Introduction

An increase in the shrinking of lakes and water salinity in arid and semi-arid lands are unavoidable processes under climate change and human activities. Fluctuations in lake level, volume, salinity, and concentration of runoff sediments—each factor possessing its own hydrologic idiosyncrasies—lead to a series of ecological changes and environmental evolutions by means of fluxes of matter and energy. Water salinity is directly influenced by the rise and fall of lake levels. If lake levels increase, salinity tends to be inversely proportional because the total dissolved solids (TDS) have more solvent in which to dissolve (Venkatesan et al. 2011a). Conversely, if lake levels decrease, TDS will increase accordingly. TDS sometimes is used as a ‘watchdog’ of environmental change. If TDS levels are high, they can affect many forms of aquatic life (Shuter et al. 1998). TDS effects are more prominent in arid and semi-arid parts of the world, and in some places the water is so saline that it has become unfit for any use (Weber-Scannell and Duffy 2007).

The dominant culture in socio-economic and political systems involves growth and expansion (Shrestha et al. 2011); as a result, water constraint problems need to be solved for the sustainable development of an eco-environmental system. Considering that climate change is already impacting flows (Sagarika et al. 2014), bringing the future demand in line with available supplies will require increasingly efficient water management practices and greater conservation of water resources (Vedwan et al. 2008). For their very existence, lakes depend upon a balance between their many sources of water and the losses that they experience. Aside from being of scientific interest, water and salt balance studies serve to reveal the dependence of each lake on particular hydrologic factors, thus enabling better management practices.

Bosten Lake is the largest inland freshwater lake in China. Between 1958 and 1988, the water level of Bosten Lake dropped from 1049 m above sea level (ASL) to 1045 m ASL. Amazingly, TDS increased from 0.56 g/L in year 1958 to its highest level at 1.87 g/L in the year 1987; further, the lake’s surface area decreased by over 20 % during this period, so that indigenous fish species were on the brink of extinction during this period (Walker and Yang 1999). In the 1990s, with affluence water flowing into the lake from watersheds, the lake level reached 1049 m ASL, and salinity declined to 1.17 g/L in 2002 gradually (Hao et al. 2008; Li et al. 2011). At the same time, environment of Bosten Lake was improved significantly. However, as a result of water conflict due to the water demand for upstream and downstream agricultural areas of the lake, the lake’s level dropped rapidly to 1045.25 m ASL in 2013; this is close to the lowest control level (1045 m ASL). Correspondingly, the lake’s salinity has shot up again (Guo et al. 2015). Regardless of great attention to this issue paid by the government and significant research over the decades, worrisome history of 1980s is recurring once again. Therefore, understanding a lake’s hydrology and its system’s behavior in a dynamic manner is a core issue for sustainable management of lake and its socio-ecosystem.

A large body of knowledge covers empirical and physical based modeling of hydrologic processes (Ahmad et al. 2010; Sivapalan 2012; Carrier et al. 2013; Puri et al. 2011; Kalra et al. 2013) and water quality changes (Melesse et al. 2011; Wang et al. 2013). The modeling approaches used in these studies have several strengths; however, the lack of ability to consider feedback mechanisms and incorporate socio-economic factors limits their usefulness to model dynamic problem of water resources planning and management. System dynamics (SD) is an approach to understand the behavior of complex systems over time, and deals with internal feedback loops while keeping track of the whole system response (Ahmad and Simonovic 2000). SD differs from other approaches to study complex systems due to the use of stocks and flows, feedback loops, nonlinearity, and time delay. These elements help in understanding how seemingly simple systems display dynamic complexity and nonlinearity (Forrester 1971). SD also provides a platform for group model building and to engage public and decision makers in scenario analysis. Over the past 50 years, numerous SD models have been developed to address water resources management issues (Winz et al. 2009; Mirchi et al. 2012). Some notable examples include models for water resources planning (Ahmad and Prashar 2010; Dawadi and Ahmad 2013; Qaiser et al. 2013; Wu et al. 2013), the water-energy nexus (Shrestha et al. 2012), salinity and water quality control (Venkatesan et al. 2011b), flood management (Ahmad and Simonovic 2004), and climate change impact on stream flow (Li et al. 2010; Dawadi and Ahmad 2012).

