Predictive skill of a subset of models participating in D-PHASE in the COPS region

The unique combination of observing and modelling efforts during the Convective and Orographically-induced Precipitation Study (COPS) and D-PHASE (Demonstration of the Probabilistic Hydrological and Atmospheric Simulation of flood Events in the alpine region) allows the systematic evaluation of models participating in D-PHASE with observations collected during COPS. This study verifies a subset of the D-PHASE models in the COPS region, including convection-permitting models as well as models with convection parametrization. Observations from the Joint D-PHASE/COPS (JDC) dataset are applied and processed with the Vienna Enhanced Resolution Analysis (VERA). The focus is set on monthly accumulated sums of the 12-hourly precipitation between 0600 and 1800 UTC and includes comparisons of horizontal distributions, averaged diurnal cycles and neighbourhood verification techniques. Two fuzzy verification scores were applied: The Fractions Skill Score (FSS) and Upscaling. The FSS analysis shows that all models are capable of forecasting the correct distribution of precipitation, especially for low precipitation thresholds. Low scores in the upscaling analysis reveal weaknesses in predicting the amounts of precipitation. The high-resolution models usually outperform their lower-resolution counterparts. This is true for the skill score analysis as well as the comparison of the horizontal distribution and the diurnal cycle averaged over the COPS period. Large differences between the models and the observations occur in regions with complex terrain, especially for models applying a convection scheme. Copyright (C) 2011 Royal Meteorological Society