Research On Retrieval Of Aerosol Optical Depth Usinghigh-Resolution Remote Sensing Image

Aerosol distributions may change at fine spatial scale in urban areas.However,the current aerosol products based on remote sensing are mostly low-resolution,which cannot reflect the finer spatial scale changes of aerosol distribution.High-resolution aerosol research can make up for the above shortcoming,so the research of aerosol retrieval based on high-resolution remote sensing image is of great significance to the aerosol research and air pollution monitoring in urban areas,and has important research value.In this paper,sentinel-2 high-resolution remote sensing images are used to perform aerosol optical depth(AOD)retrieval,in order to obtain high-resolution AOD products and meet the needs of fine spatial scale aerosol research and air quality monitoring in urban areas.According to the characteristics of sentinel-2 image,this paper proposes a new method to construct the surface reflectance database,and uses the AERONET data to determine the type of aerosol in the study area,on the basis of which a high-resolution AOD retrieval is realized.The main contents of this paper are summarized as follows:(1)Taking advantage of the short revisit period of sentinel-2,a surface reflectance estimation algorithm based on the “cleanest” image is proposed.By analyzing the data of long time series AERONET AOD,the "cleanest image" in a month's time window is determined,and the surface reflectance is obtained by accurate atmospheric correction.The surface reflectance images corresponding to the “cleanest” images constitute the surface reflectance database.In aerosol retrieval,the image to be retrieved and the surface reflectance image of the same month in the surface reflectance database and are used.(2)The aerosol model is an important factor affecting the accuracy of AOD retrieval.In this paper,by analyzing the aerosol volume distribution in Beijing,it is determined that the bimodal lognormal distribution is used to characterize the aerosol type in Beijing.On this basis,the specific parameters of the bimodal lognormal distribution of each month are obtained by statistical analysis of the three-year aerosol optical parameters of AERONET.Thus,the aerosol model of each month in Beijing are determined,which is helpful to improve the accuracy of AOD retrieval.(3)Based on the above-mentioned sentinel-2 surface reflectance database and aerosol type determination method,combined with the AOD retrieval method based on look-up table,the AOD retrieval in Beijing area is carried out,and the results are analyzed from the spatial distribution and retrieval accuracy.Firstly,the spatial distribution of AOD retrieval results is analyzed and compared with PM2.5 distribution map.The results show that AOD products obtained in this paper can well reflect the fine spatial scale changes of air quality distribution.Then the AOD retrieval results of this algorithm are verified by AERONET AOD data,and compared with that of dense dark vegetation algorithm.The results show that the Sentinel-2 AOD retrievals are highly consistent with the AERONET AOD measurements(R=0.9424),with 85.56% of them falling within the Expected Error(EE),and the retrieval effect is better than that of dense dark vegetation algorithm.(4)Based on the above-mentioned sentinel-2 surface reflectance database and AOD retrieval method,the AOD retrieval in Beijing area is carried out using Landsat8 image,and the results are verified by AOD data of AERONET.The results show that the correlation coefficient between Landsat 8 retrieval result and AOD of AERONET is 0.69,but it is significant,and 82.35% of the inversion results are falling within the Expected Error(EE).