A Trade-Off Method Between Environment Restoration and Human Water Consumption: a Case Study in Ebinur Lake

Journal of Cleaner Production 217 (2019) 732e741

Contents lists available at ScienceDirect

Journal of Cleaner Production

journal homepage: www.elsevier.com/locate/jclepro

A trade-off method between environment restoration and human water consumption: A case study in Ebinur Lake

* Hongwei Zeng a, Bingfang Wu a, b, , Weiwei Zhu a, Ning Zhang c a Key Laboratory of Digital Earth Science, Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing, 100194, China b University of Chinese Academy of Sciences, Beijing, 100049, China c Department of Geography, The Ohio State University, Columbus, OH, 43202, United States article info abstract

Article history: The shrinkage of inland lakes has become a major problem in arid region. Human activities are Received 8 June 2018 competing for water resources with sustainable lake ecosystem. Considering improving irrigation effi- Received in revised form ciency does not help on water saving at basin scale and reducing ET through mulching, zero tillage and 17 December 2018 deficit irrigation will lead to crop yield loss, this study proposed an alternative way that returning Accepted 28 December 2018 cropland to lake to achieve the balance between endorheic lake restoration and human water con- Available online 12 January 2019 sumption. The approach consists of four procedures: (1) identifying the target annual inflow (Qtar)of endorheic lake; (2) calculating available consumable water (ACW) in the upstream based on Q ; (3) Keywords: tar Endorheic lake setting ET deduction goals using water consumption balance approach; (4) estimating crop area to be Environment restoration deducted. This approach is implemented at Ebinur Lake of China under two restoration scenarios (lake 2 2 2 Available consumable water area restored to 522 km and 800 km ). The results indicate that to restore the Ebinur Lake to 522 km , Evapotranspiration the annual inflow should be increased to 602.6 106 m3, and human water consumption should be Return cropland to water reduced by 320.3 106 m3. If the restoration goal of Ebinur Lake is set to 800 km2, an annual inflow of 876.6 106 m3 is required, and the human water consumption should be reduced by 594.2 106 m3. The over-expansion of cropland and artificial forest is found to be the main reasons for the shrinkage of Ebinur Lake. Finally, the feasible solutions for ecosystem restoration of Lake Ebinur are provided for both scenarios. The methodology adopted in this research can be applied to the restoration of other inland lakes in arid regions. © 2018 Published by Elsevier Ltd.

