Validating precipitation forecasts using remote sensor synergy: A case study approach

Kurzfassung

Several types of remote sensing data are applied synergistically to evaluate the chain of microphysical processes leading to precipitation in a high-resolution numerical weather prediction model. The data provides information relating to (i) cloud-top temperature and optical depth (SEVERI), (ii) ice cloud amount (AMSUB), (iii) type and amount of precipitation particles (polarimetric radar), and (iv) surface precipitation (raingauge-calibrated radar data). Forecasts are produced by the COSMO-DE model of the German weather service, with a horizontal resolution of 2.8 km. The comparison with data is done in a model-to-observation framework, that is, forward operators are applied to the model output to produce synthetic data sets that can be directly compared to the observations. Additional diagnostics based on diurnal cycle and system tracking are also considered. Two case studies over Germany from the summer of 2006 are examined. The first case is dominated by widespread stratiform precipitation. Together the various data sets show that the model overestimates the amount of high cloud, while underestimating the concentration of ice scatterers and overestimating reflectivity and differential reflectivity (ZDR). This indicates errors in both the amount and the size distributions of cloud and precipitation particles in the model’s microphysical parameterization. In the second case a narrow band of convective precipitation is embedded in a cold front, with significant modulation by the diurnal cycle. The model fails to show a significant diurnal cycle in cloud amount, and the timing and duration of convective cells is incorrect. In this case, both the microphysical parameterization, and errors in the interaction of the simulated front with the orography of the Alps appear to contribute. These results demonstrate the potential of combinations of remote sensing data for model evaluation, although a long-term trial will be required to determine whether the errors seen in the case studies are characteristic for COSMO-DE.