基于地理加权回归克里金的中国PM2.5浓度空间制图方法

doi:10.11873/j.issn.1004-0323.2018.6.1103

遥感技术与应用

2018, Vol. 33

Issue (6): 1103-1111 DOI: 10.11873/j.issn.1004-0323.2018.6.1103

数据与图像处理

基于地理加权回归克里金的中国PM2.5浓度空间制图方法

邵彦川¹，王江浩^2，3，葛咏^2，3

（1.南京农业大学资源与环境科学学院环境科学系，江苏南京 210095;
2.中国科学院地理科学与资源研究所资源与环境信息系统国家重点实验室，北京 100101;
3.中国科学院大学，北京 100049）

Spatial Mapping of PM 2.5 Concentrationin China with Geographically Weighted Regression Kriging Model

Shao Yanchuan¹，Wang Jianghao^2，3，Ge Yong^2，3

(1.Department of Environmental Science，College of Resources and EnvironmentalSciences，Nanjing Agricultural University，Nanjing 210095，China；
2.State Key Laboratory of Resources and Environment Information System，Instituteof Geographic Sciences and Natural Resources Research，Beijing 100101，China;
3.University of Chinese Academy of Sciences，Beijing 100049，China)

全文: PDF(7321 KB)

摘要： 空气细颗粒物健康暴露风险等研究需要准确的PM2.5浓度时空分布信息作为健康评估的重要输入。然而，由于监测台站稀疏分布，通常需要融合遥感等辅助信息，通过空间制图模型得到PM2.5浓度的分布状况。如何在估计模型中将PM2.5浓度的空间分布特征融入制图模型将是提高PM2.5制图精度的关键。发展了一种融合地理加权回归和克里金插值方法的混合模型：地理加权回归克里金(Geographically Weighted Regression-Kriging，GWRK)，地理加权回归模型考虑PM2.5浓度分布的空间异质性，克里金模型对回归后的残差中存在的空间自相关性进行建模。基于该方法，利用中国空气质量监测站数据，采用遥感、模式模拟数据作为辅助信息，对2017年中国逐月的PM2.5浓度分布进行估计空间制图。交叉验证结果表明，GWRK相较于传统制图方法（最小二乘回归、地理加权回归、回归克里金）具有更高的精度，决定系数R²为0.824，平均绝对误差为6.96 μg/m³，均方根误差为10.94 μg/m³。2017年逐月的PM2.5浓度制图结果显示，在时间上，冬季是PM2.5污染最严重的时段，夏季最轻，空间上，东部经济较为发达的城市如长三角地区是污染严重区，西南地区污染程度较轻。

关键词： 地理加权回归克里金（GWRK）; PM 2.5; 空间制图

Abstract: The study of health exposure risk of fine particulate matter in air requires accurate spatial distribution information of PM 2.5 concentration as an important input for assessment.Due to the sparse monitoring stations，it is necessary to use the auxiliary information such as remote sensing to obtain the spatial distribution of PM 2.5 by spatial mapping model.One key issue to improve the accuracy of PM 2.5 concentration map isto integrate the spatial characters of PM 2.5 into model.In this paper，ahybrid method that integrate geographically weighted regression and kriging interpolation was developed as Geographically Weighted Regression Kriging (GWRK) model.Geographically weighted regression model was used to consider the spatial heterogeneity of PM 2.5 concentrationdistribution.The kriging method was adopted to model the spatial autocorrelation of the residual error.Then，we estimate monthly PM 2.5 concentration in China with GWRK by using the monitoring data fromair quality station，and auxiliary information from remote sensing and model simulation.The cross validation showed that GWRK achieved more accurate result than traditional mapping methods (least square regression，geographically weighted regression，regression kriging).The coefficient of determination was 0.824，the mean absolute error was 6.96 μg/m³，and the root mean square error was 10.94 μg/m³.The result of PM 2.5 concentration mappingshowed that winter was the most worse period，and summer was the lightest.In space，the cities with more developed economy in the east，such as the Yangtze River Delta，were polluted seriously，while the southwest was less polluted.

Key words: Geographically Weighted Regression Kriging(GWRK) PM 2.5 Spatial mapping

收稿日期: 2017-12-02 出版日期: 2019-01-29

ZTFLH:

X513

基金资助: 国家自然科学基金项目(41601427、41531174)，中国科学院A类战略性先导科技专项（XDA19040503）。

作者简介: 邵彦川（1997-），男，江苏南京人，本科，主要从事环境遥感方面的研究。Email：zhuiluosyc@gmail.com。

	服务
	把本文推荐给朋友
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章

引用本文:

邵彦川, 王江浩, 葛咏. 基于地理加权回归克里金的中国PM2.5浓度空间制图方法[J]. 遥感技术与应用, 2018, 33(6): 1103-1111.

Shao Yanchuan, Wang Jianghao, Ge Yong. Spatial Mapping of PM 2.5 Concentrationin China with Geographically Weighted Regression Kriging Model. Remote Sensing Technology and Application, 2018, 33(6): 1103-1111.

链接本文:

http://www.rsta.ac.cn/CN/10.11873/j.issn.1004-0323.2018.6.1103 或 http://www.rsta.ac.cn/CN/Y2018/V33/I6/1103

No related articles found!

Viewed

Full text

Abstract

Cited

Shared

Discussed