[August 2022]
This eUVM sub-project is about analysing and forecasting pollution levels in Berlin. A model for three different air pollutants (nitrogen dioxide, PM2.5 and PM10) will be developed by the end of 2022.
The statistical model will use so-called machine learning to establish correlations between air pollution levels, traffic volume, weather and many other factors. From this, similar to a weather forecast, the air pollution load is then predicted for the next few days.
The project continues to address the question of how the spatial coverage of air pollution measurements in Berlin can be improved. An answer to this question is provided in the sub-project for testing new sensor systems. The novel sensor systems can support the forecast model if they are placed at relevant locations. Such locations can either be so-called hotspots, i.e. places where the model repeatedly predicts high air pollutant concentrations but no measurements are available, or places where the model is particularly uncertain. Such uncertainty arises, for example, if there is no measuring station at a comparable location to date that can be used for machine learning.
Although the currently valid limit values for air pollutants are now met across the board in Berlin, an approximation to the stricter WHO guidelines for the protection of human health is desirable and necessary in the long term.
The highest pollution levels occur along the main roads due to traffic. Road users should be motivated to use public transport instead of the car, especially on days with high air pollution values. An assessment in residential areas is of particular interest to citizens, as this is where a lot of time is spent. Physical exposure to air pollutants can thus be better prevented by shifting strenuous outdoor activities, such as jogging, to times of day with lower pollution levels.
The forecast model informs Berliners in advance about upcoming days with particularly bad air. To protect human health, especially on these days public transport should be used, instead of the car.