The leap-forward development policy emphasizes water resource scarcity. Given the importance of water resources to an economy, especially in arid areas, this paper explores various management strategies that could be used in the semi-arid region of Xinjiang to increase the effectiveness and efficiency of multifunctional water projects and support sustainable ‘leap-forward’ economic growth. Embedding the logistic growth function within a system dynamics framework, our model provides a universally implementable model for sustainable water planning. The dynamic current and projected water supply and demand and water use efficiency are captured. Potential water management strategies are simulated to test system performance. The results suggest that, as water scarcity becomes more acute, the current water supply will not be able to support the leap-forward development policy with current water usage efficiency. In the short term, implementing efficient industrial water consumption technology and promoting water-saving irrigation technology may postpone the cultivated land reduction process. In the long term, developing more water-saving industrial enterprises, and planning to use more drought-tolerant plants, will be effective ways to avoid a decrease in cultivated land area.

Introduction

Over the past 30 years, mainland China's economy has grown at an annual rate of increase of 10%, a so-called ‘Chinese miracle,’ mainly caused by top-down policy making that focuses on economic development (Zhou, 2007). In arid and semi-arid regions of China, the urbanization level is quite low, and there are severe poverty alleviation problems. To accelerate the urbanization process and resolve the poverty issue, an investment-driven development strategy is essential. Xinjiang Uygur Autonomous Region is typical of regions facing severe water scarcity and poverty alleviation problems. Xinjiang is located in the north of China, and covers an area of 1,600,000 square kilometres. The average annual income per capita was just RenMinBi (RMB) 12,258 (about US$2,000) in 2009, which is relatively low compared to the Chinese average of RMB17,174.70. To promote economic development, a leap-forward development strategy to invest an extra RMB2,000 billion1 to boost the economy was introduced in 2010. Industrial output increased by 34.3% and 24.4% in 2010 and 2011, respectively.

As a result of this strategy, agriculture and industry have developed quickly, and there has been rapid urban expansion. The leap-forward policy makes sustainable water resource development a particular challenge (Davies & Simonovic, 2011; Dawadi & Ahmad, 2012; Wu et al., 2013; Dai et al., 2013). The average annual rainfall in Xinjiang is about 150 mm. The annual runoff is about 88 billion m3 of surface water plus 25 billion m3 of exploitable groundwater. With rapid industrial and population growth, it is difficult to meet municipal, agricultural, and environmental water demands simultaneously. Increasingly efficient water management practices and greater conservation of water resources will be required to bring future demand in line with constrained available water supplies and achieve successful water management (Davies & Simonovic, 2011; Deng et al., 2011). Deng et al. (2011) explored the water supply and demand regulation to support the economic and social leap-forward development. Wu et al. (2013) built a dynamic model to assess the vulnerability of regional water resources in Bayingolin, Xinjiang. Li et al. (2011) measured the spatial and temporal variability of precipitation concentration in Xinjiang. However, research into the dynamic conflict between water demand and supply in Xinjiang, specifically in the context of leap-forward development, is inadequate.

Sustainable management of water resources requires the water resource system to be integrated with the regional social-economic structure and value system (Rauch et al., 2002). Water resources are relevant to a number of research fields, including economic development, urbanization processes, and environmental protection. Unlimited population and industrial growth lead to decreasing water availability and the reallocation of water from one user or sector to another, changing overall benefits and well-being (Sterman, 2002). To correctly assess and predict water availability and match social-economic growth to the acknowledged carrying capacity of limited water resources within a complicated system, a multidisciplinary assessment modeling approach that integrates socio-economic and water factors is needed (Simonovic, 2002; De Fraiture, 2007). Among the integrated assessment models, system modeling is appropriate for addressing complex water issues (Davies & Simonovic, 2011).

In this paper, a system dynamics (SD)-based integrated assessment model is used. It attempts to represent the continuous dynamic interaction between the economy and the water resource base and explore the possibilities of balancing industry-oriented municipal development and agriculture-oriented rural development under constrained water supply within a Chinese arid area context. Through the assessment of dynamic interactions between water processes and human-mediated processes, this paper describes and analyzes the effectiveness of the investment-driven development strategy, and evaluates the effects of four possible water resources policies. Four scenarios are discussed alongside the current base scenario to give a clear understanding of the system and possible policies. Some implications of the experimental results for real-world water resources management are explored.

SD modeling in water planning

A systematic approach to solving water shortage problems, addressing both interactive and nonlinear relationships, has proven to be effective (Simonovic, 2002; Xu et al., 2002; Tidwell & Van Den Brink, 2008; Winz et al., 2009; Sánchez-Román et al., 2010; Ke et al., 2016). The endogenous structure of the system can be elucidated, the relationship among different elements of the system can be explored, and the changing relations within the system when different decisions are implemented can be assessed. Models that use a systematic approach can be used to simulate real life in space and time, to improve the understanding of the complicated internal interrelationships, influential processes and factors, the cause loops of the complex system (Sterman, 2002). Social–water SD models are appropriate for estimating the effectiveness of policy interventions within a particular social and economic context (Xu et al., 2002). Such models reveal relationships between water resources and human social systems rather than hydrological processes, water balance, water flux in agricultural systems, or climate changes. In the WorldWater model (Simonovic, 2002), the water resource limitation is first taken into consideration in a world development system.

An SD model's behavior is governed by its structure and boundaries. The model's structure determines the simulation results and the modeling effectiveness. The structural relationships between a system's individual components are important in determining aggregate system behavior (Sterman, 2002). Each SD model's structure and boundary is determined by concrete research goals and research questions (Sterman, 2002). In the thinking phase of modeling, building the model's structure and boundaries helps to define the core problem in a system. SD thus offers a unique qualitative tool to improve understanding of a complex problem (Mirchi et al., 2012). There is no universally adopted SD water model (Davies & Simonovic, 2011).

At the regional scale, Xu et al. (2002) address a coherent water–social economic system model based on a water balance model. Using the Yellow River basin as the case study, a basin-scale model provides a general dynamic water balance model. The existing and potential water supplies from surface water, aquifers and treated wastewater are estimated, and potential water demands for domestic, industrial and agricultural uses are projected. The Yellow River model provides an SD framework to evaluate water stress performance. Ten scenarios are analyzed for water-supply sustainability. Further modules have been added to this framework. Based on a water balance model, Madani & Mariño (2009) simulate explicit feedback between variables related to population, the local economy and water resources. The interrelationships among water resources and a socio-economic system are simulated in the WorldWater model (Simonovic, 2002) by an SD-based integrated assessment model. The WorldWater model was further developed to the ANEMI model by adding the carbon emission process (Davies & Simonovic, 2011). Ke et al. (2016) use a dynamic model to simulate water allocation and water reclamation in a Chinese mining city.

