基于深度学习的日间逐小时地表PM_(2.5)浓度反演  

Deep Learning-Based Daytime Hourly PM_(2.5) Concentration Inversion

作　　者：徐成康 陈斯婕 董长哲[2] 徐文韬 刘东[1] XU Chengkang;CHEN Sijie;DONG Changzhe;XU Wentao;LIU Dong(College of Optical Science and Engineering,Zhejiang University,Hangzhou 310027,Zhejiang,China;Shanghai Satellite Engineering Institute,Shanghai 200240,China)

机构地区：[1]浙江大学光电科学与工程学院,浙江杭州310027 [2]上海卫星工程研究所,上海200240

出　　处：《上海航天（中英文）》2023年第3期46-52,75,共8页Aerospace Shanghai（Chinese&English）

基　　金：国家自然科学基金(41775023);国家重点研发计划(2016YFC1400900);上海航天科技创新基金(SAST2021-051)。

摘　　要：以长三角地区作为研究区域,提出了使用深度学习算法来实现主被动遥感数据结合反演地表PM_(2.5)浓度的方法。基于MPL观测数据,使用雾霾层高度(HLH)替换了边界层高度(BLH)特征,对已有的基于气溶胶光学厚度(AOD)结合大气BLH来反演PM_(2.5)浓度的算法进行了改进。为提高数据覆盖率,对研究区域内的MAIAC AOD进行了填补与评估。利用多种机器学习算法实现了日间逐小时的PM_(2.5)浓度估算,模型验证相关性最高可达0.87。该方法能够为观测气候变化、应对大气污染提供有效帮助。This paper presents the use of deep learning algorithms to invert the surface PM_(2.5) concentration in the Yangtze River Delta region by combing the active-passive remote sensing data.Based on the observation data by micro-pulse lidar(MPL),the existing algorithm for PM_(2.5) concentration inversion by combining the aerosol optical depth(AOD)and the boundary layer height(BLH)is improved by replacing the BLH with the haze layer height(HLH).In order to improve the data coverage,the multi-angle implementation of atmospheric correction(MAIAC)AOD in the studied area is filled and evaluated.A variety of machine learning algorithms are used to estimate the daytime hourly PM_(2.5) concentration,and the obtained model verification correlation is up to 0.87.This method can provide effective assistance for observing climate change and dealing with atmospheric pollution.

关 键 词：PM_(2.5) 遥感反演 雾霾边界层(HLH) 主被动结合 深度学习 

分 类 号：P407.1[天文地球—大气科学及气象学] TP79[自动化与计算机技术—检测技术与自动化装置]