Estimating surface-level nitrogen dioxide concentrations from Sentinel-5P/TROPOMI observations in Finland

In recent decades, satellite instruments have been providing observations of air pollutants such as nitrogen dioxide (NO₂) with global coverage. Since late 2018, the TROPOspheric Monitoring Instrument (TROPOMI) on-board the Copernicus Sentinel-5 Precursor satellite has produced NO₂ vertical column densities (VCDs) with the best spatial resolution (currently 5.5x3.5 km at nadir). In order to compare satellite-based observations to traditional NO₂ surface concentration measurements available from air quality (AQ) stations, methods based on chemical-transport model (CTM) simulations can be applied to convert satellite-based VCDs to surface concentrations. Such methods have been mostly developed and applied in regions with high NO₂ concentrations at middle-low latitudes.

In this paper, we test and adapt two methods to estimate surface-level NO₂ concentrations from TROPOMI observations in Finland, which is characterised by low NO₂ concentrations at a high latitude location. Satellite-based estimates show good correlation with co-located surface NO₂ measurements from the Finnish AQ network, although the method lacking a correction for the level of NO₂ mixing within the boundary layer is more prone to underestimation. New cut-off values accounting for the level of mixing were calculated to adapt the column-to-surface conversion to specific conditions in Finland. We also present an alternate approach to reduce the underestimation based on the linear relation between in situ measurements and surface-level estimates.

We utilise chemical transport model simulations to correct our surface-level NO₂ estimates to derive full annual mean concentration estimates compatible with annual limit values defined in AQ legislation. This correction is based on the temporal sampling differences between in situ station measurements and TROPOMI observations. The annual estimates tend to overestimate concentrations compared to ground-based measurements, but can be considered as an upper estimate of surface concentrations. They can therefore complement existing AQ station measurements, especially in regions where the surface networks are sparse. Despite the uncertainties, we find that the two Finnish AQ monitoring regions currently lacking surface stations have estimated mean annual concentrations well below existing limit values, suggesting no need for new ground-based monitoring stations.

Overall, the results provide new information to complement traditional surface-based AQ measurements and to support national environmental authorities in air quality assessment and reporting.