Bishop Hill McKitrick and Vogelsang
Aug 16, 2011 Climate: Models Ross McKitrick has a new discussion paper out. This is an extension to the McKitrick, McIntyre and Herman paper that came out last year. In the earlier paper, MMH found evidence that climate models were overstating warming, particularly for observations that were right up to date. The new paper looks at the so-called "Pacific Climate Shift" of 1977-8 and develops a new method to determine confidence intervals when this shift is taken into account in the statistical model.
As our empirical findings show, the detection of a trend in the tropical lower- and midtroposphere data over the 1958-2010 interval is contingent on the decision of whether or not to include a mean-shift term at January 1978. If the term is included, a time trend regression with error terms robust to autocorrelation of unknown form indicates that the trend observed over the 1958-2010 interval is not statistically significant in either the LT or MT layers. Most climate models predict a larger trend over this interval than is observed in the data. We find a statistically significant mismatch between climate model trends and observational trends whether the meanshift term is included or not. However, with the shift term included the null hypothesis of equal trend is rejected at much smaller significance levels (much more significant).
Or in Ross's layman's terms summary:
Controlling for the 1977 Pacific Climate Shift we find the trends are insignificant from 1958-2010 and the discrepancy with climate models is highly significant.
Reader Comments (5)
This looks pretty state of the art, in that Tim Vogelsang of Vogelsang-Franses multivariate trend estimation fame (!) has "extended the theory behind the VF method to yield robust trend variances in the presence of autocorrelation of unknown form when a step-change occurs at a known point in the sample." It's amazing what could eventually flow from econometrics to climate science, albeit over the dead bodies of the present incumbents.
Quote, conclusion, "We find a statistically significant mismatch between climate model trends and observational trends whether the meanshift term is included or not. However, with the shift term included the null hypothesis of equal trend is rejected at much smaller significance levels (much more significant)."
To summarise: When tested against observational data, with or without a step change in data, modelled projections on trends are poor - hence showing the power of such an analysis.
This new paper might even encourage Climate Scientists to give more weight and credence to accurate and testable observations.
But some blokes never grow out of playing with model trains, either. They are much tidier and much more controllable than reality, which can be messy, uncomfortable and dangerous.
Ross had better stop 'controlling' for the 1977 Pacific Climate Shift and start 'explaining' it physically...
Why Ross ( an economist / statistician) ? Sounds like a job for physical scientists.