LETKF for the nonhydrostatic regional model COSMO-DE

摘要

@@ 1 Introduction Data assimilation for numerical weather prediction (NWP) at the convective scale meets with a number of challenges. They include: strongly flow dependent and unknown spatial balances between the different model variables, importance of nonlinear processes, non-Gaussian error statistics and large forecast errors in 'weather'parameters due to imperfections in the physics, in particular in the cloud and boundary layer formulations. The Local Ensemble Transform Kalman Filter (LETKF) (Hunt et al., 2007) offers some very attractive features: it is a simple algorithm, no tangent linear and adjoint versions of the prognostic model are required, and the forecast error covariance matrix is cycled and thus flow-dependent. At Deutscher Wetterdienst (DWD) it is planned to use the LETKF on the global scale (in a hybrid approach together with 3dVar) as well as on the local scale. The LETKF analysis ensemble will also serve as initial conditions for COSMO-DE EPS, a convection permitting EPS system under development at DWD. The outline is as follows: we will give a short overview on the LETKF and the NWP model COSMO-DE in sections 2 and 3. In section 4 we present the results of our LETKF experiments and we conclude in section 5.