Online Catalog

GCRIO Home ->arrow Library ->arrow Consequences -> arrow Issue -> arrow Autumn 1995 -> arrow Climate Models: How Reliable are Their Predictions? Search
U.S. Global Change Research Information Office logo and link to home
Updated 11 November 2004

Consequences Vol. 1, No. 3, Autumn 1995








Climate Models: How Reliable are Their Predictions?

by Eric J. Barron

We often hear the assertion that our extensive use of carbon- based fuels now threatens to alter the climate of the whole world: that enhanced greenhouse warming--induced by the carbon dioxide and other gases we have added to the air--will lead to a rapid and unprecedented rise in the average temperature of the Earth within the next fifty years.

We are not accustomed to long-term forecasts of anything of such consequence. Nor can it be surprising that the initial reaction of almost anyone is to question the reliability of the prediction. For what is claimed--if indeed an accurate portrayal of the future-- seems to leave few choices: do we prepare ourselves for the impacts of lasting climate change? Should we rethink our own use of coal and oil and natural gas and gasoline, when energy use, as we all know, is very much tied to economic growth?

What must trouble many decision-makers is that the sounding of this loud environmental alarm was tripped not so much by measurements as by computer models. How certain or how controversial are these largely theoretical predictions of global warming, and on what assumptions are they based? Given the potential importance of regional climate changes for the development of national policies, and the impacts of extreme, climate-related weather events such as droughts, floods, and hurricanes on agriculture and human safety, how reliable are the projections of future change? Are the uncertainties in present climate models so great that we can ignore their predictions? What elements are the most robust? What are the prospects for substantial improvements in climate models in the near future?

These questions, so often asked, were put to a group of scientists in late 1994 in response to requests from both the White House Office of Science and Technology Policy (OSTP) and from the Government Accounting Office (GAO) which was responding, in turn, to a request from Congressmen John Dingell of Michigan. The charge to the Forum, which I chaired at the request of the U.S. Global Change Research Program, was to develop a statement on the credibility of modeled projections of climate change, to provide background to the government for considering and developing national policy options. The participants included climate modelers and other knowledgeable scientists who were chosen to bring to the Forum a wide spectrum of scientific opinion regarding the potential threat of global greenhouse warming. This review provides the author's summary of the Forum report, which is listed as a reference at the end of the article.


General circulation models

Computer-run, mathematical simulations or models of the atmosphere and ocean are the principal tool for predicting the response of the climate to increases in greenhouse gases. The most sophisticated of these, called general circulation models, or GCMs, express in mathematical form what is known of the processes that dictate the behavior of the atmosphere and the ocean. GCMs include the interaction of the atmosphere with the oceans and with the surface of the Earth, including plants and other ground cover. They allow us to test, by mathematical simulation, what should happen to climate, around the world, in response to a wide variety of changes. For example, what climatic effects would follow a major volcanic eruption, or a change in the radiation from the Sun?

The great power of mathematical models lies in their ability to simulate the behavior of systems--like the atmosphere and ocean-- that are too complex or extensive for simple, intuitive reasoning. There are limits, however, to how much complexity can be handled by the computers on which the models are run. At present, models of the global climate system cannot include physical processes whose horizontal dimensions are less than several hundred miles--a constraint that imposes simplifications on how well we can model what we know and restrictions on the level of regional detail. The key is to incorporate the best possible representation of all the important processes and feedbacks necessary to characterize the climate system, while keeping within the practical capabilities of modern computers.

Our ability to evaluate the strengths and weaknesses of climate models has grown over the last two decades. A growing number of GCMs, many with independently derived components, are available for intercomparison. We have a growing store of meteorological and oceanic observations against which model predictions can be tested. We also have information on past climate change, recorded by natural processes in rocks and sediments, that allow us to assess the ability of models to replicate the known features of climates different from that of the present day. Each of these elements is the basis for debate on the reliability of climate model projections of the future climate.

