How Bad are Climate Models? Temperature

Due to building the website for The Climate Sceptics I haven’t been able to post despite some important events. My site and other files were deleted in some kind of attack, so I have had to rebuild it as well. I now have the WordPress 3.0 multiuser system which enable easy creation and management of multiple blogs, so its an ill wind eh?

The important event I refer to is the release of “Panel and Multivariate Methods for Tests of Trend Equivalence in Climate Data Series” by Ross McKitrick, Stephen McIntyre and Chad Herman (2010). Nobody is talking about it, and I don’t know why, as it has a history almost as long as the hockey stick on McIntyre’s blog (summary here), and is a powerful condemnation of climate models in the PRL.

I feel a series coming on, as these results deliver a stunning blow to the last leg that alarmists have been standing on, i.e. model credulity. Also because I have a paper coming out in a similar vein, dealing with drought models in regional Australia.

Using a rigorous methodology on 57 runs from 23 model simulations of the lower troposphere (LT) and mid-troposphere (MT) with forcing inputs from the realistic A1B emission scenario, and four observational temperature series: two satellite-borne microwave sounding unit (MSU)-derived series and two balloon-borne radiosonde series, over two time periods from 1979-99 and 1999-2009, they tested a mismatch between models and observed trends in the tropical troposphere. This represents a basic validation test of climate models over a 30 year period, a validation test which SHOULD be fundamental to any belief in the models, and their usefulness for projections of global warming in the future.

The results are shown in their figure:

… the differences between models and observations now exceed the 99% critical value. As shown in Table 1 and Section 3.3, the model trends are about twice as large as observations in the LT layer, and about four times as large in the MT layer.

The observations as a group exhibit a significant trend of 0.110 C per decade, compared to a model trend of 0.272 C decade. In the MT layer the model trend was 0.253 C decade remains significant but the mean observed trend is a statistically insignificant 0.057 C per decade. So much for the ‘hotspot’, the models exaggerate warming X2 in the lower troposphere and X4 in the middle troposphere.

The paper has many reminders of how far one must be abased to publish in this area. The most controversial statement is the last one:

As noted in the studies cited in the introduction, comparing models to observations in the tropical troposphere is an important aspect of testing explanations of the origins of surface warming.

Thankfully previous incarnations, presumably sent to different journals, have been archived. In the longest, first version they said the claims by Santer et al in 2008 that they achieved a “partial resolution” of the discrepancy between observations and the model ensemble mean trend were “unwarranted”. The problem is that Santer truncated the series to 1999 even though he (and coauthors) had the data available.

The truncation of data in this case leads to statistical insignificance, allowing Santer to claim the models could still agree with reality. McIntyre claimed that when you use all the data the models differ significantly from the observations.

Of course, truncation of inconvenient data is a useful ‘trick’ employed by people from the Climate Research Unit (where Santer did his PhD), and was mentioned in the ClimateGate emails in relation to the ‘divergence’ in proxies. Astoundingly Santer had previously been hauled over the coals for the opportunistic truncation of data in 1996 as recounted by John Daly. Old habits die hard.

This result then inspired the much quoted claim that there was “… a discernible human influence on global climate”, a remark made in the notorious Chapter 8 of the 1995 IPCC Report, a remark added later to the report after the meeting of drafting scientists in Madrid.

But I digress. McIntyre and McKitrick continued with a story of implausible denials by coauthors that, prior to ClimateGate, might have lead to Santer beating them up.

It is puzzling that S08 failed to report on the impact of using up-to-date data on their d1statistic measuring the statistical significance of the difference between model ensemble mean trends and observed trends, given the explicit discussion in their Supplementary Information of procedures that readily yield such results.

In Santer et al (2005), S08 coauthors had calculated monthly synthetic data sets for the tropical T2LT and T2 levels for 49 runs (19 models.) S08 stated that they used the same data sets. We requested this data from S08 lead author Santer, who categorically refused to provide it (see http://www.climateaudit.org/?p=4314.) Instead of supplying what would be at most 1 MB or so of monthly data collated by specialists as part of their research work, Santer directed us to the terabytes of archived PCMDI data and challenged us to reproduce their series from scratch. Apart from the pointless and potentially large time cost imposed by this refusal, the task of aggregating PCMDI data with which we are unfamiliar would create the risk of introducing irrelevant collation errors or mismatched averaging steps, leading to superfluous controversy should our results not replicate theirs.

Following this refusal by lead author Santer, we filed a Freedom of Information (FOI) Request to NOAA, applying to coauthors Karl, Free, Solomon and Lanzante. In response, all four denied having copies of any of the model time series used in Santer et al. (2008) and denied having copies of any email correspondence concerning these time series with any other coauthors of Santer et al. (2008). Two other coauthors stated by email that they did not have copies of the series. An FOI request to the U.S. Department of Energy is under consideration.

The second version is a much shorter 4 pages, but a very useful and publishable paper I think. However, it was not to be apparently, and we are left with the sanitized version that that will appear in Atmospheric Science Letters, and a story for the erudite who still follow this stuff.

But you can rest assured. The models, in important ways that were once claimed to be proof of “… a discernible human influence on global climate”, are now shown to be FUBAR. Wouldn’t it have been better if they had just done the validation tests and rejected the models before trying to rule the world with them?