Dr. McKitrick’s new paper with Lise Tole is now online at Climate Dynamics. He also has an op-ed in the Financial Post on June 13. A version with the citations provided is here. Part II is here online, and the versions with citations is here.
**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
The abstract is:
We evaluate three categories of variables for explaining the spatial pattern of warming and cooling trends over land: predictions of general circulation models (GCMs) in response to observed forcings; geographical factors like latitude and pressure; and socioeconomic influences on the land surface and data quality. Spatial autocorrelation (SAC) in the observed trend pattern is removed from the residuals by a well-specified explanatory model. Encompassing tests show that none of the three classes of variables account for the contributions of the other two, though 20 of 22 GCMs individually contribute either no significant explanatory power or yield a trend pattern negatively correlated with observations. Non-nested testing rejects the null hypothesis that socioeconomic variables have no explanatory power. We apply a Bayesian Model Averaging (BMA) method to search over all possible linear combinations of explanatory variables and generate posterior coefficient distributions robust to model selection. These results, confirmed by classical encompassing tests, indicate that the geographical variables plus three of the 22 GCMs and three socioeconomic variables provide all the explanatory power in the data set. We conclude that the most valid model of the spatial pattern of trends in land surface temperature records over 1979-2002 requires a combination of the processes represented in some GCMs and certain socioeconomic measures that capture data quality variations and changes to the land surface.
He writes on his website:
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.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
This study helps explain why the NE USA (the rust belt) has shown long term declining temperatures.
[SNIP: Rude, crude and uninformative. If you have something substantive to say, please do so. -REP]
Koutsayannis rocks through again! Could his hydrology papers prove to be the straw that ultimately breaks the back of the GCMs? Backs (to be sure) that are already strongly bent under the stress of the unremitting work done by McKittrick and McIntyre and their collaborators.
One thing is very clear. Climatologists all flunked statistics in college, or else they took just the one mandatory stats class that covers the Gaussian, erf, t-tests and so on, and never quite made it to Bayes theorem. What is the prior probability that any of the GCMs are close to correct? No better than the probability that we actually understand the climatological dynamics of the glacial-interglacial transition, since it is always good to understand the gross features of the elephant you are examining before trying to resolve the wart on its butt.
rgb
This paper, and other measured human induced climate effects, open a can of worms into the spatial statistics associated with the distribution of weather data in the land records, the subsequent samples used for analysis, and the “gridding” methodology. I’ve beat on this drum before, the underlying data mapping and analysis thereof is a problem for climate science.
Atmospheric temperature is not a metric for atmospheric heat content.
The hypothesized ‘green house effect’ is based on the notion that some gases reduce the rate of heat loss to space.
Minor changes in humidity (which has apparently been reducing) could account for all the atmospheric temperature changes by reducing atmospheric enthalpy.
However esoterically – why argue over the incorrect metric?
So, what do the two significant models project for trend GHG accumulation? What about for zero emissions?
“This study helps explain why the NE USA (the rust belt) has shown long term declining temperatures.”
Yeah, what does the surface station data look like when compared to regional CO2 levels and changes?
What specifically are the “certain socioeconomic measures” that were tested for their explanatory power?
I can’t find any mention of these details in the abstract, or in McKitrick’s comment in the Financial Post.
Why 2002? For that matter, why 1979?
Surrealclimate is not going to like this 🙂
I’m shocked, shocked I tell you, that the GCMs have so little explanatory power! I mean, they’re run on computers, right? How can computers be wrong???
/sarc
Don Keiller says:
June 21, 2012 at 10:55 am
Surrealclimate is not going to like this 🙂
xxxxxxxxxxxxxxxxxxxxxx
+10
At RealClimate, Gavin has already responded to a commenter’s question on this paper. Rather than me trying to paraphrase his points, I’m just going to quote him verbatim:
“McKitrick is nothing if predictable. He makes the same conceptual error here as he made in McKitrick and Nierenberg, McKitrick and Vogel and McKitrick, McIntyre and Herman. The basic issue is that for short time scales (in this case 1979-2000), grid point temperature trends are not a strong function of the forcings – rather they are a function of the (unique realisation of) internal variability and are thus strongly stochastic. With the GCMs, each realisation within a single model ensemble gives insight into what that internal variability looks like, but McKitrick averages these all together whenever he can and only tests the ensemble mean. Ironically then, the models that provide the greatest numbers of ensemble members with which to define the internal variability, are the ones which contribute nothing to his analysis. He knows this is an error since it has been pointed out to him before and for McKitrick and Nierenberg and McKitrick, McIntyre and Herman he actually calculated the statistics using individual runs. In neither case did those results get included in the papers. The results of those tests in the M&N paper showed that using his exact tests some of the model runs were ‘highly significantly’ (p<0.01!!) contaminated by 'socio-economic' factors. This was of course nonsense, and so are his conclusions in this new paper. There are other issues, but his basic conceptual error is big one from which all other stem. – gavin"
I’d like to see RM’s response to Gavin’s critique, which sounds a bit ummmm . . . dodgy.
