Utilizing machine learning for high resolution NO₂ total columns from PACE OCI

Abstract Satellite-based observations of nitrogen dioxide (NO 2 ) are important for monitoring atmospheric composition, estimating nitrogen oxide emissions, and informing chemistry transport models. While advancements have been made in space-based NO 2 observations and retrievals, most measurements are still unable to resolve the detailed structure of NO 2 plumes. Designed primarily for aerosol and ocean applications, the plankton, aerosol, cloud, ocean ecosystem ocean color instrument (OCI) provides a unique opportunity to retrieve NO 2 from high spatial resolution (∼1 km) 2 hyper-spectral measurements. We exploit a machine learning technique to show that OCI, with a spectral resolution of 5 nm, can provide high spatial resolution information about NO 2 when trained with high quality retrievals from the Tropospheric Monitoring Instrument (TROPOMI). This work demonstrates the potential to rapidly produce high spatial resolution NO 2 columns by making use of well validated retrievals derived from instruments with higher spectral resolution. These data can potentially enable emissions estimates with reduced uncertainties and higher spatial resolution. Additionally, the data could provide higher resolution information for exposure estimates used in epidemiological studies.