September 23, 2004

Climate Model In—Climate Model Out

Filed under: Climate History

Increased rainfall? Lower temperatures? So suggests a regional climate model. Yet, the regional climate model projecting a “warming hole” in the U.S. Midwest for the coming century suffers from the same difficulties all climate models do—the inability to accurately capture reality. So despite the fact that the net effects of the projected climate change would be overwhelmingly beneficial for the regions agriculture, it looks like this is just another in a long string of modeling studies that are more useful as an academic exercise than as a reliable indicator of our climatic future.

Here’s what our readers know: We’re continually amazed by people who say global warming will be bad in general, or particularly bad for agriculture. And we’d prefer that climate models should “work” before being used to estimate future changes.

Here’s something else you know: Carbon dioxide makes crops grow better and produce higher yields. We could list all the scientific references demonstrating that fact, except your computer is liable to choke on the download.

We’ve also informed you that midlatitude rainfall has been increasing—about 10% or so in the United States. Summer temperatures, on the other hand—when things grow—have barely budged. In fact, summer average temperatures over the Midwest and eastern Great Plains have either stayed the same or declined a bit in the last quarter-century.

Ah, but what we didn’t know is that now global warming is thought to be the cause of the increased rainfall and cooler temperatures! At least that’s the take-home lesson from a paper just published by St. Louis University’s Zaitao Pan and five other co-authors from Iowa State University.

They took a government-issue climate model, in this case the from Britain’s Hadley Centre (acronymed HadCM2) and coupled it to a smaller climate model, called a regional climate model, or RCM, for the Midwest.

That is a common game, an attempt to make up for the fact that big climate models, such as HadCM2, don’t resolve well at, say, the level of Kansas. So the overall output of the big model for, say North America or the lower 48 states is used to “drive” the smaller model, which may be the size of a Kansas or two, so as to determine how large scale changes will have smaller-scale effects.

We choose our words carefully in our World Climate Reports. Here is the one we choose to describe the input to the RCM: garbage. This was the hint we got from the “Methods” section of Pan et al.: “HadCM2 was one of the two models for the U.S. National Assessment of Climate Change.”

We tested that model against 10-year running means (i.e., 1991-2000; 1990-1999; 1989-1998, etc.) of U.S. (lower 48) temperature departures from normal in the 20th century. That is the period of changes in greenhouse gases, and it is in the second half of that century (particularly the last quarter) that most people (including us) think there is a human component to planetary temperature trends.

HadCM2 couldn’t beat a table of random numbers, a ouija board, or us at the race track. We sent that test to the “Synthesis Team” for the National Assessment, and their senior climatologist ran an even more comprehensive test (five-year, 10-year, 25-year… running means) and found the same.

So, again, we offer this well-chosen word: garbage. That’s stuff that Pan et. al. input to the RCM. It’s not their fault, though. How could they know? After all, HadCM2 was one of two models in the National Assessment (the other, the Canadian Climate Model produces similar garbage for U.S. temperatures), a document that never mentioned this slight problem with their input models.

(Anyone who wants to read the whole story about this sordid mess should buy Meltdown: The Predictable Distortion of Global Warming by Scientists, Politicians and the Media, coming out next month. You can order this gem, by WCA senior editor Pat Michaels, from right now!)

Again, we’re not faulting Pan et. al. We don’t think they read these pages, which is one of very few places where you can find out about the true nature of HadCM2 when applied to U.S. temperatures. (Come to think of it, why don’t they?)

Because we’re willing to let them off the hook, let’s dissect their result a bit more. Figure 1 is taken directly from Pan et al. It shows projected changes in precipitation, evaporation, and latent heat flux over the next fifty years for the central United States.

Midwestern Climate

Figure 1. Change in daily precipitation, evapotranspiration, and sensible heat flux in the Midwestern “warming hole” (difference between 2040s and 1990s) (from Pan et al., 2004).

Pan et al. project that warm-season (April-October) precipitation will increase a whopping 34%, or about 7.5 inches. Evaporation increases by about 3.5 inches, for a net change in soil moisture of around 4 inches.

That is a huge number for their study region. Normally the region receives about 6 inches less rain than evaporation, so there is a tendency for dry soil, which drives down agricultural yields (July and August rainfall are by far the greatest determinants of Midwestern corn and soybean yields, respectively).

Dropping the deficit to two inches will dramatically increase the likelihood of bumper crops because that will mean that in an average year there will be no prolonged period of moisture stress. In fact, the reason that Pan et al. project virtually no summer warming in the next half-century is because so much excess solar energy is directed towards evaporating all that water rather than directly heating the surface (and the fact that it must be cloudier to produce all that extra precipitation).

Pan et al. also claim that this trend has already started, as the region shows a slight summer cooling, citing a “long-term increase in precipitation in the same region” and then showing the summer temperature decline from 1976 through 2000.

But things aren’t so cut-and-wet. We looked at Kansas statewide data, pretty much in the middle of their study area. Pan et al. are projecting changes for April through October. We found no significant temperature decline for 1976-2000. We found no significant precipitation increase. We did find a significant negative correlation, though, between temperature and precipitation. In other words, the more the temperature was below normal, the greater the rainfall. Their model projects above-normal temperatures to evolve over the next 50 years.

Are we being too hard on models such as HadCM2 because they can’t beat random numbers? Is it wrong to say that these shouldn’t be input to smaller scale models because the results cannot be trusted? Are we just killjoys because we couldn’t find that the claimed increases in precipitation or decreases in temperature are statistically significant (which means, scientifically, that the changes are indistinguishable from zero)?

Or, is this just yet another example of the horrors that are happening to the process of scientific peer-review. We’ll opt for the latter, and anyone who wants to see more and more and more of this should order Meltdown.


Pan, Z., et al., 2004. Altered hydrologic feedback in a warming climate introduces a “warming hole.” Geophysical Research Letters, 31, L17109.

No Comments

No comments yet.

RSS feed for comments on this post. TrackBack URI

Sorry, the comment form is closed at this time.

Powered by WordPress