The surface energy balance algorithm for land (SEBAL) model is one of the most widely used methods for estimating evapotranspiration. Numerous physical studies have conducted to mitigate the limitations of spatial-domain dependence and the annoying uncertainties in visually identifying the dry/wet pixels in SEBAL; however, they are subject to the quality of wind speed ( u) observation, which is known as a high temporal-spatial variation and is not routinely available, especially in a heterogeneous area. In this study, we constructed a u-independent algorithm to calculate the dry/wet endpoints for each pixel. The wet endpoint was determined by assuming no turbulent heat exchange between a water-saturated surface and atmosphere. The dry endpoint was calculated using the assumption and consensus that pixels had equivalent neutral aerodynamic resistance under given atmospheric forcing and vegetation coverage. Then, we built a temperature-domain SEBAL (TD-SEBAL) model and validated it in the MUSOEXE between May and September 2012, located at desert-oasis transition zone in the middle reaches of the Heihe watershed across eight landscapes. The results showed that TD-SEBAL could provide reasonable dry/wet endpoint retrieval, and reliable latent heat flux estimates with a mean bias of -0.25 W/m², a root-mean-square error of 52.9 W/m², and a coefficient of determination of 0.91.