中国PM2.5污染与社会经济的空间关系及成因

中国高能耗的经济增长模式和不生态的城镇化模式是PM_2.5污染的主要诱因,为了弄清PM_2.5的本质,着重研究PM_2.5污染的产出机理与模型。首先用逐步回归分析法,确定对PM_2.5影响较大的变量,再对PM_2.5及其相关变量进行空间相关分析,在GIS技术与空间统计学的支持下,建立中国区域性细颗粒物空气污染评估模型。结果表明：中国PM_2.5污染具有东高西低的区域差异特点,这与中国人口分布密度特征曲线（胡焕庸线）所划分出的人口空间分布特点相一致。考虑了空间效应影响的模型拟合度（R²=0.71）优于传统统计模型（R²=0.62）。PM_2.5与总人口、人均汽车保有量的平方、第二产值比例的平方成正比,与森林覆盖率的平方成反比,其中对PM_2.5贡献率最大的是人均汽车保有量。

Social and economic drivers of PM2.5 and their spatial relationship in China

The anthropogenic reasons of particulate matter 2.5 (PM_2.5) formations are considered to be highly related to local economic development model and fast urbanization. The aim of this study is to explore the social and economic factors of PM_2.5 and their spatial and quantitative relationship in China during the period of February to December in 2013. In this paper, Geographic Information System (GIS) technology combined with spatial statistics to form a novel evaluation model for studying the PM_2.5 formation of China. The surface of PM_2.5 is obtained by using inverse distance weighted (IDW). We first design a forward-selection stepwise regression model for tracing major factors of PM_2.5 formation, and then study spatial distribution of PM_2.5 and the major factors of PM_2.5. Their G statistic indexes are calculated to detect the spatial dependence. Based on the estimated G statistic index, and Lagrange Multiplier and goodness-of-fit tests, the Spatial Lag Model (SLM) is selected to estimate PM_2.5. The results show that: (1) PM_2.5 presents an obviously uneven spatial distribution. East and most parts of northeast China have severe pollution. Especially, the provinces of Beijing, Hebei, Henan, Anhui, Jiangsu, Sichuan and Shanghai are seriously severe areas. (2) The square of per capita motor vehicle, population, the square of forest coverage, and the square of the proportion of secondary industry are the major factors of PM_2.5. (3) Because G statistic indexes of PM_2.5 and major factors confirm the presence of their spatial autocorrelation, the data of spatial autocorrelation need to be considered in the modeling process. The Spatial Lag Model has better goodness-of- fit test (R²=0.71) than classical linear regression model (R²=0.62). (4) The ranking of contribution to PM_2.5 is the square of per capita motor vehicle, spatial factors, population, the square of forest coverage, the square of the proportion of secondary industry. Positive correlation coefficients were observed between the annual mean concentrations of PM_2.5 and the square of per capita motor vehicle, population, and the square of the proportion of secondary industry. However, the correlation coefficient of the square of forest coverage was found to be negative. (5) PM_2.5 governance is an extremely urgent task. In order to reduce PM_2.5, we need to take measures in regional collaborative governance, and should terminate industrial production with high energy consumption. Optimization of the industrial structure and energy structure, as well as reforestation and afforestation can also reduce PM_2.5.