Mitigation of satellite OCO-2 CO₂ biases in the vicinity of clouds with 3D calculations using the Education and Research 3D Radiative Transfer Toolbox (EaR3T)

Abstract. Accurate and continuous measurements of atmospheric carbon dioxide (CO₂) are essential for climate change research and monitoring of emission reduction efforts. NASA's Orbiting Carbon Observatory (OCO-2/3) satellites have been deployed to measure the column-averaged CO₂ dry air mixing ratio (X_CO2) with very high precision. Although cloudy measurements are screened out, nearby clouds can still cause retrieval biases because the forward one-dimensional (1D) radiative transfer (RT) model used in the OCO retrieval algorithm does not account for the scattering induced by clouds in the vicinity of the OCO-2/3 footprints. These biases, referred to as the three-dimensional (3D) effects, can be quantified effectively using 3D-RT calculations, but these are computationally expensive, especially for hyperspectral applications (e.g., OCO-2/3). To reduce the prohibitive computational demands of 3D-RT radiance simulations across all three OCO spectral bands, this paper employs a linear approximation with two metrics (called slope and intercept) for each of the OCO bands that represent the 3D-RT perturbations on the OCO-2 spectra and accelerate the radiative transfer by a factor of 100. This is implemented by the Education and Research 3D Radiation Transfer Toolbox for OCO (EaR³T-OCO). EaR³T-OCO estimates OCO-2 satellite radiances using all available footprint-level data and imagery from the Aqua satellite, which orbits in close proximity to the OCO-2 satellite. EaR³T-OCO can calculate 3D-RT spectral perturbations for any OCO-2 footprint. These calculations can be used to spectrally adjust the OCO-2 radiance measurements with scene-dependent EaR³T-OCO perturbation calculations prior to the actual retrieval to undo cloud vicinity effects in the radiance spectra, which can subsequently be processed with the standard OCO-2 retrieval code. We find that this adjustment largely mitigates X_CO2 retrieval biases in proximity to clouds over land – the first physics-based correction of 3D-RT effects on OCO-2/3 retrievals. Although the accelerated 3D-RT radiance adjustment step is faster than full 3D-RT calculations for all OCO spectral bands, it still requires at least as much computational effort as the X_CO2 retrieval itself. To bypass 3D-RT altogether, the slope and intercept metrics are parameterized as a function of the weighted cloud distance of a footprint and several other scene parameters, all of which can be derived directly from Aqua-MODIS imagery. While this method is fastest and thus feasible for operational use, it requires careful validation for various surface and atmospheric conditions. For the case we analyzed, both the 3D-RT calculation method and the parametric bypass method successfully corrected X_CO2 biases, which exceeded 2 ppm at the footprint level, and reached up to 0.7 ppm in the regional average. We find that the biases depend most strongly on the cloud field morphology and surface reflectance, but also on secondary factors such as aerosol layers and sun-sensor geometry.