Statistical precipitation downscaling in central Sweden with the analogue method

Abstract

Most climate predictions show significant consequences globally and regionally, but many of its critical impacts will occur at sub-regional and local scales. Downscaling methods are, thus, needed to assess effects of large-scale atmospheric circulation on local parameters such as precipitation and runoff. This study aims at evaluating the analogue method (AM) as a benchmark method for precipitation downscaling in northern Europe. The predictors used in this study were daily and monthly gridded sea-level pressures from 1960 to 1997 in an area 45–75°N and 30°W–40°E with a resolution of 5×5° long-lat. Analogues for daily and monthly precipitation at seven precipitation stations in south-central Sweden were established with two techniques, principal-component analysis (PCA) and the Teweles-Wobus score (TWS). The results showed that AM downscaling on both daily and monthly basis was commonly generally much better than a random baseline but depended on the objective function used for assessment; PCA and TWS produced similar results in most cases but TWS was superior in simulating precipitation duration and intensity. Downscaling was improved when seasonality was included and when the SLP field was confined to those geographical areas that contributed most to precipitation in south-central Sweden.