The Moderate Resolution Imaging Spectroradiometer (MODIS) data has a high temporal resolution, which, at present, is an ideal data source in simulative monitoring of regional-scale changes in surface energy and water. However, the spatial resolution of its thermal infrared band is relatively low (1 km). The Landsat TM/ETM+ data have a high spatial resolution, but their single thermal infrared bands can lead to the fact that the inversion accuracy for the surface temperature is not high, and that the time resolution is low. This limits its application in the surface evapotranspiration (ET) monitoring. Combining TM/ETM + visible wave band with MODIS thermal infrared wave band, this paper discusses a multi-scale remote sensing method to estimate regional surface ET. On the basis of space enhancement method, the vegetation index estimated by TM/ETM + enhances the surface temperature scale with the inversion of MODIS to a 30 m resolution, which aims to improve the estimation accuracy of ET in the non-uniform surface mixed-pixel. The results show that this method has a higher accuracy of ET estimation compared with the method of only using MODIS or ETM+ data. Moreover, it can obtain a more obvious effect on scale correction in the uneven land surface or various surface covering types, and the corrected ET is close to the observation result.