Abstract: Land surface temperature is an important indicator of analysing the earth’s environment, and land surface emissivity is a key factor on retrieved land surface temperature from remote sensing data. In this paper according to the measured land surface temperature, we analyzed the effect of different emissivity on the retrieved land surface temperature from remote sensing data. Firstly we calculated the land surface emissivity by using NDVI threshold method and empirical formula model respectively, and then adopted a generalized single-channel method to retrieve the land surface temperature. The Landsat5 TM6 band data acquired in March 24, 2009 in Dianchi Lake watershed was taken as research example. Results show that the emissivity calculated by both NDVI threshold method and empirical formula are in significant relation with MODIS validation data, and the retrieved surface temperature has small error compared with measured temperature, in the control range. But the surface temperature retrieved from the emissivity calculated by empirical formula is abnormal in urban areas, exceeding the observed temperature by 4 ℃ -5 ℃, which does not accurately reflect the surface temperature; while the surface temperature retrieved from the emissivity calculated by NDVI threshold method is nearer to the measured surface temperature, and reflects the temperature under different land covers more accurately. Therefore, for areas with good land cover such as Dianchi Lake watershed, it is more appropriate to retrieve land surface temperature from the emissivity calculated by NDVI threshold method.

Key words: land surface emissivity, land surface temperature, NDVI threshold method, empirical formula model, generalized single-channel algorithm, remote sensing, retrieving