Many different types of water conflict have been explored using water–social SD modeling. The effects of regional agricultural projects on water resources, land degradation, agricultural pollution and demography in Southeastern Anatolia were analyzed from a systems perspective (Saysel et al., 2002). Rajasekaram et al. (2003) simulated 12 scenarios with varying fresh water availability, wastewater treatment, economic growth, population growth, energy generation and food production to explore the effectiveness of water conflict resolution. The effects of water conservation on outdoor water use in an arid region were evaluated by Qaiser et al. (2011). Climate change and carbon emission is the main concern in the global-scale integrated socio-economic and environmental model of Davies & Simonovic (2011). The risk and vulnerability of a watershed due to changing climatic and socio-economic conditions were evaluated by Prodanovic & Simonovic (2010). The effects of agricultural cropland expansion on aquifer water salinity and aquifer levels were simulated by Fernández & Selma (2004) using their New Irrigated Lands dynamic model. The sensitivity of irrigation technology to water price, the adoption of more efficient technologies and increases in dryland farming was evaluated by a nonlinear optimization model for four basins in Spain (Iglesias & Blanco, 2008). Municipal water conservation policies have been assessed in an SD model that captured the interrelationships between water availability and competing municipal, agricultural, and environmental water demands (Ahmad & Prashar, 2010).

Industrial growth is an important constraint in SD water models but industry-oriented water conflicts have not been sufficiently analyzed. Within the Chinese context, especially when assessing the leap-forward development strategy, water-economic issues should be carefully considered. Two main innovations in setting parameters and model performance criteria improve the understanding of the complicated water–social system under the leap-forward policy.

In previous social–water SD models, key water consumption parameters, such as per capita water consumption, are generally set as static (Xu et al., 2002), especially for projections (Ahmad & Prashar, 2010). The results of simulations show that the model performance is sensitive to subjectively set parameters. Per capita water consumption change is an accumulated process. Effective risk communication and reduction is grounded in a deep understanding of the accumulation process among experts, policy makers and citizens (Sterman, 2011). Our water–social SD model attempts to model the water conflicts caused by the interactions among different accumulated processes.

Water system performance is widely assessed by three criteria (Sánchez-Román et al., 2010): water stress, vulnerability, and reliability. Water stress is equal to the water withdrawal volume divided by the surface runoff (Vörösmarty et al., 2000). Vulnerability is defined as the magnitude of failure (Hashimoto et al., 1982). Reliability is calculated by the probability of the system being in a satisfactory state.

Xinjiang borders five countries. Its water supply is limited by the transnational watershed and water stress is relieved mainly through redistribution among municipal, industrial, and agricultural water consumption. Pressure on water resources directly affects agricultural water consumption and reduces the area of arable land. It is widely believed that rapid industrial development may put pressure on agricultural development. In this study, an SD model was used to simulate the accumulation processes of the water–social system, with the aim of enhancing the public's understanding of the leap-forward policy.

Water and regional development model for Xinjiang

Study area, model structures and data needed in the model

With an area of 16.6 billion km2, Xinjiang accounts for 17.2% of China. Xinjiang was originally an agriculture-oriented province. There are three mountain ranges in Xinjiang: the Altai Mountains in the north, Kunlun in the south and Tianshan in the center. Xinjiang is a water shortage region. For many years, the average precipitation was 154.50 mm, which is only 23.8% of the national average. In 2014, 94.7% of the total water supply was consumed by agriculture. It is also a poor region. The push to develop industry is urgent. With rapid economic growth in recent years, urbanization has also been rapid. Both economic development and population expansion increase water stress (Deng et al., 2011).

To explore the effectiveness of Xinjiang's leap-forward development policy, our integrated water resource–economy SD model includes four sectors: economy, population, agriculture, and water resource. These traditional categories in SD modeling are affected by urbanization (Forrester, 1971; Meadows et al., 1972; Ahmad & Prashar, 2010; Davies & Simonovic, 2011). Generally, SD models from an economic perspective omit water quality (Davies & Simonovic, 2011) and persistent pollution is not considered. Other sectors are omitted, because their dynamic inclusion will not produce any behavior modes not already contained within the existing sector (Meadows et al., 1972, p. 10).

Humans consume water for agriculture, industry, and residential (domestic) uses. These three sectors are named differently in different studies, such as municipal, agricultural, and environmental water demands (Ahmad & Prashar, 2010), or land use and irrigation, the economy, and population (Davies & Simonovic, 2011). Xu et al. (2002) projected the water demands for domestic, industrial and agricultural uses. In this paper, the sectors are named irrigation consumption, industrial water consumption and residential (domestic) water consumption. The total water resource for human water consumption includes the surface water supply and extracted groundwater (Xu et al., 2002; de Fraiture, 2007).

In Xinjiang, the industrial and residential water demands are guaranteed in advance, and the municipal water demand is relatively small (accounting for just 0.5% in 2010). The effects of limited water resource on agriculture are immediate. To evaluate the effects of booming industry on water resource scarcity, the industry sector is set to be exogenous. The criterion chosen to assess the system performance is the area of agriculture land (Figure 1).
Fig. 1.

Simplified representation of the model (PW means per capita/irrigated area/industry output water consumption).

Fig. 1.

Simplified representation of the model (PW means per capita/irrigated area/industry output water consumption).

Most model parameters and initial values are determined by the information collected. Data were collected from four sources: published government documents, published journals, informal semi-structured interviews and field observations. Most data were derived from Xinjiang Statistical Yearbooks 1990–2015 (Statistics Bureau of Xinjiang Uygur Autonomous Region, 2000–2015), and Xinjiang Water Resource reports 2000–2015.