This paper focuses an SD model developed to couple the water and salt balance for Bosten Lake. The main aim was to identify the potential impacts and sensitivity of each component of the water and salt balance of Bosten Lake. The model integrates the physical characteristics of the lake as well as the components of hydro-meteorological elements of the water and salt balance in a situation where limited data is available. Temporally, the model operated on an annual time step that encompassed the 31-year period of 1983–2013 with observed data for calibration, and 2014–2050 periods for future simulations.

Materials and methods

Study area

Bosten Lake, largest inland freshwater lake in China, is located at 57 km northeast of Korla City in the Bayin’gholin Mongol Autonomous Prefecture of the Xingjian Uyghur Autonomous Region in China. The lake is the lowest point of the Yanqi Basin; the climate type of this area is warm, temperate, continental, and arid. The Bosten Lake is divided into two parts, i.e., “Large Lake” and “Little Lake” separated by a dam. The average water depth of the large lake is 8.1 m with the deepest point of the large lake located 17 m below the water surface, and corresponding water surface area approximately 1000 km2 when lake level is about 1048 m ASL. The lake receives water inflow from a catchment area of 56,000 km2 (Wünnemann et al. 2006). There are several rivers in the Yanqi Basin which flow into the Lake (Fig. 1). The Kaidu River is the most important tributary to the Bosten lake, accounting for about 93.54 % of its average annual water inflow, while other rivers, including Huangshui river, Qingshui river, and Ushaktal river, account about 6.46 % (Wang et al. 2003). As shown in Fig. 1, the water from the Kaidu River flows into the Konqi River by means of going through Bosten Lake. The ‘Little Lake’ is a complex channel network overgrown by thick reeds. In 1983, the natural course of the Kaidu River was changed after the construction of a dam and the Xibeng pumping station. The construction of the dam, the increased inflow, and the increased discharge of the larger lake of Bosten significantly reduced the average salt concentration (Xia et al. 2003). Construction of the Dongbeng Pump Station, located 650 m east of Xibeng pumping station, further accelerated water and salt exchange of Bosten Lake by the end of 2008.

Fig. 1
figure 1

Map of Bosten Lake and its water system

Full size image

Modeling approach

The coupled SD model of the water and salt balance was developed using STELLA platform. Interested readers are referred to Sterman (2000) for more details on SD modeling. The SD model for the water and salt balance of the Large Lake and Little Lake consists of several stocks and flows. According to Bosten Lake’s water network system, the two main stocks of the SD model, i.e., water volume and salt amount, connected at the third water allocation station, a station of the Kaidu River that controls the water amount in the east and west diversions of the Kaidu River, and at the Tashdian station, located at the headstream of the Konqi River (Fig. 1). The east diversion of the Kaidu River goes through the Large Lake and the west diversion flows through the Little Lake into the Konqi River. The so-called Jiefangyiqu Canal flows into the Konqi River directly from Kaidu River. In this study, it was assumed that the Jiefangyiqu Canal does not belong to the water network of Bosten Lake. For purposes of calculation, the water amount of the Konqi River and its salinity at the Tashdian Water Station was used to estimate the outflow amount of the Little Lake and its salinity at Dawut Station. As shown in Fig. 2, the SD model for the water and salt balance of Bosten Lake was designed to demonstrate the hydrological behavior and emulate the key hydrological factors that affect water amount and salinity of the Bosten Lake.

