A quicky, since I have 20 minutes until I need to leave the house, but a new blogosphere hit has been going around, namely on the Assessment of the reliability of climate predictions based on comparisons with historical time series> by Koutsoyiannis, D., N. Mamassis, A. Christofides, A. Efstratiadis, and S.M. Papalexiou.
In their abstract they say, “we have also retrieved a number of climatic model outputs, extracted the time series for the grid points closest to each examined station, and produced a time series for the station location based on best linear estimation.” This involves spatial interpolation of the GCM outputs to infer values at the points of interest. What the authors do is pick a few long records and compare to the nearest individual grid cell.
Chris has a question: should there be a comparable link between observations at a point and model outputs at neighbouring grid cells? Changes in topography (mountains, slopes influence things), land cover, urban heat factors, and other micro-climatic variables are sure to matter. Not to be rude, but I would guess that the people who have been well receptive of the conclusions of this paper would also say that the stations which the authors use are all horribly contaminated…probably because they seen a picture. I als do not feel that the conclusions that models have little predictive ability over longer-term scales (in time or space) are just as invalid as (what the authors feel) the individual sites, compared to the nearest grid cell. It’s not something we can gather directly from their analysis, and so I’d recommend that the authors do not oversell their results.
For a comparison of various variables from IPCC models to observations since 1990, I’d recommend Rahmstorf et al 2007.
Any insight from anyone else would be good.