Extremely arid desert occupies a terminal position in the biological geochemical circulation system. These deserts harbor huge amounts of energy and mineral resources. However, a large number of ancient sites, which are very important to the modern tourism industry, also occur in such areas. Lack of water constitutes the most serious kind of land deterioration in the extremely arid deserts of the world. At present, desert land degradation, global climate warming, and human interference are intensifying the destruction of the natural desert landscape, and hence our historical and cultural heritages are facing unprecedented losses. Consequently, methods for finding available water sources for ecological restoration and protection are urgently required. It has been found that there exists a stable amount of water evaporation in the extremely arid Gobi area of the Dunhuang in China. In order to clarify the source of this evaporation, we used a greenhouse to condense and collect evaporated water and monitored the δD and δ¹⁸O values present on a weekly basis. At the same time, we monitored the δD and δ¹⁸O values of the precipitation and phreatic water present. Thus, the source of the evaporation water in the Gobi could be revealed based on the method of isotopic tracing. After monitoring for one year, our results showed that the average values of δD and δ¹⁸O in the evaporation water were -33.06‰ and -5.33‰, respectively. The equivalent results for the precipitation were found to be -66.44‰ and -8.57‰, respectively, and for phreatic water they were -72.19‰ and -9.75‰. This rules out the possibility that precipitation is the source of the phreatic water. Monitoring of the surface water in the soil at a vertical depth of 4.5 m showed that the δD and δ¹⁸O values (the average values were -22.68‰ and 8.58‰, respectively) of the soil water from 2.5 m increases due to the effect of the dry climate and reach the maximum values at 1.2 m. Above this level, the values of δD and δ¹⁸O decrease because of the influence of the precipitation. The soil moisture evaporation experiments indicated that the moisture from the dry soil shows obvious signs of an isotope fractionation phenomenon:the δD and δ¹⁸O values of the evaporation water are relatively negative, and the δ values of the remaining water are relatively positive. Therefore, the Gobi soil in the deep vadose zone selects relatively positive δ values of water in the process of moisture migration. The δD and δ¹⁸O values of the local precipitation calculated from the online precipitation isotopes are -60.00‰ and -8.50‰ and the weighted average values of the precipitation are -5.3‰ and -0.75‰. This also shows that local precipitation cannot be the source of the underground phreatic water. Values for the precipitation in the Yema Mountain region (-86‰ and -12‰), which is the water source area for the Danghe River, suggest this region is a probable source for the underground phreatic water and Gobi evaporation water, and thus there exists a clear supply channel. Phreatic water is the main source of soil water in extremely arid deserts and has a vital role in maintaining ecosystem survival in extremely arid areas. Determining the source of evaporation water in extremely arid areas lays the foundations for utilizing the phreatic evaporation. It not only has important significance in the ecological recovery of extremely arid regions but also provides a new perspective for using groundwater in arid or semi-arid areas. Moreover, it provides an important reference for research on water sources in the Mogao Grottoes.