Tracking daily NO _<i>x</i> emissions from an urban agglomeration based on TROPOMI NO ₂ and a local ensemble transform Kalman filter

Abstract. Accurate, timely, and high-resolution NOx emissions are essential for formulating pollution control strategies and improving the accuracy of air quality modeling at fine scales. Since late 2018, the Tropospheric Monitoring Instrument (TROPOMI) aboard the Sentinel-5 Precursor (S5P) satellite has been providing daily monitoring of NO2 column concentrations with global coverage and a small footprint of 5.5 km × 3.5 km, offering great potential for tracking daily high-resolution NOx emissions. In this study, we develop a data assimilation and emission inversion framework that couples an ensemble Kalman filter with the Community Multiscale Air Quality (CMAQ) model to estimate daily NOx emissions at 3 km scales in Beijing and surrounding areas in 2020. By assimilating the TROPOMI NO2 tropospheric vertical column densities (TVCDs) and taking the bottom-up inventory as prior emissions, we produce a posterior NOx emission dataset with a reasonable spatial distribution and daily variations at the 3 km scale. The proxy-based bottom-up emission mapping method at fine scales overestimates NOx emissions in densely populated urban areas, whereas our posterior emissions improve this mapping by reducing the overestimation of urban emissions and increasing emissions in rural areas. The posterior NOx emissions show considerable seasonal variations and provide a more timely insight into NOx emission fluctuations, such as those caused by the COVID-19 lockdown measures. Evaluations using the TROPOMI NO2 column retrievals and ground-based observations demonstrate that the posterior emissions substantially improve the accuracy of 3 km CMAQ simulations of the NO2 TVCDs as well as the daily surface NO2 and O3 concentrations in 2020. However, during summer, despite notable improvements in surface NO2 and O3 simulations, positive biases in the posterior model simulations persist, indicating weaker constraints on surface emissions from satellite NO2 column retrievals in summer. The posterior daily emissions on the 3 km scale estimated by our inversion system not only provide insights into the fine-scale emission dynamic patterns but also improve air quality modeling on the kilometer scale.