Predictions of climate change, such as those that may result from anthropogenic emissions of greenhouse gases, are made using General Circulation Models (GCMs). These models are very complex, deal with huge quantities of data, and require a very large number of calculations. GCMs solve mathematical equations that govern the physical processes that take place in the climate system. Starting from some initial state (i.e., the current climate conditions), the models step forward in time, computing the climatic conditions on Earth hundreds or even thousands of years into the future. GCMs are best used to present a statistical description of climate, which includes the average temperatures and precipitation for locations around the globe. This is different from global weather prediction models, which predict the expected weather conditions for each location at a specific time (e.g., Friday afternoon’s temperature and weather conditions for Tucson). Weather prediction models are used for short-term (14 days or less) forecasts, thus, they do not need to incorporate feedback processes that occur over longer time scales.

Given that global weather prediction models often produce incorrect forecasts, one may wonder how a GCM can be expected to give useful results hundreds of more years into the future. The difference is that GCMs only need to reproduce climate in a statistically averaged sense, not the weather for a given day. Several tests are used to validate GCMs. One is how well can they reproduce today's climate. For instance, a model can be initialized to the year 1900 and run through 2000 allowing for the increasing amounts of greenhouse gases during that period. The climate statistics produced are then compared to the measured statistics around the globe. A second test may be to simulate historical climate changes, such as the Ice Ages, by incorporating the long-term variations in Earth's orbit around the Sun. GCMs have also been used to simulate the climate response to short-term energy budget variations such as volcanic erruptions and El Nino. Current GCMs have handled these tests reasonably well, giving us some confidence in their ability to predict future climate changes.

Due to the complex nature of the climate system, which includes all of the potential feedbacks, GCMs are the best tools we have to predict future climate change. However, GCMs are not exact. We are still learning how to better handle some feedback processes, while others may not be included in the models at all. The current belief is that GCMs are most deficient at modeling clouds and ocean-atmosphere interactions (energy and chemical exchanges between the oceans and the atmosphere). More than a dozen GCMs are being run at various prediction centers located around the world. Each model produces different predictions of future climate largely due to differences in how they incorporate feedback processes, many of which are not fully understood. One test case that all GCMs have considered is what would happen if the amount of carbon dioxide (CO₂) in the atmosphere were doubled from its pre-industrial concentration of 280 ppmv to 560 ppmv. The models predict that the Earth’s average surface temperature would be (3-8)˚ F warmer with doubled CO₂. Because GCMs produce such a wide range in their predictions of global warming, we should not believe any one of them literally. Perhaps the true answer is somewhere within the range predicted by the various models.

The latest assessment from the Intergovernmental Panel on Climate Change (IPCC) projects that the Earth's average surface temperature will increase between 2.5 and 10.4° F between 1990 and 2100 if no efforts are undertaken to reduce the emissions of greenhouse gases ("business-as-usual" scenario). The prediction is based on a scientific review of the latest global change research studies. This is significantly higher that what the IPCC projected in 1995 (1.8 - 6.3° F). Many, but not all, scientists have confidence in the recent projections and endorse the IPCC conclusions. Of course, different projections are made if one assumes that the rate of greenhouse gas emissions will be reduced during the 21^st century.

Although the global average surface temperature is an interesting calculation, the effects of climate change will be most important on smaller scales. It is important to realize that global warming does not simply mean adding a few degrees to the average temperatures of today’s climate at each location around the globe. Climate change will likely bring changes in atmospheric and oceanic circulations, thus not every region will experience the same climate changes. Some areas will become drier (less precipitation), while others will become wetter. Some areas may become much hotter (25˚ F or more), while in other areas the average temperature may not rise at all or possibly become cooler.

The range of predictions from GCMs becomes much larger when we look at their predictions of regional climate change, i.e., changes in the temperature and precipitation for smaller areas. An example might be the climate changes predicted for the agricultural region in the United States plains. Two models that predict the same average global temperature rise may differ considerably in their predictions of regional climate change. One model may have the plains becoming hotter and drier (potentially bad for agriculture), while the other may predict little temperature change but more precipitation (potentially good for agriculture). The point is that predictions from GCMs become much more uncertain when considering regional climate changes as compared to changes in the global average temperature. Therefore many (not all) scientists have confidence in the prediction that doubling CO₂ would warm the global average surface temperature in the range (3-8)˚ F; however, little confidence should be placed in the regional climate changes predicted by a single GCM.

In order to be useful to humans, the predictions made by GCMs need to be utilized to estimate the impacts of climate change on human activities and natural ecosystems. Loosely this area of research has been given the name "impact studies" of climate change. Some of the questions to be answered include: how much will sea level rise and what effect will that have? How will fresh water resources be affected? What will be the impact on agriculture and food supply? To what extent will natural ecosystems be damaged? How may human health be affected? The answers to these questions are difficult. One reason that was mentioned above is that there will likely be a large variability in the climate changes that occur on regional scales and as of yet, we are unable to predict regional climate changes with much confidence. Another reason has to do with determining how sensitive a region or ecosystem is to change and how adaptable it may be to change. Note that both the magnitude and the rate of climate change are important in determining the sensitivity and adaptability of a region or ecosystem.

Impact studies of climate change are important. Past and current research in this area continues to advance our understanding of how human activities and natural ecosystems may respond to climate change. However, any impact projections that are based upon the climate changes predicted by a GCM should be viewed with caution. One must consider these projections as only one possible outcome of climate change, not as a factual account of what will occur. Too often the results of these studies are reported to the public without clearly stating the uncertainties associated with the predictions.