Improved NO2 column retrievals for geostationary satellites: application to GEMS and TEMPO

Satellite-based remote sensing has significantly advanced our understanding of global tropospheric nitrogen dioxide (NO₂) over recent decades. Complementing the daily global observations from low-earth-orbiting (LEO) satellites, new geostationary (GEO) missions offer high temporal resolution with multiple observations per day, enabling detailed monitoring of diurnal NO₂ variability driven by emission patterns and complex atmospheric chemistry. The emerging "Geo-Ring" constellation, comprising GEMS (Asia), TEMPO (North America), and Sentinel-4 (Europe), establishes a powerful framework for regional-to-continental air quality monitoring.

In this study, we address key challenges in NO₂ retrievals from GEO satellite observations by proposing two primary methodological improvements: (1) advanced NO₂ slant column retrievals, and (2) refined stratospheric correction techniques. An improved NO₂ retrieval algorithm is applied to GEMS and TEMPO data. First, we conduct round-robin tests of NO₂ slant column retrievals using three different approaches: classical Differential Optical Absorption Spectroscopy (DOAS), Covariance-based DOAS, and a hybrid method combining physics-based retrievals with machine learning. These advanced approaches specifically address issues related to de-striping and retrieval accuracy in inhomogeneous scenes, which is critical for GEO sensors that often lack continuous coverage of clean background regions. For the stratosphere-troposphere separation, challenges specific to GEO sensors, arising from restricted clean reference areas and pronounced diurnal variability, are investigated using two methods: an advanced reference sector approach and stratospheric NO₂ estimates derived from CAMS forecast model data.

The improved GEO NO₂ retrieval algorithm applied to GEMS and TEMPO observations is evaluated through comparisons with current operational product (GEMS v4.0.1 and TEMPO v4.0) as well as LEO satellite data from TROPOMI and GOME-2. The results demonstrate that the enhanced GEO retrieval algorithm effectively addresses challenges associated with high temporal sampling and the limited availability of clean background sectors, leading to improved retrieval accuracy and reduced uncertainties. These improvements strengthen the consistency between GEO and LEO NO₂ products and enhance the interpretation of pollution evolution and diurnal air quality variability.