Consensus predictions

All of the GCM experiments designed to assess the impact of increases of greenhouse gases point to global warming through the coming century, with accompanying changes in rainfall and other meteorological quantities. Still, the complexity of the climate system is a tremendous obstacle to predicting future climate change. Neither climatological observations nor present climate models is sufficient to project how climate will change with certainty. A workable approach is that adopted by the Intergovernmental Panel on Climate Change (IPCC) of the World Meteorological Organization and the United Nations Environment Programme, which is based on projections of the expected growth of greenhouse gases and the combined results of many GCMs. In terms of mean global surface temperature, the consensus prediction of the IPCC is for an increase of 0.5 to 2° Centigrade (about 1 to 3.5° Fahrenheit) by the year 2050, in response to an anticipated increase of 1 percent per year in CO2. The low end is a significant change; the high end, a dramatic one. Moreover, were the amount of atmospheric carbon dioxide to double, the consensus forecast is for an eventual warming of 1.5 to 4.5°C (about 3 to 8°F.)

Such changes, if realized, would represent a significant climatic change. For example, the most recent climate change of similar magnitude was the last major Ice Age that reached its peak about 18,000 years ago. The mean global temperature during that time is estimated to have been between 3 and 4° C cooler than at present. The effect of this small a change in global-mean temperature can be appreciated when we realize that during the last Ice Age, glacial ice--a mile or more deep--covered much of North America, year-round, reaching as far south as the Great Lakes and the surrounding states of present-day America. That amount of change in global-mean temperature is similar, although opposite in sign, to what is now projected due to increases in greenhouse gases. But the rate of change is not. The last Ice Age developed over thousands of years, while global greenhouse warming is projected to occur within a span of less than a century. And within the lifetime of people now living.

It is equally clear that in terms of potential impact, the difference between a 1.5° and a 4.5° C projection for future warming is very large. As a result of this uncertainty, decision- makers are confronted with a difficult question. What steps should be taken when the best indications from state-of-the-science models suggest that climate change due to human activities may be large and significant, yet the predictions are less than certain?

The scientific debate regarding these uncertainties has entered the public arena, providing considerable confusion even for those aspects of climate-model predictions that are virtually certain. The debate over how much warming--and by when, and why it hasn't yet been more clearly seen--has clouded the clearer picture that increases in carbon dioxide will increase the global-mean temperature. It has also affixed the stamp of "controversial" on almost any reference to impending global warming in the press and news media, implying, erroneously, that the general concept, and not just the details, is in serious doubt.

A method for evaluation

It is possible to get an indication of the strength of a building or other structure if we know which of its footings are solid and which are less so: in this case, to separate the aspects of predicted climate change that are virtually certain from those that are uncertain. The Forum carried out this kind of assessment of predicted global warming, to provide better illumination for policy discussions and to assist policy development.

The evaluation is divided into three parts. The first provides a basis for any discussion of climate-model predictions by identifying the foundations of the greenhouse warming theory that are most solid and robust: a series of conclusions which can be viewed as "virtually certain" based on observations, experiments, and the results of many models. The second part is a listing of specific predictions of climate models that are societally important, ranked by degree of certainty. In the last part we examine what can be done in the future to improve climate-model predictions.

The Foundation

Although the specific predictions of climate change are derived from models, the reasons for expecting significant global warming in the near future comes from a much deeper foundation that includes laboratory and field experiments, well-established knowledge of atmospheric behavior, and measurements that include worldwide monitoring of atmospheric conditions. Here we list seven of the principal scientific arguments for a global-warming prediction. Throughout, the stated conclusions are subject to little or no debate because of their level of certainty, and indeed, to some they may appear trivial.