The economic sector

Initially, industrial output is calculated by the input/output ratio. Capital is the only source of production. A part of the industry output is used for reinvestment, to further boost output (Chow & Li, 2002). Industry output growth is simulated by the Douglas-Pearce function, dependent upon capital investment and labor force. In a planned economy, in which the development process is controlled by a government development plan, industrial growth can be described by exponential or logistic functions. When there is a large gap between the current output and the ideal maximal output, the industry output growth process can be simulated by an exponential development curve. In the initial stages, new products are introduced, there are dramatic improvements in quality, costs fall sharply, and industrial output increases quickly. Eventually, when dramatic quality improvement and cost reduction opportunities are exhausted, and markets become saturated, the rate of industry output may slow. 
formula
1
where xind(t) is the output of production at time t, αind is the growth parameter of xind, and Kind is its carrying capacity.
In Xinjiang, the service industry's water consumption is included in domestic water consumption in statistical year books, in which water consumption in the economic sector only refers to primary and secondary industries. Gross domestic product, as an index of historical productivity, is chosen to measure industrial output. Industrial growth has been exponential. In 2000 and 2013, industrial output was RMB53.73 billion and RMB377.7 billion, respectively, which is relatively low compared to other regions. In 2013, industrial output in the coastal provinces, such as Guangdong and Jiangsu, reached RMB2,942.7 billion and RMB2,909.4 billion, respectively (National Bureau of Statistics of China (NBSC), 2013). The potential for Xinjiang's industry to develop is enormous. The industrial development process is modeled by function 2: 
formula
2
where xwind is the amount of water consumed by industrial production in time t, and xwindp is the per capita industrial output water consumption in time t. In 2013, xwind was 1,242.6 Mm3. Industrial output increased sharply from 1980 to 2010, mainly due to the petroleum industry.
Xu et al. (2002) modeled xwindp reduced by the scale of production increase. Simonovic (2002) simulated xwindp with an exogenous nonlinear function. Here, we calculate xwindp by a revised logistic function based on the assumption that xwindp is influenced by technology improvement and diffusion rate. 
formula
3
where αwindp is the growth parameter of xwindp, and Kwindp is its carrying capacity. xwindp was high in 2000, at about 22.27 m3/1,000 RMB. Due to technology improvement and a change in industry structure, xwindp decreased to 3.29 m3/1,000 RMB in 2013, less than the planned value of 5 m3/1,000 RMB. According to the Xinjiang Water Plan (Deng et al., 2011), xwindp should shrink to 1.88 m3/1,000 Yuan by 2030. As xwindp is already small, the gap between current xwindp and expected xwindp is small. It is simulated by a decreasing goal seeking function. In the base scenario, xwindp is set to decrease slowly with a 14.5% decrease rate to approach 1.88 m3/1,000 Yuan in 2030.

The population sector

Population is a traditional sector in water–social system studies. Population is seen as the major driving force of water exploitation and consumption, governing both municipal and agricultural water demand. Rapid population growth will be a significant challenge in the future. The logistic function was initially developed to simulate the population expansion process in 1844–1845 by Pierre François Verhulst. In water-related studies, population size is represented either as a function of water availability and economic welfare (Madani & Mariño, 2009; Bagheri et al., 2010), or water, employment and housing availability (Prodanovic & Simonovic, 2010). In the short term, population grows exponentially. The birth rate and death rate are combined as natural growth and calculated from the statistical yearbooks of Xinjiang (2001–2014) as 1.8%. 
formula
4
where xpop(t) is the population at time t, and αpop is the growth parameter of xpop.
The water consumption of the rural population is included in agricultural water consumption; the population sector concerns only the urban population. In Xinjiang, the urban population was 6,512,400 in 2000 and 10,069,300 in 2013. Compared to other regions, the population is relatively small. In 2013, the populations of Guangdong and Jiangsu provinces were 105,494,200 and 79,093,900, respectively. 
formula
5
where xwpop is the amount of urban water consumption in time t, xwpopp is per capita urban water consumption in time t and Swpopp is the minimal amount of per capita urban water consumption according to technological limitations. 
formula
6
where αwpopp is the growth parameter of xwpopp, and Kwpopp is its carrying capacity, set to 1 billion m3.

Agricultural sector

Land use changes are capable of bringing about a significant change in water consumption and the relative distribution of water demand among various competing uses. Most studies have used land use change to understand changes in water use (Ahmad & Prashar, 2010). In Xinjiang, urban, agricultural, and natural land use are the three main land use types. Agriculture land consumes most of the available water. Agriculture accounts for more than 70% of water use in Europe and 65–82% in the USA (Simonovic, 2002). In the year 2000, agricultural water demand in the Yellow River basin accounted for more than 70% of the total demand (Xu et al., 2002).

The main land stock in the agricultural sector is the area of agriculture land. Normally, agricultural water demand includes demand for farming, fisheries and livestock (Xu et al., 2002). In Xinjiang, all the farming land requires irrigation and is considered agriculture land. For simplicity, water demand for irrigation, fisheries and livestock is combined in agriculture land stock. The agriculture land demand changes with urbanization, development of new crop fields or new irrigation technology (Xu et al., 2002). In this model, the agriculture land growth process is simulated with a logistic growth function. When potential land for agriculture is large, the function changes to an exponential function. When the potential land area for agriculture is small, the function changes to a goal-seeking function.

Four data inputs are included in this sector: the initial irrigated area, the irrigated area growth rate, initial water use per unit area, and reduction rate due to improvement in irrigation techniques. 
formula
7
where xagr(t) is the area of agriculture land at time t, xwsagrp(t) is the amount of water available for agriculture, and xwagr(t) is the amount of water consumption by agriculture at time t.
The agriculture land increased quickly from 33,890 km2 in the year 2000 to 52,120 km2 in 2013. Technology improvement can improve water irrigation efficiency and enhance xwagrp, which is simulated by a decreasing goal-seeking function. With improvements in irrigation techniques, the water demand per hectare is expected to decrease with time. The decreasing trend is expressed in a nonlinear relationship. It is assumed to be a decreasing goal-seeking function. The global average xwagrp in the year 2000 (Simonovic, 2000) was 0.7–1. It was 0.8–1 in Eastern Europe, 2–2.5 in the USA, and 0.5–1.7 in Asia. 
formula
8
where xwagrp is per capita water consumption of agriculture land at time t, and Swagrp is the minimal amount of per capita irrigation water consumption according to technological limitations. 
formula
9
where αwagrp is the growth parameter of xwagrp, and Kwagrp is its carrying capacity.

