Daniel Leuenberger, MeteoSwiss

Operational Convective-Scale Ensemble Data Assimilation at MeteoSwiss

Simon Foerster, Andre Walser, Marco Arpagaus


MeteoSwiss is currently introducing COSMO-E, a new convective-scale (2.2km grid size) ensemble prediction system based on the COSMO model. Initial condition perturbations for the ensemble forecasts are provided by the Kilometric ENsemble Data Assimilation system (KENDA) based on the Local Ensemble Transform Kalman Filter (LETKF). The new system has been tested in a 16-month trial period. The skill of the ensemble and deterministic analyses and forecasts started therefrom have been compared to that of the previous operational systems. These were based on the coarser resolution COSMO-LEPS downscaling ensemble and the deterministic COSMO-2 initialized by a nudging analysis. Near-surface spread in KENDA, particularly in temperature and humidity, has been found to be too small, leading to a suboptimal use of the observations. Two methods to account for model uncertainty in the LETKF system have proven to enhance the near-surface spread: the addition of soil moisture perturbations and stochastic perturbations of physical tendencies (SPPT) improve the spread-skill relation and reduce the errors of the KENDA system. Overall, the new ensemble prediction system COSMO-E has been found to clearly outperform the COSMO-LEPS system and to be similar to or slightly better than COSMO-2 in most forecast variables and seasons.

File available here