A Fast Emission Inversion Scheme Based on Ensemble Optimal Interpolation

The emission inversion based on the ensemble Kalman filter (EnKF) is an effective method for estimating emissions and improving air quality modeling and forecasting. However, to construct the error covariance matrix between the emissions and pollutant concentrations, this method requires running the chemical transport model tens of times, which is computationally prohibitive and limits its application in updating the emissions for a real-time forecasting system. This study develops an emission inversion method based on the ensemble optimal interpolation (EnOI). The proposed method calculates the error covariance matrix from historical ensemble data and requires only a routine air quality simulation run for emission inversion from the contrasts between the observations and simulations, thereby greatly reducing the computational cost. The proposed method is applied to assimilating hourly surface observations of CO concentrations at 1107 sites over China in January 2015. During the experiment, CO emissions in January 2015 are estimated at a 15-km horizontal resolution using the historical ensemble dataset for January 2014. The total CO emission in China estimated by this scheme is only 1% higher than using an ensemble dataset for January 2015, indicating that the differences in meteorological conditions between the historical and estimated periods have a limited impact on the inversely estimated monthly CO emission. Simulations with the updated emissions reveal a decrease in the downward bias of average CO concentrations at 349 independent validation sites from 0.74 mg m⁻³ to 0.01 mg m⁻³ and a reduction of the root-mean-square error by 18%. The results suggest that the proposed method can be used as a fast emission updating scheme to lessen the uncertainties in the emission inventories.