In the Loess Plateau region, where climatic conditions are dry with severe water shortage, agricultural development is limited by water availability. Several studies have shown that due to high water consumption and poor land management, forest trees induce soil dry up and ecological degradation in Loess Plateau region. Pear jujube (Ziziphus jujuba Mill.) is one of the common tree species cultivated in this region due to its high drought tolerance and considerable economic benefits, but there are limited studies about its water consumptions. Sustainable cultivation of vast jujube plantations could prevent these ecological issues in the fragile ecological environment of Loess Plateau. Evapotranspiration (ET), which is the sum of plant transpiration and soil evaporation, is a continuous process of water loss in the soil plant atmosphere continuum. An accurate prediction of ET is necessary for developing agricultural management strategies to improve water use efficiency and ensure sustainable agricultural production. Therefore, the aim of this study on the simulation model of jujube ET would be significant. In addition, the single layer Penman-Monteith (PM) model is the most commonly used and the one recommended by FAO. However, the model is mostly used for dense canopy vegetation types as the “big leaf” assumption cannot sufficiently differentiate between vegtation and bare soil. As a measure of PM model weakness, Shuttleworth and Wallace developed a more complex two layer interactive model called the Shuttleworth-Wallace (SW) model that allows interaction between canopy and bare soil flue. The two layer SW model shows a superior performance over single layer models. In the Loess Plateau, due to water shortage, jujube forests are usually cultivated sparsely with low canopy cover, so it is necessary to select a suitable ET model for pear jujube forest. Based on the forest stand and climate condition, we parameterized the SW and PM models by using the data measured in 2012 and fulfilled the estimation of pear jujube forest ET in the fragile, semi arid, hilly Loess Plateau region of Mengcha Village in Shaanxi Province, China. Then the estimated ET by SW and PM models were validated and compared through ET computed by water balance theory, while the estimated transpiration by SW model was examined by observation from thermal dissipation probes (TDP) during 2013. The results showed that over the whole growing season, SW model performed better than PM model in ET estimation. Their accuracy was the poorest in the germination and leaf expansion stage and changed following the growth seasons. The accuracy of SW model in ET estimation met the evaluation criteria but PM model did not. The accuracy of SW model in transpiration estimation met the evaluation criteria, but it was affected by the weather condition, specifically, sunny days were better than rainy days. In the stand of sparsely cultivated jujube forest, the reliability of SW was comprehensively validated. Therefore, SW model could be the promising model applied in ET estimation to grasp the law of water consumption for jujube forest and similar stand in the semi arid Loess Plateau.