CO<sub>2</sub> column-retrieval errors arising from neglecting polarization in forward modeling of 1.6 μm band measurements

Abstract

Carbon dioxide (CO2) is a major anthropogenic greenhouse gas. Atmospheric CO2 has increased from a preindustrial value of about 280 ppmv to more than 400 ppmv, mainly due to the fossil-fuel combustion and changes in land use. Understanding the CO2 distribution and variability of sources and sinks is essential for predicting changes in future CO2 concentration. However, in-situ measurements collected at a network of surface stations are sparse to allow the CO2 sources and sinks to be inferred. Satellite remote sensing is one of the best ways to obtain such global observations. The radiances reflected or transmitted by the atmosphere contain information about the atmospheric gases through their absorption signatures. High spectral radiance in the near-infrared 1.6 μm band, observed by OCO and Chinese TanSat, can be used to derive the CO2 column densities in the atmosphere. Surface reflection and atmospheric scattering processes act to reduce the intensity of radiation polarized in the direction parallel to the principal plane, as defined by the incoming solar beam and the beam entering the instrument. Both the OCO and TanSat instruments are typical polarizing instruments that are designed to measure only the radiation perpendicular to the principal plane; this will create a disparity between the simulation and measurements if polarization effects are neglected in forward modeling. In this paper, the US standard atmospheric profile with multiple layer aerosol loading is used to evaluate the forward-modeling error made by scalar approximations in the near-infrared 1.6 μm band. Numerical computations were performed using the line-by-line and doubling-adding methods. The atmosphere is assumed to be plane parallel and the surface is set to offer Lambert reflection. The polarizations introduced by molecular Rayleigh scattering and aerosol Mie scattering are calculated respectively. Solar zenith angle, view zenith angle, surface-reflectance, and aerosol optical depth are varied one at a time to evaluate the individual errors. The results show that, except for some situations with very large observation angles, the radiation error made by the scalar approximation increases with the solar zenith angle, increase in aerosol optical depth, and decrease in the surface reflectance. The errors in the line cores are larger than those in the window area because they are affected less by the surface depolarization effect in the absorption channels. For both OCO and TanSat, the conventional nonlinear least-squares-fitting technique, by which the forward model calculation is fit to satellite observations, is used to obtain the CO2 column information. The atmospheric CO2 information comes mainly from the radiation ratio between the absorption and window channels, which means that the radiance-simulation error will be translated into errors in CO2 column retrieval. To assess these retrieval errors while ignoring the polarization effect in forward mode calculation, sensitivity studies are performed on simulated spectroscopic measurements in the 1.6 μm band. The CO2-retrieval errors arising from neglecting the polarization effect are seen to be largest when the solar zenith angle is high, the aerosol optical depth is large, and the surface reflectance is low. When the solar zenith angle is larger than 45°, the aerosol optical depth is higher than 0.3 and the surface albedo is less than 0.2; this carbon-dioxide retrieval error could be larger than 1–2 ppmv. Sometimes, this error may be as high as 10 ppmv, making it very large than what is required. The retrieval error budget could thus potentially be dominated by polarization. Thus, to reduce such errors, the forward model needs to take account of fully polarized radiation calculations. On the other hand, it is impractical to do full vector retrieval due to the computational cost. We need to devise ways to quickly complete the polarization radiative-transfer calculation.