Land Surface Emissivity（LSE） is a key parameter for determining land surface temperature because it depends on the features of the surface composition, is related to surface roughness, dielectric constant, and water content, and changes with vegetation fraction. Auxiliary data are used to calculate emissivity through single-channel algorithms, e.g., classification-based, NDVI threshold, and vegetation cover methods. However, these techniques present several limitations. For example, the classification-based method LSE is insensitive to land cover change on barren surfaces.Moreover, MODIS C5 LST products underestimate LST in an arid area of Northwest China because of the over estimate of LSE. Therefore, the precision of land surface emissivity in arid and semi-arid areas must be improved before the retrieval of land surface temperature. In this study, we developed a method for improving the accuracy of land surface emissivity for barren surface by using the latest ASTER Global Emissivity Database（GED） and Vegetation Cover Method（VCM）. ASTER GED was used to determine bare soil emissivity. The spectra of soil in ASTER spectral library were selected to fill small gaps of bare soil emissivity according to different land cover types. The vegetation emissivity of ASTER spectral library and Fractional Vegetation Cover（FVC） product were used in VCM method estimate land surface emissivity. This method can not only maintain the high precision of ASTER GED for barren surface but can also effectively characterize seasonal characteristics of vegetation emissivity over time; hence, the proposed technique an provide reliable input data for retrieval of land surface temperature. This method was evaluated with 11 scenes of ASTER land surface emissivity products of Zhangye region in 2012 and in situe missivity measurements. Results show that the biases of bands 10-12 are within 0.015 and RMSEs are less than 0.03.The biases of bands 13 and 14 are less than 0.005 and RMSEs are less than 0.01 compared with those in ASTER LSE products. The retrieved emissivity is close to the in situ measured emissivity with minor errors. One scene of Landsat 8 TIRS data was also used to analyze the influence of the proposed method on the accuracy of land surface temperature retrieval. The results show that the method can obtain more reasonable and accurate land surface emissivity than NDVI threshold method, especially for barren surface. Hence, the method can be used to produce high-precision land surface temperature products for other thermal infrared sensors, e.g., Landsat 8 TIRS,HJ-1B IRS, and FY-3 MERSI. In this study, an improved method for estimating land surface emissivity from ASTER GED products and VCM was proposed to improve the accuracy of land surface emissivity for barren surfaces. This method was evaluated with ASTER LSE products and in situ emissivity measurements. One scene of Landsat 8 TIRS data was also used to analyze the influence of this method on the accuracy of land surface temperature retrieval. The results in Zhangye city show that the proposed method can obtain reasonable and accurate land surface emissivity. However, further analysis is required for other areas. Although ASTER GED is relatively stable on bare surfaces, Hulley et al.（2010） indicated that the soil moisture at 11 and 12 micron can increase LSE by up to 0.03 times. In this regard, the effects of soil moisture on land surface emissivity must be considered in future research.

ASTER GED;land surface termpera ture;land surface emissivity;fractional vegetation cover;temperature emissivity separation