This computation involves the Level 2 processing algorithm applied to raw TROPOspheric Monitoring Instrument (TROPOMI) data. The algorithm for retrieval of methane columns from the SENTINEL-5P instrument is based on earlier developments of a CO2 and CH4 retrieval algorithm from GOSAT, called RemoTeC [Butz et al., 2009; 2010; 2011; Schepers et al, 2012; Guerlet et al, 2013a].

In order to account for the effect of aerosols and cirrus, the algorithm retrieves the CH4 column simultaneously with the aerosol/cirrus amount (column integrated particle number concentration), a parameter related to the particle size distribution, and a parameter describing the height distribution. Here, the particle size distribution is described by a power-law function [Mishchenko et al, 1999] which only has two free parameters (related to amount and size). The choice of aerosol/cirrus parameters reflects the information content of the measurements as closely as possible. The retrieval algorithm uses the Level-1B reflectance measurements in the SWIR band and additionally in the NIR band between 757-774 nm (O2 A-band). Additional fit parameters are the surface albedo and its first-order spectral dependence in the two bands, and the total columns of carbon-monoxide and water vapor, respectively.

In order to obtain a proper characterisation of the retrieved CH4 column, it is important to first retrieve a vertical profile (layer averaged number density in different layers of the model atmosphere) and use this retrieved vertical profile to calculate the vertical column. It has been chosen to provide the vertical column as a product, and not the full profile because the Degrees of Freedom for Signal (DFS) of the retrieved CH4 profile is approximately 1. The inversion is performed using Phillips-Tikhonov regularization in combination with a reduced step size Gauss-Newton iteration scheme.

The forward model of the retrieval algorithm uses online radiative transfer calculations, fully including multiple scattering. Here, the radiative transfer model developed by Landgraf et al. [2001], and Hasekamp and Landgraf [2002; 2005] is being used. This model uses the Gauss-Seidel iterative method to solve the radiative transfer equation in a plane-parallel, vertically inhomogeneous atmosphere. To avoid time-consuming line-by-line calculations, the linear-k method developed by Hasekamp and Butz [2008] is employed. Absorption cross-sections of the relevant atmospheric trace gases are tabulated in a look-up table as functions of pressure and temperature. The optical properties of aerosols are also calculated from look-up tables as described in Dubovik et al. [2006].

Cloud Filtering

As for the CO vertical column retrieval, a pre-processing step is performed to discard ground pixels contaminated by clouds. For CH4, only cloud-free ground pixels will be kept. The baseline approach for cloud/cirrus flagging is to use the cloud mask from the VIIRS instrument (on Suomi-NPP satellite, flying in close formation with SENTINEL-5P). With its 1 km x 1 km ground pixel size, the VIIRS cloud mask is much more flexible in defining the area that is required to be cloud-free than the cloud flagging from TROPOMI itself. However, in case VIIRS data are not available, the cloud mask will be obtained from TROPOMI measurements using either:

- A comparison between the apparent surface pressure in the TROPOMI Cloud Level-2 product with the "true" surface pressure in ECMWF.
- The "Two-band CH4 cloud filter" and the "Two-band H2O cloud filter".