**4. Results**

*4.1. Long-Term Changes in Lake Surface Area*

Lakes generally have seasonal variations, which possibly affect the interannual trend analysis of change in the lake surface area. Considering this effect of seasonal variations, the monthly variations of the Tuosu Lake surface area in 2003 and 2020 were extracted (Figure 2). The results show that the lake surface areas of Tuosu Lake in 2003 and 2020 were 128.2–136.1 km<sup>2</sup> and 174.0–178.7 km<sup>2</sup> with uncertainties within 1.6% and 1.2%, respectively. This indicates that the extracted lake surface areas are reliable. The maximum differences of lake surface area caused by seasonal variation are 7.9 km<sup>2</sup> and 4.7 km<sup>2</sup> in 2003 and 2020, respectively, and the relative deviation is within 4% and 2%, respectively. Therefore, the seasonal variation has only a slight effect on the interannual trend of the lake surface area.

Changes in the surface areas of Tuosu, Keluke, and Gahai Lakes from 2000 to 2020, as well as changes in lake water storage calculated from the lake surface area and slope (*I* = 2, 3, and 4), are shown in Figure 3a,b, respectively. The average surface areas of Keluke, Tuosu, and Gahai Lakes were 54.2, 146.5, and 33.1 km2, with uncertainties of within 1.3%, 1.7%, and 1.6%, respectively. Keluke Lake remained stable because it is connected to Tuosu Lake through the Lianshui River. When the water level of Keluke Lake rises, excess water flows into Tuosu Lake through the river. Therefore, changes in lake surface area mainly occurred in Tuosu Lake, with three distinct trends: continuous shrinkage (before 2003), slow expansion (2003–2011), and rapid expansion (after 2011). Before 2003, Tuosu Lake shrank at a rate of 2.50 km2/year and then gradually increased at a rate of 1.54 km2/year from 2003 to 2011, with an average increase in lake water storage of 1.45 × 10<sup>8</sup> m3/year (Figure 3b). After 2011, Tuosu Lake expanded rapidly at a rate of up to 3.38 km2/year, which was much higher than the overall increase of 2.19 km2/year (R<sup>2</sup> = 0.88) after 2003. At this time, the average increase in lake water storage was as high as 5.75 × 10<sup>8</sup> m3/year (*I* = 3). Gahai Lake shrank gradually prior to 2000 and then expanded steadily after 2000 at a rate of 0.45 km2/year (R<sup>2</sup> = 0.96), with an average increase in lake water storage of 0.29 × 10<sup>8</sup> m3/year. Although detailed lakeshore slopes were not obtained in the calculation of lake water storage changes, the gentle lakeshore zone allowed changes in water storage to be largely reflected in the lake surface area rather than the lake height, which supports our subsequent analysis of the water budget based on surface area changes.

Similar lake expansion occurred at the edge of the Tibetan Plateau. Taitema Lake in the north of the Altun Mountains reappeared in 2003 after prolonged drying over many years, and then expanded rapidly [43]. Similarly, Qinghai Lake in the northeast margin of the Tibetan Plateau expanded rapidly at a rate of 8.67 km2/year after 2003 (https://hydroweb.theia-land.fr/hydroweb, accessed on 27 November 2021), with a simultaneous increase in groundwater level in the Hexi corridor in the north of the Qilian Mountains [44], which may imply similar lake response patterns.

### *4.2. Lake Evaporation Calculated by the Improved Penman–Monteith Model*

Lake expansion is a direct reflection of the water budget of a lake. Regional groundwater head distribution determines groundwater inflow and groundwater outflow in the water balance of the endorheic lake. The groundwater level in the study area gradually decreases from the mountains to the plains, and the groundwater flows from the mountains to the lakes and eventually contributes to the lakes. For endorheic lakes, the input components of the lake water budget mainly include precipitation, runoff, and groundwater inflow, while the output component is mainly lake evaporation, and groundwater outflow can be neglected. In this study, lake evaporation was quantified by using the improved Penman–Monteith model and used to analyze the influence of the major output component (lake evaporation) on lake expansion in the study area. Based on the monitoring data of Delingha station, evaporation values for Tuosu, Keluke, and Gahai Lakes were 1233–1476 mm/year, 1164–1379 mm/year, and 1407–1700 mm/year, with average values of 1342, 1274, and 1542 mm/year, respectively (Figure 4). These results show that evaporation varied significantly between lakes with different lake surface areas and depths. All three lakes exhibited stable interannual evaporation values, with insignificant variation trends and relative deviations of 9.9%, 9.6%, and 10.2%, respectively. Therefore, evaporation did not cause significant changes in lake surface area and cannot explain the rapid expansion of the lakes since 2003. This suggests that climate change processes such as increased temperature and precipitation did not significantly affect lake evaporation and were not the main factors affecting the lake water budget during the study period.

