Evaluation of the OCO-2 and OCO-3 ACOS data products against TCCON

The Orbiting Carbon Observatory-2 (OCO-2), launched in 2014, is NASA’s first satellite dedicated to measure sources and sinks of carbon dioxide (CO₂) in Earth’s atmosphere on regional scales. Since 2019, measurements from the Orbiting Carbon Observatory-3 (OCO-3) have complemented OCO-2’s data record. In addition, OCO-3’s Snapshot Area Mapping (SAM) mode observations over emission hotspots like cities, power plants, and volcanoes provide a novel data set for carbon cycle studies on local scales. Data from both instruments are analyzed with the Atmospheric Carbon Observations from Space (ACOS) retrieval algorithm to estimate column-average dry-air mole fractions of carbon dioxide (XCO₂) in Earth’s atmosphere. Evaluating these space-based estimates of XCO₂ against independent validation data sets provides information about the quality, potential biases, and errors in the OCO-2/3 data record. Here, we present comparisons of the ACOS V10 XCO₂ from OCO-3 and the new and improved ACOS V11 XCO2 from OCO-2 against ground-based measurements from the Total Carbon Column Observing Network (TCCON).

For both instruments, the root-mean-square error (RMSE) is below 1 ppm for all observational modes when compared to collocated TCCON observations. The OCO-3 V10.4 data version, an improvement over the initial vEarly data version, reduces an XCO₂ time-dependent bias that was present in the previous OCO-3 data record. Consequently, data from both instruments does not indicate any significant time-dependent bias over the span of several years. Further, we evaluate differences between OCO-3 and TCCON related to different local overpass times. On average, OCO-3’s equator crossing time occurs about 20 minutes earlier every day. Comparisons against TCCON indicate no significant local time of day bias in the OCO-3 XCO₂ data, however, comparisons over individual TCCON sites indicate a dependence which provides insight into potential airmass and viewing geometry related biases. The improved OCO-2 V11 data version reduces the mean bias against TCCON to ~0.15ppm from previously ~0.4ppm in V10. We find the largest reduction in RMSE over ocean due to an improved ocean glint surface treatment in V11. Both OCO data products are of comparable quality and are an improvement over earlier OCO data versions. Finally, we analyze how well OCO-2 captures the mean seasonal cycle amplitudes and growth rates over selected TCCON sites.