Bishop Hill McKitrick's new paper
Jun 21, 2012 Climate: Models Climate: Surface Ross McKitrick writes:
Lise Tole and I have published a paper in Climate Dynamics testing the ability of climate models to reproduce the spatial pattern of temperature trends over land. This builds on previous work of mine looking at the correlation between indicators of industrial development over land and the spatial pattern of warming trends, a relationship that is not predicted by models and is supposed to have been filtered out of the surface climate record. The paper is
- **McKitrick, Ross R. and Lise Tole (2012) “Evaluating Explanatory Models of the Spatial Pattern of Surface Climate Trends using Model Selection and Bayesian Averaging Methods” Climate Dynamics, 2012, DOI: 10.1007/s00382-012-1418-9
Preprint here; data and code archive here; university press release here. We apply classical and Bayesian methods to look at how well 3 different types of variables can explain the spatial pattern of temperature trends over 1979-2002. One type is the output of a collection of 22 General Circulation Models (GCMs) used by the IPCC in the Fourth Assessment Report. Another is a collection of measures of socioeconomic development over land. The third is a collection of geopgraphic indicators including latitude, coastline proximity and tropospheric temperature trends. The question is whether one can justify an extreme position that rules out one or more categories of data, or whether some combination of the three types is necessary. I would describe the IPCC position as extreme since they dismiss the role of socioeconomic factors in their assessments. In the classical tests, we look at whether any combination of one or two types can "encompass" the third, and whether non-nested tests combining pairs of groups reject either 0% or 100% weighting on either. ("Encompass" means provide sufficient explanatory power not only to fit the data but also to account for the apparent explanatory power of the rival model.) In all cases we strongly reject leaving out the socioeconomic data. In only 3 of 22 cases do we reject leaving out the climate model data, but in one of those cases the correlation is negative, so only 2 count--that is, in 20 of 22 cases we find the climate models are either no better than or worse than random numbers. We then apply Bayesian Model Averaging to search over the space of 537 million possible combinations of explanatory variables and generate coefficients and standard errors robust to model selection (aka cherry-picking). In addition to the geographic data (which we include by assumption) we identify 3 socioeconomic variables and 3 climate models as the ones that belong in the optimal explanatory model, a combination that encompasses all remaining data. So our conclusion is that a valid explanatory model of the pattern of climate change over land requires use of both socioeconomic indicators and GCM processes. The failure to include the socioeconomic factors in empirical work may be biasing analysis of the magnitude and causes of observed climate trends since 1979.
I have written a pair of op-eds to explain the work. The first part appeared in the Financial Post on June 21. A version with the citations provided is here. Part II is here online, and the versions with citations is here.
Reader Comments (10)
"in 20 of 22 cases we find the climate models are either no better than or worse than random numbers": and that's probably an upper bound on the merits of climate models. Suppose that McKitrick & Tole have overlooked some socioeconomic factors that matter (being humans, and not Climate Scientists, they might make such an oversight). There will then be some slack available in their predictors which might, by fluke, be taken up by a couple of climate models. It's plausible, one might guess, that perfect insight into the relevant socioeconomic factors would perhaps reduce the role of climate models close to nil. Who knows?
Even if 2 out of 22 models were to survive such a hypothetical test, it still wouldn't demonstrate that the models were a good physical account of what's going on - they still might be little more than accidental suppliers of correlates.
One thought occurs: is there any feature of the two blessed models that distinguishes them qualitatively from the dud nineteen? Indeed, is there any feature of the one cursed upside-down model that distinguishes it qualitatively from the dud nineteen, or the blessed two?
I'm just waiting for "The Teams" response before I can comment! :)
Regards
Mailman
Another little something for you on our friend Richard Black Bish:
http://blackswhitewash.com/2012/06/21/richard-black-so-what-is-he-actually-doing-in-rio/
dearieme: The best model (identified by both the classical and bayesian tests) is a Chinase model IAP FGOALS 1.0g. The other two (identified by either one) are NCAR CCSM 3.0 and Russian INM CM 3.0. I don't know anything about these models, but one could easily speculate that these models may have been independently developed (from the "main stream" models). Does anyone have any real knowledge about the issue?
Also it is worth noticing that although the other model of NCAR was identified, the other one (NCAR PCM 1) was not Maybe that gives a hint to a knowledgeable person.
One wonders how much of the statistical significance attributed to the two models was pure chance, especially given that the third model was *negatively* correlated with reality; which, if nothing else, supports the hypothesis that significant correlation is possible as a matter of chance alone.
The models in question are, after all, designed to produce outputs that are at least climate-like, regardless of whether they possess explanatory power or not ... the point being that they're not just white noise generators or some other random process for which the probability of correlation with real climate data would be of an extremely low order of probability.
In this case, one of the twenty two happened to be negatively correlated, which, to my mind at least, brings into question the significance of the fact that two out of the twenty two happened to be positively correlated, unless the correlation was much stronger in the case of the two positively correlated models than in the case of the single negatively correlated model.
Two of twenty-two. Isn't that 90% confidence? Best go look for my Montgo or my Bethea.
==========
One thought occurs: is there any feature of the two blessed models that distinguishes them qualitatively from the dud nineteen?
===============
The two models that performed were from Russia and China. The other models, including the negative model were from the "west".
Well, ferd, what can one say but that those communist bastards must be out to disturb a cosy academic club?
Why does RossMcK always provide links to his data and his code -- doesn't know that's not the climate science way?
Mosher has found that McKitrick's data is full of nonsense: http://judithcurry.com/2012/06/21/three-new-papers-on-interpreting-temperature-trends/#comment-211553
Maybe it doesn't change the thrust of his conclusions. But as Mosher says - funny how people don't check the workings when they like the conclusions.
For an undogmatic noob trying to get his head around this stuff, the propaganda & politicisation on both sides of the debate is tedious & childish.