Source attribution of particulate matter in Berlin

The exposure to ambient particulate matter in metropolitan areas is a major health problem. A prerequisite for formulating effective mitigation strategies it to understand the origin of particulate matter in terms of source regions and sectors. We performed a source attribution of particulate matter (PM) for the Berlin agglomeration area covering the period from 2016 to 2018 using the LOTOS-EUROS chemistry transport model. The (3 year-) mean modelled urban background PM2.5 concentration (10.4 μg/m³) is largely explained by households (3.2 μg/m³) and industry & energy (2.0 μg/m³), while the remaining source sectors contribute the other half. The modelled annual mean urban increment for PM2.5 is mainly attributed to households (1.6 μg/m³) and traffic (0.5 μg/m³). With respect to its relative shares the PM10 source attribution looks similar to that of PM2.5 throughout the year, but with enhanced natural contributions. From a geographical perspective the main source area for the PM2.5 in Berlin is Germany (5.1 μg/m³) itself, followed by the contributions from transboundary transport (3.4 μg/m³). The German sources could be further split into Berlin (2.6 μg/m³), Brandenburg (0.7 μg/m³) and remaining states of Germany (1.8 μg/m³). About one third of the foreign shares can be attributed to Germany's neighbouring countries Poland and Czech Republic. During episodes these contributions can significantly differ, e.g. in February 2017 the Polish contribution is about 1/3rd. The sectoral contributions agree with previous findings except that our study indicates lower contributions for traffic. The model's underestimation of total PM is largely caused by an underestimation of the coarse mode PM. Both the coarse mode urban increment as well as the regional background concentrations are underestimated by the model, especially during summer. We suggest that the enhanced coarse material (in the city) during warm seasons is predominated by (road) resuspension processes which need more of our attention to further improve our models