Comparison of model-derived carbon dioxide datasets with the Orbiting Carbon Observatory 3 (OCO-3) observations

Abstract

Multiple satellites are currently in orbit around the Earth, providing reliable and consistent estimates of the column-averaged dry-air mole fraction of CO₂, i.e., XCO₂. However, the satellite datasets suffer spatiotemporal gaps due to the narrow swath widths and the influence of clouds and aerosols. Model-derived CO₂ datasets can fill these gaps; however, these model datasets lack the consistency of the satellite observations. Scientists are continually refining their methods to improve the accuracy of the model datasets. Therefore, regular evaluation of the model estimates against precise datasets is imperative to confirm their reliability. In our study, we extensively evaluated the performance of two widely used models, the National Oceanic and Atmospheric Administration (NOAA) CarbonTracker and the Copernicus Atmosphere Monitoring Service (CAMS), by comparing their CO₂ datasets with the Orbiting Carbon Observatory 3 (OCO-3) XCO₂ retrievals utilizing three years of data from 2020 to 2022. The results showed that overall, the CarbonTracker dataset was underestimated by -0.08 ± 0.38 ppm with an RMSE of 0.98 ppm, and the CAMS estimates were overestimated by 0.34 ± 0.43 ppm with an RMSE of 1.05 ppm. For a more detailed assessment, we compared the model and the satellite datasets separately over 10 regions of the world in terms of spatial distribution, monthly changes, seasonal variations, latitudinal distribution, and annual XCO₂ growth rates. The model datasets exhibited good consistency with the satellite observations in most regions. However, significant discrepancies were observed in areas such as the Tibetan Plateau, the Himalayan Mountain ranges, equatorial Africa, and the Amazon.