Analysis, prediction and mitigation of exposure to vehicular air pollution based on multi-source urban data

An increasing amount of vehicular emissions in urban air pollution create a health risk for urban residents. Meanwhile, calculation and analysis of vehicular pollution using GPS trajectories and microscopic models is getting more popular as this method proves to be more useful and reliable in comparison to other methods. However, GPS-trajectory-based estimations suffer from the lack of GPS data and absence of validation/calibration of estimated emission amounts. Another problem is in the assessment of pollution levels using GPS trajectories as previous studies only consider changes in total vehicular emissions and ignore air quality guideline levels. In this paper, the methodology and preliminary results of experiments conducted for imputation of missing emission data are reported. An existing graph convolutional network model which is designed to predict traffic flows is adopted to estimate vehicular emissions in Pisa. This approach is based on the assumption that the same model can predict traffic emissions as a traffic flow and resulting emission are correlated. In the end of the paper, there is a discussion of future research directions planned to be taken during my PhD period to address issues in the estimation, analysis and mitigation of exposure to vehicular emissions in cities.