Fig. 2
figure 2

Dynamic hypotheses for the water and salt dynamics of Bosten Lake; the model is simplified in this figure regarding the relationships among key elements and the feedback loops of the Large Lake and Little Lake

Full size image

The model variables included precipitation, evapotranspiration, water volume (storage) of lake, ground water flows, including drainages and seepage, and surface water flows and their corresponding salinity. The main factors that influence the salt balance in the lake included reed production, sandstorms, the salt amount with precipitation, total suspended particulates (TSP), and lake sedimentation. The rectangles in Fig. 2 represent stocks for the water volume and salt mass in the Large Lake and Little Lake. The flows that control water volume (stock) in the lake consisted of inflows—surface water inflows from rivers, precipitation as well as groundwater and drainage inflows—and outflows, including transpiration, evaporation, seepage, and surface water outflow from the lakes. According to Water Balance Principle, the common water balance equation of the large lake and the little lake can be written as:

$$V^{t} \; = \;V^{t - 1} \; + \;\left( {\left( {\mathop \sum \limits_{1}^{n} R_{n}^{t} } \right)\; + \;P^{t} A^{t} \; + \;I^{t} \; + \;D^{t} } \right)\; - \;(T^{t} W^{t} \; + \;E^{t} A^{t} \; + \;S^{t} \; + \;O^{t} )$$
(1)

where V t is water volume of the lake at time t, V t−1 is the water volume of the lake at the initial time, \(R_{n}^{t}\) is the surface water inflow of river n to the lake at time t, P t is the precipitation amount at time t, A t is the water surface area of lake at time t, I t is the groundwater inflow to the lake at time t, D t is the drainages water inflow to the lake at time t, T t is the transpiration amount per area of the reeds at time t, W t is the reeds-based wetland area of the large lake or the little lake, E t is the amount of evaporation per lake surface area at time t, \(S^{t}\) is seepage from the lake at time t, and O t is the surface water outflow from the lake at time t. For the \(W^{t}\), reeds area around the large lake (\(W_{1}^{t}\)) and the little lake (\(W_{2}^{t}\)) at time t was calculated by following Eqs. (2) and (3).

The growth of reeds in Bosten Lake depends on salinity as well as the hydrological and nutritional situation of the wetlands (Liu 2004). Our study discovered that the area of the reeds, extracted from Landsat images,has a large correlation to the average lake level in the previous year. The estimation functions for the reeds area of the Large Lake and Little Lake in this study were calculated in Eqs. (2) and (3), respectively, as:

$$W_{1}^{t} \; = \;5.9611\; + \;\frac{29.64472}{{1\; + \;10^{{(1046.67675\; - \;H_{1}^{{\left( {t - 1} \right)}} )\; \times \;1.36577}} }} R^{2} \; = \;0.96$$
(2)
$$W_{2}^{t} \; = \;269.38875\; + \;\frac{16.84145}{{1\; + \;10^{{(1047.32795\; - \;H_{2}^{{\left( {t - 1} \right)}} )\; \times \;7.95781}} }} R^{ 2} \; = \;0. 9 9$$
(3)

where \(W_{1}^{t}\) and \(W_{2}^{t}\) are the reeds area of the large and Little Lakes at time t, respectively, and \(H_{ 1}^{{\left( {t - 1} \right)}}\) and \(H_{2}^{{\left( {t - 1} \right)}}\) are the average level of the Large and Little Lakes 1 year previous, respectively.

Because most of the salt in the lake are mass transported by water inflow and outflows, the salt balance equation was derived from the water balance equation. However, the amount of salt from atmosphere as well as salt sedimentation on the lake bed was also considered. Therefore, the common equation of the salt balance for the Large and Little Lake was written as:

$$S^{t} \; = \;S^{t - 1} \; + \;\left( {\left( {\mathop \sum \limits_{1}^{n} R_{n}^{t} C_{n}^{t} } \right)\; + \;S_{a}^{t} A^{t} \; + \;I^{t} C_{g}^{t} \; + \;D^{t} C_{d}^{t} } \right)\; - \;(S_{r}^{t} W^{t} \; + \;S^{t} C_{s}^{t} \; + \;S_{s}^{t} A^{t} \; + \;O^{t} C_{0}^{t} )$$
(4)

