Comments
New Paper by McKitrick et al on Tropical Troposphere Trends
TROPICAL TROPOSPHERE: The 2006 CCSP report pointed to a mismatch between models and observed trends in the tropical troposphere as a “potentially serious inconsistency.” In short, the climate models need to get the tropical troposphere right, since it’s a vast region where the models all show a relatively enhanced and rapid response to greenhouse gases. If the models and data don’t agree in that region, there might be a deep problem with the way the models represent the climatic system’s response to greenhouse gases. Or at least there is an issue that needs to be sorted out. An important question, therefore, is whether the apparent mismatch between models and observations is statistically significant, or just random noise. In 2007 Douglass et al. looked at the data and said Yes, the mismatch is significant: if you adjust the models so that they agree at the surface, the resulting profile of tropospheric trends are too high to match the observations. In 2008 Santer et al. said No, the mismatch is not statistically significant. They argued that the Douglass results were biased due to a failure to deal with autocorrelation, and in a model-observation comparison on data ending in 1999 they could not reject a null hypothesis of trend equality. The Santer paper figured prominently in the EPA endangerment finding research and in other places where the validity of climate models is at issue. In a new paper:
* **McKitrick, Ross R., Stephen McIntyre and Chad Herman (2010) “Panel and Multivariate Methods for Tests of Trend Equivalence in Climate Data Series” in press at Atmospheric Science Letters.
we critically examine the common methods used for trend comparisons, and explain some modern econometric methods that provide improved handling of the complex error structures in these data sets. We then apply the methods to the tropical troposphere issue, based on an up to date (1979-2009) time series comprising the full suite of IPCC climate models and 4 observational data series. We conclude that observed trends in the lower troposhere (LT) are significant but those in the mid-troposphere (MT) are not; that on average the balloon and satellite observational data sets agree with each other, though the RSS and UAH satellite series exhibit significant trend differences; that the model-predicted trends are two to four times larger than observed trends and the model-data discrepancy is statistically significant in both the LT and MT layers. See also Supplementary Information; Data/code archive.
See also:
Climate Audit: McKitrick et al (2010) accepted by Atmos Sci Lett