Global climate models cannot predict impacts for even large regions accurately; yet a new study uses them to forecast a 30% drop in winter precipitation for California alone due to Arctic ice melt.
Time and time again, researchers publish papers that use global climate model output at the regional scale to make august pronouncements about our dire future. Yet, as we recently pointed out, scientists were quoted in Nature as saying “privately” that regional climate models—those applied to areas as big as, say, the lower 48 states—were “oversold.”
Yet in that same issue, Nature published a study using global models to forecast major melting in Greenland. And now, Geophysical Research Letters has published a paper that has caused quite a stir by predicting a massive drop in California’s winter rain and snowpack caused by—ready for this?—melting of ice in the Arctic Ocean.
In the Nature news article, scientists pointed to a prime example of “overselling”: the infamous U.S. “National Assessment” of global warming, a Clinton Administration document that has been roundly criticized for its shoddy science projecting the future with models that can’t even simulate the present (see http://www.co2andclimate.org/wca/2004/wca_16b.html)
There’s no doubt that University of California-Santa Cruz researchers Jacob Sewall and Lisa Sloan did exactly what Nature reported was “oversold”: applying a global model for regional climate projections with huge policy implications.
According to the school’s press office, “Sloan and graduate student Jacob Sewall used powerful computers running a global climate model developed by the National Center for Atmospheric Research to simulate the effect of reduced Arctic sea ice.” When the ice was reduced, precipitation along the California coast fell by as much as 30%.
What’s really questionable here is that most of this has to do with winter precipitation, which is the rainy season in California. Projected changes in winter Arctic ice are in fact very small, compared with summer.
So most GRL peer reviewers’ eyebrows should have gone up like an infield fly.
Scientists are taught to be skeptical whenever a tiny cause is said to produce a large and significant effect. Every graduate student is hectored into testing shaky hypotheses before running to the word processor or the school press office.
In this case, a person could easily test whether reduced Arctic ice is associated with reduced precipitation along the West Coast. That’s because it’s been warm up there for the past few decades—as warm as it was 75 years ago when 20th-century Arctic temperatures hit their first maximum.
Figure 1 performs that test. There is obviously no correlation between Arctic ice extent, supplied by NASA, and winter precipitation out West.
Figure1. The history of winter precipitation in two regions of the western United States (top) shows no correspondence at all to Arctic ice extent measurements (bottom) available since 1979 (http://polynya.gsfc.nasa.gov/seaice_projects_image_10.html).
Now, someone may complain that we only compared total Arctic ice extent with western U. S. winter precipitation. Or someone may imply that the model used by Sewall and Sloan could perhaps be sensitive to regional variations in ice cover. But know this: We gathered ice extent data from nine subregions of the Arctic. We performed both individual correlations and a multiple comparison between the regional ice variations and winter precipitation in the western United States. We compared the annual ice extent to the precipitation the following winter (i.e., winter 1998 precipitation was compared with the 1997 average ice extent anomaly [the average of 12 monthly anomalies]).
For precipitation in the California-Nevada region there were no significant correlations with ice extent in any of the Arctic subregions. For precipitation in Washington-Idaho-Oregon, there was one significant correlation, and it was negative. In other words, less ice produced more precipitation.
For anyone who is counting, we found a total of one significant correlation out of 18 comparisons—pretty close to the 1-in-20 expectation from random number using the normal 95% confidence level.
Here are the correlation results (1979-2002):
|Sub-Region||CA-NV Prec.||WA-ID-OR Prec.|
|Sea of Okhotsk||-.02||.12|
|Hudson Bay||.02||-.42 *significant|
|Gulf of St. Lawrence||.02||-.28|
Or, alternatively, the authors could have saved themselves the work by simply consulting these archives, namely Vol. 8, No. 11 of World Climate Report . In response to an analogous paper that used a global model to relate U.S. drought to global warming, we tested that hypothesis, too. It’s in Figure 2. Global temperatures have been on a warming trend since about 1975. U.S. precipitation continues to go up, not down.
Figure 2. Despite rising global temperatures (top), U.S. precipitation has continued to climb (bottom).
In recent World Climate Reports, we have been remarking upon the incredibly slipshod nature of the scientific review process when it comes to papers touting dire projections from global warming. Indeed, we’ve uncovered so many examples that the senior editor of these pages has written a book about it, scheduled for September publication. It’s called Meltdown: The Predictable Exaggeration of Global Warming by Scientists, Politicians and the Media.
Gregory. J.M., P. Huybrechts, and S.C. Raper, 2004. Climatology: Threatened loss of the Greenland ice-sheet. Nature, 428, 616.
Schiermeier, Q., 2004. Modellers deplore “short-termism” on climate. Nature, 428, 593.
Sewall, J.O., and L.C. Sloan, 2004. Disappearing Arctic sea ice reduces available water in the American west. Geophysical Research Letters, 31, L06209. 4 pp.