Timely Updates of Anthropogenic Emission With Satellite and Ground Observations: Comparison of Multiple Top‐Down Methods Against Bottom‐Up National Emission Inventories

Abstract Due to the COVID‐19 pandemic and changes in technology and energy consumption, recent years saw substantial variations in air pollution emissions, highlighting the importance of timely emission updates. In the United States, the level of surface ozone in many regions is close to the National Ambient Air Quality Standard, making it important to understand even small changes in emissions. Subsequently, model simulations of ozone are less tolerant of uncertainties in emission estimation, such as those caused by outdated emission updates. Here, we propose and assess four emission refresh methods: state‐level emission adjustment (ADJ), ADJ with budget‐aware urban adjustment (BA), ADJ with TROPOMI‐derived city emissions (SatNOx), and an approach with a machine learning product generated by Variational AutoEncorder (VAE). The observed temporal trends from ground measurements Air Quality System (AQS) and TROPOMI products are employed to provide updates of anthropogenic nitrogen oxide (NO x ) emissions from a base year (2019) to the targeted model period (2020). The adjusted emissions effectively reflect the decrease in summer from 2019 to 2020. Comparisons of the ozone forecasts using the base and refreshed emission inputs show that the emission adjustments have reduced model mean biases by 19%–21% and enhanced the model's capability to forecast O 3 surface concentrations. The nitrogen dioxide (NO 2 ) predicted with updated emissions shows more underestimation at the surface and aloft due to lower baseline emissions. This work demonstrates the potential of using near‐real‐time ground and satellite observations to rapidly update emissions to reduce a key source of emission uncertainty in chemical transport modeling.