First, as confirmed in laboratory experiments, certain gases that are naturally present in small amounts in the atmosphere play an active role in maintaining the Earth's temperature. They do this by absorbing energy (infrared radiation) emitted from the land, ocean, clouds, and the atmosphere itself, and then re-emitting it. The most important of these so-called greenhouse gases are, in order, water vapor, carbon dioxide, and methane, followed by nitrous oxide, ozone, and chlorofluorocarbons (or CFCs), which are manmade compounds of chlorine, fluorine and carbon.

Second, because they absorb radiated energy, increased concentrations of greenhouse gases will inevitably raise the Earth's temperature. The extent of the warming will depend on possible amplifying or damping mechanisms (feedback processes), particularly those involving water vapor and clouds, that are major players in controlling the natural greenhouse effect. Such feedbacks can change the magnitude of the warming, but there are no known cases where they bring about an opposite, cooling effect. Thus that greenhouse-gas increases will produce warming is not in question. The heart of the greenhouse debate concerns the nature and timing of temperature increase, and the associated changes and impacts of other climatic quantities, not the fact that increases in greenhouse gases will lead to a rise in global temperature.

Third, the amounts of carbon dioxide, methane, nitrous oxide and chlorofluorocarbons present in the air today are significantly higher than their "pre-industrial" levels--that is, the amount that was present, naturally, before the intensive use of energy that began with the Industrial Revolution about 200 years ago. For example, the amount of carbon dioxide that is measured in the air throughout the world today is about 30 percent greater than that found in years before about 1800, as determined from the chemical analysis of air trapped in well-dated, polar ice cores. Similar findings apply to methane (which has increased by more than 100 percent) and to other greenhouse gases, with the possible exception of water vapor. The increases can be tied directly to human activities that include fossil-fuel burning (as for heating, or in internal combustion engines), the burning of trees to clear land, and certain agricultural and industrial practices.

Fourth, it would take hundreds of years for the concentration of carbon dioxide to fall back to pre-industrial levels, even if the amount emitted were immediately and substantially reduced around the world. The reason is the slow pace of the natural processes that remove carbon dioxide from the atmosphere. Further, the projected growth in world population and energy use in the developing countries make it highly unlikely that any substantial reductions in total global carbon-dioxide emissions will take place over the next several decades. Thus the atmospheric concentration of carbon dioxide is expected to continue to rise well into the 21st century. Similar arguments apply to most other greenhouse gases.

Fifth, there are many more microscopic, airborne particles (known as aerosols) in the atmosphere than were present in pre-industrial times, concentrated in and downwind of areas of intensive human activity. Aerosols are present naturally in the atmosphere in the form of windblown dust from cultivated soils, hydrocarbons from vegetation and forests, and soot from forest and grassland fires. What has increased is the anthropogenic or human-made contribution: soot, sulfate aerosols and other particles found downwind from regions of intensive fossil-fuel combustion and biomass burning.

Sixth, laboratory and atmospheric measurements demonstrate that sulfate aerosols (containing compounds of sulfur and oxygen) that come either from volcanic eruptions or fossil-fuel combustion exert a cooling influence on the climate, by reflecting some of the incoming solar radiation back into space. The increase in airborne particles cited above could thus offset some of the warming expected from the buildup of greenhouse gases, although the magnitude and extent of aerosol cooling is not known and is difficult to quantify, in part because the regional distribution and character of past and future emissions of aerosols are poorly known.

Seventh, the globally averaged temperature at the surface of the Earth has risen about 1°F (or 0.5°C) in the last 100 years. Because of the natural variability of climate, the change cannot yet be ascribed unambiguously to the increase in greenhouse gases over the same period. Nor is the recorded temperature rise as great as that expected, based on climate-model results, from greenhouse warming, although some or all of the difference may be due to the cooling effect of aerosols, noted above, or to the action of other competing long-term effects.

These seven findings form the basis for the conclusion of a vast majority of scientists that human activities are now modifying the energy balance of the Earth system. Less certain are the magnitude and the timing of the associated climate changes, which are derived from models and which are the subject of considerable debate.

