In a recent paper by Ruosteenoja et al. the simple-pattern-scaling idea, described first by Santer et al. has been tested for average temperature and precipitation change signals, and proven to perform extremely satisfactorily. This method allows different scenarios to be run by intermediate models, far less complex and expensive and much faster, but limited to producing climate change signals only in terms of global or zonal averages. These signals can then be combined to geographical patterns of climate change, estimated on the basis of a full AOGCM run under one particular scenario. The hypothesis underlying this method is that such geographical patterns remain somewhat constant across different scenarios, and are just modulated by the average climate change signal.
We apply this method to runs of the NCAR PCM-DOE model performed under different SRES scenarios, and we study average surface temperature and precipitation patterns of change between the two periods of 1961-1990 and 2070-2099, separately for the four seasons.
Our work addresses two important aspects of the application of pattern-scaling to climate-change projections, that have been insufficiently studied so far. The first one is the evaluation of the scaling method for different horizontal scales of resolution, i.e. for different scales of regional aggregation. The second is the computation of uncertainty measures, to be attached to the scaled values, in order to provide a degree of confidence in the robustness of the scaled signal.
Our findings suggest that the method is accurate down to very fine spatial scales, and the error has the characteristics of a gaussian random field, whose parameters can be estimated by standard statistical methods and consitute the basis for the characterization of the uncertainty in the scaled signals.
Supplementary URL: