CR 19:45-55 (2001)  -  doi:10.3354/cr019045

ABSTRACT: Two dynamically and statistically downscaled precipitation scenarios for Sweden are compared with respect to changes in the mean. The dynamically downscaled scenarios are generated by a 44 km version of the Rossby Centre regional climate model (RCM). The RCM is driven by data from 2 global greenhouse gas simulations sharing a 2.6°C global warming, one made by the HadCM2 and the other by the ECHAM4 general circulation model (GCM). The statistical downscaling model driven by the same GCMs is regression-based and incorporates large-scale circulation indices of the 2 geostrophic wind components (u and v), total vorticity (ξ) and large-scale humidity at 850 hPa (q850) as predictors. The precipitation climates of the GCMs, RCMs and statistical models from the control runs are compared with respect to their ability to reproduce the observed seasonal cycle. Great improvements in the simulation of the seasonal cycle by all the downscaling models compared to the GCMs significantly increase the credibility of the downscaling models. The precipitation changes produced by the statistical models result from changes in all predictors, but the change in ξ is the greatest contributor in southern Sweden followed by q850 and u, while changes in q850 have greater effects in the northern parts of the country. The temporal and spatial variability of precipitation changes are higher in the statistically downscaled scenarios than in the dynamically downscaled ones. Comparisons of the 4 scenarios show that the spread of the scenarios created by the statistical model is on average larger than that between the RCM scenarios. The relatively large average spread is mainly due to the large differences found in summer. The seasonally averaged difference of the dynamical and statistical scenarios for the ECHAM4-based downscaled scenarios is 12%, and for the HadCM2 downscaled scenarios 21%. The differences in annual precipitation change are smaller, on average 4.5% among the HadCM2-based downscaled scenarios, and 6.9% among the ECHAM4-based downscaling scenarios.

KEY WORDS: Statistical downscaling · Dynamical downscaling · Precipitation · Sweden