Climate Model Predictions

Predictions of future climate are imperfect because they are limited by significant uncertainties that stem from: (1) the natural variability of climate; (2) our inability to predict accurately future greenhouse-gas and aerosol emissions; (3) the potential for unpredicted or unrecognized factors, such as volcanic eruptions or new or unknown human influences, to perturb atmospheric conditions; and (4) our as-yet incomplete understanding of the total climate system. The reliability of climate-model predictions depends directly upon each of these.

With this in mind we list below, in order of certainty, the major policy-relevant predictions of present climate models.

Calculated changes in climate variables will obviously depend upon the assumptions made regarding the future concentrations of greenhouse gases in the atmosphere, which are a function of projected population growth and associated economic expansion. The modeled results that are given here assume that greenhouse- gas concentrations in the atmosphere will continue to increase in coming decades. For purposes of simplicity, the climate model used considers only carbon dioxide and assumes that it will increase 1 percent each year, which, for purposes of calculation, replicates the effect of the anticipated increases in the concentrations of all other greenhouse gases.

A Ranked List

In ranking its conclusions, the Forum adopted a system of four levels of certainty, as these terms are defined in general usage: virtually certain, very probable, probable, and uncertain.

Virtually Certain:

(1) The temperature of the stratosphere--an upper region of the atmosphere that extends from about ten to fifty kilometers (six to thirty miles) above the surface of the Earth--will be significantly cooled. This cooling comes about through the combined effect of increases in carbon dioxide and the observed depletion in stratospheric ozone, and the manner in which the two gases absorb and re-emit energy. Opposite in sign to what is expected near the ground, the change had been predicted by models and has now been observed. As such, it provides potential early evidence of greenhouse warming.

Very Probable:

(2) The surface temperature of the Earth will continue to rise through at least the middle of the 21st century. The prediction is based on (a) projected, continued increases in greenhouse-gas emissions; (b) the results from a host of model calculations; and (c) the analysis of past climates of the Earth. The best available estimate, from the international assessment by the IPCC and based on the range of available model predictions, is that the global-mean surface temperature will increase by about 0.5 to 2 °C (roughly 1 to 3.5° F) over the period from 1990 to 2050 (Fig. 1). For comparison, an increase of 0.5° C--the lower limit--is equal to the warming that has taken place in the past 100 years. Beyond the year 2050, the carbon dioxide concentration is expected to reach twice that of pre- industrial times. When that level is reached, and after the climate has reached equilibrium, the best estimate for the resulting climate change is a warming of 1.5 to 4.5°C (about 3 to 8°F). The IPCC considers a mid-range increase of 2.5°C (5°F) the most probable result. The model calculations assume that the present levels of sulfur aerosol emissions (for example, from the burning of soft coal) will to some degree diminish in years ahead; if they do not, the temperature increase will be somewhat less. The actual temperature change could fall outside the ranges given here should natural climate variations happen to be large during the period of the prediction.

(3) Higher surface temperatures will cause an increase in the average precipitation over the globe. This comes about because temperature affects the rate at which surface water is evaporated (to return to the ground in the form of rainfall and snow). While the connections between temperature and precipitation rates are well understood, the distribution of changes in precipitation over the Earth is less certain.

(4) The amount of sea ice in the Northern Hemisphere will be diminished. Studies of past climates provide evidence for the polar amplification of either global warming or global cooling. The reason is a positive feedback loop that connects warming, reduction of sea ice, replacement of the highly reflective sea ice with a darker and more absorbing ocean surface, and hence additional warming (Fig. 2). For these reasons, it is very probable that the extent of sea ice in the polar regions of the Northern Hemisphere will be reduced by melting. Changes in corresponding areas of the Southern Hemisphere are less certain due to differences in ocean circulation and the presence of the Antarctic continent.

(5) Land areas in the Arctic should experience amplified wintertime warming . The positive feedback loop noted in point 4 above also applies to the land. The magnitude of the surface warming there will also depend on how the normal transfer of heat by the atmosphere from the equator to the pole responds to global warming, and this point is uncertain.