Water supply

The total water resource supply includes the surface water supply and groundwater extraction (Xu et al., 2002; de Fraiture, 2007). Xinjiang is an arid region whose water mainly comes from mountain areas. In 2013, the average surface water runoff was about 90,090 Mm3, and groundwater runoff was 56,130 Mm3. The surface water supply was 47,659 Mm3, and groundwater supply was 11,038 Mm3 (Xinjiang Uygur Autonomous Region Water Conservancy Bureau, 2013). The surface water resource volume in the base scenario is an extension of the present trend. From 2005 to 2014, the surface water supply was relatively stable, in the range of 42.497 and 47.787 billion m3. The average surface water supply was 45.52 billion m3. The surface water supply in the model is set to be constant, at 45 billion m3.

Groundwater is an important part of the model. The groundwater water resource is modeled as an aggregated unit in the form of a single aquifer (Fernández & Selma, 2004). It is usually idealized as a wedge to estimate aquifer capacity. A single aquifer represents the volume of available groundwater accessible by pumping, and constrained by groundwater capacity. The groundwater supply is represented by the volume of the aquifer multiplied by a storage coefficient. In this model, the amount of groundwater withdrawn is represented by a logistic function. When the initial amount of groundwater withdrawn is small, it is an exponential function. When the amount of groundwater withdrawn is large, a goal seeking function is used. 
formula
10
where xgws(t) is the groundwater supply at time t, αgws is the growth parameter of xgws, and Kgws is its carrying capacity.

In 2000, the extracted groundwater volume was 54.2 billion m3. After 2005, the extracted groundwater volume increased sharply. In 2007, 2008 and 2009, the extracted groundwater volumes were 67.8 billion m3, 77.7 billion m3 and 87.4 billion m3, respectively. Along with the increasing amount of groundwater, over-exploration resulted in a falling groundwater table and a worse environmental system. αgws is 0.185, calibrated by historical data for xgws(t). As the volume of available groundwater is 23,610 Mm3, the maximal groundwater extraction should be 11,100 Mm3 when the extraction coefficient follows the international standard of 0.47. In 2013, the volume of groundwater extraction was 11,467 Mm3, indicating overexploitation. An extraction coefficient of over 1.2 indicates severe overexploitation (Deng et al., 2011). This model assumes that overexploitation might continue for several years. Kgws is set to be 13,760 Mm3, 20% higher than standard maximal groundwater extraction.

Model calibration and scenarios

The time horizon of the model is 30 years from 2000 to 2030 operated on an annual time step. The model build is based on the assumption that the model is a projection of historical development processes. Xinjiang's economic development process began around 2000. The period from 2000 to 2014 is used for calibration and validation of model relationships and verification of model performance (Figures 24).
Fig. 2.

The calibration of the exogenous indexes from year 2000–2014: (a) population size, (b) the amount of industry output, (c) the amount of groundwater extraction (line: simulated data; points: historical data).

Fig. 2.

The calibration of the exogenous indexes from year 2000–2014: (a) population size, (b) the amount of industry output, (c) the amount of groundwater extraction (line: simulated data; points: historical data).

Fig. 3.

The index for validation from year 2000–2014 (line: simulated data; points: historical data).

Fig. 3.

The index for validation from year 2000–2014 (line: simulated data; points: historical data).

Fig. 4.

The parameters of per capita water consumption from year 2000–2014: (a) the trend of irrigation water consumption per area of agriculture land, (b) per capita urban water consumption, (c) per capita industry output water consumption (line: simulated data; points: historical data).

Fig. 4.

The parameters of per capita water consumption from year 2000–2014: (a) the trend of irrigation water consumption per area of agriculture land, (b) per capita urban water consumption, (c) per capita industry output water consumption (line: simulated data; points: historical data).

Tests of model behavior evaluate the adequacy of behavior generated by the structure (Forrester, 1971; Sterman, 2002) including behavior reproduction and behavior prediction, extreme condition tests and numerical integration tests. The family of behavior reproduction tests examines how well behavior generated by the model matches the observed behavior of the real system. Main variables are examined through statistical parameters of mean of the relative error (MRE), coefficient of determination (R2) and point-by-point comparisons, which represent a widespread ‘accepted symptom generation test’ (Sterman, 2002). The MRE shows the difference between simulated data and the observed data. R2 measures what percentage of the variables' historical behavior can be simulated by the model. The variables tested are Agricultural Irrigation Area, Gross Industrial Product, and Total Population.

Historical data and the simulation results show the agricultural, industrial and population development processes. The comparison indicates the model's validation (Figure 2). All the standard errors of estimation are within 10%. The population in the model is 21.2 million in the year 2014. Compared to the statistical data (21.8 million), there is a 0.6 million gap. The comparison of the historic data and simulated behavior of the amount of agriculture land, industry output, and population demonstrates the reliability of the simulation. A sensitivity test, extreme condition test and numerical integration test are carried out. The results show that the model is robust. The model is therefore effective and efficient for projection. The dynamic model can project the social, economic, population, and water exploitation processes.

Four scenarios were created to simulate the effects of the leap-forward development policy and the corresponding strategies to overcome the water shortage problems. The possible future sustainable dynamics of the system under each scenario were explored. The simulated results provide plenty of information in all sectors of the model and cover the whole time zone. Scenarios were projected over the time horizon for which the model was designed, covering the period 2011–2030.

Scenario 1. Base scenario

Scenario 1 is a business as usual scenario, the status quo scenario. It provides a reference against which to compare scenarios incorporating the leap-forward development policy and corresponding strategies. It simulates a future in which the historical trends are projected, but is not a prediction of what will happen. All model parameters remain unchanged, assuming the continuation of current trends over the next 20 years. The base scenario's assumptions, including the growth rates of the population (2.95%), economy (13%) and per capita water demand (0.35%), the water extraction and distribution rules and the technology implications, are common to all scenarios. The industry development rate is calibrated as 16.2%. This assumes that industry's increasing rate is exogenous, and not affected by water stress.

Scenario 2. Leap-forward development policy

The second scenario is a simulation of leap-forward development policy modification, and assumes a significant stimulus in industrial capital investment. The industry output increase rate, 16.2%, is relatively high in the base scenario. Under the leap-forward policy it might be either amplified or lessened. In scenarios 2.1 and 2.2, the industry output increase rates are set to be 25% and 10% decrease. In scenarios 2.3 and 2.4, the industry output increase rates are set to be 10% and 5% increase. Scenario 2 is intended to examine the potential effects on the economy and on water balance of implementing the leap-forward development policy.

