基于随机森林算法的华北平原灰霾区域识别研究  

Study on Haze Region Identification in North China Plain based on Random Forest Algorithm

作　　者：张丽坤 潘一凡 赵楚雯 邱国梁 周佩 陈翔 汪洋 ZHANG Likun;PAN Yifan;ZHAO Chuwen;QIU Guoliang;ZHOU Pei;CHEN Xiang;WANG Yang(Institute of Geography,Fujian Normal University,Fuzhou 350007,China;College of Geography Science,Fujian Normal University,Fuzhou 350007,China)

机构地区：[1]福建师范大学地理研究所,福建福州350007 [2]福建师范大学地理科学学院,福建福州350007

出　　处：《遥感技术与应用》2023年第3期752-763,共12页Remote Sensing Technology and Application

基　　金：国家自然科学基金项目(42001295);福建省公益类科研院所专项(2020R11010009-1);遥感科学国家重点实验室开放基金(OFSLRSS202011)资助。

摘　　要：大气灰霾污染是我国近年来面临的重大环境问题之一。灰霾监测是灰霾污染治理体系的重要组成部分,而遥感手段可实现大范围、长时序的动态监测,在一定程度上弥补了传统地面站点监测手段的不足。基于MODIS/Terra数据,以华北平原七省市为研究区域,利用大气灰霾的光谱和空间特性搭建10维特征空间,并使用随机森林算法构建灰霾区域识别模型。该模型不仅实现了灰霾区域的识别提取功能,同时还实现了轻、重霾区域的区分功能。经地面站点PM_(2.5)浓度监测数据的验证,灰霾区域识别的总体精度为87.82%,Kappa系数为0.75。轻、重霾区域识别的总体精度为86.81%,Kappa系数为0.73。研究结果表明:该模型针对卫星影像中的灰霾区域具有较好的识别效果,可为大气污染监测提供数据支撑,且对其他大气污染物的监测研究也具有一定的借鉴意义。Atmospheric haze pollution is one of the major environmental problems facing China in recent years.Haze monitoring is an important part of haze pollution governance system,and remote sensing can realize largescale and long-term dynamic monitoring,which to a certain makes up for the shortcomings of traditional site monitoring methods.Based on MODIS/Terra data,this study takes seven provinces and municipalities in North China Plain as the study area,built a 10-dimensional feature space by using the spectral and spatial char⁃acteristics of atmospheric haze,and built a regional identification model of haze using random forest algorithm.This model not only realizes the function of haze region identification and extraction,but also realizes the func⁃tion of light and heavy haze region classification.Through verification of PM_(2.5) monitoring data from ground sta⁃tions,the overall accuracy of haze region identification is 87.82%,and the Kappa coefficient is 0.75.The over⁃all accuracy of light and heavy haze region identification is 86.81%,and the Kappa coefficient is 0.73.The study results show that the model has a good identification effect on haze region in satellite images,which can provide data support for air pollution monitoring,and has certain reference significance for monitoring other atmospheric pollutants.

关 键 词：大气污染 灰霾监测 遥感识别 随机森林算法 华北平原 

分 类 号：X513[环境科学与工程—环境工程] X87