Correcting 3D cloud effects in X_CO₂ retrievals from the Orbiting Carbon Observatory-2 (OCO-2)

Abstract

The Orbiting Carbon Observatory-2 (OCO-2) makes space-based radiance measurements in the oxygen A band and the weak and strong carbon dioxide (CO2) bands. Using a physics-based retrieval algorithm these measurements are inverted to column-averaged atmospheric CO2 dry-air mole fractions (XCO2). However, the retrieved XCO2 values are biased due to calibration issues and mismatches between the physics-based retrieval radiances and observed radiances. Using multiple linear regression, the biases are empirically mitigated. However, a recent analysis revealed remaining biases in the proximity of clouds caused by 3D cloud radiative effects (Massie et al., 2021) in the processing version B10. Using an interpretable non-linear machine learning approach, we develop a bias correction model to address these 3D cloud biases. The model is able to reduce unphysical variability over land and sea by 20 % and 40 %, respectively. Additionally, the 3D cloud bias-corrected XCO2 values show agreement with independent ground-based observations from the Total Carbon Column Observation Network (TCCON). Overall, we find that the published OCO-2 data record underestimates XCO2 over land by −0.3 ppm in the tropics and northward of 45∘ N. The approach can be expanded to a more general bias correction and is generalizable to other greenhouse gas experiments, such as GeoCarb, GOSAT-3, and CO2M.