Remote Sensing Retrieval Of Aerosol Optical Depth And Single Scattering Albedo In China

In recent years,the frequent occurrence of air pollution incidents in China with aerosols as the primary pollutant has attracted increasing attention from the public.Atmospheric aerosols can lead to a decrease in atmospheric visibility and adversely affect the ecological environment and human health.Moreover,they play an important role in the global climate change through direct and indirect radiative forcing.Aerosol optical depth(AOD)and single scattering albedo(SSA),which characterize the extinction and absorption ability of aerosols,respectively,are key optical parameters of aerosols.High-quality satellite AOD and SSA products can help reduce the uncertainty in the studies associated with climate change,and monitor the spatio-temporal distribution and regional transportation of aerosol pollutants.However,the existing AOD and SSA retrieval algorithms still suffer from inaccuracy and low efficiency of retrievals,which limit the further application of satellite aerosol remote sensing products.This study focuses on developing high-precision retrieval algorithms for AOD and SSA.Firstly,a stable and reliable AOD retrieval algorithm based on China’s polar-orbiting satellite FY-3D and Japan’s geostationary satellite Himawari-8 is developed.Secondly,the SSA retrieval algorithm is constructed based on MODIS satellite observations applicable to visible-NIR satellite observations.The main results of this study are as follows.(1)High-precision retrieval algorithm for aerosol optical depthA high-precision algorithm to retrieve AOD suitable for the entire land areas(except snow/ice and inland water)is developed.Considering both the accuracy and retrieval efficiency,a new cost function is constructed.The retrieval method based on the new cost function avoids iterative operations and achieves fast retrieval of AOD and dynamic determination of the aerosol type,and is nearly 50 times faster than most current methods using iterative calculations.An a priori surface bidirectional reflectance distribution function(BRDF)database is built to provide the surface reflectance in different directions to reduce the errors caused by the surface Lambertian hypothesis.Considering the characteristics of regional distribution of aerosols,eight candidate aerosol models in different natural zones of China are constructed based on AERONET aerosol products from 2010 to 2021.The new method is applied to FY-3D and Himawari-8 observations to retrieve AOD during 2020-2021 almost covering the land of China.The validation results based on the ground-based measurements show that MERSI-II AOD retrievals agree well with the ground-based AOD with correlation coefficient(R),root mean square error(RMSE),and relative mean bias(RMB)of 0.913,0.123,and 0.955,respectively.In addition,72.19 % of AOD matchups fall within the expected error(EE)envelopes.The accuracy of the Himawari-8 AHI AOD retrievals based on the new algorithm is significantly improved compared with its official AOD product,with 0.16 improvement in correlation coefficient,31.78% reduction in RMSE,and about 67% of AOD matchups within the EE envelopes.In addition,this study analyzes the application scenarios of geostationary satellite AOD products in monitoring extreme aerosol pollution events and shows that the AOD products retrieved by the new algorithm can effectively monitor the spatial distribution and regional transport of extreme aerosol pollutants.(2)Aerosol single scattering albedo retrieval algorithm based on basic aerosol componentsA new method is proposed to optimize the aerosol optical model based on four basic aerosol components,namely black carbon,brown carbon,ammonium sulfate-like,and mineral dust,to indirectly retrieve the aerosol single scattering albedo.Firstly,the accuracy and stability of the new method are demonstrated by the sensitivity analysis of SSA retrieval considering the uncertainties of apparent reflectance,AOD,and surface BRDF parameters;Then,the retrieval test based on AERONET aerosol products demonstrates that the basic aerosol components used in this algorithm are more representative compared with those used in the 6SV model;Finally,the new algorithm is applied to MODIS observations to retrieve the SSA at 470 and 550 nm wavelengths,and the results showed that the SSA retrieved by this algorithm has better agreement with AERONET SSA than the MODIS official SSA product,and the RMSE is less than 0.03.(3)Retrieval of aerosol SSA in combination with the machine learning methodA fast retrieval method of satellite SSA is proposed by combining the machine learning method and aerosol optical model constructed with mixing aerosol components.In this study,firstly,satellite apparent reflectances are simulated based on the randomly generated data of volume mixing ratio of four basic aerosol components,angles and AOD,and a simulation dataset for model training and validation is constructed;Then,machine learning models for SSA retrieval are constructed based on the theoretical basis of retrieval SSA,and the sensitivity of the accurate SSA retrieval to the input parameter errors is analyzed.The results show that BPNN based SSA retrieval models can replace the iterative optimal solution process to a certain extent and realize the fast retrieval of satellite SSA;Finally,the retrieval model for retrieving MODIS SSA is determined by comparison test based on real observations,and the spatially continuous and reasonably distributed satellite SSA products are retrieved based on the new method and MODIS observation data.The retrievals show good consistency with AERONET ground-based SSA products,with more than 70% of the samples within the error range of 0.03.