Aerosol Retrieval Algorithm On Bright Surface For MERSI Onboard Fengyun-3D

The Medium-Resolution Spectral Imager（MERSI）sensor mounted on China’s Fengyun-3 series satellite is the same type of sensor as MODIS（Moderate-Resolution Imaging Spectroradiometer）of NASA（National Aeronautics and Space Administration）.Based on the idea of dark target algorithm,the author’s team developed an improved MERSI-Ⅱland aerosol inversion algorithm for Fengyun-3D satellite,and the validation accuracy of retrieval result is close to that of MODIS aerosol products.Algorithmic inversion is currently limited to dark surfaces with low reflectivity,however,there are no retrieval results for bright surfaces with high reflectivity（such as bare land,desert and other arid and semi-arid areas）.To this end,based on the validation and evaluation of the product results of the two existing MODIS deep blue（DB）and multiple angle implication of Atmospheric Correction（MAIAC）algorithms with retrieval capabilities over bright surface,a new surface prior knowledge based on band reflectance ratio was used to construct a MERSI aerosol inversion algorithm for bright surface based on Dark Target（DT）algorithm,and the bright surface in Northwestern China was taken as the study area.The main research work and conclusions of this paper are as follows:（1）Validation of existing MODIS aerosol products in Northwestern China.Based on observations from 23 AERONET and CARSNET sites,MODIS DB and MAIAC AOD products over the bright surface of Northwestern China from 2002 to 2014 were systematically validated and evaluated in a long-term and systematic manner.The results show that the retrieval accuracy of DB and MAIAC Aerosol Optical Depth（AOD）in Northwestern China is generally low,the correlation coefficient R is about 0.8,and the proportion of verification matching points falling within the expected error（EE=±（0.05+0.2））range is less than 54%.The proportion below EE is lower than 56%,and the mean deviations of DB and MAIAC(MB_DB:-0.148,MB_MAIAC:-0.095)are negative,indicating systematic underestimation.From the perspective of a single site,validation results at almost all site are significantly underestimated.（2）Construction of bright surface estimation and aerosol inversion algorithm in Northwestern China.The fixed linear ratio relation of DT algorithm for dark targets is no longer applicable to bright surface,and it can be seen from the above validation results that MODIS DB and MAIAC algorithm cannot be directly transplanted to MERSI.Therefore,this paper establishes a new surface estimation method for bright surface.Based on MODIS surface albedo product MCD43 data,a surface reflectance library with a resolution of 10 km×10km is constructed through monthly synthesis.Then,the reflectance ratio of blue/red and red/short-wave near-infrared bands can be taken as prior knowledge,and the reflectance of blue and red bands can be calculated by using MERSI’s2.1μm band reflectance.Then,the DT algorithm is extended from the dark surface to the bright surface,and a MERSI aerosol inversion algorithm over the bright surface is constructed.（3）Validation and evaluation of MERSI aerosol retrieval results.Based on the MERSI data in 2019 and 2020,the MERSI AOD inversion results with a resolution of10km×10km over the bright surface in northwest China are obtained.Compare and verify them,and the results show that,overall,compared with MODIS DB and MAIAC AOD,the MERSI retrieval result has a higher correlation coefficient of 0.693（while DB is 0.559,MAIAC is 0.581）and a lower mean deviation of 0.099（while DB is 0.354,MAIAC is0.333）;the proportion of validation matching points falling within the EE is 59.97%,slightly higher than DB（56.11%）,and slightly lower than MAIAC（75.41%）.The fitting slope of MERSI validation scatter plot is 0.927,which is closer to 1,significantly better than MODIS DB（0.298）and MAIAC（0.286）.From the analysis of satellite AOD retrieval deviations under different aerosol depth,MODIS DB and MAIAC AOD have obvious negative deviations when the aerosol depth increases,but MERSI AOD has positive deviations.From the analysis of AOD monthly mean,the distribution of the three products was similar.The AOD value was generally lower in Inner Mongolia,Ningxia and Gansu,while the high value was mainly distributed in the Taklimakan Desert.Furthermore,MERSI was higher than MODIS DB and MAIAC.In addition,MERSI AOD is overestimated in individual sites,and the algorithm needs to be further improved.In summary,this paper improves the DT algorithm by constructing new surface prior knowledge,and realizes the retrieval of the aerosol optical depth over the bright surface of MERSI in Northwestern China,and the retrieval results are slightly better than MODIS DB and MAIAC products.The results of this study are expected to provide a reference for the improvement of aerosol inversion algorithm and product development of MERSI and even similar sensors over the bright surface.There are 27 figures,7 tables and 90 references.