(6) Global warming will cause sea level to rise. This is expected as a consequence of three temperature-related changes: the physical expansion of sea water as the ocean temperature increases, the partial melting of mountain glaciers, and changes in the extent and thickness of the Antarctic and Arctic ice sheets. The expansion of sea water can be determined from the projected temperature change described above. Reasonable estimates of the retreat of mountain glaciers are also available, but calculations of the changes expected in polar ice caps are far less certain. Based on calculations of sea-water expansion and the retreat of mountain glaciers, and ignoring the possible long-term response of the polar ice caps or any potential catastrophic collapse of the west Antarctic ice sheet, it is estimated that global sea level will rise from 5 to 40 centimeters (2 to 16 inches) by 2050. This projection compares to an anticipated rate of sea-level rise of 5-12 centimeters (5 inches) if currently observed rates of rise over the past century continue.

(7) The climatic effect of any changes expected in the amount of energy radiated from the Sun in the course of the next fifty years is much smaller than that from increased concentrations of carbon dioxide and other greenhouse gases. Based on current knowledge, the Sun's energy output varies by about 0.1 percent over the eleven-year sunspot cycle, and this variation can affect the surface temperature of the Earth. However, the effects of anticipated greenhouse warming are four to seven times greater than those that could result from these short-term changes in the total flow of energy from the Sun. Were the Sun's radiation to fall to the lowest levels yet measured and to remain there through the next fifty years, it could diminish the expected effects of greenhouse warming by about 25 percent. Were solar radiation to remain abnormally high throughout this time, which is about equally likely, it could add to the effect by at most the same amount.


(8) Continental dryness will increase at middle latitudes in summer in the Northern Hemisphere. The basis for this prediction is the fact, cited earlier, that higher temperatures lead to much higher rates of evaporation: in net effect, the increase in evaporation will on a regional basis exceed the accompanying increase in rainfall. The amount of drying is qualified, however, by several factors that are not well represented in models. These include the movement of evaporated moisture from place to place through atmospheric circulation; the effects of changes in ground cover due to the response of vegetation to increased carbon dioxide; the role of aerosols; and interactions between the land surface and the atmosphere, including the storage of wintertime precipitation in soils.

(9) Rain and snow at high latitudes will increase as the amount of moisture in the atmosphere is increased. The freshwater that is added there by precipitation could alter the deepwater circulation of the oceans, which is driven in part by differences in the salt content of different parts of the ocean and which in turn affects climate. Additional precipitation could also affect the size of the polar ice caps, and hence perturb sea-level.

(10) The Antarctic and North Atlantic Oceans will warm more slowly than the global average. Changes in sea-surface temperature are moderated in regions such as these because of the regular mixing of the surface water with the deeper, cooler water of the ocean. They are thus the most logical sites for slower-than- average warming. The sea-surface temperatures there, however, will also depend on accompanying changes in precipitation and freshwater inputs that can change the rate of vertical mixing of the oceans.

(11) The occasional eruption of a major volcano will temporarily diminish global warming, but for no more than a few years. Historical records indicate that the solid particles introduced into the stratosphere by volcanic eruptions can cool the mean temperature of the Earth by a few tenths of a degree C for up to a two or three years, during which time the particles are removed from the upper atmosphere. Such changes would constitute transient interruptions in the longer-term trend of greenhouse warming.


(12) Changes in climate variability will occur. However, the exact nature of changes in climate variability due to greenhouse warming is as yet not well defined. All models predict a possible reduction in wintertime variability in warmer climates; it is also commonly predicted that thunderstorm activity should increase as a result of the increased moisture content of the atmosphere. The frequency of El Niñ o events could change as a result of a global warming, as could the frequency of atmospheric "blocking" events that set up persistent weather patterns that last weeks to months at a time.

