Local differences in feature importance for the prediction of PM10, O3 and NOx using Dresden as example

Due to high yearly casualties, there has been a great effort to model air pollutants with high spatiotemporal resolution. This study uses Dresden as a case study. Additionally to the commonly used meteorological data, the VAMOSlab provides traffic data with high resolution. We investigate the connections with correlation coefficients, mutual information and random forest feature importance. We identify that the traffic volume is an important predictor for the air pollutants, especially NOx. The findings provide a robust foundation for the development of sophisticated spatio-temporal models aimed at enhancing air quality in urban environments.