where S t is the total amount of salt in the lake at time t, S t−1 is the amount of salt in the lake at the initial time, \(C_{n}^{t}\) is the salinity of the river n at time t; \(S_{a}^{t}\) is the amount of salt per lake area from the atmosphere, including salt from dust storms, precipitation; \(C_{g}^{t}\) is the salinity of the groundwater into the lake, \(C_{d}^{t}\) is the average salinity of drainages into the lake, \(S_{r}^{t}\) is the amount of salt lost per reeds area, \(C_{s}^{t}\) is the salinity of the seepage, \(S_{s}^{t}\) is the amount of sedimentation per lake bed area at time t, and \(C_{0}^{t}\) is the salinity of the surface water outflow from the lake at time t.

Data sources

Historical flow and TDS values, 171 pairs of Lake volume-level and volume–area curve data were provided by the Water Resources Bureau of Bayin’gholin Mongol Autonomous Prefecture in Xinjiang. Metrological data were acquired from Bagrash weather station and Dahekou weather station, which are located around the Bosten Lake (Fig. 1). Other components’ data were collected from various sources. For example, the average annual salt amount with precipitation is 0.03 × 105 ton, and salt from sandstorms on the lake is 1.966 kg/s (Cheng 1995). The salt amount per lake area per year from the atmosphere is estimated as 0.153 kg/m2/a (Zhao 2005). The sedimentation rate on the Large Lake’s bed is about 1.3 kg/m2/a (Zhang et al. 2004), and includes superfluous dissolvable salt, chemical deposits, and organic matter (Wünnemann et al. 2006). However, no research has been done about how much salt mass enters the lake’s water body from the lake’s bed and how much is their interchange rate. Therefore, in this study, by means of fitting the observed and calculated average salinity of the Large Lake, STELLA’s graphical function was used to estimate some unknown factors of the salt balance, including the salt mass that entered from the lake bed and conversion coefficient for actual evaporation. The value of salt mass that entered from the lake bed was estimated to be between 0.9 and 1.2 kg/m2/a. Xia et al. (2003), using measured pan evaporation data, has estimated that the average conversion ratio of 0.47 can be used to obtain actual evaporation amount from pan evaporation data. They also reported that the average transpiration rate of reeds is 1.3 times of the potential evaporation rate. In this study, estimated conversation coefficient for actual evaporation using STELLA’s graphical function for calibration is 0.44 to 0.62 (Fig. 3a).

Fig. 3
figure 3

Comparison of observed and simulated data for a the Large Lake level (m ASL.) and b the Large Lake’s average salinity (g/L) from 1983 to 2013

Full size image

Results and discussion

Calibration and evaluation of the model

Model calibration is the adjustment of the model’s parameters and constants to fit the simulated results with the observed data for all state variables (Rykiel 1996). Proper model calibration is important in hydrologic modeling studies to reduce uncertainty in model simulation. Parameters were selected from a range of feasible values, then tested in the model and adjusted until satisfactory agreement between predicted and observed variables was obtained (Li and Yakupitiyage 2003). In this study, the calibrations of the model provided new estimations of the evaporation rate, the salt amount from the lake bed, and possible evolution of the lake level and salinity with time. The estimated conversion coefficient of actual evaporation vs. potential evaporation (which is observed by evaporation pan of 20 cm) varied between 0.44 to 0.62. Depending on the salinity of the Large Lake, the salt amount from the lake bed varied from 0.9 to 1.2 kg/m2/a. With these parameters, the SD model was used to capture the interrelationships of the lake volume and its salinity with interrelating hydrological and meteorological factors, such as inflow and outflow volume and salinity, evaporation, transpiration, drainage flow, groundwater flow, seepage flow, precipitation, salt from atmosphere, salt lost with vegetation, sedimentation, and salt input from lake’s bed.

