industrial leaders came together to seriously consider future global climate. Their goal: to negotiate limits on the future release into the atmosphere of carbon dioxide (CO2), the most plentiful greenhouse gas, in order to prevent further human-driven climate change. The conferees took to heart the latest scientific findings on global climate processes that have resulted, in large part, from climate models.
Climate models are a major tool that scientists use to understand the complex web of climate mechanisms. However, important uncertainties remain. Scientists at Lawrence Livermore are actively seeking answers to the questions of how much climate is changing and how much CO2 is stored naturally in the atmosphere and oceans, ultimately helping to resolve the larger uncertainty: what effect human activities have on our environment.
Climate modeling began in the 1960s, when the National Oceanic and Atmospheric Administration initiated development of general circulation models--complex mathematical descriptions of the physics and dynamics of the atmosphere and oceans. Since then, climate scientists here and elsewhere have been refining these models and formulating an entire hierarchy of other models to elucidate global climate. Their ultimate goal is a virtual global climate system, a model that will allow them to confidently predict future global and regional climate.
Development of such a model is far from complete. But with global climate change at the forefront of public attention and policymaking agendas, climate scientists must provide input to policymakers based on today's imperfect models while pressing forward to improve them. Indeed, they are working under some urgency, for the world now appears ready to take action against global warming.
One important question that models of the carbon cycle answer is what the future atmospheric concentrations of CO2 may be. Carbon dioxide, a major greenhouse gas, is important to Earth's energy balance and climate. Because CO2 concentrations have been increasing rapidly, climatologists have pinpointed CO2 as a major cause of the increase in global temperatures observed since 1860, when complete temperature records began. It is no wonder, then, that the global carbon cycle has become a vigorous area of study for the Climate System Modeling Group in Lawrence Livermore's Earth and Environmental Sciences Directorate. (See S&TR, October 1996, for an overview of climate work in the Atmospheric Sciences Division, much of which stems from Livermore's multidisciplinary capabilities in scientific computation, nuclear testing, and atmospheric science.) The group also collaborates with Department of Energy sponsors and universities.

Determining Carbon's Fate
In nature, carbon is plentiful and dynamic. Its natural cycle is thought to have been in a relatively steady state for thousands of years until the industrial revolution. Then, humans perturbed the carbon cycle by clearing forests and burning fossil fuels (coal, oil, and natural gas), which increased the CO2 content of the atmosphere. How has this anthropogenic change disrupted Earth's climate? To find out, scientists are obtaining a better understanding of how the climate system functions and tracking CO2 released into the environment by humans.
Philip B. Duffy, group leader of the Climate System Modeling Group, says that their work includes modeling how a portion of anthropogenic CO2 remains in the atmosphere and how the rest of it is taken up by the oceans and terrestrial ecosystems. Tracking CO2 is neither simple nor easy: measurements of CO2 in the ocean and terrestrial biosphere are scarce; CO2 released by humans is indistinguishable from naturally occurring CO2; and the natural fluxes of carbon between ocean, atmosphere, and terrestrial biosphere are poorly understood.
Lawrence Livermore modelers have been formulating and testing a suite of models to reproduce carbon absorption, transport, and storage processes. The models cannot possibly incorporate all climate factors everywhere--even Livermore's advanced supercomputers cannot provide that resolution--so modelers must select the most important climatic factors and influences and represent them as well as possible.
The goal of the carbon cycle models is to predict the future behavior of carbon in the climate system. The models are tested by "hindcasting" the behavior of the carbon cycle from the historical past to the present. If model simulations match past records and observations, then model calculations predicting the future will merit confidence.

The Ocean Carbon Cycle
The Climate System Modeling Group's work involves both terrestrial and ocean carbon cycles, with current emphasis on the ocean. The ocean is an important regulator of climate as well as a major absorber of excess CO2. Livermore modelers, having access to powerful, next-generation computers from the Department of Energy's Accelerated Strategic Computing Initiative, are able to formulate models that represent ocean currents, eddies, and interactive processes with more fidelity than previously possible. Their most recent ocean models reflect the advance of climate simulation techniques and skills. The projects described here demonstrate the approaches that modelers use to test hypotheses about carbon cycle dynamics, understand process interactions, and refine representations of the processes to derive accurate and useful models.

