Application of geographically weighted regression (GWR) in the analysis of the cause of haze pollution in China

Abstract

Haze pollution is an increasingly serious problem in China. Based on the PM2.5 data from 283 prefecture-level cities in China, we combine the stochastic impacts by regression on population, affluence, and technology (STIRPAT) model and environmental Kuznets curve (EKC) hypothesis and use a geographically weighted regression (GWR) method to evaluate the effects of different factors influencing haze pollution in different regions. Nighttime light data from NOAA was used in place of gross domestic product (GDP) in a robustness test. From a global analysis perspective, the EKC is established, and the inflection point is approximately 40000 yuan per capita; however, the GWR estimate is better than the ordinary least squares (OLS) estimate with an improvement in R2 from 0.20 to 0.75. The GWR estimation results show that different environmental protection inputs have different effects in different regions: 1. Technical inputs tend to increase productivity rather than reduce emissions in most areas except Guangdong and Jiangsu. 2. The industrial output values in Guangxi and Yunnan have a greater impact on pollution than those in other regions. 3. In the central and eastern regions with dense populations, comprehensive public transportation can effectively reduce haze pollution. In terms of environmental protection measures, park green areas can reduce pollution, and due to the current status of industrial waste recovery in China, increasing the recovery rate of industrial waste is not conducive to reducing pollution. This article analyzes different causes of haze in different regions and provides suggestions for implementing different environmental policies in different regions.