Data assimilation is the term given to the process which provides the initial atmospheric conditions, also known as the “analysis”, which is used as the starting point for a model forecast. For this purpose, previous model condition forecasts and real-time observations of temperature, pressure, humidity, wind and precipitation that are distributed irregularly throughout the forecasting region, are combined in such a way as to arrive at a best estimate of the current atmospheric conditions on the model grid.
In the case of ensemble systems such as COSMO-1E and COSMO-2E, a host of slightly differing initial conditions is required. This analysis ensemble is calculated at MeteoSwiss with the help of the KENDA (kilometer-scale ensemble data assimilation) system. For high-resolution, convective-scale ensemble forecasting, the analysis ensemble on the COSMO-1E grid with 1.1-km grid spacing and 80 layers is updated with new measurement data in an assimilation cycle on an hourly basis. The method employed - the Local Ensemble Transform Kalman Filter (LETKF), evaluates observation data and model conditions together with the relevant uncertainties to arrive at the best possible estimate of the current atmospheric conditions. LETKF also provides a multitude of slightly differing analyses that represent the uncertainty of the analysis. Using this analysis ensemble as a basis, the COSMO-1E forecast is generated once every three hours (00:00, 03:00, ... 21:00 UTC). To obtain the initial atmospheric conditions for the COSMO-2E forecasts, the analysis ensemble is interpolated every six hours (00:00, 06:00, 12:00 and 18:00 UTC) by upscaling from the higher-resolution COSMO-1E grid to the COSMO-2E grid.