Errors in Retrieved Gases and Inferred Fluxes Arising from Non-uniform Scene Illumination: A Case Study for the GeoCarb Mission

Abstract:

Previous work (e.g. Hu et al., 2016, and Landgraf et al., 2016, for the TROPOMI mission) has shown that inhomogeneous illumination within the field of view (FOV) can introduce retrieval bias due to distortions in the actual instrument spectral response function (ISRF) relative to the ISRF that is measured in pre-flight calibration and characterization. Laboratory studies have shown that this distortion grows with wavelength. Various techniques have been employed to ameliorate these issues, including continuously scanning in the across-slit direction, spatial and temporal oversampling, and spatial defocusing of the telescope. More recently, hardware solutions have been developed, referred to as slit homogenizers.

In this work, we demonstrate the impacts of the scene brightness inhomogeneity on retrievals of XCO₂, XCH₄ and XCO using the case study of the GeoCarb instrument, which measures hyperspectral radiances in the 0.76um, 1.6um, 2.05um, and 2.3um spectral regions. We simulate sub-scene spectra and perform retrievals using an assumed instrument line shape function, such as would be measured in pre-flight calibration, for three scenarios: a “control case” in which there is no subslit variation, a second case in which the scene inhomogeneity is included, and a third case in which a slit homogenizer is added to mitigate the nonuniformity effects. Without a slit homogenizer, we find that the posterior scatter and bias are significantly larger than in the case of uniform scene illumination, and the number of good quality XCO₂ and XCH₄ soundings is reduced by ~30%, in the presence of realistic scene inhomogeneity. Through rigorous quality filtering and bias correction, we are able to achieve a mean XCO₂ error scatter of 1.4ppm in the presence of scene inhomogeneity, which does not meet the accuracy requirements for GeoCarb. The introduction of a slit homogenizer dramatically reduces these effects, as well as the dependence on the bias correction, with a mean error scatter of 1.1ppm, which is within the GeoCarb requirements. Fitting for an ISRF correction factor reduces the errors even further, such that when paired with a slit homogenizer, the effects of inhomogeneous illumination within our simulations become almost negligible.