Evaluation of precipitation product characteristics over Germany for hydrologic model forecasts

Abstract

As a primary component of the Earth’s hydrological cycle, precipitation plays a central role in many environmental processes and human activities. The availability of reliable precipitation data is essential for many sectors and applications, such as water resources management, flood and drought risk analysis, or hydrological modeling. In this study, we evaluate the characteristics of different precipitation datasets based on distinct methodologies and sources. This is in the context of high-resolution hindcasts and prototypical daily forecasts with the integrated hydrological model ParFlow over a central European model domain, where precipitation is a first order driver as part of the atmospheric forcing. Our objective is to determine, how closely precipitation from the ECMWF HRES numerical weather prediction matches in-situ observations, and how HRES compares to other precipitation products, some of which might be suitable for a bias adjustment of the hydrological model inputs. The European Climate Assessment & Dataset (ECA&D) in-situ daily precipitation observation dataset of 5072 stations in our ParFlow model domain serves as the reference. The time span of the comparison is from 2014 to 2022. Aside from ECMWF HRES, the evaluation includes at present data at different spatio-temporal resolutions: The ERA5 reanalysis as a background dataset, the HYRAS interpolated hydrometeorological raster data from the German Weather Service (DWD), the meteorological radar data product OPERA, a European composite dataset from EUMETNET, and the radar data product RADOLAN from DWD. Due to the spatial coverage of some datasets, the analysis is restricted to Germany constituting a subset of the hydrological model domain. The initial part of this evaluation uses only daily data, and precipitation products are compared at station locations. The spatial distribution and temporal variability is assessed with annual and seasonal sums, mean errors, and spatial correlation coefficients. Precipitation intensity is analyzed through the spatial distribution of the typical climate indices. The temporal characteristics of precipitation is determined through the precipitation fraction, i.e., the number of moderately wet days (75th percentile), very wet days (95th percentile), and consecutive wet days. Perkin's skill score is used for the comparison of the empirical distributions. While preliminary results indicate that HRES agrees well with the observational reference data, some form of bias adjustment may still be necessary.