Editor's note: Jonathan Gilligan is associate professor of earth and environmental sciences at Vanderbilt University. As an addendum to the Scene's Aug. 26, 2010 cover story ("Climate change isn't coming to Nashville — it's already here. And the future could be hotter than we can handle"), Gilligan prepared this statement in order to provide some background on climate models, their foundations and their limitations.
Why should we trust climate models? We shouldn't trust everything about them. For some things, they're extremely reliable. For others, they're not reliable at all.
In short: In predicting global temperature, climate models show a pattern of temperature changes at different times and places, and these patterns are very close to what's been seen in thousands of measurements of temperature, both on the surface and high in the atmosphere, over the last 40 years. If there were something very wrong with the models, we'd expect that the patterns they calculate would be different from what we actually see. For example, models that assume climate change is due to changes in the sun's brightness, instead of greenhouse gases, calculate temperature patterns that disagree dramatically from actual measurements. Regarding predictions, perhaps the most famous climate model prediction was published in 1988 by James Hansen at NASA's Goddard Institute for Space Studies. Since 1988, measured temperatures have closely matched Hansen's prediction.
At greater length:
First, about climate and weather: Very simply, we can make much better predictions about climate than about weather. Your weight fluctuates a lot from day to day, so it's very difficult to predict accurately how many pounds you will weigh at exactly noon one week from today. Depending on when you eat, go to the bathroom, and exercise, your weight at that moment might be as much as several pounds more or less than it is at this instant. However, if you add 500 calories per day to your diet without dramatically changing your exercise habits, I can reliably predict that you will be a lot heavier five years from now than you are today, regardless of the daily ups and downs. Weather and climate are a lot like that. I can't predict the exact weather next July 24, much less July 24, 2100 — but I can predict confidently that in Nashville, July 24, 2100 will be a lot warmer than Jan 24, 2100. The detailed condition of a specific hour on a specific day is weather. General patterns (e.g., Alaska is colder than Florida, July is warmer than January) are climate. It's much easier to make reliable predictions about climate than about weather.
So what can and can't we trust about models?
Models of the greenhouse effect are very reliable. They use only very solid scientific principles that have been extensively tested since the 19th century, and which serve as the basis of all sorts of technology — including heat-seeking missiles, spy satellites, and the infrared thermometers used in doctors' offices and industrial furnaces — that would not work if there were any significant errors in the theory. A large part of the raw data used to calculate the greenhouse effect was gathered by the U.S. Air Force for the purposes of designing military technology.
Models of the greenhouse effect are rock solid. We can check greenhouse effect models by calculating what the temperature should be on other planets, such as Venus, Mars, and Jupiter; these calculations agree well with measurements by space probes, which gives us great confidence that we understand the greenhouse effect very well.
Models of other aspects of the climate are less well understood. We don't fully understand the interactions of the atmosphere and the oceans; we don't fully understand the behavior of clouds and some aspects of global winds. For these, we need some way to test the full climate models, not just the greenhouse effect parts of those models.
One thing scientists do is to look not just at one number, such as a prediction of how much the global temperature will change in the next 100 years, but at patterns in space and time of the way temperatures will change at different times and places. These let us see whether models get the behavior of the earth system correct.
Measured temperatures around the world show a clear, but small, warming in the last 150 years, with most of the warming occurring in the last 40 years or so. (Note that more than half the increase of greenhouse gases in the atmosphere over the past several thousand years happened since 1970.) Most climate scientists say this warming happened because the growth of greenhouse gases caused the planet to warm up, but some people continue to believe the warming was due to the sun getting brighter.
Both changes in the amount of greenhouse gases in the atmosphere and changes in the sun's brightness can cause warming, so how can we test whether models based on the greenhouse gas theory are right? One way is to compare different models' predictions about the patterns of warming. Solar-brightness models predict that the earth will warm most at the times and places where the sunlight is brightest: in the tropics, in daytime hours, and in summer months. Greenhouse effect models predict the opposite: temperatures should rise most at times and places where there's the least amount of sunlight: in the polar regions, in nighttime hours, and in winter months. Actual measurements of temperatures around the world disagree with the solar models and agree with the greenhouse models very nicely: the Arctic has warmed much more than the tropics; nights have warmed more than days; winter months have warmed more than summer.
Another prediction relates to the different layers of the atmosphere. The solar models predict that a brightening sun would make both the troposphere (the bottom 6 miles or so of the atmosphere) and the stratosphere (from about 6 to 30 miles up) warm at the same time. The greenhouse models predict that a growing greenhouse effect would cause the stratosphere to cool off when the troposphere warms up. Data from satellites and weather balloons clearly shows that over the last several decades the troposphere has warmed and the stratosphere has cooled. There have been many other tests, but these are the easiest to explain simply.
Thus we can draw two conclusions: First, if we compare solar models to greenhouse models, measurements and observations of the atmosphere over the last several decades contradict the solar models and agree with the greenhouse models, so the climate change we've already observed is much more likely due to greenhouse gases than to a brightening sun.
Second, there's always the possibility that neither set of models is correct, so ruling out the sun doesn't prove greenhouse gases are responsible. But when we look at all these patterns — arctic vs. tropics, summer vs. winter, day vs. night, troposphere vs. stratosphere — we see that most of the observations come very close to the patterns predicted by the basic greenhouse theory. (The Antarctic does not follow this very simple description, but for reasons too complicated to explain here, its behavior doesn't contradict the greenhouse theory.) If something else were causing the climate change we observe, then you'd expect that many of the observed patterns would be very different from those predicted by the models.