**Figure 4.** Plots exhibiting the evaporation of Tuosu (**a**), Keluke (**b**), and Gahai Lakes (**c**) calculated by the improved Penman–Monteith model.

### *4.3. Annual Hydrometeorological Trends*

Considering the lack of significant changes in the main output component of the lake water budget (lake evaporation), rapid lake expansion may instead have been caused by the input components, which include precipitation, runoff, and groundwater inflow. Rainfall and runoff data were derived from long-term monitoring at Delingha meteorological and hydrological stations. Because of the complexity and hidden nature of the groundwater runoff process, the groundwater contribution to lakes is difficult to directly quantify and observe long-term. Therefore, we first analyzed the long-term variation trends of rainfall and runoff.

The Mann–Kendall test was used to inspect the long-term trends of hydrometeorological data and identify abrupt changes in the time-series data. An abrupt change in the precipitation time-series occurred in 1989 (α = 0.05), with average precipitation before and after this change equal to 156 mm/year and 215 mm/year, respectively, representing an increase of more than 37.7%. However, this abrupt change in precipitation occurred much earlier than the beginning of lake expansion in 2003. This indicates that, although precipitation increased annually, it was not the dominant factor influencing lake expansion. Additionally, precipitation in the plain area is only 50 mm/year; thus, its contribution to the lake water budget can be neglected (Comprehensive Investigation Committee of Chinese Academy of Sciences, 1984).

Conversely, there was no obvious abrupt change in the runoff measured at Delingha station, although there was a significant difference in average runoff before and after 2002 (3.19 × 10<sup>8</sup> m3/year and 4.55 × 10<sup>8</sup> m3/year, respectively), representing a difference of 42.6% (Figure 5b). Runoff from Zelinggou station represents glacier meltwater, whereas Delingha station is located at the boundary between the upper and middle reaches of the Bayin River; thus, runoff is derived from both glacier meltwater and precipitation in mountainous areas. A comparison of the runoff values between the two stations during 1957–1983 revealed strong similarity, with a correlation coefficient of up to 0.88 (Figure 6). Runoff at Zelinggou station accounts for 90% of that at Delingha station, which indicates

that glacial meltwater is the main source of the Bayin River. The rate of glacial melt is controlled by the average air temperature, which changed significantly (α = 0.05) in 1997, from 4.1 ◦C to 5.1 ◦C, representing an increase of 23.3% (Figure 5d). This increase in air temperature likely accelerated glacier meltwater.

**Figure 5.** Plots depicting the change of runoff at Delingha and Zelinggou stations (**a**), results of MK test for runoff (**b**), precipitation (**c**), and temperature (**d**) at Delingha station.

**Figure 6.** Graph representing the comparison between runoff at Delingha and Zelinggou stations from 1957 to 1983.

#### *4.4. Isotopic Characteristics of Surface Water, Groundwater, and Spring Water*

To trace groundwater processes and sources, the stable isotopes (2H, 18O) of collected water samples were analyzed (Figure 7 and Table 2). Samples R12, R09, and R07 represent glacial meltwater, as discussed earlier. The δ2H and δ18O of glacial meltwater, phreatic groundwater, and river water lie in the range of −57.1 to −56.7‰ and −8.8 to −6.6‰, −65.4 to −42.0‰ and −9.9 to −8.3‰, and −60.1 to −50.8‰ and −9.4 to −6.6‰, respectively. The points of river water were distributed along with the least square fitting line in Figure 7,

i.e., δ2H = 3.43 δ18O −27.63 (R<sup>2</sup> = 0.77), showing the evaporation characteristics of river water. Most of the river points fell within the range of phreatic groundwater. This implies a strong interaction between river water and phreatic groundwater and that the river was recharged by both phreatic groundwater and glacial meltwater. The fitting line of lake water was δ2H = 5.1883 δ18O − 12.599 (R<sup>2</sup> = 0.81), and its intersection (−54.7‰, −8.1‰) with the global meteoric water line (GWML) fell within the range of river water, suggesting that phreatic groundwater contributed to the lake after flowing into the river.

**Figure 7.** Plot illustrating δ18O vs. δ2H diagram of groundwater, lake water, river water, and spring water in the study area of the Qaidam Basin, China. Squares represent samples from the literature [11].

Zhang [45] also collected precipitation samples from Delingha from September 1991 to December 1992 and derived a weighted average of δ2H and δ18O in precipitation of −44.2‰ and −6.8‰, respectively. The δ2H and δ18O values of confined groundwater ranged from −71.1‰ to −58.2‰ and from −10.6‰ to −8.8‰, respectively. Thus, the confined groundwater was more depleted in deuterium and oxygen-18 than phreatic groundwater and local meteoric precipitation. The concentrations of tritium in river water, spring water, and confined water were 11.6–17.8 TU, 3.7–11.9 TU, and 1.3–15.6 TU, respectively.
