Sensitivity of PM Assimilation Results to Key Parameters in the Ensemble Kalman Filter

Abstract

To study and forecast atmospheric tracer concentrations at ground level, an assimilation system is available around the LOTOS-EUROS model based on the Ensemble Kalman filter technique. For applications focusing on air-quality related to aerosols, the available observation data is usually limited to ground based observations of total PM2.5 or PM10, and model uncertainty is specified for the emissions. In this study, the key parameters of the assimilation system have been varied: the assumed temporal variation in the emission uncertainty, the amplitude of the representation error, the localization length of the analysis, the averaging period of the observations, and the number of ensemble members in the filter. Although in theory these parameters are all important, the most important parameters are those related to the representation error between simulations and observations.