Fine particulate matter (PM2.5) remains one of the most relevant pollutants affecting air quality and human health worldwide. While satellite-derived aerosol optical depth (AOD) is commonly used to estimate surface PM2.5 concentrations, most existing approaches rely heavily on auxiliary meteorological model data. This study presents a novel approach that derives PM2.5 for Germany and neighboring countries for the year 2022 based on TROPOMI satellite observations by applying a Random Forest (RF) algorithm. In addition to AOD, various TROPOMI products related to atmospheric composition are included to assess their added value for improving model performance. A comparison with CAMS forecasts is performed to demonstrate that the satellite-based model can more realistically reproduce both spatial patterns and temporal dynamics of PM2.5. Furthermore, with a case study for March 2022 the model’s ability to capture pollution peaks during high-pollution events, which are particularly relevant for public health assessments, is illustrated. The TROPOMI-based RF model achieves high accuracy despite the absence of meteorological input and successfully captures the spatiotemporal variability of PM2.5 concentrations. The results of the study highlight the potential of TROPOMI data for near-real-time PM2.5 monitoring and underline its value as an independent, observation-based alternative to chemical transport model forecasts. As part of the DLR project INPULS, the proposed approach provides an important step toward the development of an operational daily satellite-based PM2.5 product from the atmospheric Copernicus Sentinel missions and contributes to improving air quality surveillance, both under common and extreme pollution conditions.
Handschuh et al. (2026) studied this question.