1. Introduction lake on earth, has decreased to 12% of its original size during the past decades (AghaKouchak et al., 2015). As the former largest lake Inland lakes are rich in aesthetic, cultural, economic, recrea- in the world, Aral Sea lost 82% of its size from 1960s to 2011 tional, scientific and ecological values. They are the major surface (Micklin, 2007). water resources for arid regions (Bai et al., 2012), and serve as the The shrinkage of inland lakes is due to inflow reduction caused major food source for migratory birds (AghaKouchak et al., 2015). by climate change and human activities, especially the agriculture However, the shrinkage of inland lakes has become a major prob- water usage in the upstream (Liu et al., 2013). For example, the lem in arid region. For example, about 121 lakes in the semi-arid dramatic shrinkage and salinization of Aral Sea was caused by the region of China that bordering the Asian Gobi desert were dried expanded irrigation and the huge irrigation systems in the desert up from the 1970s to 2000s (Liu et al., 2013). The Lake Chad in Africa (Micklin, 2007). The water loss of Lake Urmia was attributed to the decreased more than 90% in area from the 1960s to 2000 (Coe and construction of dams, irrigation projects and overuse of surface Foley, 2001). The Lake Urmia, which is one of the largest saltwater water and groundwater (AghaKouchak et al., 2015; Yamaguchi et al., 2012). It has been estimated that agricultural water consumption accounted for about 75% to 85% of global water usage (Perry, 2007 Foley et al., 2005). The over-expansion of cropland to * Corresponding author. State Key Laboratory of Remote Sensing Science, Insti- ensure food security has profoundly increased agriculture water tute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing, consumption and led to insufficient water supply for natural eco- 100101, China. E-mail address: [email protected] (B. Wu). systems (Xue et al., 2017). In addition, the lake shrinkage in turn https://doi.org/10.1016/j.jclepro.2018.12.284 0959-6526/© 2018 Published by Elsevier Ltd. H. Zeng et al. / Journal of Cleaner Production 217 (2019) 732e741 733 enhances the usage of groundwater for irrigation and accelerates are (1) how much cropland should be reduced to achieve the bal- the drop of the groundwater table (Liu et al., 2013). In general, the ance between lake restoration and human water consumption? (2) shrinkage of inland lakes in arid region can cause severe economic, Is returning cropland to water a feasible way? environmental and health problems (AghaKouchak et al., 2015; This paper is organized in four sections to demonstrate our Yamaguchi et al., 2012), including a decline in waterfowl, water approach and results. In the next section (“Method”), the whole set salinization (Micklin, 2007), dust storm (Micklin, 1988) and of methodology is introduced, including (1) identifying the target epidemic diseases (Hashizume et al., 2010). annual inflow with restoration goals; (2) calculating available Human activities are competing for water resources with sus- consumable water (ACW) in the upstream; (3) estimating the tainable lake ecosystem, To stop lake shrinkage, or to restore the returned water area from cropped land. A study case of applying lake size in endorheic basins, the inflow of inland lake should be our methodology to Lake Ebinur and its results are provided in increased (Gao et al., 2018), and agriculture water usage should be “Case study and Results” section. Finally, we discussed the accuracy, reduced (Xue et al., 2016). This is a trade-off of water allocation driven forces caused water shrinkage and implication of our between human (agricultural) water consumption and maintaining methods and conclusions were drawn in the final section. healthy lake ecosystem. How to reduce the agricultural water usage in the upstream? 2. Methods Enhancing irrigation efficiency (IE) and reducing evapotranspiration (ET) are two commonly applied strategies. Other strategies Fig. 1 illustrates the general framework of this study. It consists include changing cropping patterns and improving crop cultivars. of 4 major processes, (ⅰ) identifying the target annual inflow (Qtar) Enhancing IE has been widely encouraged to reduce agricultural of endorheic lake under two restoration goals; (ⅱ) calculating ACW water usage (Perry et al., 2009). For example, Xue et al. (2017) in the upstream according to Qtar;(ⅲ) setting ET deduction goals studied the trade-off between environmental inflows and agricul- based on water consumption balance; (ⅳ) estimating crop area to tural water usage and suggested to adopt drip irrigation. However, be deducted to achieve the ET deduction goal. more evidences support that increasing IE only save water at farm scale rather than at watershed or basin scale (Grafton et al., 2018). 