Prediction of air pollutants on roadside of the elevated roads with combination of pollutants periodicity and deep learning method

Abstract

The socio-economic effects of urban elevated roads have been well-documented in previous studies. Nevertheless, the environmental impacts of the elevated road location are rarely considered. In this study, we quantified such impacts on four traffic pollutants (e.g., NO, NO2, CO, and O3) in Shanghai, using the two-year observation data from Shanghai roadside air quality monitoring stations—under the elevated road and on the side of the elevated road. Auto-correlation analysis was employed to identify the periodic patterns of air pollutants. The results showed that the severer traffic pollution and more notable daily periodic characteristics of air pollutants were observed under the elevated road. This result may be attributed to the elevated “cap” structure, which has become a barrier to prevent the pollutant diffusion in street canyon. Further, a novel Long Short-Term Memory model with the identified periodicity was proposed to predict air quality. The proposed model achieved higher goodness of fit and lower prediction error in prediction of four pollutants compared to other baseline models, and that the forecasting accuracy was higher under the elevated road. These findings ascertain the effect of the elevated road location on the variations of atmospheric pollutants and could provide implications in taking control measures to mitigate traffic pollution under the elevated road.