Model evaluation involves running a model using input parameters measured or determined during the calibration process, and reproduction of historical records of the model performance (Li and Simonovic 2002). Moriasi et al. (2007) recommended several methods for evaluation of accuracy in hydrological models including quantitative statistics and the graphical techniques, and established corresponding evaluation criteria for model performance. We used both the graphical techniques and ratio of the root mean square error (RMSE) to the standard division of measured data (RSR) to evaluate the performance of system dynamics model of water and salt balance during calibration in this study. According to the objectives of this study, the Lake level and average salinity of water were selected as key variables for evaluation of SD models regarding the behavior and structure of water and salt balance from 1983 to 2013. The values of the coefficient of determination (R 2) between the observed and simulated values are found to be high with 0.95 and 0.98 for lake level (Fig. 3a) and lake average salinity (Fig. 3b), respectively.

RMSE is one of the commonly used error index statistics, and RSR standardizes RMSE using the observation’s standard division. It is commonly accepted that the lower the RMSE, the lower the RSR, and the better the model performance. The calculated RMSE error and RSR are 0.26 and 0.04 for lake level, respectively; the calculated RMSE error and RSR are 0.06 and 0.03 for lake average salinity, respectively. The evaluation of the results suggests that the developed SD model can accurately simulate the dynamics of water and salt balance components. According to Moriasi et al. (2007), which established performance ratings of hydrological model evaluations, the evaluation results of lake level and average water salinity of Bosten Lake can be judged as “very good”.

In summary, the calibration and evaluation of the system dynamics model for water and salt balance of Bosten Lake showed that the simulated lake level and average lake salinity stay within the bounds of measurement error, and fit the observed data quite well.

Qualitative analysis of the water and salt balance

According to the earliest water salinity records from the Xinjiang Comprehensive Investigation Team of the Chinese Academy of Sciences in 1958, the average salinity of Bosten Lake was about 0.6 g/L. At that time, Bosten Lake was a freshwater lake. However, since the 1970s, the salinity of the lake water generally has been over 1.0 g/L; thus, Bosten Lake has become slightly a saltwater lake (Xia et al. 2003). During 1960 and 1970, the existing irrigation system and arable area in this region was expanded and the population increased rapidly in the Yanqi Basin, similar to other regions in China (Brunner 2005). The impact of these practices of consuming a huge amount of water for arable area continued until 1980s. As a result, a large amount of surface water was used for irrigation with poor irrigating efficiency. Table 1 shows that the surface water inflow to Bosten Lake decreased as did its water level; at the same time, its water salinity increased. However, in the 1990s and early 2000s, the surface water inflow to Bosten Lake has increased, its water level has gradually increased, and its water salinity has decreased.

Table 1 Observed average values of Bosten Lake’s water and salt balances (Large Lake)
Full size table

Generally, a lake’s salinity decreases when the water exchange rate increases or level of the lake increases. Water exchange rate refers to the replacement of lake water and is related to average residence time, i.e., water exchange rate is high when average residence time is low. The high water exchange rate along with low-salinity water coming as inflow decreases water salinity in lakes. The increase or decrease in water level is a result of the water balance, and is affected not only by the inflow but also by the outflow. Moreover, if the inflow and the water level of the lake increases, and the outflow from the lake decreases, the water area of the lake will expand, resulting in an increase in unproductive water lost from the lake.

According to the model’s calculations, the annual unproductive water lost (11.48 × 108 m3) is close to the annual surface water inflow (16.96 × 108 m3) of the Large Lake from the Kaidu river east division from 1983 to 2013. As time passes, the quantity of accumulated salts in the lake gradually increased and so did the water salinity of the lake. During 1983 to 1987, the corresponding average salinity was about 1.8 g/L. After 1990, the annual runoff increased rapidly in the Kaidu River watershed (Li et al. 2011) and the efficiency of irrigation improved in the Yanqi Basin (Brunner 2005); as a result, surface water inflow to the Bosten Lake increased from 1988 to 2002 (Table 1). During this period, water volume increased and average salinity of the Bosten Lake decreased to about 1.2 g/L. The surface water outflow from the lake increased after Dongbeng Station was built; however, surface water inflow maintained its decreasing trend during 2000 to 2013, and then the water exchange rate of Bosten Lake increased. Consequently, the average salinity of Bosten Lake decreased to about 1 g/L.