2.1. Target annual inflow estimation The adoption of advanced irrigation technologies actually reduces the return flows of the lake and slowed down aquifer recharge at Environment healthiness and biodiversity of endorheic lake basin scale (Ward and Pulido-Velazquez, 2008). Therefore, it is depends on the size of water body, which is greatly affected by the impossible to increase the inflow of endorheic lake at basin scale annual inflow from upstream. The relationship between lake area through enhancing IE. and annual inflow is complicated, and it is impacted by climate, Reducing ET is another promising way to save water for lake terrain, shape of lake, and human factors. Xue et al. (2016) proposed restoration, considering ET is the dominate source of water losses an approach to calculate the environmental flow in oasis areas within a basin (Hellegers et al., 2010; Perry et al., 2009). More and using Bayesian networks (BNs). However, this approach is complex more studies confirmed that water management should focus on and has great uncertainty related to the availability of socio- reducing ET for the basin with high water depletion (Kendy et al., psychological variables (Poppenborg and Koellner, 2014). In this 2004; Nakayama et al., 2006). In water resource management, ET study, the curve fitting approaches, including Linear, Polynomial, can be decomposed to uncontrollable ET (ETunc), which is the ET ExpDex, Boltzmann, Log3P1, Bradley, Power1, and Logistic fitting from the natural environment, and controllable ET (ETcon), which is functions are adopted using SPSS to quantify the relationship be- the ET from human interventions (Wu et al., 2014). Owing to the tween lake area and annual inflow based on long-term measure- uncontrollable characteristic of ETunc, reducing ETcon in the up- ments. The fitting approach with the highest coefficient of stream is a more applicable way of reducing water consumption. determination (R2) between the fitted and observed values is The common methods of reducing ETcon include mulching, zero selected as the optimal model. The optimal model is then used to tillage, and regulated deficit irrigation. Studied have revealed a estimate the target inflow (Qtar) with the target lake area under positive linear relationship between crop yield and ET (Ward and different restoration scenarios. Pulido-Velazquez, 2008), thus reducing ET is associated with less available water and a loss in crop yield (Perry, 2011). For example, 2.2. Available consumable water (ACW) soil tillage can reduce ET by 9%, but reduce yield by 20% at the same time (Yan et al., 2015). Regulated deficit irrigation can significantly Endorheic basin consists of two parts, the endorheic lake and it’s reduce ET but at the meanwhile reduce yield by 9% (Howell, 1990; upstream. The water balance of endorheic lake is described as: Steduto et al., 2007; Yan et al., 2015). Considering enhancing IE does not help on water saving at basin Qin ¼ Pe e ETe e DSe (1) scale and crop yield loss always comes along with reducing ET, we are here to seek alternative way to balance water requirement from where Qin,Pe,ETe and DSe are respectively the annual inflow, pre- agriculture and endorheic lake. Yan et al. (2015) suggested cipitation, evapotranspiration and water storage change of lake. returning cropland to water is a potential way to solve the water The Qin equals to the outflow (Qout) from the upstream of lake with: deficit in the Hai Basin. Xue et al. (2017) agreed that returning cropland to water is a feasible solution if the food security of the Qout ¼ Pup -ETup eDSup (2) basin is preferentially guaranteed. In Heibei plain, about 9 3 1.6 10 m water were conserved by reducing winter wheat sown where Pup,ETup, DSup respectively stands for the precipitation, area from 1998 to 2010 (Wang et al., 2014a). In addition, a shift from evapotranspiration and water storage change in the upstream of crop rotation to fallow rotation has been taken place in China, and endorheic lake. Thus, a decrease in the total ET in the upstream the land fallow policy is strongly encouraged by the Chinese gov- (ETup) can increase the Qin of endorheic lake. ET can be further ernment for water saving in the regions with serious groundwater divided into controllable ET (ETcon) and uncontrollable (ETunc)(Wu depletion (Wang et al., 2016). Therefore, in this study we proposed et al., 2014)as an applicable trade-off approach by returning agricultural land to water area to achieve the lake restoration. Our research question ET ¼ ETcon þ ETunc (3) 734 H. Zeng et al. / Journal of Cleaner Production 217 (2019) 732e741

