Why regional climate forecasts are meaningless
Regional assessments of future climate change. That first word, “regional,” should strike fear in the hearts of anyone who believes factual evidence is an essential prerequisite for any informed policy decision.
Such assessments are impossible. How can anybody assess the impact of “global warming” on, say, Ottumwa, Iowa, and its maizey environs? How will Ottumwa’s climate change by the year 2075? Should Ottummies start preparing for the conversion of farmland from corn to oranges? Should they build more catchment ponds or move to higher ground? What’s a regional planner in Iowa to do?
The truth is, no one can speak to such climate change specifics. Yet the federal government has mandated the production of a report on the future climate of various regions of the United States. Climate Change Impacts on the United States, commonly known as the U.S. National Assessment—which has consumed years of taxpayer-financed person-hours of federal scientists and nonscientists alike—has been released in dribs and drabs over the past several months.
After years of keeping track of various researchers’ efforts to forecast future regional climates, my scientific colleagues and I have noticed they all seem to take one of two naive approaches.
Naive Approach No. 1: Take Ottumwa’s current climate (the last 50 years should suffice), and raise the temperature every day by one, two, five, or 10 degrees, and then see what happens to livestock, water, corn, disease, ticks, fleas, slugging percentage, and so on. Never mind that climate does not work that way. In fact, it is not physically possible for Ottumwa’s future climate to be anything like that. Nevertheless, Naive Approach No. 1 does have the significant and not-to-be-taken-lightly advantage of simplicity.
Naive Approach No. 2: Use one, or better yet, several, general circulation climate models (GCMs) to see what is supposed to happen in Ottumwa in 2075. Of course, all GCMs come stamped with a warning: These Projections Should Not Be Used To Make Regional Predictions. Yet those forecasts are nevertheless provided as a service to the “user-community.” So you look up the grid cell or cells closest to Ottumwa, find the model year closest to 2075, see that the temperature is projected to increase by 1.895726443°C in March, and use this number to investigate the impact on livestock, corn, disease, ticks, fleas, and so on.
Naive Approach No. 2 has the significant advantage of not appearing to be naive, given that it’s based on the prognostication of a sophisticated model that has lots of equations, variables, parameterizations, feedbacks, El Niños, albedos, and the like and have cost us about a billion dollars.
The problem is that those models do not work at that level of specificity. Many well-informed climatologists would tell you with some confidence that, by 2075, the globe will probably be a bit warmer (more so in nighttime minima than in daytime maxima), and that precipitation is likely to be about the same or perhaps higher over the globe. But such a general and very large-scale forecast provides no useful information for Ottumwa’s regional planner.
Nevertheless, GCMs serve as the basis for much of the Regional Assessment–related hype. For example, the latest issue of the journal Climate Research is devoted entirely to the Mid-Atlantic Regional Assessment of the impacts of climate change. The sections on forestry, water resources, coastal impacts, and health were all based, to varying degrees, on GCM regional forecasts for the mid-Atlantic.
All of this provides an interesting context for two papers that recently hit the dusty shelves of the university library nearest you. First is a paper by German researchers Zorita and Gonzalez-Rouco in which they compared an atmospheric feature called the Arctic Oscillation (AO) in two GCMs. That oscillation is important because it is pretty strongly related to winter climate in the Northern Hemisphere. When the AO is strong, for example, Eurasia has milder-than-normal winters.
They compared AO forecasts using two models: the Hadley Center GCM and the model from the Max-Planck Institut of Meteorology. First, both models agree with each other in reproducing the mean Northern Hemisphere winter circulation patterns and their variability. But when the models are forced by increasing greenhouse gas levels, one model predicts significant increases in the AO while the other produces both upward and downward trends for different model runs. The authors write:
The different AO trends should also have an impact on the simulated regional air-temperature change. A negative AO trend . . . should weaken the [predicted] temperature increases over Eurasia and Southeastern USA and reinforce temperature increases over Greenland and Western Canada; positive trends . . . should show opposite tendencies.
They conclude, “the predictions of the intensities of the main patterns of atmospheric circulation, even at planetary scales, are either not yet reliable or they depend strongly on internal model variability.” In other words, at present, general circulation models provide little information about future general circulation.
The second bit of bad news for regional climate soothsayers emanates from two climate scientists at the National Center for Atmospheric Research. Giorgi and Francisco assembled the output of five different GCMs for 23 terrestrial regions across the globe and compared model predictions for temperature and precipitation for the years 2070–2099 relative to the 1961–1990 baseline period. Furthermore, they determined how good each model was at reproducing the 1961–1990 baseline climate.
Obviously, the latter comparison is more important, for if the models fail to reproduce current climate, then what they say about the future is irrelevant. In general, while some models are pretty close to the zero line (no error) in some regions, in general there is oodles of scatter. In some cases the temperature error is more than 5 degrees and the precipitation errors approach 200 percent but are generally less than 100 percent, at least from June to August. No one model does much better than any other across all regions.
Given their inability to map present conditions, it’s hardly worthwhile to even consider their projections for the future—which vary so much from projection to projection as to be irrelevant. Indeed, some models show large precipitation decreases in regions where others show large increases.
So what’s a well-meaning Ottumwa regional planning official to do about future climate change? Our advice is simple: Flip a coin, write a report, and take the afternoon off to play golf.
Robert E. Davis, Ph.D., is an associate professor of environmental science at the University of Virginia.
Giorgi, F., and R. Francisco, 2000. Evaluating uncertainties in the prediction of regional climate. Geophysical Research Letters, 27, 1295-1298.
Zorita, E., and F. Gonzalez-Rouco, 2000. Disagreement between predictions of the future behavior of the Arctic Oscillation as simulated in two different climate models: Implications for global warming. Geophysical Research Letters, 27, 1755-1758.