The conducted research demonstrates the potential of using satellite data from the Sentinel-5P mission to support the monitoring of nitrogen dioxide (NO2) pollution. The study showed that NO2 concentrations observed near the Earth's surface are strongly influenced by meteorological factors and local environmental as well as socio-economic conditions. It was emphasized that both mesoscale variables and local anthropogenic pressure must be taken into account when interpreting changes in air quality. Various modelling approaches were applied to estimate ground-level NO2 concentrations from satellite data, including classical linear regression models and machine learning methods (random forests and support vector machines). The results confirmed that incorporating additional meteorological, geographic, and socio-economic variables significantly improves estimation accuracy. Machine learning models achieved accuracy levels of 63-69%, outperforming linear regression approaches. It was also shown that temporal averaging of data improves the accuracy of results. Among the analysed methods, random forests were identified as the most effective and practical tool for further analyses. At the same time, the study highlighted the limitations of satellite data, mainly related to the column-based nature of measurements and their sensitivity to cloud cover. Despite these limitations, integrating satellite data with meteorological and socio-economic information was shown to enable reliable spatial estimations of NO2 concentrations. The analysis indicated that the most influential variables for model performance were boundary layer height, solar radiation, road density, population density, and night-time light intensity. Research on the representativeness of the current measurement network showed that part of the population, both urban and non-urban, lives in areas exceeding the annual recommended NO2 concentration levels that are not covered by representative measurements.
Specific locations for new monitoring stations were proposed, the addition of which would significantly increase the country's monitoring coverage, especially in highly urbanised areas. This demonstrates the practical potential of satellite data for planning and optimising the national air quality monitoring system. In conclusion, the study confirmed that Sentinel-5P data, combined with meteorological and socio-economic information, can be an effective tool for supporting air quality monitoring in Poland. They allow not only for assessing pollution levels on a regional scale, but also for supporting administrative decision-making regarding the location of monitoring stations or responding to crisis situations. The study thus confirms the importance of an interdisciplinary approach that combines Earth and environmental sciences with the analysis of socio-economic phenomena in research on air quality and its impacts on human health and living conditions.