Improving Convection Models
The ocean is a large "sink" for anthropogenic CO2. A model that can accurately simulate how the ocean absorbs anthropogenic CO2 is prerequisite to predicting future atmospheric CO2 concentrations and rates of global warming. But according to climate modelers Duffy and Ken Caldeira, current ocean models may overpredict the amounts of CO2 that oceans absorb. One reason is that models do not accurately represent convection, the rapid vertical mixing of water that occurs when dense water overlies less dense water. They saw the effects of that inaccuracy in models involving sea-ice formation, which is a major cause of ocean convection.
When sea ice is forming, it expels salt into surrounding surface water.
The saltier, denser water triggers convection, sinking the surface water and CO2 it contains into the depths of the ocean. Surface ocean concentrations of CO2 are thus reduced, and more CO2 transfers from the air to the sea.
In the real ocean, convection occurs in horizontal regions that are meters or hundreds of meters across. Because ocean models cannot accurately represent such small features, some significant inaccuracies can arise in their simulations. Duffy and Caldeira thought that the modeled transfer of CO2 from air to sea was probably too great because the models simulated excessive convection. If the instabilities resulting from expelled salt were treated more carefully, model results might better represent reality.
The two modelers tested their hypothesis by performing a pair of simulations that compared the standard treatment of convection to a "test" simulation in which the model's convection mechanism was partially suppressed. In the standard "control" simulation, salt released during sea-ice formation is placed in the model's top layer (which is 25 meters thick), and the model's convection mechanism mixes it into the rest of the ocean. Because the convection mechanism is clumsy, excessive convection occurs. This in turn causes excessive ocean absorption of carbon.
In their test simulation, Duffy and Caldeira dispersed expelled salt uniformly over a broader area--from the surface down to a depth of 160 meters-- and suppressed the model's convection mechanism (Figure 1). This simulation produced much more realistic results for simulated convection, salinity, circulation, and absorption of chlorofluorocarbons (CFCs).

Then CFC uptake was simulated
as an indirect test of the model's ability to simulate ocean uptake of anthropogenic CO2. No direct method is possible because anthropogenic CO2 cannot be reliably distinguished from natural CO2 in the ocean. By contrast, CFCs have no natural background concentration in the ocean. Moreover, CFC uptake is very closely related to the ocean's absorption of human-induced CO2, as Figure 2 shows.
Because of this close relationship, Caldeira and Duffy reasoned that the standard model treatment of ice formation, which results in excessive simulated ocean uptake of CFC, also produces excessive uptake of anthropogenic CO2. In addition, the improved treatment of ice formation, which produces greatly improved simulated uptake of CFCs, should also produce more accurate calculations of the uptake of anthropogenic CO2. Further simulations are planned to verify these results.

Modeling Marine-Biology Effects
Marine biological processes play a large role in ocean carbon cycle dynamics, but they have been incorporated only recently into climate models. The inherent difficulties of mathematically describing their spatial and temporal effects have been a challenge. As a result of inadequate modeling of these processes, climate scientists have not been able to study the influences of marine biology and feedback responses to them. They have not known how the processes would affect the ocean carbon cycle.
Caldeira has a project under way to investigate some of these processes and clarify some of the uncertainties surrounding ocean and marine-biology interactions. He is using recently collected remote sensing and satellite data to model the interplay between sunlight, plankton (barely moving plant and animal aquatic organisms), and the ocean's absorption of CO2.
Earlier models have established connections among solar radiation, plankton, and ocean circulation dynamics. With newly available data, Caldeira is studying these intertwinings further, considering feedback interactions in the process. For example, when sunlight penetrates the ocean layers and heats the deeper waters, convective mixing results and causes two other effects. First, CO2 in the surface waters downwells and must be replaced by CO2 transferred from the air to the sea; second, nutrients from the depths well up to the surface. Additional nutrients increase plankton growth. In time, the increased volume of plankton blocks solar penetration, so the sunlight heats only the surface waters. This inhibits CO2 downwelling, nutrient transport to the surface, and plankton growth. Does the ocean then return to its steady state, or does the cycle continue through other ecosystem dynamics? And what is the cumulative effect on the ocean's absorption of CO2?
Caldeira's study will systematically simulate the interactions among solar radiation, plankton, and ocean dynamics; the feedback resulting from those interactions; and the impacts of the feedback on the predicted response of the ocean carbon cycle to climate change. The simulations will replicate the time frame from the preindustrial ocean carbon cycle (before 1765) to the present. The simulations will also calculate future atmospheric CO2 content based on several emission scenarios that may result when CO2 stabilization policies are implemented.