The analysis based on the data in Table 1 reveals that surface water inflow, outflow, and unproductive water lost from the lake’s surface are the main factors related to water balance of Bosten Lake. However, the water balance is a result of the complex interaction among these factors. The change in water salinity of Bosten Lake is affected by these factors simultaneously, and is a complicated dynamics process to simulate and explain.

Simulation results and discussion

Sensitivity analysis

To ensure the integrity of the system dynamics model and demonstrate the manner in which the model structure and feedback connections determine simulated behavior, this study took the calibration period as a base case and performed a sensitivity analysis by changing the values of one variable at a time. The responses of selected target variables to the parameter changes were compared. Following Li and Simonovic (2002), in this study, selected parameters were changed by about ±10 % of the baseline value in the sensitivity analysis. The following five important parameters in the model were selected for sensitivity analysis of the Large Lake: (1) total surface water inflow (\(\sum_{1}^{n} R_{n}\)); (2) surface water outflow (O); (3) groundwater inflow (I); (4) seepage (S); and (5) evaporation (EA). Important eco-hydrological parameters—such as the level of the Large Lake, the average salinity of the Large Lake, unproductive water lost, including water lost through evaporation and transpiration, and the reeds area—were selected as target variables for the sensitivity analysis. As shown in Table 2, the influence of variations in the selected parameters on the Large Lake level caused these factors to vary. The variation rankings for each selected parameter are as follows:

Table 2 Sensitivity Analysis of the Model (%)
Full size table

For the Large Lake level: \((\sum_{1}^{n} R_{n} ) > EA\; > \; O \; > \;I \; > \; S\).

For average salinity of the Large Lake: \(O\; > \;( \sum_{1}^{n} R_{n} )\; > \; EA \; > \; I\; > \; S\).

For unproductive water lost: \((\sum_{1}^{n} R_{n} ) \; > \; O\; > \; EA\; > \; I\; > \; S\).

For the reeds area: \((\sum_{1}^{n} R_{n} )\; > \; EA \; > \;O \; > \;I \; > \;S\).

Overall, the variation in the value of I and S had little influence on selected target parameters. On the other hand, the selected target parameters were much more sensitive to (\(\sum_{ 1}^{n} R_{n}\)), O and EA. It is worth noting that the reeds were the most sensitive parameter than the others. This attests to the fragile characteristic of the ecosystem in arid and semi-arid lands and their excessive sensitivity to water. It is generally recognized that inflow and outflow of lakes play important roles in the water and salt balance, and this study shows a similar pattern. Besides (\(\sum_{ 1}^{n} R_{n}\)) and O, the variation in value of EA had a distinct influence on the selected target parameters, indicating that evaporation is also an important variable in the ecological and hydrological process in arid and semi-arid lands (Qaiser et al. 2011). Results of a study by Brunner (2005) also demonstrate that the exchange of groundwater between the Bosten Lake and its aquifer is negligible.

