The proof of concept for profiles retrievals from CrIS radiances using a deterministic method

Different aspects of the remote sensing inverse problem in operational environment are thoroughly discussed and various ambiguities in prevalent stochastic method namely optimal estimation method (OEM) are pointed out. To overcome these obscurities, a deterministic method namely regularized total least squares (RTLS) is proposed. For demonstration purposes, water vapor profiles retrievals from simulated Suomi-NPP CrIS hyperspectral measurements are considered. Synthetic CrIS radiances are generated using a line-by-line radiative transfer model (GENSPECT) with realistic radio-sonde profiles and US standard atmosphere as inputs. Our findings show that the current stochastic method, even with additional deterministic constraints (truncated singular value decomposition) applied on top of OEM, is unable to produce useful retrieval results, i.e., posterior error is more than the a priori error. In contrast, RTLS is able to produce deterministically unique results commensurate with the available information content in the measurements, which could result in a paradigm shift in operational satellite inversion.