摘要
气溶胶光学厚度(AOD)是气溶胶最重要的参数之一,现有的遥感AOD产品受云、积雪等因素的影响空间缺失严重,因此,生成空间覆盖完整的AOD具有重要意义.本文融合MODIS的MAIAC AOD和Himawari-8的AHI AOD,结合气象数据和高程数据,提出一种集成反距离权重插值(IDW)和CatBoost模型的时空连续AOD重构方法(命名为IDW-CatBoost).将此方法应用于京津冀和台湾岛的AOD重构,并与IDW、CatBoost方法对比,重构结果利用地基监测AERONET AOD进行验证,其中,京津冀的验证数为352个,台湾岛的验证数为641个.结果表明:在空间分布上,IDW AOD存在星点状特征,CatBoost、IDW-CatBoost的AOD具有空间连续分布的纹理特征;精度上,经地基监测AERONET AOD验证,京津冀地区IDW AOD与IDW-CatBoost AOD接近;台湾岛IDW-CatBoost AOD相比于IDW、CatBoost结果,R^(2)分别提高了10%和5%.经过多传感器AOD融合,与单传感器AHI L2、L3、MAIAC AOD相比,IDW-CatBoost重构AOD精度显著提升,在京津冀地区,R^(2)分别提高了15%、35%和12%,RMSE分别下降了25%、38%和22%;在台湾岛,R^(2)分别提高了14%、76%和76%,RMSE分别下降了6%、24%和24%.因此,基于多传感器AOD融合的IDW-CatBoost模型用于重构AOD产品,不仅空间覆盖完整,而且具有更高的精度.
Aerosol optical depth(AOD)is one of the most important parameters for aerosols.Existing AOD products derived from remote sensing data are seriously affected by clouds,snow,and many other factors.It is thus of great significance to generate AOD with large spatial coverage.This paper proposes a spatio-temporal continuous AOD reconstructed method(i.e.,IDW-CatBoost)based upon Inverse Distance Weight interpolation(IDW)and CatBoost models.It fuses the MAIAC AOD of MODIS,the AHI AOD of Himawari-8,and meteorological and elevation data.The IDW-CatBoost model was applied to the AOD reconstruction of Beijing-Tianjin-Hebei(BTH)and Taiwan Island,and compared with IDW and CatBoost methods.We validated reconstruction results using the ground-based monitoring AERONET AOD,where 352 samples were used for BTH and 641 for Taiwan Island.Results showed that the AOD obtained by IDW had star-dotted features in spatial distribution,while CatBoost and IDW-CatBoost AODs exhibited texture features of continuous spatial distribution.The AOD results of IDW were close to those of the IDW-CatBoost method in the BTH region when verifying by ground monitoring AERONET AOD data.Compared with IDW and CatBoost methods,the AOD results of the IDW-CatBoost in Taiwan Island were improved by 10%and 5%,respectively,for the R^(2) measure.Compared with single-sensor AHI L2,L3,and MAIAC AODs,the accuracy of the IDW-CatBoost AOD fused with multi-sensor data was significantly improved,leading to an R^(2) improvement of 15%,35%and 12%,and a RMSE decrease of 25%,38%and 22%,respectively,in the BTH region.Moreover,in Taiwan Island,the R^(2) was improved by 14%,76%and 76%,and the RMSE decreased by 6%,24%and 24%,respectively.We conclude that the IDW-CatBoost method fused with multi-sensor data is suitable to reconstruct accurate AOD products for large areas.
作者
张晨
汪小钦
邬群勇
郑汉捷
王涵菁
尹延中
ZHANG Chen;WANG Xiaoqin;WU Qunyong;ZHENG Hanjie;WANG Hanjing;YIN Yanzhong(Key Laboratory of Spatial Data Mining&Information Sharing of MOE,Fuzhou University,Fuzhou 350108;The Academy of Digital China(Fujian),Fuzhou University,Fuzhou 350108;National&Local Joint Engineering Research of Satellite Geospatial Information Technology,Fuzhou 350108)
出处
《环境科学学报》
CAS
CSCD
北大核心
2023年第5期353-365,共13页
Acta Scientiae Circumstantiae
基金
国家自然科学基金项目(No.41471333)
福建省科技计划引导项目(No.2021H0036)。
作者简介
张晨(1998—),男,E-mail:cheouing@163.com;责任作者:汪小钦,E-mail:wangxq@fzu.edu.cn。
