This computation involves the Level 2 processing algorithm applied to raw TROPOspheric Monitoring Instrument (TROPOMI) data. The retrieval algorithm requires several input fields:
• The measured Earth radiance and solar irradiance spectra including noise estimate, solar and viewing
geometry, and information of geo-location.
• ECMWF temperature, water vapor, and pressure profiles, and geo-potential height.
• An estimate of the CH4 field using a chemistry transport model, e.g. Transport Model 5 (TM5, [RD33]).
• An estimate of the CO column from a chemistry transport model (e.g. TM5).
The retrieval is performed in two steps: first, as part of the SWIR preprocessing module, the vertically integrated amount of methane is retrieved from a dedicated fit window of the SWIR band between 2315 and 2324 nm using a non-scattering radiative transfer model. The extent of lightpath shortening and enhancement due to atmospheric scattering by clouds and aerosols can be indicated by comparing the retrieved CH4 column with a priori knowledge. If the difference ∆CH4 exceeds a certain threshold, observations are strongly contaminated by clouds and are rejected. In a second step, the SICOR full physics retrieval approach is used to infer CO columns from the adjacent spectral window, 2324-2338 nm. Here, the methane absorption features are used to infer information on atmospheric scattering by clouds and aerosols, which passed the cloud filter, together with the atmospheric CO and H2O abundances, surface albedo, and spectral calibration of the reflectance spectrum. The scattering layer has a triangular height distribution of fixed geometrical thickness, and its optical depth and height are parameters to be retrieved. This step of the retrieval relies on accurate a priori knowledge of CH4 which will be provided within an accuracy of ±3 % by a dedicated methane forecast using the TM5 atmospheric transport model. The atmospheric scattering is described by a two-stream radiative transfer model. Finally, the retrieval product consists of a CO column estimate including a column averaging kernel and a random error estimate. For more information on the processing algorithm please look at the ATBD document on the TROPOMI CO webpage.