Lake Area(Larea) Curve Optimal curve Identifying Q tar Fitting function Qtar Q

Precipitation(P)

Calculating ACW ACW = P – ETunc -Qtar ACW ETunc

Setting ET ETsol_con ET = ETcon-ACW Deduction Goal Qm ETcon ET Qb

Setting Crop Area Crop Mask ET Area = Area Deduction Goal CropET Crop ET ET

Fig. 1. General framework of this approach.

Where, ETcon is the ET that is or can be affected by human in- ETcon ¼ ETarg_con þ ETwin_con þ ETurb_con þ Qm þ Qb (7) terventions and ETunc represents evapotranspiration from various natural landscapes without impacts of direct human activities. ETarg_con ¼ ET - ETfal (8) Considering ETunc is uncontrollable, it is operationally targeted to decrease the total ETcon in the upstream to increase Qin of lake. ETwin_con ¼ ET - ETwin (9) The concept of ACW was proposed by Wu et al. (2014), which is the maximum water that can be consumed for human activities ETurb_con ¼ ET - ETurb (10) without groundwater over-depletion. ACW can be described in eq. (4): ¼ GDPind i Q m i Q m 0 (11) GDPind 0 ACW ¼ P-ETunc-Q (4)

Popi Where P is the precipitation within the watershed, Q is the surface Q ¼ Q (12) b i b 0 Pop water that cannot be drawn under the current conditions of the 0 basin. In this study, Q is set to the estimated target inflow (Qtar)in in which, ETagr_con,ETwin_con and ETurb_con is calculated individually different scenarios to estimate the ACW of Ebinur Lake. ETunc can be as the difference between ET and ETa of the corresponding com- further decomposed by two landscapes, the natural landscape (ETn) ponents (eq. (8) to eq. (10)). ETfal,ETwin and ETurb are respectively and the artificial landscape (ETa) the ETa from fallow cropped land, wind-break forest, and residential and industrial land. Qm_i is industrial water consumption in year ETunc ¼ ETn þ ETa (5) i, while Qm_0 is estimated industrial water consumption in reference year (both available from industrial water consumption sur- ETn include ET from natural forest, woodlands, grasslands, wet- vey). GDPind_i and GDPind_0 are respectively the industrial Gross lands, water surface and bare land. ETa include ET from fallow Domestic Product (GDP) in year i and reference year. Qb_i and Qb_0 cropped land (ETfal), wind-break forest (ETwin), and residential and are respectively the domestic water consumption in year i and industrial land (ETurb). ETn can be estimated by overlapping the reference year (both available from domestic water consumption annual ET estimation from ETWatch model (Wu et al., 2012) with survey). Popi and Pop0 are respectively the annual population in the land cover data (Wu et al., 2018). ETa can be determined by year i and reference year. Finally, the difference (DET) between multiplying gross precipitation with an efficiency factor u (Gerrits ETcon and ACW is our target ET to be reduced under different et al., 2009). restoration scenarios.

ETa ¼ u*P (6) DET ¼ ETcon - ACW (13)

Under the extreme dry condition, all precipitation is lost through evaporation without human intervention, and u can be taken as 1.

2.4. Crop area deduction goal 2.3. ET deduction goal Saving water consumption from agriculture is essential for ETcon includes cropped land (ETagr_con), wind-break forest water body recovery. Return cropland to water area may be one (ETwin_con), residential land (ETurb_con), industrial process (Qm) and feasible way to increase lake inﬂow. Assuming the crop type pro- domestic process (Qb). The total ETcon within watershed can be portion remains the same, the cropping area to be deducted described as: (DArea) can be estimated by eq. (14)-(15): H. Zeng et al. / Journal of Cleaner Production 217 (2019) 732e741 735

