Trace gas concentration retrieval from short-wave infrared nadir sounding spaceborne spectrometers

Kurzfassung

The remote sensing of short wave infrared (SWIR) radiation reflected from the Earth allows to infer atmospheric trace gas concentrations by solving the inverse problem. The retrieval algorithm BIRRA (Beer InfraRed Retrieval Algorithm) has been developed at the DLR (Deutsches Zentrum für Luft- und Raumfahrt) Remote Sensing Technology Institute (IMF) since around 2005 and is one of multiple algorithms to infer molecular concentrations from calibrated radiance spectra. BIRRA's forward model is based on the Generic Atmospheric Radiation Line-by-line Infrared Code (GARLIC) which has also been developed at the DLR-IMF. First, the BIRRA retrieved carbon monoxide (CO) columns from SCIAMACHY (SCanning Imaging Absorption spectroMeter for Atmospheric CHartographY) 2.3 micrometer observations from 2003-2011 were validated against eighteen stations from the ground-based networks TCCON (Total Carbon Column Observing Network) and NDACC (Network for the Detection of Atmospheric Composition Change). The BIRRA inferred CO concentrations were found to be circa 10% low biased which is in large agreement with other similar studies. Next, the latest updates from the radiative transfer code GARLIC were incorporated in BIRRA's forward model and the physical results of both, the old (but validated) and the latest (updated) BIRRA algorithms were verified and found to be numerically consistent for SCIAMACHY input data. Subsequently, the forward model was extended by upgrading its capabilities with respect to spectroscopy, i.e., enhanced line models were incorporated in order to utilize latest spectroscopic information from line lists such as the SEOM-IAS (Scientific Exploitation of Operational Missions - Improved Atmospheric Spectroscopy). More specifically, 'beyond Voigt' line profiles were implemented and the impact of the SEOM-IAS spectroscopy was studied with respect to latest compilations of HITRAN (HIgh-resolution TRANsmission molecular absorption database) and GEISA (Gestion et Etude des Informations Spectroscopiques Atmosphériques) for a large set of SCIAMACHY measurements. It was found that the SEOM-IAS line data and corresponding line models have significant impact on the spectral fitting: the residuals become smaller and the retrieved CO concentrations are also slightly different. The same methodology was then applied to study the spectroscopic impact on CO from S5P/TROPOMI measurements. The impact of the SEOM-IAS spectroscopy revealed to be even more pronounced, in particular with respect to the fitting residuals and smaller retrieval errors (higher precision) of the CO and co-retrieved parameters. Overall, the TROPOMI results are in agreement with that found for SCIAMACHY. A subsequent part of the thesis examines instrument spectral response functions (ISRF), in particular appropriate parameterizations for the TROPOMI's SWIR band responses. A first assessment with tabulated instrument profiles indicates that the parameterized variants can mimic the tabulated responses within circa 3-6%, depending on the instrument model and spectral position. The positive impact of the SEOM-IAS spectroscopy on the spectral fitting residuals could also be identified with the parameterized response functions. Moreover, the presented instrument profiles are considered promising candidates for the description of responses from upcoming sensors due to their flexibility. Finally, the co-retrieval of aerosol parameters in the CO fit is presented. Based on a simple model for the aerosol optical thickness the feasibility to co-retrieve aerosol extinction was investigated. In this context two different inverse solvers, namely the 'classical' nonlinear least squares and separable least squares, were examined with respect to convergence. First results show a stable CO retrieval for the separable least squares solver, however, the co-retrieved aerosol and reflectivity parameters indicate issues due to degeneracies. This thesis improved the retrieval of CO from SCIAMACHY observations. Moreover, the upgraded BIRRA algorithm successfully retrieved CO concentrations from cloud-free TROPOMI measurements. Many aspects investigated in this study are also relevant for the retrieval of other atmospheric constituents, such such CO2 or CH4. The study does hence provide a proven basis for further developments.