Scenario analysis

The system dynamics model couples together the water and salt balance. The relationships among the corresponding parameters are quantitatively expressed. These relationships can be used as the basis to predict the change in parameters of interest to this study. According to the sensitivity analysis shown in Table 2, main factors that resulted in fluctuations of the water and salt balance of the Bosten Lake were the surface water inflow, evaporation and surface water outflow. Actually, Bosten Lake’s water cycle and corresponding salinity fluctuated according to both climate change and the water demand of upstream and downstream area. Xia et al. (2003) validated that dead water level, normal water level and flood level for the large lake are approximately 1045.0, 1046.5 and 1048.0 m ASL, respectively. Li et al. (2011) have projected an increase in temperature and decrease in rainfall in the Bosten Lake watershed area from 2020s to 2080s. However, the studies have acknowledged that there is a large uncertainty in climate projections (Forsee and Ahmad 2011). To capture the effect of climate change and water demand of upstream area, two different situations, i.e., wet season and dry season, were considered as: (a) water amount of high surface water inflow (I h ) period is calculated as 17.63 × 108 m3 and resulting lake level is higher than normal water level (1046 m ASL); (b) likewise, water amount of low surface water inflow (I l ) period is calculated as 16.12 × 108 m3 with resulting lake level at less than normal water level. With respect to water demand due to socio-economic development, Guo et al. (2015) revealed that driving factors of water demand in the downstream area of Bosten lake are increasing population and resulting economic development and expansion of agricultural irrigated area. Actual delivered water to the downstream of the Bosten Lake equals to the observed water amount at Tashdian water station, and 90 % of the diverted water from the Konqi River is used for agricultural irrigation. According to the socio-economic development plan of Bayin’gholin Mongol Autonomous Prefecture in the future 50 years, Chen et al. (2013) assumed three main variation trends viz. high water demand (O h ), medium water demand (O m ) and low water demand(O l ),taking into account the increased irrigated area and modern irrigation techniques for the future 50 years (Fig. 4). To simulate the water level and salinity dynamics of the lake with regard to various changes in the future, a period from 2014 to 2050, six scenarios were designed with different rates of inflows (\(I_{h} \;{\text{and}}\; I_{l}\)) and outflows (O l , \(O_{m} \;{\text{and}} \;O_{h}\)), as shown in Table 3.

Fig. 4
figure 4

Different situations of scenario determination related total surface water inflow (\(I_{h} \;{\text{and}}\;I_{l}\)) from upstream area and water demand of downstream area (O l , \(O_{m} \;{\text{and}}\;O_{h}\))

Full size image
Table 3 Scenario assumptions and result of main factors for Bosten Lake
Full size table

The simulated results from six scenarios indicate that the lake level and average salinity of the Large Lake, the total amount of water of Konqi River and its salinity at the Tashdian station, the total reeds area, and unproductive water lost (including evaporation and transpiration from the Large Lake and Little Lake) have fluctuations that vary in manner under different inflow and demand scenarios (Fig. 5). The scenario simulations in Fig. 5 reveal that the amount of surface water inflow to the lake is a crucial factor to variations in levels of Bosten Lake (Fig. 5a and Table 3) and also determine the amount of reeds and unproductive water lost (Fig. 5e, f). There is a connection among the lake level, the reeds area, and unproductive water lost; therefore, it is worthwhile to consider the efficient water management practices and greater conservation of water resources in the upstream area of Bosten Lake.

Fig. 5
figure 5

Simulated scenarios for a the Large Lake level (m ASL.), b the Large Lake’s average salinity (g/L), c the amount of water at Tashdian Station (108 m3), d the water salinity at Tashdian Station (g/L), e total reeds area (km2), and f total unproductive water lost (108 m3) from 2014 to 2050

Full size image

It is common knowledge that water salinity is influenced by the rise and fall of lake levels. This study argues that except fluctuation of lake level, water exchange rate also affect the salinity of lake. For example, Fig. 5d and Table 3 show that mean salinity of the Large Lake will maintain a downtrend due to the increasing water exchange rate resulting from water demand in the Konqi River area. This happens because some dissolved solids are transported to the downstream by water outflows from lake. The more the outflow, the more the transported dissolved solids; it results in decreasing water salinity of the lake. Thereby, it is clear from these scenario analyses that increasing the water exchange (Scenarios 3 vs. Scenarios 4) is beneficial to both the decreasing salinity of the lake and increasing water to the downstream area of Bosten Lake. This means that enlarging the agricultural area in the Konqi River basin rather than in the Kaidu river area helps to decrease the lake’s salinity and the total salt amount in the lake as well. Therefore, Scenarios 2 and 4 are propitious to sustainable development of the Bosten Lake and its ecosystem.

