Estimation Of Vegetation Coverage In Desert Steppe Based On Sentinel-2 MSI Multiple Endmembers

Fractional vegetation cover(FVC)is an important index to measure the growth status of vegetation in an area and has been widely used in the dynamic monitoring of vegetation and ecological environment evaluation.As Inner Mongolia Autonomous Region is an important ecological barrier in the north of China,there are a large number of desert steppe areas in the territory,so it is important to obtain the fractional vegetation cover of desert steppe area quickly and accurately for ecological environment monitoring and management.This paper took the desert steppe area in the central part of Siziwang Banner,Ulanqab City,Inner Mongolia Autonomous Region as the study area,and estimated the FVC based on Sentinel MSI images.The linear spectral mixture analysis(LSMA)and the multiple endmembers spectral mixture analysis(MESMA)were used to estimate the FVC of the study area,and the accuracy of the model estimation results was verified by the measured vegetation cover in the field.The main research contents and the conclusions of this paper are as follows:After the preprocessing of the Sentinel-2 MSI images of the study area,the MNF transform was applied to them,and then the pure pixel index was calculated.The spectral curves of the three Endmembers in the study area were extracted based on the results of the pure pixel index calculation results.Then the FVC of the study area was estimated by using LSMA,and the estimated results were correlated to the measured values with determination coefficient R~2=0.28,which indicated that they were correlated,but the correlation was weak,and this method has not been suitable for the estimation of fractional vegetation cover in desert steppe area.230 pure pixels were selected on the remote sensing images of the study area based on the result of pure pixel index calculation and combining the spectral curves for various types of ground features and Google Earth high spatial resolution images of the study area,and 75optimal end members were selected using the Count-based Endmember Selection,Endmember Average RMSE,Minimum Average Spectral Angle,respectively.The correlation analysis between the estimated results and the measured values showed that the determination coefficient R~2=0.63,which indicates that the MESMA is more suitable for the estimation of FVC in desert steppe than the LSMA.Using the apparent reflectance of the study area as the data source,the FVC estimation was performed using MESMA,and the estimated results were compared with the surface reflectance estimation results,and which showed great differences.This indicates that the atmospheric disturbance can greatly affect the vegetation cover estimation results of the MESMA without atmospheric correction.