Abstract
The knowledge of uncertainties resulting from simulations of the hydrological cycle in meteorological models is crucial for the interpretation of model results. In order to gain confidence in statements about future changes, it is necessary to assess the model skill in the representation of the hydrological cycle. This study focuses on the evaluation of the atmospheric part of the hydrological cycle in two mesoscale meteorological models, MM5 and COSMO-CLM (CCLM). By using high resolution observations from the General Observation Period (GOP) performed in the German Priority Program on Quantitative Precipitation Forecasting, the representation of integrated water vapour, total cloud cover and precipitation is evaluated. Model runs were performed for the period of 2007 and 2008 within the model domain covering Germany with a spatial resolution of about 18 km. Both models are forced by reanalysis data of the National Centers for Environmental Prediction (NCEP1). The performance of the models is evaluated concerning their annual cycles, space-time structures and diurnal cycles of the model simulations. Error structures of the three considered key variables are very different: Concerning integrated water vapour, errors are mainly due to the large scale forcing in both models. MM5 exhibits a systematic wet bias. In contrast, errors in predicted total cloud cover are dominated by shortcomings of the parametrizations concerning convection and clouds. Precipitation errors are influenced by the orography and depend on the convection parametrization. Interestingly, the wet bias in integrated water vapour of MM5 does not result in a positive precipitation bias.
DOI Link to Publication