(13) Changes in the climate of regional-scale areas (from the size of large metropolitan regions to the scale of states or small countries) are likely to be quite different from the global average. We have only a very limited capability to estimate changes expected in the climate of any specific region. The spatial resolution of climate models is, as yet, too coarse to incorporate effects such as regional land characteristics, surface contours, and local hydrologic conditions, even though these factors are known to be important. Regional changes in climate can differ from global changes, but the nature of the probable differences is uncertain.

(14) The intensity of tropical storms, including hurricanes, may increase. This can occur as a result of the effects of the higher sea-surface temperatures that are associated with global warming, because tropical storms derive their energy from temperature differences. However, there are simply too many unresolved issues--such as how possible changes in the poleward transport of heat may influence the amount of tropical warming--to predict more precisely what the effect will be. Whether the number of such tropical storms will also increase is also uncertain, in part because GCMs are not run at spatial resolution fine enough to simulate hurricane formation.

(15) Forecasts of climate change over the next twenty-five years are as yet uncertain. Although such forecasts are much to be desired, present uncertainties in the factors that control the natural variability of climate, in the model simulations, and in expected changes in atmospheric chemistry make it extremely difficult to predict decade-to-decade changes in climate. In any given decade, the changes in temperature and related variables could be substantially less than or more than the predicted long-term trend. Warming estimates in terms of degrees per decade and the use of these trends to analyze a single decade are unwarranted and misleading.

(16) Interactions between climate and vegetation may modify the magnitude of predicted greenhouse warming, but whether these effects will amplify or diminish climate change is as yet uncertain. The limited assessments that have been made suggest possible feedbacks due to climate-induced changes in vegetation, such as the replacement of high-latitude tundra by vegetation more characteristic of temperate latitudes, or the displacement of forests by grassland. Other climate impacts can result from the direct effects of enhanced carbon dioxide on plant growth, from impacts of tropical deforestation, and from the effects of plant productivity on atmospheric chemistry.

Steps to Reduce the Uncertanties in Present Models

The uncertainties cited in the list above can provide a set of ordered priorities for improving present climate models. Given the compelling need for clearer answers, we can count on continued improvements. Yet, for many reasons--including the need for additional observational data-- significant reductions in many of the uncertainties will require sustained efforts over a decade or more.

Much of the research effort of the multi-agency U.S. Global Change Research Program is designed to address the uncertainties cited here, including those that involve cloud-radiation-water vapor interactions, ocean circulation, aerosols, natural climate variability, land-surface processes that include vegetation changes and chemical cycling, the frequency and intensity of high-impact events such as hurricanes, factors that create the potential for surprises, and the interaction between chemistry and climate.

Seven areas of improvement are described below that are likely to reduce uncertainties in GCM predictions over the coming decades. These should be viewed as opportunities for significant improvement in climate- model predictions.

(1) The use of finer spatial resolution in climate models (see Fig. 3).Many of the uncertainties associated with the results of climate models stem from the relatively coarse spatial resolution that they employ--that is, the smallest element of the landscape for which input can be provided, or results obtained. In the present use of GCMs this is often 5° in latitude by 5° in longitude, or a square approximately 350 miles on a side, which is about the size of New Mexico or all of New England. Many advantages accrue at higher resolutions: storms and circulation patterns, for example, are significantly better represented--in part because of improved representation of land contours and characteristics, and in part because of the capability of including major weather-system processes. The same arguments apply to the ocean component of climate models.

The use of finer resolution comes at the costs of longer computing time and greater data-handling requirements. Switching from a 5° x 5° grid to one with a resolution of 2.5° x 2.5° requires substantially more computing time. Each time the spatial resolution is doubled, eight times more computer time is required on the same machine. A calculation with 5° resolution might typically take ninety hours of continuous running on today's fastest supercomputers; the higher-resolution run would tie up the same supercomputer, night and day, for almost a month! It is not only cost that holds back the use of higher resolution: in many cases not enough is known about detailed processes to utilize a finer grid. The combination of increased availability of computer resources and of studies that elucidate the physical processes at finer resolution are very likely to bring substantial improvements in climate-model capability. They will also provide the opportunity to tailor predictions to specific regions.

