Do you fear the force of the wind,
The slash of the rain?
Go face them and fight them,
Be savage again.
—Hamlin Garland, “Do You Fear the Wind?”
Changes in temperature and precipitation are linked in the climate system. So we should not be surprised to learn that climate models of a future world say that changes in precipitation patterns will occur in response to increasing levels of carbon dioxide in the air.
The IPCC 1996 Summary for Policymakers stated that the expected scenarios in a greenhouse gas-warmed world call for “more severe droughts and/or floods in some places and less severe droughts and/or floods in other places.”
From a scientific viewpoint, the IPCC statement is odd because it is empty of specificity. It combines results from many climate models, none of which is validated. For this reason, all of the predictions can be regarded as equally probable—or improbable—and perhaps none is right.
By combining all the outcomes, we must somehow believe that the true result miraculously emerges from the unproven models. And alongside the truth, the models also produce confusing debris of incorrect results. Perhaps listing an ensemble of results from unproven models gives them a ring of veracity. But how do we know which specific outcome is correct, or that the correct prediction even lies among the many scenarios they produce?
The answers must come from the application of the scientific method, which requires testing the models against good measurements from the real world. A model can make correct predictions if accurate observations validate it. Even then, however, its predictions may be faulty. Still, a good model is a necessary first step in making a credible prediction.
One important feature of the climate system is the hydrological cycle—its pattern of precipitation. Studying changes in, for instance, tropical precipitation is a useful way of testing modeled knowledge of the hydrological cycle, for two reasons. First, tropical precipitation is a driver of global climate change. Second, results for tropical precipitation are known to vary among different models.
B.J. Soden has made just such a comparison of tropical precipitation results, using an ensemble of 31 atmospheric models. He compared year-to-year changes among several key climate parameters. Carrying the comparison over several years is a good approach because it covers the important El Niño-Southern Oscillation cycle, which is a source of major climate influence over a period of years in the tropics as well as over the globe.
Compare the observations and model outputs for the tropics. On the bright side, three of five observed parameters are well modeled: the amount of water vapor (our main greenhouse gas), tropospheric temperature (the temperature of the lower atmosphere), and outgoing long-wave radiation (the heat the Earth emits back out into space).
But the simulated precipitation change ranges from 0.03 to 0.10 millimeters per day (which averages to 0.06 mm), while the observed range in precipitation is a factor of three to four larger. The models also underestimate the amount of change in the absorbed long-wave radiation.
How can the models predict atmospheric temperature change correctly when precipitation, an important factor influencing temperature change, is so far off the mark? There are only three possibilities: Either the observations are wrong, or the models are wrong, or both the models and the observations are wrong.
Wrong models?
Soden argues that if the observations are adequate, then the 31 models are fundamentally flawed. After considering several climate processes, Soden focuses on the models’ inability to explain the observed amount of long-wave radiation absorbed at the surface of the Earth. Such an error might arise, for example, from a poor simulation of low-lying clouds, which closely govern radiation balance in the climate system.
Another problem with the 31 models is their common procedure of specifying sea-surface temperature and then calculating atmospheric response. A. Kitoh and O. Arakawa point out that this process neglects the coupling between air and sea and produces unreliable results. Their bottom line? A model whose sea-surface temperature is fixed can produce a mean state of climate in the tropics very different from that produced by a coupled model.
Wrong observations?
Enough model criticism. What if the observations are wrong? If we do not have an accurate picture of current climate, then it is impossible to validate the models of future climate, so the models’ predictions are not credible.
That the ensemble of models simulates year-to-year changes in the tropical temperature fairly well, yet gives incorrect results for precipitation, leads to two conclusions. First, modeled temperature change is insensitive to model inaccuracies, so it’s a poor way to diagnose systematic errors. And second, precipitation change is a good way to diagnose modeling errors and offers some hope of building credible climate models in the future.
Back to science.
References
Kitoh, A., and O. Arakawa, 1999. On overestimation of tropical precipitation by an atmospheric GCM with prescribed SST, Geophysical Research Letters, 26, 2965–2968.
Soden, B.J., 2000. The sensitivity of the tropical hydrological cycle to ENSO, Journal of Climate, in press.
Sallie Baliunas, Ph.D., and Willie Soon, Ph.D., are colleagues at the Harvard-Smithsonian Center for Astrophysics. This bimonthly contribution to Environment & Climate News is made possible by the George C. Marshall Institute, Washington, DC, where Baliunas is senior scientist and Soon is a visiting fellow.
