Estimation of Power Plant Emissions With Unscented Kalman Filter

Abstract

Emissions from power plants constitute a major part of air pollution and should be adequately estimated. In this paper, we consider the problem of estimating nitrogen dioxide (NO $_X$ ) emission of power plants by developing an inverse method to integrate satellite observations of atmospheric pollutant column concentrations with species concentrations and direct sensitivities predicted by a regional air quality model, in order to discern biases in the emissions of the pollutant precursors. Using this method, the emission fields are analyzed using a “bottom-up” approach, with an inversion performed by an unscented Kalman filter (UKF) to improve estimation profiles from emissions inventories data for the Sydney metropolitan area. The idea is to integrate information from the original inventories with tropospheric nitrogen dioxide (NO $_2$ ) emissions estimated during one month from the air pollution model-chemical transport model, and then, for validation, to compare the resulting model with satellite retrievals from the ozone monitoring instrument (OMI) above the region. The UKF-based estimation of NO $_2$ emissions shows better agreement with OMI observations, implying a significant improvement in accuracy as compared with the original inventories. Therefore, the proposed method is a promising tool for estimation of air emissions in urban areas.