The Central Europe Refined analysis version 2 (CER v2): evaluating three decades of high-resolution precipitation data for the Berlin-Brandenburg metropolitan region

The Central Europe Refined Analysis (CER) was developed in 2016 as a high-resolution, reanalysis-based, gridded data set for Central Europe and the Berlin-Brandenburg metropolitan region of Germany in particular. The data set was successfully used for investigations of near-surface air temperatures, but showed inaccuracies in the simulated precipitation compared to station measurements. In this study we characterize the development of the second version of this data set (CER v2), which focused primarily on improving the performance of precipitation products. This new version uses an updated version of the WRF model and new ERA5 forcing data. Comprehensive sensitivity studies were carried out to optimize the physical parameterization of daily precipitation results. The combination of the Kain-Fritsch cumulus and the Thompson microphysics scheme was selected for the CER v2 due to the reduction of the domain average Mean Deviation (MD) by 77% and the Root Mean Squared Deviation (RMSD) by 18% when compared to the original CER setup. The validation of 30 years (1991–2020) of the CER v2 precipitation data against station data by the German Meteorological Service (DWD) revealed that the domain median RMSD was the lowest during the winter with seasonal median RMSD of 0.24 mm d-1 and the highest during the summer with 0.71 mm d-1. The comparison against 20 years of radar data (2001–2020) identified the highest seasonal RMSD during the summer along the western and southern border of the model domain and in the northeast of Berlin with values above 1 mm d-1. CER v2 data was compared to the CER v1 and ERA5 data for the time period of 2001–2018 on a resampled 0.25 ° grid. In terms of the domain median RMSD and MD, the CER v2 outperformed the CER v1 and the ERA5 during the winter, spring and autumn. However during summer, the domain median CER v2 RMSD was 46% higher than for ERA5. One of the biggest advantages of the data set is the substantial reduction in the domain median annual MD, which was about 94% lower than for the ERA5 forcing data. Due to its longer available time series and increased performance compared to the previous version, the CER v2 could provide important insights about the local- to mesoscale precipitation dynamics of the study region and serve as a foundation for data-driven hydrological models.