Abstract: Abstract: Accurate calculation of land surface evapotranspiration is meaningful for the rational utilization of water resources. Penman-Monteith (PM) theory is a classic method to calculate evapotranspiration (ET) of land surfaces. Although Mu et al. (2011) improved ET estimates in Mu et al (2007) by adding nighttime ET and wet canopy surface ET component, it is difficult to acquire the 1km resolution spatial distribution of daily minimal air temperature (Tmin) for a regional scale, which is used as input variable (25km resolution) in Mu et al. (2007, 2011)'s global ET algorithms. Yuan et al. (2010)'s ET algorithm is modified from Mu et al (2007), they set invariant model parameters across the various vegetation types and Tmin is not used, therefore it is suitable for regional application. Soil resistance is largely controlled by the soil moisture. However, directly monitoring large scale soil moisture is always a challenge for remote sensing. In this study, we developed an ET estimation algorithm by incorporating a soil moisture index (SMI) derived from the improved surface temperature-vegetation cover feature space, denoted as the PM-SMI algorithm. The PM-SMI algorithm was compared with the triangle ET algorithm and another Penman-Monteith based algorithm (PM-Yuan) that calculated soil evaporation using relative humidity. Three ET algorithms were compared and validated by Bowen Ratio measurements at 12 sites in the Southern Great Plain (SGP) that were mainly covered by grassland and cropland with low vegetation cover. For instantaneous latent heat flux, although R2 of PM-SMI was lower than that of triangle and PM-Yuan for some sites (EF2, EF4 and EF12), the RMSE and bias was the lowest across almost all the sites. PM-Yuan obviously underestimated LE with bias of -82.41 W/m2. Triangle overestimated LE with 48.2 W/m2. PM-SMI algorithm showed the best performance with RMSE, bias and R2 of 53.67, 6.83 and 0.86 respectively. For daily latent heat flux, the bias of triangle was lower than PM-SMI for some sites, however, the RMSE was greater and R2 was lower across almost all sites. Similarity, the R2 of the PM-Yuan algorithm was greater for some sites, however the RMSE and bias was greater across almost all sites. Overall, the PM-SMI performed the best, with the highest R2 (0.87) and the lowest RMSE (39.07 W/m2) and bias (-4.04 W/m2). The PM-Yuan algorithm significantly underestimated LE. The results showed that the PM-SMI algorithm performed the best among the three ET algorithms both on the instantaneous scale and the daily scale. PM-SMI is more reliable for estimation of ET over regional scale.