中国东部沿海四省一市PM_(2.5)浓度遥感估算方法研究  

Research on Methods of Remotely Sensed PM_(2.5) Concentrations Estimation in Four Provinces and One City along the East Coast of China

作　　者：杨立娟 张建霞[1] 林木生 Yang Lijuan;Zhang Jianxia;Lin Musheng(Department of Surveying and Mapping Engineering of Minjiang University,Fuzhou 350118,China)

机构地区：[1]闽江学院测绘工程系,福建福州350018

出　　处：《遥感技术与应用》2021年第6期1408-1415,共8页Remote Sensing Technology and Application

基　　金：闽江学院优秀引进人才科研启动项目(MJY20001);福建省自然科学基金项目(2021J05204)。

摘　　要：卫星遥感反演的气溶胶光学深度(AOD)产品已被广泛应用于近地面PM_(2.5)浓度的估算。已有研究表明通过构建AOD和PM_(2.5)之间的高级统计模型—线性混合效应模型(LME)可以有效获取近地面PM_(2.5)浓度的空间分布,但由于引入了大量的气象和土地利用等因子,使得模型对变量的解译能力有所降低。为此,基于MODIS AOD(空间分辨率:3 km),以我国东部长江三角洲—福建—广东(YRD-FJ-GD)为研究区,构建了两种非参数机器学习模型,即支持向量机(SVM)和随机森林(RF)模型,来估算2018年YRD-FJ-GD地区的近地面PM_(2.5)浓度,并将其与线性混合效应模型(LME)的估算结果进行对比。研究发现,3种模型估算的PM_(2.5)浓度与地面实测值之间的R^(2)均高于0.6,其中,RF模型的估算精度最优,模型拟合的R2高达0.91,比SVM模型(R^(2)=0.79)和LME模型(R^(2)=0.64)的估算结果分别提高了13%和30%;且RMSE(~9.07μg/m^(3))也远低于LME(~19.09μg/m^(3))和SVM模型(~17.29μg/m^(3))。此外,由随机森林(RF)模型估算的2018年YRD-FJGD地区的PM_(2.5)空间分布显示,长江三角洲(YRD)地区的年均PM_(2.5)浓度最高(>46μg/m^(3)),其次为广东省(GD),福建地区(FJ)的年均PM_(2.5)浓度最低(<37μg/m^(3));4个季节的平均PM_(2.5)浓度则呈现冬季(46.32μg/m^(3))>春季(38.80μg/m^(3))>秋季(36.15μg/m^(3))>夏季(30.16μg/m^(3))的分布格局。研究结果表明:与高级统计模型(LME)和机器学习(SVM)相比,随机森林(RF)模型能更好地应用于YRD-FJ-GD地区的PM_(2.5)浓度估算。The Aerosol Optical Depth(AOD)derived from remote sensing imageries has been widely used in estimating ground-level PM_(2.5) concentrations in large areas.Previous studies that focused on PM_(2.5) estimation have reported high predictability of PM_(2.5) concentrations when using AOD and the advanced statistical model(i.e.,Linear Mixed Effects model(LME)).However,the interpretation ability of the LME model was lowered,as it introduced many meteorological and land use variables in the model,and the importance of each variable to PM_(2.5) concentrations was hard to interpret.Therefore,this study developed two nonparametric machine learning methods,i.e.,Support Vector Machine(SVM)and Random Forest(RF),to estimate the ground-level PM_(2.5) concentrations.The eastern Yangtze River Delta-Fujian-Guangdong(i.e.,YRD-FJ-GD)region in China was employed as our study case,and we also compared the predictability of these two models with the LME model.The results showed that the overall R^(2) between estimated and observed PM_(2.5) concentrations exceeded0.6 for three models,where RF received a R2 of 0.9,i.e.,13% and 30% higher than SVM(R^(2)=0.79)and LME(R^(2)=0.64)model,respectively.The RMSE values were 9.07,17.29 and 19.09μg/m;for RF,SVM and LME model,respectively.In addition,the spatial distribution of PM_(2.5) concentrations estimated from the optimal model(i.e.,RF)illustrated high annual PM_(2.5)in YRD(>46μg/m^(3)),and GD ranked the second.FJ exhibited a relatively low annual PM_(2.5)(<37μg/m^(3)).The seasonal PM_(2.5) concentrations presented a distribution pattern as winter(6.32μg/m^(3))>spring(38.80μg/m^(3))>autumn(36.15μg/m^(3))>summer(30.16μg/m^(3)).Our results revealed that the AOD and RF model could be a good proxy for estimating PM_(2.5)concentrations in YRD-FJ-GD region.

关 键 词：LME SVM RF PM 估算 YRD-FJ-GD 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]