Multiple driving factors and hierarchical management of PM2.5: Evidence from Chinese central urban agglomerations using machine learning model and GTWR

Abstract

In the fast-developing urban agglomerations (UAs), it is of importance to make accurate judgments concerning the multiple driving factors, and establish hierarchical joint management policy. The impact of weather conditions on daily PM2.5 concentrations in the Chinese central UAs was studied using machine learning algorithm, and the analyzed results were integrated into “the proportion of day numbers with negative weather conditions (PDNW)”. Geographically and temporally weighted regression (GTWR) was used to analyze the driving factors of PM2.5 pollution. Results showed that PM2.5 pollution in central China decreased from north to south, and spatial gathering was becoming increasingly prominent. The PM2.5 predicted values decreased smoothly, with barometric pressure and humidity exerting a large effect, and wind speed and direction having a complex effect. Meteorological conditions had a small effect on the annual scale, but the timing of the effect varied in each city. The distribution of PDNW ranged from 23.3% to 55.6%. The proportion of the tertiary industry's GDP (mean − 0.191), education expenditure (mean − 0.057), and the greening rate of urban built-up areas (mean − 0.295) were found to be negatively correlated with PM2.5 pollution. Transportation, urban greening, innovation, and entrepreneurship were driving factors with obvious spatial differences.