基于AOD数据和随机森林模型估算河南省ρ(PM_(2.5))  

Estimation ofρ(PM_(2.5)) in He’nan Province based on AOD data and random forest model

作　　者：田智慧[1] 吴文秀 魏海涛[1] TIAN Zhi-hui;WU Wen-xiu;WEI Hai-tao(Joint Laboratory of Ecological Meteorology,Chinese Academy of Meteorological Sciences and Zhengzhou University,College of Earth Science and Technology,Zhengzhou University,Zhengzhou 450052,China;School of Water Conservancy and Environment,Zhengzhou University,Zhengzhou 450052,China)

机构地区：[1]郑州大学地球科学与技术学院中国气象科学研究院郑州大学生态气象联合实验室,郑州450052 [2]郑州大学水利科学与工程学院,郑州450052

出　　处：《兰州大学学报（自然科学版）》2022年第2期172-178,共7页Journal of Lanzhou University(Natural Sciences)

基　　金：国家重点研发计划项目(2018YFB0505000);河南省重点研发与推广专项(科技攻关)项目(192102210124)。

摘　　要：基于融合的深蓝/暗目标气溶胶光学厚度数据集构建了随机森林(RF)、地理加权回归(GWR)和线性混合效应(LME)模型估算河南省的ρ(PM_(2.5)),通过十折交叉验证评价模型性能.结果表明,RF模型估算的河南省ρ(PM_(2.5))精度优于其他模型,其中RF模型月数据的R^(2)比GWR和LME模型分别高0.26、0.27,R_(MSE)分别低6.56、5.32μg/m^(3);RF模型季节数据的R^(2)比GWR和LME模型均高0.24,R^(2)MSE分别低3.13、4.31μg/m^(3).利用RF模型估算河南省2017年的ρ(PM_(2.5)),冬季最高,R最高,R_(MSE)最大;春、秋季次之;夏季最低,R^(2)和R_(MSE)最小,空间分布特征为北高南低.A random forest(RF)model,geographically weighted regression(GWR)and linear mixed effects(LME)model were constructed to estimate theρ(PM_(2).5)in He’nan Province based on the fused deep blue/dark target aerosol optical depth data set.The model performance was evaluated by a 10-fold crossvalidation.The results showed that the performance of the RF model was better than the other two models.The monthly model R^(2)was 0.26 and 0.27 higher than the GWR and LME models,and R_(MSE)=6.56 and 5.32μg/m^(3)lower than the other two.The R^(2)of seasonal data of RF model was 0.24 higher than that of the GWR and LME models respectively,and the R_(MSE)was 3.13 and 4.31μg/m^(3)lower respectively.ρ(PM_(2).5)in He’nan Province for four seasons was estimated by the RF model.The highest R^(2)and maximum R_(MSE)was in winter,and the lowest R^(2)and minimum R_(MSE)in summer,with the spring and autumn being in the middle.The highestρ(PM_(2).5)was in winter,and the lowestρ(PM_(2).5)in summer.The spatial distribution was high in the north and low in the south.

关 键 词：气溶胶光学厚度 PM_(2.5)质量浓度 随机森林模型 河南省 

分 类 号：X513[环境科学与工程—环境工程]