Towards a climatological background error covariance matrix for a regional, convective-scale NWP system
Additive covariance inflation (ACI) is an established method to account for inherent model error in Ensemble Kalman Filter methods, thus improve their performance. Random perturbations drawn from a climatological background error covariance B-matrix are added to the ensemble members in the data assimilation step. ACI is used operationally at Deutscher Wetterdienst (DWD) and MeteoSwiss as part of the KENDA assimilation system. In both systems, the climatological B-matrix from which perturbations are drawn was obtained from the global ICON model (DWD). Our objective is a new climatological B-matrix tailored to the convection-scale COSMO model on a limited area. This B-matrix should describe model error correlation structure and amplitude in terms of the variables geopotential, stream function, and velocity potential, as well as important cross-correlations.
We study error correlations from COSMO forecast differences (NMC method) as a proxy for model error correlations. Correlation patterns are compared with results obtained from the global IFS model to analyse differences between global and regional models. We furthermore show first results of error cross-correlations between the control variables geopotential, stream function and velocity potential obtained from COSMO. This study aims at investigating if balances accounted for in modelling global climatological B-matrices also hold on regional scales.
Dr. Claire Merker MeteoSwiss