Scenario 3. Industrial water-saving strategies

The industrial structural transformation and technology improvement rate might affect xwindp and thus the amount of water consumed by industry. In 2009, the xwindp of mainland China was 11.62 m3/1,000 RMB. The xwindp of the most efficient city, Tianjiang, was 1.18 m3/1,000 RMB (National Bureau of Statistics of China (NBSC), 2010). In Xinjiang, industrial development may increase or decrease water consumption. The minimal amount of per capita industrial output water consumption in the future, Kwindp, is a determinant factor in the water balance system. Scenarios 3.1 and 3.2 suppose a −50% and +50% change in Kwindp. The change rate of per capita industrial output water consumption, αwindp, represents the technology diffusion rate. It is a measurement of how fast industry applies water-saving technology. Scenarios 3.3 and 3.4 set a −10% and +10% change of αwindp.

Scenario 4. Agricultural water-saving strategies

The third scenario emphasizes agricultural water-saving strategies. Changes may result from increased irrigation technology. Historically, flood irrigation has been the most common irrigation method. Currently, this technology is gradually being substituted by drip irrigation to enhance water resource utility. Scenarios 4.1 and 4.2 suppose a −20% and −10% change of Kwagrp. These scenarios simulate the optimal and pessimistic irrigation technology improvement respectively. Scenarios 4.3 and 4.4 simulate the change of technology diffusion rate and assume −10% and +10% change in αwagrp.

Scenario 5. Sustainable strategies

To achieve sustainable development, multifunctional strategies would be implicated. Scenario 5 is a combination of all the former scenarios. Scenario 5.1 sets a 10% decrease of the minimal amount of per capita water consumption of agriculture land in the future, and a 50% decline of the minimal amount of per capita industrial output water consumption in the future. Scenario 5.2 is scenario 5.1 combined with a 10% increase of industry output growth rate.

Model results

Scenario 1: Base scenario of future projection

Prior to considering any future policy options, a model simulation is performed by extending current trends over the simulation period. Figure 5 shows the variation in the main system state variables over the simulation period of 2015 through 2030 for all the scenarios. The main system state variables are industrial output and agriculture land area. The current trend indicates that industry increases exponentially over time. The simulated results show that industrial output will reach RMB1,082.8 billion in 2020, and RMB4,859.9 billion in 2030. This increase coincides with the historical trend. Xinjiang's industry output has generally doubled every 5 years.
Fig. 5.

The forecasting of key indexes from year 2015–2030 (scenario1 (sc1) is a business-as-usual scenario; sc2.1, sc2.2, sc2.3 and sc2.4 set the industry output increase rate as −25% decrease, −10% decrease, 10% increase and 5% increase individually; sc3.1 and sc3.2 set a −50% and 50% change of the minimal amount of per capita industrial output water consumption in the future. sc3.3 and sc3.4 set a −10% and 10% change of the decline rate of amount of per capita industrial output water consumption. sc4.1 and sc4.2 set a 5% decrease and 10% decrease of the minimal amount of per capita water consumption of irrigation in the future. sc4.3 and sc4.4 set a 10% decrease and 10% increase of the decline rate of amount of per capita water consumption of irrigation. sc5.1 sets a 10% decrease of the minimal amount of per capita water consumption of irrigation in the future, and a 50% decline of the minimal amount of per capita industrial output water consumption in the future. sc5.2 is sc5.1 combined with a 10% increase of industry output growth rate).

Fig. 5.

The forecasting of key indexes from year 2015–2030 (scenario1 (sc1) is a business-as-usual scenario; sc2.1, sc2.2, sc2.3 and sc2.4 set the industry output increase rate as −25% decrease, −10% decrease, 10% increase and 5% increase individually; sc3.1 and sc3.2 set a −50% and 50% change of the minimal amount of per capita industrial output water consumption in the future. sc3.3 and sc3.4 set a −10% and 10% change of the decline rate of amount of per capita industrial output water consumption. sc4.1 and sc4.2 set a 5% decrease and 10% decrease of the minimal amount of per capita water consumption of irrigation in the future. sc4.3 and sc4.4 set a 10% decrease and 10% increase of the decline rate of amount of per capita water consumption of irrigation. sc5.1 sets a 10% decrease of the minimal amount of per capita water consumption of irrigation in the future, and a 50% decline of the minimal amount of per capita industrial output water consumption in the future. sc5.2 is sc5.1 combined with a 10% increase of industry output growth rate).

The main variable in the agricultural sector is the annual agriculture land area. Growth in the area of agriculture land follows an inverted U shape. The amount of agriculture land is limited by the amount of water available for agriculture. The amount of water available for agriculture is controlled by the amount of water used by industry, population water use, and total water supply. The amount of groundwater extracted will continue to increase and approach maximal groundwater extraction. The amount of industrial water consumption will continue to increase. An extension of this trend over time could severely affect agricultural production. The fraction of water resource used in the agriculture sector will decrease. Combining these two effects, the indicator shows a steady increase in the area of agriculture land in the first 7 years of the simulation period. The agriculture land area reaches a peak (58,797.9 km2) in 2022, and then begins to decrease. It declines to 53,815.3 km2 in 2030.

Scenario 2: Leap-forward development strategy

Compared to the base scenario, when the industry development rate decreases by −25% and −10% (scenarios 2.1, 2.2), industrial output might be 41.3% and 19% lower in 2030, at RMB2,854.5 billion and RMB3,937 billion, respectively. In these scenarios, water stress is released by constraining industry water consumption. The decrease in agriculture land is delayed for several years, with agriculture land area reaching its peak of 59,536.02 km2 in 2025 and 59,098.47 km2 in 2022, respectively. In 2030, the area of agriculture land is 57,794.80 km2 (scenario 2.1) and 55,629.40 km2 (scenario 2.2.) (Table 1 and Figure 3).

Table 1.

