Evaluation of daily precipitation characteristics in the CLM and their sensitivity to parameterizations

This paper presents a detailed evaluation study of the regional climate model CLM (climate version of the "Lokal-Modell"), driven by ERA-40 re-analysis data, with respect to summer precipitation characteristics over Germany. Since extremes might become more important in a changing climate, characteristics other than mean values need to be analyzed in more detail. Two aspects are investigated, namely the capability of the CLM to simulate different statistical characteristics of daily precipitation and the sensitivity of these characteristics to changes in the physical parameterizations. Our evaluation is based on 132 model runs, encompassing I I summers with 12 different model configurations each, and relates to observations of a dense station network. In addition to its frequency distribution, different characteristics of daily precipitation, including extremes, are assessed by means of several indices. We analyze these characteristics in four different regions of Germany with a size of about 220x220 km(2), by assessing the model skill with respect to both the spatial average and the spatial structure within these regions. The range of the model uncertainty due to modified parameterizations is derived for all indices and regions. Our results indicate that: (i) the CLM underestimates both observed total precipitation and wet-day frequency, although the wet-day precipitation intensity is simulated quite well, (ii) the model deviations from observations of the indices describing extreme events are comparable to those of the total precipitation amount, (iii) the frequency distribution is reasonably well simulated by the CLM, (iv) the CLM is unable to adequately reproduce the observed spatial structure of precipitation indices within the selected regions, (v) in the southernmost region which is characterized by complex terrain the CLM performance is generally low, and (vi) the uncertainty due to the modified physical parameterizations is considerable. All in all, the model performance differs for different characteristics, regions or model configurations, which leads to the conclusion that application-specific evaluation studies are indispensable prior to using the model results from climate change projections.