(2) Improved representations of the lowest layer of the atmosphere (the often turbulent and so-called boundary layer in which the temperature and contours of the ground surface affect the moving air) and of the distribution of water vapor throughout the atmosphere. Significant improvements in the first of these depend, at present, on a better understanding of the science of how air moves under the particularly complicated conditions at the air- land interface: where, in a sense, the rubber meets the road. How much water vapor is in the air and how it is distributed geographically and with height above the surface is another major source of uncertainty in model predictions, because of its dominant role in determining the temperature of the atmosphere. The distribution of water vapor is highly variable, temporally and spatially; improvements in modeling the effects of water vapor await observations more extensive and more accurate than those currently available.

(3) Improved representations of the connections that link the atmosphere, the ocean, and the surface of the land. In Nature, each of these affects--and is affected by--the other two. In most of today's models, the real connections that link them together are approximated by arbitrary adjustments or are characterized by large uncertainties. Were we able to include more-accurate representations of these processes, we could use climate models to explore the causes and characteristics of natural climate variability on all time scales. The focus of GCMs was initially on atmospheric processes. Corresponding improvements in the representation of the land surface and the connections between the ocean and the atmosphere are likely to result in substantial model improvements.

(4) More explicit representation of the land surface, including vegetation, soil characteristics, and effects of enhanced levels of carbon dioxide and ozone on plants. Modeled estimates of soil moisture, summertime continental drying, and regional climate change depend very much on how accurately the land surface is represented, including more explicit treatment of vegetation of all kinds. All of the known processes that link the atmosphere with the biosphere, or the atmosphere with the soil, are interactive or two- way connections, in the sense that each controls, to a degree, the other.

(5) Continued comparisons of models with observational data and with other models. Important new data sets--for example, new, long-term, consistent observations from NASA's Earth Observing System and supplementary data sets now being developed that span the current century--can provide more critical tests of the accuracy of climate models. Progress in model development and improvement can be accelerated by comparisons of this kind, and through the continued intercomparison of the results of different models.

(6) Demonstrated capability of climate models to simulate global changes of the past. The ability of a model to predict the climate of the future can be measured by its success in simulating what is known to have happened in the past. Data that describe significant global changes of the past-- including the coming and going of the major ice ages and the climate changes of the last 1000 years--have been obtained through the analysis of tree-rings and the sediments deposited in dated ice and in lake and ocean cores. They provide an invaluable test of the reliability of climate models. Greater emphasis on the analysis of past climates can help assess model projections of climate sensitivity and variability, and lead to enhanced model credibility.

(7) Improved representation of the interactions that link climate and vegetation with the concentration of greenhouse gases, and of the effects of aerosols on climate. Temperature, wind, and rainfall are involved in the ongoing exchange of chemical elements and compounds among the air and the water and the solid earth, thus affecting the concentration of greenhouse gases in the atmosphere. Changes in vegetation also affect the continual exchange of chemical elements among air and water and land, and the distribution of water. Aerosols, through their indirect influence on clouds and atmospheric chemistry, can also influence climate.


Three major conclusions can be drawn from this three-part examination of the capabilities and limitations of climate models. The first is that we know very well how greenhouse gases affect the energy balance of the Earth, and with similar confidence that the concentrations of these gases are now increasing due to human activities, and that these increases should result in global warming. At issue is not whether the Earth will warm, but by how much, where, when, and with what consequences for society and ecosystems.

Second, the level of confidence in the results from present climate models depends very much on the spatial and temporal specificity of the prediction. The most certain are those that pertain to the Earth as a whole and that apply to a roughly fifty-year period. Regional predictions, predictions on a decade-by-decade basis, and predictions of higher-resolution phenomena such as hurricanes, are considerably less certain. For the decisions that we face as individuals, it would be much better were it otherwise, although a highly confident, general prediction with expectations of improved detail can provide a useful guide for broad policy decisions.

