Numerous modeling approaches have been proposed to estimate evapotranspiration (ET) and its partitioning between evaporation from soil (E) and transpiration from vegetation (T) over the last several decades. Although these ET models claimed to give reasonable E and T partitioning, few studies have compared their modeling results with direct E and T observations. In this study, a hybrid dual source scheme and trapezoid framework based evapotranspiration model (HTEM) fed with MODIS data was applied in a Chihuahuan Desert shrubland during the growing season of 2003 and validated with direct ET measurement using the Bowen-ratio technique and T measurement using scaled-up sap-flow measurements. Results show that the HTEM is capable of decomposing the remotely sensed land surface temperature into temperature components (soil and canopy temperatures) and providing accurate E and T estimates. At satellite overpass time, the root-mean-square error (RMSE) of estimated latent heat flux (LE) is . The agreement between estimated and simulated LE was largely improved when observed net radiation and ground heat flux were used (). At daily scale, the RMSE of estimated daily ET, E, and T are 0.52, 0.36, and , respectively.