Water is one of the most important factors affecting vegetation distribution in terrestrial ecosystems, especially in arid region, water will be the key factor to restrict growth of plant speies. Plant species have different availability to absorb water from different sources. Water sources of plant species directly determined the patters of species distribution. Therefore, quantifying water use of dominant species from certain sources are critical important to define and predict spatio-temporal distribution patterns of vegetation under global changes. Stable isotopes of oxygen and hydrogen were considered to be a good tools and have been used to identify the source of water, because different sources of water possess different oxygen or hydrogen isotope signatures. Until now, several approaches have been developed to quantify the contribution of different water sources to plants based on the isotopic mass-balance principal. In this study, we used four different models to estimate the sources of water used by Nitraria Tangutorum, the dominant species in an alpine desert steppe in Golmud on the Tibetan Plateau. Furthermore, the advantages and disadvantages of the four models were compared. The four models were as follows: a three-compartment linear mixing model, a multiple sources linear mixing model, a model of mean depth of water uptake by plants and a dynamic model. Results showed that N.tangutorum mainly used the below-ground water which concentrated in the 50 to 100 cm depth (mostly from the depths of 70 to 100 cm). The four models we used in this study have their own range of application and limitations. The three-compartment linear mixing model was used only in the situation that only less than three water sources were available. If more than three water sources were available for plants, the water sources should be reduced within three sources by removing those with less contribution to plant growth. The multiple sources mixed-linear model can remedy the defects of the three compartment linear mixing model. Therefore it can be used to quantify more than three water sources as well as the range of each source. However, either of the two models can not be used to calculate the mean depth of the water from which N. tangutorum can take up. However, the model of mean depth of water can solve this problem. For the dynamic model, control experiments are often needed. This study ultimately failed to use this model because of the rainfall time and the precipitation. Nonetheless, control experiment and the dynamic model should play a great role in predicting spatio-temporal distribution of dominant plant species under future precipitation patterns in this region. By comparison, the multiple sources mixed-linear model and the model of mean depth of water can work very well only with hydrogen or oxygen isotope. The result obtained by the oxygen isotope should be more precise, because hydrogen isotope fractionation should be more easily influenced by environmental factors. In view of the different application scopes, selection of models should be crucial and need to be pay more attention. Stable isotope method is a promising approach, although there are some defects in distinguishing the source of water of the plant species using stable isotopic hydrogen and oxygen. It is suggested that stable isotopic measurement on hydrogen and oxygen should be in combination with measurement on soil water potential and deuterium in different water components.