[4-2]

Experience with Kilometer Scale Ensemble Data Assimilation for the COSMO Model (KENDA)

Hendrik Reich (DWD), Christoph Schraff (DWD), Andreas Rhodin (DWD), Annika Schomburg (DWD), Klaus Stephan (DWD), Africa Perianez (RIKEN), Roland Potthast (DWD)

An ensemble Kalman filter for kilometre-scale or convective scale data assimilation (KENDA) was developed for the COSMO model. The KENDA system includes a fully developed local ensemble transform Kalman filter (LETKF) including a deterministic analysis based on the Kalman gain for the analysis ensemble mean. The KENDA software suite includes tools for adaptive localization, multiplicative and additive covariance inflation, relaxation to prior perturbations, as well as adaptive observation errors. We assimilate conventional data (radiosonde, aircraft, wind profiler, surface station data). Also, latent heat nudging of radar precipitation has been integrated into the KENDA system to be applied to the deterministic analysis or to all ensemble members. The performance of the different system components was investigated using a basic cycling environment (BACY) for a period of 6 days with 24h forecasts. The deterministic KENDA forecasts were compared with forecasts based on the current operational nudging data assimilation scheme at the German Weather Service (Deutscher Wetterdienst, DWD), also for a more extensive 12-day period. For our experiments lateral boundary conditions for the regional model are given by the pre-operational global ensemble Kalman filter for the ICON model. The performance of the KENDA system proves to be overall superior to the forecast quality of the operational nudging scheme.