The attempt to estimate annual variability of NOx emission in Poland using Sentinel-5P/TROPOMI data

Abstract

Information on the annual variability of nitrogen oxides (NOx) emissions is needed in air quality modeling. Emission inventories like EMEP or E-PRTR provide only the total amounts of NOx emission per year. Satellite data on atmospheric composition allows for the quantification of air pollutants emissions and the monitoring of their changes in time. In this study the Sentinel-5P/TROPOMI NO2 tropospheric vertical column (TVC) data from the area of Poland and its border areas from the period of May 2018–November 2021 were used to estimate the NOx lifetime, as well as total NOx emission from Poland area, and NOx emissions from 28 large point sources (LPSs), with their annual variability. The box model (BM) was used to estimate NOx emissions. A novelty is the use of NO2 TVC data in a circular area around each LPS with a radius depending on the wind speed. Furthermore, a new search streak method (SSM) for LPS emission estimation was proposed, and the results were compared with these obtained by the BM method assuming the 5th percentile of NO2 TVC as the emission background (BM5). The estimated monthly and hourly NOx emissions were compared with the stack measurement data for 4 power plants and with the E-PRTR data for all analyzed LPSs. The annual variability of NOx lifetime estimates, higher in winter and lower in summer, is roughly consistent with earlier estimates based on SCIAMACHY and OMI data for mid-latitude sites. The results showed difficulties in estimating NOx emissions for the winter months, whereas for the other seasons the inaccuracies and errors of estimation were much lower. When estimating NOx emission from the LPS using TROPOMI data, the BM5 method gives better results when the LPS is the dominant emission source, while the SSM method works better in an area with many significant emission sources. Comparison of the annual average NOx emission estimates obtained by the BM5 and SSM methods for 28 LPSs with the E-PRTR data showed that the average normalized BIAS is strongly correlated with the emission background around the LPS. The introduction of an additional TROPOMI data selection criterion related to the wind direction sector significantly improves the quality of the results.