Efficient parameterization for radiative transfer calculations in the infrared spectrum for information extraction of trace gases

Kurzfassung

The radiance spectrum can be simulated by solving the radiative transfer equation (RTE), while the retrieval of CO2 and CH4 parameters is achieved through an inverse model that uses the RTE as the forward model. However, numerically solving the RTE is computationally intensive due to the inclusion of the multiple-scattering term. To address this, using approximate radiative transfer models offers a significant advantage for retrieval, as they can be solved much more quickly. For instance, FALCAS, a fast Python-based solar radiation measurement model currently being developed at DLR, accelerates computations by approximating the multiplescattering term. It does this by considering the radiative transfer process within an optically thin, isotropic scattering layer, that mimics the influence of multiple scattering. In this way, the computing time can be significantly reduced. In this thesis, an algorithm for retrieving CO2 and CH4 concentrations is designed based on FALCAS. The accuracy of CO2 and CH4 parameter retrievals in the wavelength range of 1500 nm to 1750 nm is investigated using the radiative transfer model PYDOME as the reference model, wherein the RTE is solved using the discrete ordinate method with matrix exponential. Ten retrieval scenarios for different geometries and aerosol optical thickness are considered. For finding the objective function minimum, the BFGS algorithm is used. In the absence of aerosols, convolution, and noise, the retrievals of the concentrations for both CO2 and CH4 using FALCAS are closely matching the true values. The final retrievals of CO2 and CH4 concentrations are 1.5% and 4.5% more accurate, respectively, than those obtained using the Beer model, a simplified approach that accounts only for absorption and ignores scattering. As aerosol concentration increases, both CO2 and CH4 retrieved concentrations rise by up to 2.0%. When aerosol and convolution conditions are varied, the CO2 and CH4 retrievals remain nearly identical to those obtained without convolution. However, when aerosol, convolution, and noise are all considered, the retrievals fluctuate with increasing noise levels. At the highest noise level (SNR = 50), the error of retrieval is within 5 % for both trace gases.