Projected results of main indexes in each scenario.

 xagr in year 2030 (km2) xagr in year 2030 compared to sc1 (km2) Changed % xagr peak year xagr peak area (km2) xagr peak area compared to sc1 (km2) Changed % xind in year 2030 (billion) xwind in year 2030 (Mm3) 
sc1 53,815.35 – – 2022 58,797.89 – – 4,859.90 10,038.23 
sc2.1 57,794.77 3,979.42 7.39 2025 59,536.02 738.13 1.26 2,854.46 5,895.96 
sc2.2 55,629.43 1,814.08 3.37 2022 59,098.47 300.58 0.51 3,937.02 8,132.004 
sc2.3 51,640.88 −2,174.47 −4.04 2021 58,521.31 −276.58 −0.47 5,981.64 12,355.22 
sc2.4 52,777.01 −1,038.34 −1.93 2021 58,642.54 −155.35 −0.26 5,393.61 11,140.00 
sc3.1 57,864.03 4,048.68 7.52 2024 59,754.56 956.67 1.63 4,859.90 5,905.69 
sc3.2 49,766.67 −4,048.68 −7.52 2021 58,145.29 −652.6 −1.11 4,859.90 14,170.78 
sc3.3 53,536.1 −279.25 −0.52 2022 58,684.19 −113.7 −0.19 4,859.90 10,296.93 
sc3.4 54,039.36 224.01 0.42 2022 58,902.34 104.45 0.18 4,859.90 9,834.198 
sc4.1 55,914.77 2,099.42 3.90 2023 60,183.47 1,385.58 2.36 4,859.90 10,038.23 
sc4.2 58,184.64 4,369.29 8.12 2024 61,812.05 3,014.16 5.13 4,859.90 10,038.23 
sc4.3 53,579.45 −235.9 −0.44 2022 58,560.75 −237.14 −0.40 4,859.90 10,038.23 
sc4.4 54,021.21 205.86 0.38 2022 59,023.99 226.1 0.38 4,859.90 10,038.23 
sc5.1 62,562.04 8,746.69 16.25 2026 63,496.58 4,698.69 7.99 4,859.90 10,038.23 
sc5.2 61,147.32 7,331.97 13.62 2025 62,879.01 4,081.12 6.94 5,981.64 12,355.22 
 xagr in year 2030 (km2) xagr in year 2030 compared to sc1 (km2) Changed % xagr peak year xagr peak area (km2) xagr peak area compared to sc1 (km2) Changed % xind in year 2030 (billion) xwind in year 2030 (Mm3) 
sc1 53,815.35 – – 2022 58,797.89 – – 4,859.90 10,038.23 
sc2.1 57,794.77 3,979.42 7.39 2025 59,536.02 738.13 1.26 2,854.46 5,895.96 
sc2.2 55,629.43 1,814.08 3.37 2022 59,098.47 300.58 0.51 3,937.02 8,132.004 
sc2.3 51,640.88 −2,174.47 −4.04 2021 58,521.31 −276.58 −0.47 5,981.64 12,355.22 
sc2.4 52,777.01 −1,038.34 −1.93 2021 58,642.54 −155.35 −0.26 5,393.61 11,140.00 
sc3.1 57,864.03 4,048.68 7.52 2024 59,754.56 956.67 1.63 4,859.90 5,905.69 
sc3.2 49,766.67 −4,048.68 −7.52 2021 58,145.29 −652.6 −1.11 4,859.90 14,170.78 
sc3.3 53,536.1 −279.25 −0.52 2022 58,684.19 −113.7 −0.19 4,859.90 10,296.93 
sc3.4 54,039.36 224.01 0.42 2022 58,902.34 104.45 0.18 4,859.90 9,834.198 
sc4.1 55,914.77 2,099.42 3.90 2023 60,183.47 1,385.58 2.36 4,859.90 10,038.23 
sc4.2 58,184.64 4,369.29 8.12 2024 61,812.05 3,014.16 5.13 4,859.90 10,038.23 
sc4.3 53,579.45 −235.9 −0.44 2022 58,560.75 −237.14 −0.40 4,859.90 10,038.23 
sc4.4 54,021.21 205.86 0.38 2022 59,023.99 226.1 0.38 4,859.90 10,038.23 
sc5.1 62,562.04 8,746.69 16.25 2026 63,496.58 4,698.69 7.99 4,859.90 10,038.23 
sc5.2 61,147.32 7,331.97 13.62 2025 62,879.01 4,081.12 6.94 5,981.64 12,355.22 

Note: sc1 is a business-as-usual scenario; sc2.1, sc2.2, sc2.3 and sc2.4 set the industry output increase rate −25% decrease, −10% decrease, 10% increase and 25% increase individually; sc3.1 and sc3.2 set a −50% and 50% change of the minimal amount of per capita industrial output water consumption in the future. sc3.3 and sc3.4 set a −10% and 10% change of the decline rate of amount of per capita industrial output water consumption. sc4.1 and sc 4.2 set a 20% decrease and 10% decrease of the minimal amount of per capita water consumption of irrigation in the future. sc4.3 and sc4.4 set a 10% decrease and 10% increase of the decline rate of amount of per capita water consumption of irrigation. sc5.1 sets a 10% decrease of the minimal amount of per capita water consumption of irrigation in the future, and a 50% decline of the minimal amount of per capita industrial output water consumption in the future. sc5.2 is sc5.1 combined with a 10% increase of industry output growth rate.

In scenarios 2.3 and 2.4, 10% and 5% increases in industry output development rate result in 23% and 10.98% industrial output increase in 2030, reaching RMB5,981.6 billion and RMB5,393.61 billion, respectively. Correspondingly, water consumption for agriculture land decreases from 45 billion m3 in scenario 1 to 42.7 billion m3 in scenario 2.3 and 43.94 billion m3 in scenario 2.4. In both scenarios, the area of agriculture land reaches its peak in 2021, one year earlier than in the base scenario. The peak area amounts are 58,521.02 km2 and 58,642.54 km2, respectively, for scenarios 2.3 and 2.4. Compared with the base scenario, there is a 276.58 km2 and 155.35 km2 decrease in agriculture land area, respectively. In 2030, the area of agriculture land in scenario 2.3 is 51,640.88 km2 and in scenario 2.4 it is 52,777.01 km2. This represents a loss of 2,174.47 km2 and 1,038.34 km2 compared to the base scenario (Table 1 and Figure 3)

Scenario 3: Industrial water-saving strategies

Scenario 3.1 shows that a 50% decrease in the maximal amount of per capita industrial output water consumption (Kwindp) leads to a 40% decrease in industrial water demand in 2030. The area of agriculture land increases from the base scenario by 7.52% to 4,048.68 km2. Scenario 3.2 shows the opposite results. The industrial structure determines the amount of industrial water consumption. In the long term, developing more water-saving industrial enterprises is an effective way to avoid a decrease in cultivated land area.

