The verification of precipitation forecasts from numerical models on short time-scales over Germany with pattern-oriented error measures

点击翻译按钮获取中文摘要

The verification of precipitation forecasts from numerical models is essential for the future improvement of quantitative precipitation forecasts (QPF). There are various options when defining a QPF verification strategy, concerning for instance the temporal resolution, the choice of the evaluation data set, the verification domains, the weather-type stratification and the applied error measures. In this project new techniques will be developed to verify QPFs from the weather prediction models LM, GME and ECMWF over Germany on very short time scales (typically 1 hour). A currently developed technique is used to derive an evaluation data set of hourly precipitation analyses from rain-gouge measurements and radar composites. The verifications will be done for the catchment areas of the major river systems and the QPFs will be categorized into different synoptic weather patterns. Finally, new error measures will be implemented that consider the spatial patterns of precipitation, in contrast to more traditional measures that evaluate the similarity of two fields on a point-by-point basis. Such an pattern-oriented approach provides novel and meaningful information about the qualities and deficiencies of a particular QPF. The results will be directly useful for the interpretation of model output in operational weather forecasting and for their quantitative use in impact applications such as runoff forecasting.