Actually 1979-2002 in this paper, 1979-2009 for M,M&H; and 1959-2010 for McKitrick and Vogelsang (not “Vogel”). How can the excuse about models not being meaningful on a short time scale apply to all those different intervals? And when they thought Santer et al. had shown consistency between models and observations over the even shorter 1979-1999 interval, they were all happy to declare the matter resolved.
We tested each model individually and in every possible linear combination, including an average of them all. But if Gavin really believes that all a model run shows is “strongly stochastic” interval variability, then how would an accumulation of dozens or hundreds of such runs yield any information, especially if he condemns the average? Lots of people are using GCM runs to make forecasts of climatic changes at the local level. Either these are meaningful or not. If not, fine, let’s say so and throw all those studies in the trash. But if we are supposed to take them seriously, then let’s first test the models against observations and see how well they do.
I’ve responded to published criticism, but this new claim says, in effect, that GCMs should never be compared to observations on any time scale. It is not one I’ve encountered before. If Gavin thinks he has a legitimate point he should send a comment in to the journal.
The cases Gavin showed in his IJOC paper did not correct for spatial autocorrelation (after he had spent so much ink complaining about the problem in my results, ignoring the fact that it is not a problem in my model but was a real factor in his). And his “significant” coefficients in the model-generated data took the opposite sign to the coefficients estimated on observations, so it is a significant FAIL not a significant replication of the effect. Finally, Gavin ignores the fact that not every model gets equivalent posterior support in the data. The Bayesian Model Averaging accounts for this.
Thanks, Ross.
Here’s something interesting: I give you CAUW:
http://houston.cbslocal.com/2012/06/21/climatologist-urban-heat-islands-could-increase-mortality-rates/
“But if we are supposed to take them seriously, then let’s first test the models against observations and see how well they do.”
Arguing in favor of empirical validation is a surefire sign of a crank. It’s a wonder you get anything published when spewing such pseudoscientific nonsense.
JC-why 1979? Because that is when the satellite records of RSS and UAH start. Why 2002? Because some of the socioeconomic data was originally available only to then.
Gavin Schmidt has said,
“Any single realisation (of a GCM) can be thought of as being made up of two components – a forced signal and a random realisation of the internal variability (‘noise’). By definition the random component will uncorrelated across different realisations and when you average together many examples you get the forced component (i.e. the ensemble mean).”
At best, the internal variability is the modellers estimate of natural climate variability.
Otherwise, the socioeconomic factors are likely proxies for aerosol emissions.
for McKitrick and Nierenberg and McKitrick, McIntyre and Herman he actually calculated the statistics using individual runs. In neither case did those results get included in the papers. The results of those tests in the M&N paper showed that using his exact tests some of the model runs were ‘highly significantly’ (p<0.01!!) contaminated by 'socio-economic' factors.
What Gavin is claiming is that if you run the model enough times you will eventually get an accurate representation of the actual climate over whatever period.
Well, Duh. Thats what randomness will do. More subtly he is suggesting that the models accurately model natural variability, but can’t come out and say this, because he is on the record as saying they don’t.
Philip Bradley says: ‘Otherwise, the socioeconomic factors are likely proxies for aerosol emissions.”
If so, then the aerosol effects must be to warm, not cool, since many of the models are negatively correlated with the observed spatial pattern, and those models use aerosols…
The pathetic tentatives to find meaning in a big pile of noise and chaos with too many variables to be understandable continue…
timetochooseagain: Thanks.
timetochooseagain says:
June 21, 2012 at 5:10 pm
If so, then the aerosol effects must be to warm, not cool, since many of the models are negatively correlated with the observed spatial pattern, and those models use aerosols…
Reduced aerosols warm the surface by increasing solar insolation. And they have reduced over most of the world land area over last 40 years or so. Unfortunately satellite measurements only go back about 15 years (about the time warming ended) and miss the main reductions from the 1970s. But still show modest aerosol reductions over most land.
http://www.atmos-chem-phys-discuss.net/12/8465/2012/acpd-12-8465-2012.html
That GISS and other models incorporate an increasing cooling effect from aerosols over this period is suspect to say the least.
http://data.giss.nasa.gov/modelforce/
Interesting historical analysis of black carbon and organic carbon emissions.
http://www.cee.mtu.edu/~nurban/classes/ce5508/2008/Readings/Bond07.pdf
Note table 4. large reductions in coal emissions in Europe, N America and the former USSR since 1950, with a huge increase in China.
There have been similar reductions in vehicle emissions since the mid-1970s.