Climate modeling turkey shoot, western style

Western U.S. Precipitation Extremes—How Did This Turkey Get Published?

By Dr. Patrick Michaels

When it comes to changes in future precipitation across the United States, climate models projections are all over the map. In other words, they provide no useful guidance for the future. But that doesn’t stop people from trying to sell them. Now comes a paper which clearly demonstrates a systematic failure of precipitation models and still calls the results “useful”. Reviewers…halloo??

A recent example of this comes from a research team led by Francina Dominguez of the University of Arizona who published the results of their study “Changes in winter precipitation extremes for the western United States under a warmer climate as simulated by regional climate models” in the journal Geophysical Research Letters.

Angels fear to tread on “Regional Climate Models”. In fact, Roger Pielke Sr. has been going to town on them relentlessly (see, for example, here) and has recently published a Forum article in the journal Eos titled “Regional Climate Downscaling: What’s the Point?” where he concludes that:

“It is therefore inappropriate to present [regional climate model] results to the impacts community as reflecting more than a subset of possible future climate risks.”

Dominguez et al. apparently didn’t take Roger’s advice to heart when describing the motivation for their work:

“We have analyzed future changes in extreme precipitation events as simulated by dynamically downscaled GCM projections with the goal of providing useful estimates for engineering design of water management structures.”

What is somewhat unusual about this is that unfamiliarity with Roger’s work cannot be an excuse, because one of the co-authors of the Dominguez work is Chris Castro, who got his PhD under Roger at Colorado State University a few years back. Even weirder is that Castro was the lead author of several papers which took a dim view of claims that regional climate models such as those used in the Dominguez study could add skill over that provided by full blown General Circulation Models (e.g., Castro et al., 2005). Which leads back to Roger’s question—what’s the point in trying to do so?

This question is even more germane in the cases where the regional climate models clearly fail to replicate observed reality—which is the situation with the Dominguez et al. study.

Consider Figure 1 (below) taken from the Dominguez paper. The top row shows two versions of the observed mean winter precipitation rate (in mm per day) over the western U.S. (the two leftmost panels) and the same thing produced by an ensemble of regional climate models (right hand panel). The bottom row shows the results for extreme precipitation—the precipitation amount from a 1-in-50 year daily storm event. Grossly the model pattern seems to resemble the observations, but grossly turns out not to be closely when you get down to specifics.

Figure 1. Top Row: Winter 1979–1999 precipitation climatology (mm/day) for two observed datasets (a) and (b), and the multi-model ensemble of eight downscaled simulations (c). Bottom Row: The daily precipitation amount for the 50-yr return period event from the same two observed data sets (d) and (e) and for the multi-model ensemble of eight downscaled simulations (f). (Source: Dominguez et al., 2012).

Figure 2 shows the results of spatial averaging the data in Figure 1. We colored the observations in red. The other symbols are the values from the individual regional climate models. Winter mean precipitation rate is in the left hand panel, the precipitation amount from a 50-yr storm in is the right panel. The amounts have been averaged for four subregions of the study area in Figure 1 (NW, NE, SW, SE) and over the entire region (All). Notice two things: 1) the models all produce more average precipitation than is observed, in some cases by more than 100%, and 2) the models produce extreme precipitation amounts that, in some cases, are routinely 2 to 3 times as heavy as the actual observations show.

Figure 2. Area-averaged mean (left) and 50-year return period (right) winter precipitation for the historical period (1979–1999) for the eight regional climate model simulations (black and white symbols) and the observations (red circle) (Adapted from Dominguez et al., 2012).

Clearly, the regional climate models fail at correctly simulating the processes responsible for producing heavy precipitation events in the western U.S.

Faced with these facts, one should conclude that the regional climate models serve no useful purpose, as Roger Pielke Sr. has been espousing.

Damn the data! Full speed ahead!

Dominguez et al. press on with their futile analysis and use the regional climate models to produce projections of the changes in winter precipitation extremes across the western U.S. for the period 2038-2070.

And you’ll never guess what they found–an increase in the model projected precipitation extremes!

Figure 3 shows the details. The solid black dots are the model average for the percentage change in winter precipitation (left) and in the amount falling in the 50-yr daily storm event (right). The changes in total winter precipitation projected by the individual regional climate models are all over the place—some predict 5, 10, or even 20% more precipitations in some areas, others project similar sized declines. But when it comes to extreme precipitation, virtually all the models project increases, some upwards of 20 to 30%.

Figure 3. Change in the area-averaged mean (left) and 50-year return period (right) winter precipitation between the future (2038-2070) and the historical period (1968–1999) for the eight regional climate model simulations (black and white symbols). The model ensemble mean is indicated as well (solid black circle). (Adapted from Dominguez et al., 2012).

This was a pointless exercise, since the models cannot even replicate observed precipitation extremes.

This is also another example of the shoddy state of peer review in climate science. Repeat after us, if models cannot accurately and robustly simulate the observed climate, they are worthless.

References:

Castro, C.L, R.A. Pielke Sr., and G. Leoncini, 2005. Dynamical Downscaling: Assessment of value retained and added using the Regional Atmospheric Modeling System (RAMS). Journal of Geophysical Research, 110, D05108, doi:10.1029/2004JD004721.

Dominguez, F., E. Rivera, D. P. Lettenmaier, and C.L. Castro, 2012. Changes in winter precipitation extremes for the western United States under a warmer climate as simulated by regional climate models. Geophysical Research Letters, 39, L05803, doi:10.1029/2011GL050762.