Verification of precipitation from regional climate simulations and remote-sensing observations with respect to ground-based observations in the upper Danube catchment

Abstract

An evaluation of precipitation fields for four selected months simulated by the regional climate model AtmoMM5 and provided by the satellite retrieval method AtmoSat is presented. As reference, observations at 5 km resolution on a daily and monthly basis are used. We applied conventional verification tools (root mean square error, grid-point based categorical error scores, etc.) as well as the new error score SAL, which separately considers aspects of the structure, amplitude and location of the precipitation field in a predefined area. We also discussed the advantages and disadvantages of each of the scores. The aim of our evaluation was to unfold the strengths and weaknesses of AtmoMM5 and AtmoSat to calculate daily and monthly high resolution precipitation. As a result we found that the catchment averaged monthly mean precipitation is simulated with an acceptable accuracy by both methods. The spatial pattern of the monthly precipitation (typically with a precipitation maximum in the alpine foreland) can only be reproduced by AtmoMM5. Regarding the daily precipitation, our evaluation revealed that both methods still need improvement. The deviations to the observations increase with decreasing precipitation amount resulting in large uncertainties in case of very dry conditions. Overall, we can conclude that AtmoMM5 is better suited to simulate precipitation at 5 km resolution on a daily basis than AtmoSat.