met.no report 8, 2004

Rasmus E Benestad

Empirically downscaled SRES-based climate scenarios for Norway

RegClim results

The results for 2000-2100 from six global climate models following the latest SRES B2 emission scenarios were examined in terms of common Empirical Orthogonal Functions (cEOFs). These common EOFs were also used as a basis for empirical downscaling, employing a stepwise multiple regression and a number of different predictors and domains.

Statistics is presented for linear trend estimates and their associations with GCMs, predictors and domains. The sea level pressure (SLP) is not considered as an appropriated predictor for temperature, as the warming signal is not well captured by the SLP. The downscaling analysis indicated a general warming of the local climate, however, there were a few cases where negative temperature trends had been obtained. An explanation for these negative trends is that the predictor domains chosen were not appropriate.

The GCMs were generally not able to reproduce the observed annual cycle in the precipitation for interpolated locations. However, the downscaling analysis suggested generally good skill for the models using large-scale precipitation as predictor. Through empirical downscaling, local precipitation series with a realistic annual cycle can be constructed, but the large scatter in local seasonal precipitation variations interpolated from GCMs calls for the question whether current GCMs are able to predict how precipitation patterns will change under an enhanced greenhouse warming. There was no clear signal or consensus regarding future precipitation trends in traditional SLP based downscaling, but new downscaling models employing anomalous large-scale precipitation rates point to slight future trends in the precipitation for a selection of locations.

Keywords: Climate change empirical downscaling monthly mean temperature monthly precipitation

Download pdf-file (ca 5,5 Mb).