CO2 by Radiocarbon Proxy
Radiocarbon, or 14C, has provided much of our knowledge about the rates of carbon exchange from the atmosphere to the oceans and land, and it yields important information about the ocean circulation. The radioactive half-life (5,730 years) of 14C is comparable to the time span taken by surface ocean water to circulate to the ocean bottom and back. This means that the spatial distribution of 14C in the water yields significant knowledge about ocean circulation--water that has not been near the surface recently has significantly lower 14C concentrations because of radioactive decay.
For climate modelers, data on 14C distributions took on additional usefulness as a result of the atmospheric nuclear tests conducted mostly in the period from 1954 through 1963. The concentration of 14C in atmospheric CO2, which had remained fairly constant for the previous thousand years, was doubled by those nuclear tests in less than 10 years. Since the end of the tests, 14C concentrations have been decreasing as atmospheric CO2 moves into other carbon reservoirs. This infiltration of nuclear-testing 14C into carbon reservoirs can provide valuable information on carbon exchange. In particular, if the rate of nuclear-testing 14C transfer from the atmosphere to the oceans and land could be accurately predicted, climate scientists would be able to predict the rate at which the oceans and land absorb human-induced CO2 because essentially the same physical rules govern the transfer processes. Thus, understanding 14C transfer is another route to determining rates of future greenhouse warming.
Models to simulate this transfer process make use of the estimated amounts of 14C created by nuclear tests; they also make use of nuclear-testing 14C data that were actually measured in the troposphere and stratosphere. If the models are accurate, the total amount of nuclear-testing 14C transferred to the oceans, land, troposphere, and stratosphere should equal the estimated inventory of nuclear-testing 14C, and the amounts transferred to the troposphere and stratosphere should approximate the measured data.
Early models simulating this transfer matched observed data only in part. In 1995, Duffy, five collaborators from Lawrence Livermore, and one from the University of Illinois tackled the problem of transfer. They used different and relatively more sophisticated multidimensional models of the ocean, stratosphere, and land, which they ran using observed data on tropospheric concentrations of 14C as the model boundary conditions. Their simulations, for the period from 1955 through 1990 (Figure 3) are well within previously recognized uncertainties. This exercise proved that contemporary models can accurately account for all the bomb-produced 14C in the climate system.

Radiocarbon by Salmon Proxy
To validate models of how the ocean absorbs nuclear-testing 14C, modelers must be able to distinguish between how much of the total 14C now in the ocean is bomb-produced and how much is natural. The modelers therefore need to know the natural 14C concentrations before the weapons tests and how quickly the 14C entered the oceans. The problem is that few open-ocean measurements of that period have been taken.
Tom Brown of Livermore's Center for Accelerator Mass Spectrometry, working with the Livermore ocean modelers and ocean and fishery scientists from the University of Washington, came up with a way to provide the "pre-bomb" 14C data. In 1997, they measured the 14C content of archived salmon scales using Livermore's accelerator mass spectrometry capability (see S&TR, November 1997), which can measure minute isotopic quantities with high precision. The salmon-scale measurements are proxy indicators for the 14C levels, and even the 13C-to-12C ratios, of the surface waters of the oceans during the time period the salmon dwelled in the waters.
Fishery scientists have been collecting and archiving salmon scales for nearly a hundred years for various research purposes. The scales are particularly suitable for developing time-history data of ocean waters because the salmon's seasonal migration patterns are known. And because salmon live and feed on plankton and small fishes in the uppermost surface waters of the ocean, the scales can be equilibrated with the 14C content of the surface waters. Furthermore, the sections of the scales that grow while salmon live in the open ocean are identifiable--they appear as bands of different thickness roughly corresponding to seasons (Figure 4).

By selecting and measuring appropriate sections of the scales, Brown obtained estimates of 14C content of North Pacific surface waters, averaged over the very large areas of the salmon's seasonal migration patterns and over the 1- to 2-year time spans represented by the sections. The 14C measurements show excellent agreement with the few direct, open-ocean measurements available of 14C content and clearly show the rise in 14C content from the atmospheric nuclear tests (Figure 5). By providing rare estimates of "prebomb" values and the initial increase of ocean 14C concentrations, Brown's measurements are a valuable help to climate scientists trying to predict future uptake of CO2 by the ocean as well as future climate.

