by Patrick Michaels
A critique of Judah Cohen’s recent cover article in Science linking February’s disastrous cold outbreak in Texas to global warming
I’ve always had trouble with the notion that warming causes cooling. It leaves me with the squeasy feeling I get when my country neighbors insist that putting hot water in the ice cube tray results in quicker ice cubes. That’s actually an experiment you can run, and I can assure that it most certainly does not (despite the arguments that are likely to follow in the comments section).
But it’s much harder to run a similar experiment on, say, the hypothesis that an anomalous and costly ($200 billion) cold outbreak in Texas last February was caused by global warming. Leaving out that much of the damage had to do with remarkably unprotected generation equipment—both conventional and renewable—it was very cold and windy, even by Texas blue norther standards. You just can’t stick a slightly warmer Texas in the fridge to see if it now freezes faster.
Predictably, champions of the warming-causes-cold-anomalies have come forward, with Judah Cohen, a consulting atmospheric scientist, with his theory that sea-ice changes in the arctic and snow-driven October changes in Siberia conspire to stretch the stratospheric polar vortex down to, say, Texas. Somehow his stuff always makes it into The New York Times, which is likely not a measure of its quality, but rather yet another thing to turn on their climate change alarm (which it rarely turns off).
Cohen concludes:
“Therefore, Arctic change is likely contributing to the increasing of SPV [Stratospheric Polar Vortex] stretching events, including one just prior to the Texas cold wave of February 2021.”
How he reached this conclusion is a conventional story. First, break down some target variable (in this case,100mb-heights) into characteristic patterns, and then use a General Circulation Model (GCM) to explain its behavior. While Cohen and his four coauthors said the patterns were from “a machine learning technique”, it was actually good old-fashioned cluster analysis, something that has been around physical geography since the ice age.
Guess what. Amplitudes of some of the clusters are going up, others are going down and, 40% have no statistically significant changes. Cohen then correlated these changes to October Eurasian snow cover.
Given that Cohen has had some success in correlating October Siberian snow amount and geographic advance across with cold outbreaks into the U.S. (along with reductions in ice cover in the Arctic Ocean), he sought to “prove” the relationship with “a simplified GCM…well suited for isolating the atmospheric response to idealized heating perturbations”. The model is acronymed MiMA, for Model with an idealized Moist Atmosphere.
The word idealized isn’t defined, nor is the related reasoning, so we have to consult Chaim Garfinkel, the fourth author of the Cohen paper, and the first author of a paper describing MiMA, where we find out that it’s “idealized” because the extant GCMs are “tuned” so much that they become unstable:
“These comprehensive [general circulation] models, however, tend to be less flexible and tuned such that removing too many relevant forcings leads to unstable behavior.”
A good guess as to what’s “tuned” in the GCMs that leads to unstable behavior might be what’s left out of MiMA – it has no clouds. The albedo (think of “reflectivity”) of clouds exerts a net cooling particularly over latitudes away from the tropics. MiMA artificially decreases the earth’s albedo because of its lack of clouds, from constant 27% down to about a constant 20% (in reality it is never constant), which represents a massive 25% increase in solar radiation heating the earth’s surface.
So, to this simulated climate, Cohen et al. change (raise) the albedo of Siberia and east Asia in the early fall, to compensate for an increase in October snow cover that has been detected since 1979, as well as raise the temperature of the model’s Arctic Ocean to get it to lose more ice.
And, presto-chango, the modified model stretches its wintertime polar stratospheric vortex to somehow get to Texas in February 2021. How useful this is for his company that makes money by selling in-advance winter forecasts. Just think of how many billion dollars (and lives) could be saved the next time he makes such a forecast!
Indeed, Cohen goes on to note: “Third, our analysis is informative for policymakers”. He finishes by noting that it’s unwise to prepare for “only a decrease in severe winter weather” (there is some evidence Texas did this, judging from the performance of their backup gas plants, which were too cold to fire up), when the stratospheric vortex might stretch all the way down to the Lone Star State, as shown by his cloudless, constant-albedo model of what can only charitably be related to the earth’s climate.
So does Cohen actually get a better handle on Texas cold outbreaks in an atmosphere with no clouds and a constant albedo? Except for Siberia, which he did brighten, which, everything else being equal, will become colder from increased snowfall precipitated by a cloudless atmosphere. This allows the big, seasonal cold Siberian high-pressure systems to get larger, increasing the likelihood that the vortex will transport some of its cold air down to Texas.
If you’re scratching your head after reading this, think of how much hair I lost reading Cohen’s paper. It’s got a lot of pretty pictures that look seductive until you get into the details as to how they were ultimately applied by the MiMA model.
The bottom line is that Cohen et al. are going to have to be lot more convincing before I believe that a single month’s snowfall in Siberia drives the weather thousands of miles and several months away.