An improved statistical-dynamical downscaling scheme and its application to the alpine precipitation climatology

Abstract

An improved statistical-dynamical downscaling method for the regionalization of large-scale climate analyses or simulations is introduced. The method is based on the disaggregation of a multi-year time-series of large-scale meteorological data into multi-day episodes of quasi-stationary circulation. The episodes are subsequently grouped into a defined number of classes. A regional model is used to simulate the evolution of weather during the most typical episode of each class. These simulations consider the effects of the regional topography. Finally, the regional model results are statistically weighted with the climatological frequencies of the respective circulation classes in order to provide regional climate patterns. The statistical-dynamical downscaling procedure is applied to large-scale analyses for a 12-year climate period 1981-1992. The performance of the new method is demonstrated for winter precipitation in the Alpine region. With the help of daily precipitation analyses it was possible to validate the results and to assess the different sources of errors. It appeared that the main error originates from the regional model, whereas the error of the procedure itself was relatively unimportant. This new statistical-dynamical downscaling method turned out to be an efficient alternative to the commonly used method of nesting a regional model continuously within a general circulation model (dynamical downscaling).