et al., 2006). Therefore, it is urgent to reduce the lake bed expo- DET DArea ¼ (14) sure and restore the ecological environment. CropET Examined by the Normalized Difference Water Index (NDWI) (McFeeters, 1996), the area of Ebinur Lake was reduced from 2 2 ¼ ETarg con 790.64 km in 2003 to 416.07 km in 2013. Fig. 2 (c) and (e) CropET (15) CropArea compared the water body of Ebinur Lake in 2002 and 2013. It is obvious that the northwest part of the lake (indicated by the Where, CropET, CropArea are respectively the water consumption per hatched region) had completely dried out in 2013 and the southern hectare crop, and the area of agricultural land. part of the lake had shrunk significantly compared with that in 2002. Previous studies (Bao et al., 2006) indicated that the core environment of Ebinur Lake can be prevent from damage if lake 3. Case study and results area reaches to 522 km2. If lake area is above 800 km2, there will be no lake bed exposure and the environment and biodiversity can be 3.1. Study area completely restored (Bao et al., 2006). Therefore, 522 km2 and the 800 km2 are set as the target lake areas in this study for two Ebinur Lake is the largest saltwater Lake in Xinjiang Uygur restoration scenarios, the “core restoration” scenario and the Autonomous Region in China. It is located in the downstream of “complete restoration” scenario. Alashankou, a famous desert with maximum wind speed up to 55 m/s (Ma et al., 2007; Zhang et al., 2014) and strong winds 3.2. Data (Beaufort wind scale of eight or higher) 164 days annually (Wu, 2004). The Ebinur Lake basin (Fig. 2 (a, b)) had an area of Due to the limitation of data availability, the case study focuses 2 50,621 km (Wang et al., 2014a). It is composed of Bortala, Jinghe, on the change of Ebinur Lake in recently decades, and only data and Kuitun watersheds, in which the Kuitun watershed alone from Bortala and Jinghe watersheds were used. The variables contributes 44.87% of water resource for Ebinur Lake basin. adopted in this case include annual precipitation, ET, land cover, Due to the construction of reservoirs for agricultural irrigation, lake area, annual inflow, and temperature. fl the Kuitun River did not provide any in ow to Ebinur Lake since Annual precipitation is the accumulation of daily gridded pre- 1977 (Su, 2006), and the Bortala River and Jinghe River became the cipitation from China Meteorological Data Service Center (CMDC). fl unique source of in ow for Ebinur Lake. In recent decades, the Daily gridded precipitation was produced by the optimal interpo- fl in ow from Bortala River and Jinghe River were gradually reduced lation based on climate observations from 2419 stations over China. as the expansion of irrigation agriculture in these two watersheds. The original data were resampled from 0.25 0.25 to 1 km reso- The area of Ebinur Lake decreased by 31.4% in 2013 compared with lution using bilinear interpolation. The Monthly and yearly ET with that in 1972 (Zhang et al., 2014). The biodiversity and eco- spatial resolution of 1 km from 2009 to 2014 were derived from environment of Ebinur Lake is facing increasing challenge. The ETWatch model (Wu et al., 2012) based on multi-source remote Ebinur Lake National Wetland Nature Reserve was once rich in sensing and meteorological data. Land cover for 2010 were biodiversity with 595 species of plants, 327 vertebrates, 233 birds, extracted from the China Cover dataset at 30 m (Wu et al., 2017) fi and 10 kinds of shes (Muhtar et al., 2017), but now the habitats of and resampled to 1 km using majority resampling. China Cover animals and plants are severely degraded because of water area dataset was generated using Landsat TM/ETM and Chinese HJ-1 shrinkage. Compared to the 1950s, 60% of the vegetation by the satellite data. Its first level classes include forest, grassland, lakeside had degenerated and biomass had sharply reduced (Bao wetland, cropland, residential land, and other natural lands with overall accuracy larger than 94% (Wu et al., 2017). The overall accuracy of its second level classes is larger than 86% (Ouyang et al., 2016). Annual lake area, inflow of Ebinur Lake from 1998 to 2015, annual runoff of Jinghe at Jingheshankou station and annual runoff of Bortala at Bole hydrological station were provided by the Bozhou Water Conservancy Bureau of Xinjiang Uygur Autonomous Region (BWCB). The monthly and yearly precipitation and mean temperature of Wenquan, Bole, Alashankou, and Jinghe meteorological stations from 1959 to 2013 were provided by CMDC.

