Forty-eight lake water sample sites were established around Bosten Lake. Sixteen surface water sample sites were established along the Kaidu River, and seven surface water sample sites were established along the Huangshuigou River. Eleven groundwater sample sites were established, including seven groundwater sample sites along the rivers and four groundwater sample sites near the lake. Groundwater samples were collected from groundwater wells that were more than 20 m depth. One ice-snow-melt sample site was located in a snow pack of an alpine mountain in Bayinbuluke, and one agricultural and industrial wastewater sample site was placed close to the main sewage outlets to the Bosten Lake. Water samples were collected from these sample sites in January, March, June, July, August, October and November 2011. Precipitation samples were obtained from (by event) the Bayinbuluke weather station (in the alpine mountain) and the Yanqi weather station (close to Bosten Lake) from January to October 2011. The sampling site locations are illustrated in Figure 1.

Once the samples were collected in dry and clear 5 ml glass blue bottles, they were stored in a refrigerator (4˚C) as soon as possible until analysis. In addition, we collected another sample at each sample site to analyze total dissolved solids (TDS) by weighing.

Oxygen-18 and deuterium in the water samples were measured at the State Key Laboratory of Desert and Oasis Ecology, Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences using a laser absorption water isotope spectrometer analyzer (Model DLT-100, Los Gatos Research Inc.). Isotopic concentrations were expressed as the difference between the measured ratios of the sample and compared with the measured reference ratio of standard mean ocean water (SMOW), expressed using δ notation [13] :

The analytical errors for the entire procedure were ±0.1‰ and 0.3‰ for δ¹⁸O and δD, respectively.

For hydrograph separation, a widely used multi-component method, we used the following the analysis equation [10] :

where Q_t is the total lake water discharge; Q_m is the discharge of component m; and C_jm is the tracer j incorporated into the component m. C_jm should be constant during the calculation period, such as a rainfall-runoff process, and this value should be different between each component. A steady-state model adequately represents the watershed conditions.

In this study, δ¹⁸O and TDS were used as the tracers for the isotopic hydrograph separation. The runoff from the Kaidu River and the Huangshuigou River can be separated into the three components (ice-snow-melt, alpine precipitation and groundwater), and the lake water can be separated into four components (river runoff, lake precipitation, groundwater, and agricultural and industrial drainage).

Comparing the stable isotope (δ¹⁸O and δ²H) of different water samples using the global meteoric water lines (GMWL) and local meteoric water lines (LMWL) was useful for clarifying the water source and isotopic fractionation for the study of regional hydrological processes [13] [14] . Because isotopic data on precipitation in Bosten Lake Basin (including Kaidu River Basin and Kongque River Basin) are insufficient to form LMWL, we used the data from the Urumqi weather station (43.78˚N, 87.62˚E, 918 m above sea level), which is also located on Tianshan Mountain, to construct the LMWL. The isotopic data on the precipitation from the Urumqi weather station were collected from the Global Network of Isotopes in Precipitation (GNIP) from 1986 to 2003 (the construction process see [1] .

The δ¹⁸O values of the water samples collected along the Kaidu River ranged from −8.99‰ to 13.33‰, with a mean of −10.14‰. The δD values ranged from −66.02‰ to 90.42‰, with a mean of 72.36‰ (Figure 2(a)). The δ¹⁸O values of the Huangshuigou River ranged from −9.13‰ to −10.82‰, with a mean of −9.83‰. The δD values ranged from −56.96‰ to 74.12‰, with a mean of −64.13‰ (Figure 2(a)). These values fit with or were above the GMWL and LMWL (Figure 2(a)), indicating that the runoff of the Kongque River and the Huangshuigou River originated from ice-snow-melt water and precipitation in the alpine mountain at low tempera-

ture and absolute air moisture content and that little evaporation was involved in the flows in the Kaidu River and the Huangshuigou River. Simultaneously, the isotope values of the groundwater samples had been covered by the river water samples, demonstrating that the river flow and groundwater had a similar water source and that the groundwater recharged the runoff [3] [10] . The differences in δ¹⁸O and δD in the precipitation samples from individual rainfall events were significant at the local area. Reference [14] and Reference [10] suggested that the isotope composition of precipitation is typically controlled by air temperature, moisture and precipitation history. In this study, the precipitation samples were collected from different months in 2011.Therefore, these samples were potentially affected by the temperature.

The δ¹⁸O values of the surface water samples collected around Bosten Lake ranged from −0.28‰ to 9.61‰, with an average of −4.10‰. The δD values ranged from −14.12‰ to −66.25‰, with an average of -29.15‰. The stable isotope values of the water from Bosten Lake were increased compared with the water from the Kaidu River and Huangshuigou River runoff (Figure 2(b)), suggesting that the isotope composition was significantly enriched in the lake and depleted in the rivers [1] . The water samples of Bosten Lake were significantly below the LMWL and GMWL but fit the local evaporation line (LEL), which was regressed by the below equation using the 48 water samples from Bosten Lake:

These data indicate that significant evaporation processes occurred in Bosten Lake [15] . Reference [16] demonstrated that an average of approximately 8.74 × 10⁸ m³/yr, which was greater than 40% of Bosten lake inflow, was being depleted by evaporation every year. In addition, most of the isotope composition values of the groundwater samples overlapped with those of the river samples, suggesting a very close relationship between groundwater and runoff. Moreover, the agricultural and industrial wastewater overlapped the lake water, proving that all of the agricultural and industrial waters were from Bosten Lake. However, the groundwater was similar to the river runoff mainly originating from the alpine precipitation and ice-snow-melt water. Additionally, the isotopic composition of river water, groundwater, agricultural and industrial wastewater, and precipitation in the Yanqi Basin lay near the LEL, indicating that all four water sources recharge to Bosten Lake. In particular, the samples in the intersection of the LEL with the LMWL, groundwater and river water corresponded to the average isotopic composition of the water entering the lake [17] .

