Study on Fine-Grained Traffic Pollution Monitoring in Urban Area

Fine-grained monitoring methods such as mobile sensing and grid-based network provide an effective approach to describe the traffic pollution distribution in urban areas. However, how to optimize the monitoring performance is still a problem to be solved. Based on the field data, adopting Kennard-Stone (KS) algorithm, Kriging, Natural Neighbor Tessellation (NNT), and Inverse Distance Weighting (IDW), this paper further studies the scientific mobile monitoring sampling frequency and monitoring network placement strategy applied to infer the surrounding traffic pollution. Experimental results illustrated that different pollutants are specific to mobile sampling frequency and estimation method. For PM2.5, NNT performs better than Kriging and IDW. The accurate distribution pattern of PM2.5 can be inferred when the mobile sampling frequency is every 60 m a sample (RMSPE = 9.98%). For CO2, IDW reaches higher inferring quality, which basically gets access to the surrounding distribution pattern at a sampling frequency of every 10.59 m a sample (RMSPE = 14.63%). Meanwhile, the monitoring network placement strategy for different pollutants is consistent. A spatially uniform distribution strategy can preferably infer the surrounding traffic pollution distribution as fewer detectors are planned to be deployed in the region. The conclusions of this paper may provide guidance for fine-grained traffic pollution monitoring and facilitating strategies in preventing and controlling traffic pollution in smart cities.