A novel convective-scale regional reanalysis COSMO-REA2: Improving the representation of precipitation
Atmospheric reanalyses are a state-of-the-art tool to generate consistent and realistic state estimates of the atmospheric system. They provide a synthesis of various heterogeneous observational systems and model simulations using a physical model together with a data assimilation scheme. Current reanalyses are mainly global, while regional reanalyses are emerging for North America, the polar region, and most recently for Europe. However, deep convection is still parameterized even in the regional reanalyses. A novel convective-scale regional reanalysis system for Central Europe (COSMO-REA2) has been developed by the Hans-Ertel Center for Weather Research – Climate Monitoring Branch. The system is based on the COSMO model and uses observational nudging for regional data assimilation. In addition to conventional observations, radar-derived rain rates are assimilated using latent heat nudging. With a horizontal grid-spacing of 2 km, the model runs without parameterization of deep moist convection. COSMO-REA2 produces horizontal wind fields that represent a realistic energy spectrum for horizontal scales above 14 km. COSMO-REA2 is currently available for seven years from 2007 to 2013.This study illustrates the improved representation of local precipitation over Germany by the convective-scale reanalysis COSMO-REA2 compared to coarser gridded European and global reanalyses. A systematic verification using rain gauge data reveals the added value of high-resolution regional atmospheric reanalyses on different time scales. On monthly to annual time scales, regional reanalyses yield better estimates of the spatial variability of precipitation patterns which can not be provided by coarser gridded global models. On hourly to daily time scales, the convective-scale reanalysis substantially improves the representation of local precipitation in two ways. On the one hand, COSMO-REA2 shows an enhanced representation of observed frequencies of local precipitation, especially for high precipitation events, which is mainly due to the high spatial resolution. On the other hand, the assimilation of radar data largely improves the spatial and temporal coherence between model output and local observations. Both lead to reanalyzed precipitation values which are comparable to point observations.
No Supplementary Data
No Article Media
Document Type: Research Article
Publication date: January 1, 2017