The largest area of desert riparian forest is distributed in the Tarim Basin, northwest China. It controls the structure and function of the desert ecosystem; determines landscape patterns, vegetation processes, and land desertification; and changes the ecological environment of oases in southern Xinjiang. Populus euphratica Oliv. is the dominant species and is distributed widely across desert riparian forests. It protects biodiversity and desert ecosystem stability, and serves as a natural barrier to curb desertification and sandstorms. Since the 1950s, land has been reclaimed and the water resources have been utilized in the area. This has resulted in a reduction in the groundwater level, weak vegetation growth, and high desert plant mortality, which has affected the sustainable development of agriculture and the economies of the affected oases. The aim of this study was to understand the ecological responses of P. euphratica communities to changes in groundwater levels (GWL), to determine the rational GWL needed for the survival of desert riparian forest, and to provide a theoretical basis for population conservation and vegetation restoration in the Tarim Basin. Nine belt transects and 45 plots were established to monitor the GWL and vegetation in the upper reaches of the Tarim River and the ecological responses of the P. euphratica community to the GWL. The relationship between the vegetation and the GWL gradient were analyzed using multi-analysis methods, and Hill and β-diversity index analyses. The Ward cluster analysis divided the 45 sampling plots into three types and the quantitative characteristics (density, DBH, canopy diameter, coverage, mortality) of the P. euphratica population and the GWL gradients were significantly correlated (P < 0.01). The P. euphratica growth vigor decreased and the population structure transitioned from expanding to declining as the GWL decreased. Species diversity varied considerably in the three different plot types as the GWL decreased. In particular, when the GWL ranged from 4 to 6 m, the Hill-diversity index decreased sharply, and the diversity curve became a smooth line. Species diversity began to decline at GWL > 4 m. Furthermore, the Cody (β_c) and Whittaker (β_w) values increased and the Sorensen (β_s) value decreased as the GWL fell. The β_s, β_c, and β_w indexes, and the differences in groundwater levels were significantly correlated (P < 0.01). At GWL > 4 m, the β_c and β_w values increased, but the β_s value decreased considerably. The above analyses suggested that the P. euphratica communities had similar structures and many common species, and that the P. euphratica population grew most efficiently at GWL < 4 m, which was the most suitable ecological water table for P. euphratica population survival. At GWL > 4 m, species diversity and the common species in the community decreased rapidly, differences in species components and environmental heterogeneity increased, the quantitative characteristics of the P. euphratica population varied significantly, and the community structure became simplified. When the GWL was about 6 m, vegetation became degraded and the dominant population declined considerably. Therefore, the rational ecological GWL for restoring degraded desert vegetation is about 4 m in extremely arid areas.