Multi-angular Polarized Overland Cloud Detection And Aerosol Retrieval Using Optimal Grouped Residual Method

Aerosol is an important component of Earth's atmospheric environment.It has profound influence on many fields,such as atmospheric transmission,climate simulation,environmental research,remote sensing application,pollution monitoring and etc.Aerosol particles are closely related to the process of cloud condensation and change.The interaction among aerosol,cloud,and radiation draws the interest of many researchers in the climate field.At present,there are two types of remote sensing methods of aerosol and cloud:ground-based remote sensing and satellite remote sensing.In the satellite remote sensing,the multi-angular and polarized method,has higher precision over traditional optical remote sensing method in the issue of ground-atmosphere decoupling.Therefore,it is widely applied and rapidly developing in the field of cloud and aerosol.Typical multi-angular polarized satellite payloads such as POLDER?Polarization and Directionality of Earth's Reflectance,France,1996²013?,MAPI?Multi-angular Polarization Imager,China,Tiangong-2,2016²019?and DPC?Directional Polarized Camera,China,Gaofen-5,2018-present?,have accumulated a large amount of data for the quantitative remote sensing study of cloud and aerosol.At present,aerosol retrieval algorithms generally adopt the strategy of single angle look-up-table method,which ignores some correlation of multi-angular data.The physical threshold method is often used for cloud detection,which is weak in utilizing the multi-angular polarized scattering characteristics.In addition,the research on cloud and aerosol of two domestic payloads,MAPI and DPC,is still in the initial stage,and the demand is very urgent.In this work,we study the overland aerosol retrieval and cloud detection based on multi-angular polarized satellite data using the OGRM?Optimal Group Residual Method?.This work includes the following four parts.Firstly,OGRM is proposed.According to the Mie scattering theory,the scattering characteristics of the single spherical particle are simulated.Then the scattering phase function and polarized scattering phase function of aerosol are calculated by combining with the size distribution function.The polarized reflectance of aerosol is obtained by the polarized scattering phase function.The polarized reflectance of atmospheric molecules and surface are calculated by the empirical formulae.Then,according to the atmospheric radiation transmission theory,the apparent multi-angular polarized reflectance of the top of the atmosphere based on the hypothesis of single scattering is simulated.The RMSE?root mean squared error?was below 10^-3,and the numerical simulation accuracy is acceptable.Secondly,the overland cloud detection results are obtained using the OGRM method and supplemented by reflectance criterion and NDVI?Normalized Differential Vegetation Index?criterion,based on the multi-angular polarized data of POLDER,MAPI and DPC.The unique information of multi-angular polarized data is introduced,and a new three-threshold cloud detection method is proposed.The accuracy of these results reached approximately 90%compared with the official cloud mask product of POLDER and the cloud product of MODIS?Moderate Resolution Imaging Spectrometer?.The results are also verified by the remote sensing image of MAPI and DPC.The results have practical value to be applied as a pretreating product of aerosol retrieval.Thirdly,the overland AOD?Aerosol Optical Depth?results are retrieved using the OGRM method,based on the POLDER and DPC data.The exact geographical matching and regression analysis are carried out between the retrieval results of POLDER and the reference dataset,MYD04?MODIS aerosol products?.The consistency is verified by the regression slopes that were close to 1,and R² that were larger than 0.69.Correlations were found by comparing the results with AERONET?AErosol RObotic NETwork?aerosol product at two stations in Beijing,China and Kanpur,India.The regression slopes of the results of DPC and MYD04 reference data were also close to 1,but R² were lower than that of POLDER results.Fourthly,in order to verify the reliability of the method in this study from a broader space-time dimension,the AOD products are synthesized in long time series.The Asia-Pacific region has a good representativeness due to the diverse types of surface and aerosol.Therefore,the temporal and spatial distribution of AOD in this region was analyzed and summarized.The results showed that the southeast of China and the north of India are two typical large AOD regions,which reached the average level of 0.4-0.8.The AOD in South Asia is larger in spring?February to May,>0.6?,and lower in other seasons?<0.3?.The average level of summer AOD in the whole Asia-Pacific region is the lowest?⁰.1-0.5?.The temporal and spatial distribution of AOD is in good agreement with the existing data.