基于地理加权模型的我国冬季PM2.5遥感估算方法研究

为了分析冬季我国区域范围内近地面PM_(2.5)质量浓度时空分布特征,根据卫星遥感反演PM_(2.5)质量浓度的基本原理,综合考虑我国不同地区的PM_(2.5)污染特征的空间差异性,基于卫星遥感、气象模式资料及同期地面观测的PM_(2.5)质量浓度数据采用地理加权模型进行回归分析,研究构建了我国区域范围内近地面PM_(2.5)遥感反演模型.结果表明:在冬季暗像元反演AOD算法受限制的情况下,深蓝算法产品可以一定程度上弥补暗像元算法的不足,将二者有效融合能同时提高AOD产品的精度和空间覆盖度;利用地理加权回归模型进行全国区域PM_(2.5)遥感估算,既能体现全国PM_(2.5)时空分布的全局变化特性,又能从局部体现全国PM_(2.5)组分、污染程度及垂直分布结构特征的空间差异特性,基于地理加权回归模型的PM_(2.5)遥感反演结果(R2=0.7)明显优于多元线性回归模型(R2=0.56);2013年12月—2014年2月份全国PM_(2.5)空间分布呈现明显的区域特征,PM_(2.5)浓度较高的地方主要分布在华北南部、长三角中部和北部、华中东部及四川东部等地,西部和北部地区PM_(2.5)污染相对较轻;从时间变化来看,全国冬季12月份PM_(2.5)污染最重,1月份次之,2月份相对最低.这可为全国PM_(2.5)区域联防联控提供有力的信息支撑.

Estimations of PM2.5 concentrations based on the method of geographically weighted regression

In order to analyze ground-level PM_2.5 concentration of China during winter, a PM_2.5 retrieval model was built based on satellite remote sensing, meteorological model and ground-based PM_2.5 observation data. Considering spatial variations of PM_2.5 in China, PM_2.5 concentration was retrieved using a geographically weighted regression procedure and cardinal principle of PM_2.5 retrieval in this model. The results showed that deep blue algorithm can make up for the limitation of AOD retrieval from dark object algorithm, and the combination of the two algorithms can improve both AOD accuracy and space coverage. The estimated PM_2.5 concentration of China using geographically weighted regression model can not only keep the overall variation characteristics, but also reflect the regional differences of PM_2.5 component, pollution level and horizontal vertical distribution characteristics. The PM_2.5 result from geographically weighted regression model (R²=0.7) is better than that from multiple linear regression model(R²=0.56). From PM_2.5 result between December 2013 and February 2014 we can find that the distribution of PM_2.5 in China shows obvious regional characteristics, with high concentration distributed mainly over the southern part of North China Plain, central and northern Yangtze River Delta, eastern part of central China, eastern Sichuan and other places, while western and northern China see a lower concentration. PM_2.5 has the highest concentration value in December, followed by January and February. This study provides great scientific support for joint prevention and control of PM_2.5 over China.