Scenarios 3.3 and 3.4 slow down/speed up the agriculture land contraction process although the effects are relatively limited. With a 10% technology diffusion rate increase, the xagr peak year is 2022, one year later than in the base scenario. The area of agriculture land increases by 0.42% in 2030. There is a 279.25 km2 agriculture land increase in scenario 3.3 compared with the base scenario (224.01 km2 agriculture land decline in scenario 3.4) (Table 1 and Figure 3). Implementing efficient industrial water consumption technology may postpone the cultivated land reduction process, although the contribution of improving the technology diffusion rate is relatively small and its effect is short term.

Scenario 4: Agricultural water-saving strategies

A 20% decrease (scenario 4.1) and 10% decrease (scenario 4.2) in the minimal amount of per capita water consumption of agriculture land, expand the area of agriculture land by 2,099.42 km2 (a 3.9% increase) and 4,369.29 km2 (an 8.12% increase), respectively. The peak year occurs 1 year later (scenario 4.1) or 2 years later (scenario 4.2) than in the base scenario. The effects of scenarios 4.3 and 4.4 are moderate. The agriculture land area is decreased by 235.71 km2 in scenario 4.3 and increased by 205.86 km2 in scenario 4.4 in 2030. The changes of growth rate of amount of per capita water consumption of agriculture land, scenarios 4.3 and 4.4, have no effect on xagr peak year (Table 1 and Figure 3). Similar to scenario 3, in the short term, improving irrigation technology may postpone the cultivated land reduction process. In the longer term, planting crops with improved drought tolerance may improve water usage efficiency more effectively.

Scenario 5: Sustainable strategies

Scenario 5.1, to decrease Kwindp, the minimal amount of per capita industrial output water consumption in the future, by 50% and to decrease Kwagrp, the minimal amount of per capita industrial output water consumption in the future, by 10%, may expand the area of agriculture land to 62,562.04 km2. It may extend the peak year to 2026. Compared to the area in the peak year of the base scenario, the area of agriculture land in 2030 in scenario 5.1 expanded by 3,764.15 km2. Therefore, the expanded area would support the implication of scenario 5.2. To combine scenario 5.1 with a 10% increase of industry output growth rate may extend the peak year to 2005, and expands the area in year 2030 to 61,147.32 km2. Compared to the area in the peak year of the base scenario, the area of agriculture land in 2030 in scenario 5.2 expanded by 2,349.43 km2.

Combining the results of the five scenarios, suggests the leap-forward strategy might influence the area of agriculture land by changing the water distribution structure. The leap-forward development policy enforces the stress of water resource scarcity. Water scarcity becomes more acute, and the current water supply cannot support the leap-forward development policy with current water usage efficiency. In the short term, implementing efficient industrial water consumption technology and promoting water-saving irrigation technology may postpone the cultivated land reduction process. In the long term, developing more water-saving industrial enterprises, and planning for more drought-tolerant plants, may be effective for avoiding a reduction in cultivated land area.

Conclusion and discussion

Leap-forward investment in China is popular. Most Chinese regions are facing severe water constraints. Given the importance of water resources in an economy, appropriate water management is needed to achieve sustainable regional development under water constraints. An SD model was built to simulate the water constraints in a leap-forward developing region. Embedding the logistic growth function within an SD framework, the model provides a universally implementable model for sustainable water planning. The dynamic current and projected future water supply, demand, and water use efficiency are captured. The model boundary, structure and parameter setting processes and rationale are described in detail. Potential water management strategies can be simulated to test system performance in the future.

The effectiveness and efficiency of the model in capturing the dynamic supply and demand development process is tested empirically in Xinjiang, China. Four potential water management strategies are simulated. The effectiveness and efficiency of strategies implementing multifunctional water projects to support sustainable economic growth are projected. The results suggest that when water scarcity becomes more acute, the current water supply cannot support the leap-forward development policy under current water usage efficiency. In the short term, implementing efficient industrial water consumption technology and promoting water-saving irrigation technology may postpone the cultivated land reduction process. With a 10% industrial water-saving technology diffusion rate increase, the xagr peak year is 2022, one year later than in the base scenario. The area of agriculture land increases by 0.42% in 2030. A 10% increase in the growth rate of per capita water consumption of agriculture land increases the agriculture land by 0.38% in 2030. In the long term, developing more water-saving industrial enterprises and planning for more drought-tolerant plants are effective ways to avoid a decrease in cultivated land area. A 50% decrease in the maximal amount of per capita industrial output water consumption (Kwindp) leads to a 40% decrease in industrial water demand in 2030. The area of agriculture land increases by 4,048.68 km2 (7.52%) compared to the base scenario. A 10% decrease in the minimal amount of per capita water consumption of agriculture land may expand the area of agriculture land by 8.12%, and postpone the peak year by 4 years. To combine both industry and agriculture structure transformation policies, a relatively sustainable future would be achieved.

The results of the model provide possible outcomes under different scenarios and help to improve the understanding of the complicated water system. The model would help water managers in Xinjiang to implement efficient water strategies and promote sustainable water management planning. The model provides a practical and an analytical tool for water resource managers to evaluate the effects of specific policies from the demand side. Changing economic and agricultural structures to be more water saving, and improving water-saving technology would relieve water stress in the long term, and improving the technology diffusion rate would be helpful in the short term. Given limitations of time, resources and data, the model is an idealization of the complex water–social system. In the future, water pollution, and water's temporal and spatial distribution, could be simulated by an expanded model.

Acknowledgements

This study is supported by the National Natural Science Foundation of China (No. 41601611), a project of Xinjiang Key Laboratory of Water Cycle and Utilization in Arid Zone, Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences (XJYS0907-2012-03), Sun Yat-sen University young teacher cultivation project (No. 16wkpy19). The authors would also like to thank the editor of Water Policy and the anonymous reviewers for their precious time and constructive comments on the paper.

1

The term billion refers to 109, here and throughout the paper.

