Assimilation of SO2, CO, HCHO and O3 satellite data with Optimal Interpolation implemented in Atmospheric Modelling System MINNI

SO₂, CO, HCHO, NO₂ and O₃ satellite data measured by the TROPOspheric Monitoring Instrument (TROPOMI) instrument on board the Copernicus Sentinel-5 Precursor satellite launched in October 2017 provides a comprehensive atmospheric composition dataset at high horizontal resolution (5.5 km -7 km). Here, we used total column SO₂, CO, HCHO and tropospheric O₃ L2 products for testing Optimal Interpolation (OI) algorithm implemented in the atmospheric modelling system MINNI. The three-dimensional hourly concentrations fields produced with the chemical transport model FARM at European scale were adjusted with satellite retrievals of pollutants simultaneous and, separately, for O₃. The three-dimensional optimal interpolation (OI) scheme developed for satellite data assimilation consider the spatial and temporal error structures of the background field through the background error covariance matrix (B). This matrix was estimated in the same way for all five pollutants and, for simplicity, was assumed diagonal considering thus that the retrievals at different points do not influence each other. Besides, model data were corrected only where the observations were available.

This simple OI scheme is computationally feasible but not effective in the same way for all pollutants. There are also differences in assimilating a single species or all together for those formed in the atmosphere such as O₃. The differences in the three-dimensional modelled concentrations without and with assimilation for all pollutants and single species are discussed as well as their performances in comparison with ground observations for evaluating the impact of assimilation.

To understand the limitations of the implemented OI algorithm, several experiments were run to investigate the effects of using different definitions of B in different states of atmospheric composition. The comparisons and quantitative evaluations were performed both horizontally and vertically, analysing 2D fields and point time series.

The preliminary results show that the assimilation can improve modelled NO₂ and capture SO₂ volcano eruptions which are not present in anthropogenic emission inventories. However, the assimilation of the short-lived species like NO₂, HCHO and O₃ poses many problems, in particular due to their interdependencies, therefore more research is needed.