Fossil Fuel Affects 14C Fluxes
Burning fossil fuel may appear to have no effect on atmospheric radiocarbon content because fossil fuel contains no 14C. However, global processes are rarely that simple or linear. Caldeira and Duffy collaborated with Greg Rau from the University of California at Santa Cruz to formulate a model to quantify radiocarbon fluxes. Their model predicted significant, though indirect, effects of fossil-fuel burning on global distributions of 14C. More important, if its prediction of increased 14C levels by 1998 is correct, this model will soon become a test for global carbon-cycle models.
The model's calculations of changes induced by land clearing, fossil-fuel burning, and atmospheric nuclear tests show that, in the very near future, 14C in the atmosphere will begin to increase. Even though 14C, which increased between 1954 and 1963, has been steadily declining, that decline will be reversed if humans do not change their habits in burning fossil fuel.
Caldeira, Duffy, and Rau used a simplified model of the atmosphere, land, and ocean. The model--driven by carbon fluxes from land clearing, fossil-fuel burning, and atmospheric nuclear tests--was used to simulate changes that would occur in the face of isotope decay, continuing CO2 emissions, radiocarbon exchanges into the oceans and atmosphere, and estimates of future biomass. Nuclear test radiocarbon inventory in 1975, as well as data on both natural and bomb radiocarbon collected through GEOSECS (Geochemical Ocean Sections Study, an ocean data collection program), were used as reference points for adjusting model parameters.
Simulations were performed to separately assess the changes caused
by (1) land clearing and fossil-fuel burning, (2) land clearing only, (3) fossil-fuel burning only, and (4) both plus observed atmospheric data until 1990, which includes nuclear-testing 14C levels. The last simulation indicates the model's ability to portray past trends and predict future 14C fluxes.
The simulations indicate what percentages of increases in atmospheric CO2 can be attributed to deforestation and fossil-fuel burning. The data indicate that, prior to about 1910, most of the carbon entering the ocean resulted from deforestation; since that date, the carbon flux has been dominated by the effects of fossil-fuel burning.
The modelers' interpretation of the trends led them to the unexpected prediction that 14C levels in the atmosphere will begin to increase as a result of fossil-fuel burning. The modelers explain it this way: When atmospheric CO2 content increases, the ocean's absorption of it increases. In the case of fossil-fuel-caused increases, because fossil fuel contains no radiocarbon, the ocean is absorbing CO2 that consists primarily of 12C, a weak acid. A more acidic ocean tends to reject carbon in all its isotopic forms. So the 14C component of ocean CO2 is rejected along with the other carbon isotopes, adding to the atmospheric 14C content and reversing the decline that began in the 1960s after the end of atmospheric testing. The model indicates that 14C levels will begin increasing as early as 1998 and, by 2015, the fossil-fuel-induced radiocarbon flux out of the ocean will exceed the nuclear-explosion radiocarbon flux into the ocean, so the ocean's 14C mass will then begin to diminish.
The Caldeira, Duffy, and Rau model is noteworthy because the prediction that the 14C flux into the ocean will be reversed early in the next century indicates that human impacts on the global carbon cycle are significant on geologic, not just human, time scales.

Closing In on Global Climate
The growing consensus that fossil-fuel use is causing climate change and the recent effort to formulate international treaties to limit greenhouse- gas emissions lend urgency to understanding how carbon moves within the climate system. If indeed humans have been responsible for changing the climate, climate models must accurately and conclusively portray this cause and effect. Then we will have the understanding needed to begin mitigating the effects and assuring a better future for the environment.
--Gloria Wilt

Key Words: carbon cycle, carbon dioxide (CO2), climate change, climate model, fossil-fuel burning, global warming, greenhouse gas, marine biology, mass spectrometry, ocean carbon cycle, ocean convection, proxy data, radiocarbon (14C).

For further information contact Philip B. Duffy (925) 422-3722 ( or Ken Caldeira (925) 423-4191 (

PHILIP B. DUFFY is a physicist at the Laboratory, where he is group leader for the Climate System Modeling Group in the Atmospheric Science Division. Duffy worked in strategic defense systems when he joined the Laboratory in 1986. Prior to that, he received his Ph.D. and M.S. in astrophysics in 1986 and 1981 from Stanford University and an A.B. in astronomy and astrophysics in 1979 from Harvard University. Duffy has published research on astronomy, atomic physics, and numerical modeling of ocean circulation.

KENNETH G. CALDEIRA joined the Laboratory's Atmospheric Chemistry Group as a physicist in 1993 and has been an environmental scientist in the Climate System Modeling Group since 1995. He received his Ph.D. and M.S. in atmospheric science from New York University in 1991 and 1988 and his B.A. in philosophy from Rutgers University in 1978. He also served as a postdoc at Pennsylvania State University's Earth System Science Center. Caldeira has published many papers, for example, on climate stability of early Earth and the global carbon cycle as it has been affected by human activity over millions of years.

Back to March 1998