Diffusion factors of street canyon pollutants in the cold winter of Xi’an city based on back propagation neural network analysis

Abstract

The factors affecting pollutant diffusion in street canyons during the cold winter of Xi’an City, China, were analysed by measuring the near-surface atmospheric parameters at the height of 1.5 m from ground surface and the aspect ratios (ARs) of 40 street canyons. SPSS software was adopted to obtain eight multiple regression models, including the city street locations, street orientations and street widths. A back propagation neural network was employed to obtain the weight of the factors and their sensitivities. The near-surface atmospheric parameters, leaf area index (LAI) and street canyon AR in the pedestrian area could have an effect on the pollution in street canyons. CO diffusion is strongly influenced by near-surface atmospheric parameters, such as wind velocity, air temperature and relative humidity (RH), together with the CO emission source intensity, street canyon AR and LAI. Their weights were 0.360, 0.275, 0.137, 0.113, 0.068 and 0.047. In particular, near-surface atmospheric parameters were shown to be affected by the street orientation in the city.