Results indicate that keeping the lake level high by means of decreasing outflow is not a wise choice. Otherwise, decreases in the water exchange rate will worsen the quality of lake water and ecosystem, for example, reeds area (e.g., scenario 1 and 4 in Fig. 5). Accordingly, when surface water inflow decreases or maintains a certain amount due to climate change and human activities, in the short term, it is better to increase the water outflow from the lake helps in both maintaining the good water quality in the lake and decreasing unproductive water lost through evaporation. Because of the large open water surface, greater evapotranspiration occurs in arid lands. For example, the results of Scenarios 5 and 6 regarding the lake’s mean salinity and total reeds area are almost the same; however, big differences in the results occur for the unproductive water lost. According to the model, the average unproductive water lost (11.59 × 108 m3) was larger than half of the average surface water inflow (16.80 × 108 m3) of the Large Lake.

The objective for optimal water management in a long term is to lower salinity and unproductive water lost and to maintain a higher lake level. According to the objective of minimizing the average salinity of the Large Lake, Scenario 3 is the best and Scenario 4 the worst. The results indicate that it is possible to maintain the Large Lake as a freshwater lake (salinity ≤1 g/L) with a quite high lake level (≥1046 m ASL), for example., Scenarios 2 and 3. However, because of increasing water demands due to socio-economic development in the Konqi river area and Korla City, it is very difficult to fulfill Scenario 1. Through modern water conservation technologies and by enlarging the agricultural area in the Konqi River area rather than the Kaidu River area, it is possible to keep the Bosten Lake and its ecosystem within a good quality and is also possible to decrease the lake’s salinity. In contrast, enlarging the agricultural area in the upper area of Bosten Lake might result in shrinking Bosten Lake or its disappearance altogether. As a result, downstream area of Bosten Lake will encounter water crisis.

Conclusions

The coupled system dynamics model to simulate the water and salt balance of Bosten Lake demonstrated that the simulated lake level and salinity, the amount of the flux of Konqi River, and the lake’s salinity fit historical data quite well. The results of the sensitivity analyses demonstrated that important eco-hydrological parameters—such as the lake level, average salinity of the lake, unproductive water lost, and the reeds area—all have a high correlation with both the differences in total inflow and outflow of surface water as well as evaporation from the lake’s open water surface. Bringing the future demand in line with available water supplies of Bosten Lake will require increasingly efficient water management practices of these most sensitive parameters.

Considering the potential water demand for the socio-economic development and the possible future uncertainty in flows due to climate change, six scenarios were developed. Main conclusions drawn from this study are:

  1. 1.

    It is a better practice to increase the water outflow from the lake to maintain good water quality. Additionally, this practice decreases the loss of unproductive water by means of evaporation. The larger open water surface of lake predicates greater loss of unproductive water in arid lands.

  2. 2.

    Increasing the water exchange of the lake promotes both decreasing average salinity of the lake and increasing water supply to the downstream area of the lake. It is propitious to sustainable development of the Bosten Lake’s ecosystem and socio-economy in the downstream area.

  3. 3.

    Enlarging the agricultural area in the Konqi River area rather than Kaidu River basin favors the decrease of the lake’s salinity and also good for decreasing the total salt amount in the lake.

To summarize, fluctuations in lake level and environment quality are unavoidable under climate change and anthropogenic pressures. Model results demonstrate future water management challenges that Lake Bosten will face. Climate change probably will aggravate the situation, and will require a reduction in human water consumption. Thus, water management practices that can reduce demand by using efficient irrigation practices or increase supply by reducing evaporation losses will become increasingly important. This implies that increasingly efficient water management practices and greater conservation of water resources are required to save Bosten Lake and its ecosystem as well as create a sustainable socio-economy both in the Yanqi Basin and the downstream region of the Konqi River.