References

References
Chow
G. C.
Li
K. W.
(
2002
).
China's economic growth: 1952–2010
.
Economic Development and Cultural Change
51
,
247
256
.
Dai
S.
Li
L.
Xu
H.
Pan
X.
Li
X.
(
2013
).
A system dynamics approach for water resources policy analysis in arid land: a model for Manas River Basin
.
Journal of Arid Land
5
,
118
131
.
doi:10.1007/s40333-013-0147-1
.
Davies
E. G. R.
Simonovic
S. P.
(
2011
).
Global water resources modeling with an integrated model of the social-economic-environmental system
.
Advances in Water Resources
34
,
684
700
.
doi:10.1016/j.advwatres.2011.02.010
.
Dawadi
S.
Ahmad
S.
(
2012
).
Changing climatic conditions in the Colorado River Basin: implications for water resources management
.
Journal of Hydrology
430–431
,
127
141
.
doi:10.1016/j.jhydrol.2012.02.010
.
De Fraiture
C.
(
2007
).
Integrated water and food analysis at the global and basin level: an application of WATERSIM
.
Water Resources Management
21
,
185
198
.
doi:10.1007/s11269-006-9048-9
.
Deng
M. J.
Li
X. Q.
Long
A. H.
Zhang
Y.
Wang
D. X.
(
2011
).
Regulation of supply and demand structure of the water resources and support economic and social leap-forward development of protection measures
.
Arid Land Geography
34
,
379
390
(in Chinese)
.
Fernández
J. M.
Selma
M. A. E.
(
2004
).
The dynamics of water scarcity on irrigated landscapes: Mazarrón and Aguilas in south-eastern Spain
.
System Dynamics Review
20
,
117
137
.
doi:10.1002/sdr.290
.
Forrester
J. W.
(
1971
).
World Dynamics
.
Wright-Allen Press
,
Cambridge, MA
.
Iglesias
E.
Blanco
M.
(
2008
).
New directions in water resources management: the role of water pricing policies
.
Water Resources Research
,
44
(
6
),
259
269
.
Li
X.
Jiang
F.
Li
L.
Wang
G.
(
2011
).
Spatial and temporal variability of precipitation concentration index, concentration degree and concentration period in Xinjiang, China
.
International Journal of Climatology
31
(
11
),
1679
1693
.
Madani
K.
Mariño
M.
(
2009
).
System dynamics analysis for managing Iran's Zayandeh-Rud River Basin
.
Water Resources Management
23
,
2163
2187
.
doi:10.1007/s11269-008-9376-z
.
Meadows
D. H.
Meadows
D. L.
Randers
J.
(
1972
).
The Limits to Growth
.
Chelsea Green Publishing
,
New York
.
Mirchi
A.
Madani
K.
Watkins
D.
Ahmad
S.
(
2012
).
Synthesis of system dynamics tools for holistic conceptualization of water resources problems
.
Water Resources Management
26
,
2421
2442
.
doi:10.1007/s11269-012-0024-2
.
National Bureau of Statistics of China (NBSC)
(
2010
).
China Statistic Year Book 2010
.
Chinese Statistics Express
,
Beijing, China
.
National Bureau of Statistics of China (NBSC)
(
2013
).
Chinese Statistics Yearbook 2013
.
Chinese Statistics Express
,
Beijing, China
.
Prodanovic
P.
Simonovic
S. P.
(
2010
).
An operational model for support of integrated watershed management
.
Water Resources Management
24
,
1161
1194
.
doi:10.1007/s11269-009-9490-6
.
Qaiser
K.
Ahmad
S.
Johnson
W.
Batista
J.
(
2011
).
Evaluating the impact of water conservation on fate of outdoor water use: a study in an arid region
.
Journal of Environmental Management
92
(
8
),
2061
2068
.
Rajasekaram
V.
Simonovic
S. P.
Nandalal
K. D. W.
(
2003
).
Computer support for implementation of a systemic approach to water conflict resolution
.
Water International
28
(
4
),
454
466
.
Rauch
W.
Bertrand-Krajewski
J. L.
Krebs
P.
Mark
O.
Schilling
W.
Schütze
M.
Vanrolleghem
P. M.
(
2002
).
Deterministic modelling of integrated urban drainage systems
.
Water Science and Technology
45
,
81
94
.
Sánchez-Román
R.
Folegatti
M.
Orellana González
A.
(
2010
).
Water resources assessment at Piracicaba, Capivari and Jundiaí River Basins: a dynamic systems approach
.
Water Resources Management
24
,
761
773
.
doi:10.1007/s11269-009-9470-x
.
Saysel
A. K.
Barlas
Y.
Yenigün
O.
(
2002
).
Environmental sustainability in an agricultural development project: a system dynamics approach
.
Journal of Environmental Management
64
(
3
),
247
260
.
Simonovic
S. P.
(
2002
).
World water dynamics: global modeling of water resources
.
Journal of Environmental Management
66
,
249
267
.
doi:10.1006/jema.2002.0585
.
Sterman
J. D.
(
2002
).
All models are wrong: reflections on becoming a systems scientist
.
System Dynamics Review
18
,
501
531
.
doi:10.1002/sdr.261
.
Sterman
J. D.
(
2011
).
Communicating climate change risks in a skeptical world
.
Climatic Change
108
(
4
),
811
826
.
Tidwell
V. C.
Van Den Brink
C.
(
2008
).
Cooperative modeling: linking science, communication, and ground water planning
.
Ground Water
46
,
174
182
.
doi:10.1111/j.1745-6584.2007.00394.x
.
Vörösmarty
C. J.
Green
P.
Salisbury
J.
Lammers
R. B.
(
2000
).
Global water resources: vulnerability from climate change and population growth
.
Science
289
,
284
288
.
Winz
I.
Brierley
G.
Trowsdale
S.
(
2009
).
The use of system dynamics simulation in water resources management
.
Water Resources Management
23
,
1301
1323
.
doi:10.1007/s11269-008-9328-7
.
Xinjiang Uygur Autonomous Region Water Conservancy Bureau
(
2013
).
Xinjiang Water Resource Annual Report
. .
Xu
Z. X.
Takeuchi
K.
Ishidaira
H.
Zhang
X. W.
(
2002
).
Sustainability analysis for Yellow River water resources using the system dynamics approach
.
Water Resources Management
16
,
239
261
.
doi:10.1023/a:1020206826669
.
Zhou
L. A.
(
2007
).
Governing China's local officials: an analysis of promotion tournament model
.
Economic Research Journal
7
,
36
50
(in Chinese)
.