An ensemble Kalman filter for short‐term forecasting of tropospheric ozone concentrations

An air‐quality forecasting system based on the pair ‘NWP model MM5–chemistry transport model CAMx’ is proposed. A version of the ensemble Kalman Filter has been developed. The model‐error covariance matrix is parametrized with the help of a covariance function and represented by an ensemble formed as a random selection from leading eigenvectors. The performance of the system is tested on the case of an ozone episode in June 2001. As a source of observations, the AirBase database has been used. Starting the forecast from analysed concentration fields improves the quality of forecast of the next day's ozone concentration maxima.