Joint prediction of CO, NOx, NMHC pollutant concentrations in urban area

Abstract

Primary air pollutants could directly and indirectly (through generating secondary air pollutants) threaten natural or human systems. In particular, urban air pollution issue becomes more and more significant in the recent decades, as a result of rapid urbanization. However, the estimation of multiple-pollutant concentrations is limited by high spatial and temporal variations, which hinders the accuracy of mechanistic modeling of air pollution. In this study, we employed an Artificial Neural Network (ANN) model to jointly predict multiple primary pollutants, including CO, NOx, and nan-methane hydrocarbons (NMHC), over the urban area of Italy. The results showed that performances of the ANN model (MSE and Pearson correlation) in joint prediction cross multiple pollutants were much better than in prediction of any single pollutant individually, indicating the joint measurements of multiple pollutants could favor the machine-learning model by providing useful information from one pollutant to predict another.