3.3. Results

3.3.1. Target inflow estimation Based on annual lake area and inflow measurements of Ebinur Lake from 1998 to 2015, the curves fitting results with fitting equations are shown in Fig. 3. The dependent variable (Y) is annual water area of Ebinur Lake, and the independent variable (X) is annual inflow of Ebinur Lake. According to Fig. 2, Bradley curve is identified as the optimal fitting function with maximum R2 of 0.629, and the custom equa- tion is Y ¼ 2480.28ln (0.20375ln(X)). Based on this approach, if the lake area were to be recovered to 522 km2 and 800 km2, the target Fig. 2. (a) Topography and (b) location of Ebinur Lake watershed. Subplots (c), (d) and annual inflows are estimated to be 602.6 106 m3 and (e) respectively denote the remote sensing image of water surface of Ebinur Lake in 6 3 2002, 2013, and the difference (hatched area) between the two years. The underlined 876.6 10 m respectively. In previous study, Bao et al. (2006) 6 3 6 3 texts in Fig. 2 (a) denote the names of weather stations. estimated the target inflow as 511 10 m and 792 10 m 736 H. Zeng et al. / Journal of Cleaner Production 217 (2019) 732e741

ET is very important for accurate estimation of ACW. The accuracy of CMDC monthly gridded precipitation was assessed by comparing it with the gauged precipitation obtained from Alashankou, Bole, Jinghe and Wenquan meteorological stations. Fig. 4 shows the scatter plot of the two datasets from January 2010 to December 2013. It is seen that the CMDC data slightly overestimate the precipitation compared with the station measurements. However, there is an overall good agreement between the two datasets with R2 of 0.97, 0.98, 0.97, 0.89 and RMSE of 1.90, 2.93, 1.01, and 2.93 mm month 1 over four sites. The accuracy of remotely sensed ET was evaluated by the measurements from Eddy Covariance (EC) tower (Fig. 5, left), which was installed in Jinghe irrigation agriculture experiment station. Fig. 3. Relationship between annual inflow and lake area of Ebinur Lake. The monthly ET measurements were recorded from the EC towers by the EdiRe software and were used to validate the monthly ET estimated from ETWatch model with 1 km spatial resolution. The respectively using the same two scenarios. Although our estimation validation was carried out from July 2014 to August 2015. The re- 6 3 6 3 ’ are 91.6 10 m and 84.6 10 m higher than Bao et al. (2006) s sults indicate there is a strong correlation between the measured estimation, they are within the reasonable range considering ET and the ETWatch estimated ET (Fig. 5, right). The adjusted R2 different approaches adopted and the complicated relationship between estimated and measured monthly ET is 0.95 and the RMSE fl between in ow and water area. In reality, the lake area may not is 3.59 mm month 1. present perfect positive correlation with the annual inflow. For example, in 2002 the annual inflow was 1211 106 m3 with water 3.3.3. Water consumption to be reduced surface of 873 km2, while in 2003 the annual inflow was only The water consumption for industrial (Q ) and domestic (Q ) 754 106 m3 but the water surface increased to 884 km2. m b usage were respectively 3.26 106 m3 and 2.87 106 m3 based on a survey over 60 factories and 74 communities in year 2011. Surveyed 6 3.3.2. Available consumable water estimation industrial GDP over 60 factories was 2718.2 10 CNY in 2011, Based on eq. (4), the ACW of Bortala and Jinghe watershed are accounting for 55.97% of the total industrial GDP in the study re- calculated as 1058.7 106 m3 and 784.7 106 m3 respectively if gion. The population over the 74 communities was 133,485, ac- annual outflow reached to 602.6 106 m3 and 876.6 106 m3. The counting for 27.44% of total population. Considering the proportion detailed information on annual precipitation (P), ACW and each of surveyed GDP and population to the total in the region, Qm and 6 3 component of ETunc are given in Table 1. It is revealed in Table 1 that Qb of the study region in 2011 were estimated to be 3.8 10 m 6 3 the annual precipitation over the two watersheds (Bortala and and 10.5 10 m , respectively. Industrial and domestic water Jinghe) varied greatly from 2009 to 2014, with the average of consumption for the rest of years were estimated according to eqs. 7588.0 106 m3. The maximum precipitation was 9577.0 106 m3 (11) and (12) and the results are listed in Table 2. 6 3 in 2011, and the minimum was 5633 10 m in 2012. The mean Table 2 indicated the mean annual ETcon over Bortala and Jinghe 6 3 6 3 e annual ETunc during the period 2009 to 2014 was 5926.7 10 m , River watershed was 1378.9 10 m during 2009 2014, with 6 3 and the mean ETfor,ETgra,ETwet,ETbar,ETfal,ETwin and ETurb were maximum of 1578.1 10 m in 2012, and the minimum of 6 6 6 6 6 6 3 343.0 10 , 2782.7 10 , 62.2 10 , 2031.0 10 , 548.2 10 , 1256.3 10 m in 2011. The mean annual ETcon of industrial, do- 152.3 106, and 7.3 106 m3 respectively, accounting for 5.79%, mestic, cropland, wind break forest and residential area were 5.3, 6 3 46.95%, 1.05%, 34.27%, 9.25%, 2.58%, 0.12%, and 0.12% of total ETunc. 10.5,1040.9, 280.5 and 41.7 10 m , respectively. In the real world, Natural grassland, natural bare land are the major land cover types in Bortala and Jinghe Watershed, which accounts for 81.22% of ETunc in research basin. In the “core restoration” scenario that lake area reaches 522 km2, the average annual ACW was 1058.7 106 m3, with minimum of 547.4 106 m3 in 2009, and maximum of 1980.4 106 m3 in 2010. In the “complete restoration” scenario where lake area restores to 800 km2, the average annual ACW was 784.7 106 m3 with the minimum of 273.4 106 m3 in 2009, and maximum of 1706.4 106 m3 in 2010. Fig. 4. Comparison between CMDSS gridded monthly precipitation and gauged According to eq. (4), ACW for human activities is greatly affected monthly precipitation at Wenquan, Bole, Jinghe, Alashankou weather stations from Jan by precipitation and ET. Therefore, the accuracy of precipitation and 2010 to Dec. 2013.

