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.
作者
徐成康
陈斯婕
董长哲
徐文韬
刘东
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)
出处
《上海航天（中英文）》
CSCD
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)。