Third, substantial opportunities now exist, given consistent and long- term research endeavors, to improve the specificity of climate-model predictions. These include efforts to refine spatial resolution, to improve the physical representation of the lowest layer of the atmosphere, to provide more realistic representation of non- atmospheric components of the climate system, and to provide more critical tests of models, both among themselves and against new and more comprehensive observations of present climate and also data from the past.

Interestingly, the current focus on the policy relevance of climate models in the U.S. has often been negative: a view that climate models are far too uncertain to be used in setting costly economic or national security policy. The widely-publicized scientific debate over these uncertainties has resulted in considerable confusion, even in cases where the conclusions from climate models are robust. For policy decisions, it would seem far more helpful to understand the degree of certainty or uncertainty associated with the different elements that enter climate predictions.

Accurate near- and long-term climate forecasts carry the potential of tremendous economic and humanitarian value. An example of the potential value of advanced prediction is the recent successful use of coupled ocean- atmosphere models for El Niño forecasts to limit the impact of these shorter-term changes in climate on food production and fisheries. The prediction of long-term climate change using GCMs is more challenging than an El Niño forecast, although improvements in the former are likely to be of even greater economic value. Indeed, the importance of advance knowledge may well be the reason that Japan and many countries in Europe are developing strong environmental observation and prediction efforts regarding impending climate change. The improvements in models, outlined above, can be viewed as a path toward more accurate climate predictions. In some cases, these predictions may serve as warnings in areas of societal vulnerabilities, such as food production or the adequacy of freshwater. In many cases, they will clearly contribute to economic vitality.

For Further Reading

An Introduction to Three-Dimensional Climate Modeling, edited by W. M. Washington and C. L. Parkinson. Oxford University Press, Oxford, UK, 422pp, 1986.

Climate Change: The IPCC Scientific Assessment, edited by J. T. Houghton, G. J. Jenkins and J. J. Ephraums, Cambridge University Press, Cambridge, UK, 364pp, 1990.

Climate Change 1992: The Supplementary Report to the IPCC Scientific Assessment, edited by J. T. Houghton, B. A. Callander and S. K. Varney, Cambridge University Press, Cambridge, UK, 200pp, 1992.

"Earth's Shrouded Future: The Unfinished Forecast of Global Warming" by E. J. Barron, in The Sciences (a publication of the New York Academy of Sciences), Sept/Oct issue, pp. 14-20, 1989.

Forum on Global Change Modeling. USGCRP Report 95-02, 26pp, Washington, D.C., 1995. Available from GCRIO User Services, 2250 Pierce Rd., University Center MI 48710.


Dr. Michael Schlesinger is Professor of Meteorology and director of the Climate Research Group in the Department of Atmospheric Sciences at the University of Illinois in Urbana. His interests are in modeling, analyzing and simulating climate and climate change, with particular emphasis on determining the sensitivity of the climate system, integrated assessment of climate change, and the simulation and understanding of past climates.

Dr. Brian P. Flannery pursued research in astrophysics before joining Corporate Research, Exxon Research and Engineering Company in 1980. Since then he has participated in scientific and technical studies of global climate change and has been actively involved on behalf of business and industry in national and international programs that address the causes and effects of environmental changes.

Scientific reviewers provide technical advice to the author and Editor, who bear ultimate responsibility for the accuracy and balance of any opinions that are expressed.

Return to Table of Contents

U.S. Global Change Research Information Office, Suite 250, 1717 Pennsylvania Ave, NW, Washington, DC 20006. Tel: +1 202 223 6262. Fax: +1 202 223 3065. Email: . Web: Webmaster: .
U.S. Climate Change Technology Program Intranet Logo and link to Home