We can also compare model predictions to specific events, such as volcanic eruptions. When Mt. Pinatubo erupted in 1991, it had a significant effect on global climate for a couple of years because it shot so much ash and other aerosol particles into the stratosphere that it blocked a lot of sunlight from reaching the surface. Comparing what actually happened to the atmosphere to model calculations allowed scientists to check their models, and in particular served to provide a solid confirmation of the way models calculated the interactions between water vapor and climate.
Another big success for models was James Hansen's 1988 prediction of climate change over the following 30 years. Hansen didn't know what would happen regarding greenhouse gas emissions, so he laid out three different predictions. The worst-case prediction assumed huge increases in greenhouse gas emissions with no compensating cooling effects. The most-likely-case prediction made more moderate assumptions of greenhouse gas emissions and included the cooling effects from volcanoes and aerosol pollution. The best-case prediction assumed significant cuts in greenhouse gas emissions. Since 1988, global temperatures have been very close to Hansen's "most-likely" prediction.
None of this proves that the climate models can be trusted to predict the next 100 years or more, and the more detail you ask from the predictions, the less reliable things are, so let me lay out a list of predictions in order of decreasing reliability:
Most reliable (pretty much totally trustworthy): Increasing greenhouse gases WILL cause the planet to warm up. This is such basic physics that for the planet not to warm up would mean something is very wrong with the basic laws of thermodynamics from the 1800s. I can't see how these laws could be so badly wrong, when they have been so extensively tested and when so much technology based on those laws works so well.
Very reliable, but not completely certain: How much will the planet warm up over the next century? First, we need to break this into two questions. First: What will happen to greenhouse gases over the next century? Second: How much will the changing greenhouse gases warm up the planet? The first question is not a scientific question. It's a political and economic question. (What laws regarding greenhouse gas emissions will we enact? how quickly will the world's economy grow? How quickly will the third world industrialize? And so on.) So we can ask a more clearly scientific question: If we double the amount of greenhouse gas in the atmosphere, how much would the world warm up? (This number is what scientists call "climate sensitivity.")
The different models differ on the climate sensitivity. The most common answer is about 5-and-a-half degrees Fahrenheit, but the number is uncertain and different models predict different sensitivities. The range of values goes from around three and a half degrees to eight degrees Fahrenheit. This range represents more than a factor of two difference, and since it would be very different to deal with 3-and-a-half degrees of warming than 8 degrees, the fact we can't be more certain in this prediction is frustrating.
There are also very solid reasons to believe that as the planet warms, severe drought will become widespread. We see this both from theoretical predictions of the future and from geologic records of climatic conditions during warmer periods in earth's past: for instance, for several centuries during the so-called Medieval Warm Period, much of North America suffered through droughts more severe than the dust bowl of the 1930s, which lasted as long as 150 years at a time.
Down at the bottom of the list (i.e., the least likely predictions): Here we get to detailed predictions of climate change in specific places (e.g., will there be lots more floods in the Southeastern U.S.), predictions of sea-level rise and predictions of hurricane activity.
I've discussed regional climate change above.
Sea level rise comes from many sources: Water expands as it warms up, and this causes the sea level to rise. This part can be calculated very reliably. However, there's another part: If ice on land were to melt, it would flow into the oceans and cause them to rise. This is very uncertain. If all the ice on Greenland and Antarctica were to melt, sea level would rise by more than 200 feet. It seems very unlikely that all the ice in Greenland would melt in the next century or two, and much less likely still that Antarctica would melt. But some of the ice in both places is already melting and scientists don't have any good way to predict how much ice will melt or how quickly, so the IPCC and a more recent panel assembled by the U.S. Climate Change Science Program judged that we don't know enough to make even a crude estimate of how much sea level might rise in the next century due to the Greenland and Antarctic ice caps melting.
Sea level is currently rising faster than the IPCC's "worst-case" scenario, which leads many scientists to worry that the consequences of global warming might be much worse than models predict, although I will emphasize that what I mean by "much worse" is sea-level rise in the range of two to six feet by 2100, not dozens of feet. And sea-level rise of 30 feet or more, which would drown most of the world's largest cities, is a real danger, but it's important to understand that if even if things are much worse than we currently think, it would take centuries for the oceans to rise that high.
Hurricanes are another very uncertain prediction. Some aspects of global warming are likely to make hurricanes more frequent and stronger because hurricanes draw their energy from warm water on the surface of the ocean. As the water warms up, there is more energy for hurricanes and a larger area of the ocean surface that's warm enough to sustain a hurricane. On the other hand, global warming is also likely to create wind patterns similar to El Niño, which tend to break up or inhibit hurricanes, so this might cause hurricanes to become less frequent or weaker. No one knows which effect will win out — the warmer ocean surface or the disruptive wind patterns — so there are fierce arguments among scientists about how global warming is likely to affect hurricanes.
So what I would say to the public on this is that you can trust models very well that temperatures will rise significantly over most of the planet and that there will be big changes to weather and climate patterns, but the models can't predict the exact details with a lot of confidence.