To aid the analysis of the effect of atmospheric aerosol for the detection and quantification of CO2 plumes from space-borne observations, datasets of synthetic satellite observations are generated. To this end, a new parameterization is developed that estimates expected random and systematic errors for the retrieval of the column-averaged dry-air mole fraction of CO2 (XCO2) for CO2 instrument flying in constellation aboard six satellites in a sun-synchronous orbit. The parameterization takes the sun-satellite geometry as well as spectrally resolved surface albedo and aerosol optical thickness as input and estimates the corresponding XCO2 retrieval errors using two artificial neural networks (ANN). The parameterization has been trained using a global dataset of simulated satellite measurements and corresponding XCO2 retrievals representing a large variety of geophysical scenarios. Random (noise) XCO2 errors are parameterized with almost perfect precision (R=0.99). For the systematic errors, representing the deviation between retrieved and true XCO2, the precision is slightly lower (R=0.87), given the many aerosol properties that contribute to the systematic XCO2 errors like aerosol type, shape, amount, size distribution and vertical distribution.
The new XCO2 error parameterization is used to generate two datasets of synthetic satellite observations for two domains. The first domain, focusing on Europe, covers a geographical area ranging from approx. 33°N to 66°N and -26°E to 53°E (see e.g. Figure 5). The second domain, focusing on the city of Berlin and its surroundings, covers an area ranging from approx. 49°N to 55°N and 7.6°E to 19°E (see e.g. Figure 9). For the European domain, synthetic satellite observations are generated for the entire year of 2015, whereas for the Berlin domain, observations are limited to February and July 2015. Using a satellite orbit simulator developed by SRON, satellite orbits that intersect with the two domains are simulated for the six satellites. With a satellite Level-2 product generator developed by EMPA, high-resolution aerosol (and XCO2) data simulated with the LOTOS-EUROS model, as well as MODIS surface albedo data are projected onto the simulated satellite grids such that the corresponding random and systematic XCO2 errors can be estimated for each orbit and satellite pixel using the XCO2 error parameterization. For each satellite orbit intersecting the respective domains a netcdf-file is generated containing the synthetic satellite observations for the given orbit segment including also the input data (albedo and aerosol optical thickness) used by the XCO2 errors parameterization. The collection of these orbit files, ca. 7800 and 270 orbits for the European and Berlin domains, respectively, constitute the deliverable D2.5 Synthetic satellite datasets of the CO2 Human Emissions (CHE) project.