For any given lake, δ¹⁸O_lake and δ²H_lake will rely on the hydrological balance between inputs (groundwater, precipitation, surface and stream inflows) and outputs (groundwater loss, evaporation, surface and stream outflows) [18] [19] . Therefore, the spatial and temporal characteristics of the isotope composition in lakes must be identified to illuminate the water cycle, such as water sources, residence time of water and evaporation, to operate water management. δ¹⁸O was used in this study to analyze the spatial and temporal variability due to its enhanced stability compared with δD [20] . A contour plot of δ¹⁸O in Bosten Lake water was made using surfer V8.0 software and the 48 samples collected around the whole river in July 2011 (Figure 3). The δ¹⁸O values of the Bosten Lake surface water samples ranged from −0.27‰ to −9.99‰ with an average of −3.53‰. The water level of Bosten Lake was relatively low (the deepest level is 16 m ); most of the δ¹⁸O values of the water samples varied around −3‰ - to 2‰. The low δ¹⁸O values of the water were concentrated in the southwestern corner where the river recharged to the lake. The inflow from the river diluted the lake water, thus resulting in lower δ¹⁸O in the southwestern lake. The high δ¹⁸O values were concentrated in the northwestern corner where the agricultural and industrial wastewater, especially for the irrigation drainage, recharged into the lake. The wastewater, which mixed with surface water and groundwater after use, experienced a strong evaporation, finally resulting in its significantly enriched isotope composition. Generally, low δ¹⁸O diffused from south to west, and high δ¹⁸O diffused from north to west. Then, these factors met and mixed to gradually stabilize in Bosten Lake.

The isotope (δ¹⁸O) of Bosten Lake water exhibited an obvious change with the change in seasons (Figure 4). The average δ¹⁸O values of the Bosten Lake water were relatively high in summer (from June to August) but low in winter. The seasonal trends in δ¹⁸O of Bosten Lake water were consistent with the seasonal changes of the Kaidu River runoff [1] , further proving that the Kaidu River is one of the main water sources recharging into Bosten Lake. However, the average δ¹⁸O value of Bosten Lake water (−5.17‰) was increased compared with the river water (approximately −11‰, [1] ), especially in July, when the mean δ¹⁸O values of Bosten Lake and

Kaidu River were −2.50‰ and −10.08‰, respectively. Combined with the analysis of the spatial and temporal distribution of δ¹⁸O in Bosten Lake water, these results indicate that the water cycle in the lake was very slow and that the evaporation was strong, resulting in the high δ¹⁸O of the lake water.

Isotope hydrograph separation is widely applied to quantify the replenishment of rivers for water resources management, especially in ungauged areas [21] . The distribution relationships between δ¹⁸O and TDS of the different water sources in the Kaidu River, the Huangshuigou River and Bosten Lake are presented in Figure 5. For the Kaidu River and Huangshuigou River, the alpine precipitation data were the mean values from the Bayinkuluke station located in the alpine mountain close to the base of the snow pack, the groundwater data were the mean values collected along the lower reaches of the Kaidu River and the Huangshuigou River, and the ice-snow-melt water data were collected from the Tianshan Mountain in Bayinbuluke. For Bosten Lake, the river water data were the mean values from the Kaidu River and Huangshuigou River water samples, the groundwater data were the mean values collected from Yanqi County located near Bosten Lake, the lake precipitation data were the means obtained from the Yanqi weather station, and the agricultural and industrial wastewater were collected from the main irrigation drainage channel. According to Figure 5, the water sources of the Kaidu River and Huangshuigou River runoff can be separated into three components as follows: ice-snow-melt water, alpine precipitation and groundwater; the water sources of Bosten Lake could be separated into four components as follows: river runoff, lake precipitation, groundwater, and irrigation and industrial drainage (wastewater). As shown in Figure 5, we can obtain the ratios of each component contributing to the Kaidu River, the Huangshuigou River and Bosten Lake based on Equation (3).

The results of hydrograph separation of the rivers and lake are presented in Table 1. The ratios of ice-snow- melt water and groundwater, accounting for the majority of the Kaidu River, were 43% and 37%, respectively. The ratios of alpine precipitation and ice-snow-melt water of the Kaidu River were increased compared with the Huangshuigou River. The reasons are because the birthplace of Kaidu River is the Hargat and Jacsta Valleys in the Sarming Mountains of Tianshan Mountains and because the river runoff was fed mainly by glacier melt and snowmelt from the alpine zone, as well as by precipitation from the mountain zone [22] [23] .

Moreover, the water sources of the groundwater were similar to the runoff mainly originating from the seepage of ice-snow-melt water and alpine precipitation. However, the ratios of direct ice-snow-melt water and alpine precipitation of the Huangshuigou River diminished gradually due to the far distance from the mountains. Therefore, the groundwater, which was slowly recharged by the alpine precipitation and ice-snow melting in the mountain, accounted for a larger contribution. For Bosten Lake, groundwater and river runoff mainly accounted for 66% of the water sources forming the lake water, followed by wastewater and lake precipitation. Moreover, the groundwater had a similar replenishment as the river runoff. Therefore, both the rivers and lake in the

Bosten Lake Basin are sensitive to the climate change of the local region. This finding is consistent with the results from analyzing the relationship of historic trends between climate variables and runoff [16] . Studies have indicated that the main factor causing Bosten Lake to gradually become a mild salt lake from a fresh lake was the irrigation and industrial wastewater drainage [24] - [26] . Our results provide further evidence using stable isotopes, proving that wastewater has significantly affected the TDS of Bosten Lake.