The 8th EnKF Data Assimilation Workshop

Additive Covariance Inflation in an operational, convective-scale NWP Ensemble Kalman Filter Assimilation System

Daniel Leuenberger; Claire Merker
MeteoSwiss, Switzerland

Poster: 20180305_led_KENDA_ACI_EnkFWorkshop.pdf

Ensemble Kalman Filter data assimilation for Numerical Weather Prediction (NWP) applications typically needs some sort of covariance inflation to account for model errors and to ensure that the observations are given enough weight in the update step. The MeteoSwiss operational, convective-scale COSMO-LETKF-based ensemble data assimilation system uses adaptive multiplicative covariance inflation, relaxation to prior perturbations (RTPP) and perturbations of soil moisture in the update step and stochastic perturbations of physical tendencies (SPPT) in the forecast step of the analysis cycle. Spread-skill inspection has shown that especially in winter periods, the ensemble is still under-dispersive, i.e. resulting in too little spread and not enough weight given to the observations. Therefore, an additional covariance inflation based on the climatological B-matrix of the global ICON model of Deutscher Wetterdienst has been tested in assimilation experiments.
In this contribution we show results of these experiments and compare analyses and forecasts with and without additive inflation. We further investigate the structure of these climatological perturbations and compare them with flow-dependent variances and covariances of the data assimilation ensemble.