Table 1 Available consumable water (ACW) for human activities with 522 km2 and 800 km2 lake area goals (unit: 106 m3 yr 1).

2 2 Year P Uncontrollable ET (ETunc) Area ¼ 522 km Area ¼ 800 km

ETfor ETgra ETwet ETbar ETfal ETwin ETurb Sum Qtar ACW Qtar ACW 2009 6785.0 318.0 2698.0 53.0 1964.0 443.0 153.0 6.0 5635.0 602.6 547.4 876.6 273.4 2010 8489.0 353.0 2723.0 75.0 1960.0 583.0 204.0 8.0 5906.0 602.6 1980.4 876.6 1706.4 2011 9577.0 407.0 3407.0 77.0 2460.0 670.0 233.0 10.0 7264.0 602.6 1710.4 876.6 1436.4 2012 5633.0 264.0 2011.0 47.0 1574.0 383.0 84.0 5.0 4368.0 602.6 662.4 876.6 388.4 2013 8787.0 425.0 3510.0 75.0 2505.0 705.0 153.0 9.0 7382.0 602.6 802.4 876.6 528.4 2014 6257.0 291.0 2347.0 46.0 1723.0 505.0 87.0 6.0 5005.0 602.6 649.4 876.6 375.4 Ave. 7588.0 343.0 2782.7 62.2 2031.0 548.2 152.3 7.3 5926.7 602.6 1058.7 876.6 784.7 H. Zeng et al. / Journal of Cleaner Production 217 (2019) 732e741 737

nearly 3.5 times more than the average level of China. This indicates the crop production in this region is enough for feeding the population even with 33.72% cropped land fallowed. If the water size were to reach 800 km2, 141, 342 ha (z62.55%) cropland should be fallowed. It is not impossible to fallow 141, 342 ha land in Bortala and Jinghe watershed and feed the population at the same time. An alternative way is to assign the fallow land proportion according to the water resource proportion to each watershed. The Kuitun watershed accounts for 44.87% of total water resources in the Ebinur Lake, and the Bortala and Jinghe watershed together con- Fig. 5. Eddy covariance (EC) station and the scatter plot between estimated monthly tributes 55.13% of water resources. Therefore, if assigned 44.87% of ET from ETWatch and the eddy covariance (EC) observations at Jinghe irrigation target fallow land to Kuitun watershed and the rest portion to agriculture experiment station from July 2014 to August 2015. Bortala and Jinghe watershed, the Kuitun needs to reduce cropland by 63,420 ha and the reduction share for Bortala and Jinghe the wind-break forest was planted around the cropland to reduce watershed is 77,922 ha. the production loss from strong wind, which indicates ETcon from wind-break forest (ET ) can be deemed as part of agricultural win_con 4. Discussions water consumption. It is obvious from Table 2 that ETcon from cropped land (ET ) and wind break forest (ET ) are the agr_con win_con 4.1. Accuracy assessment and feasibility of trade-off method major water consumption in human activities, since they together account for 95.83% of total ET . Compared with the large fluctu- con According to eq. (13), ACW and water consumption from human ation of precipitation and ACW, the ET was relatively stable con activities (ET ) are two major factors in determining the ET during the study period with standard deviation of 117.2 106 m3 con deduction goal and the amount of cropland to be reduced, thus yr 1. accurate estimation of these two parameters are important for By comparing Tables 1 and 2, it is found that the gap between decision support. ACW is calculated by precipitation (P), ET and annual ET and ACW is 320.3 106 m3 and 594.2 106 m3 for the con target annual inflow (Q ) (eq. (4)). Section 3.3.1 indicated a two scenarios, which indicates the ET has to be reduced by tar con reasonable estimation of target inflow and the results in section 320.3 106 m3 and 594.2 106 m3 respectively if the lake area 3.3.2 revealed high accuracy of the precipitation product and esti- were to reach 522 km2 and 800 km2. Although ET and agr_con mated ET, which all guaranteed the reliability of our ACW results. In ET are the major water consumption from human activities, win_con addition, the ACW estimation conducted in this study is an ET takes less than one third of ET (Table 2). Therefore, win_con agr_con improvement of the similar study from Wu et al. (2018), where the reducing ET or reducing the size of cropland is a more agr_con environmental restoration is not considered. In this study, two effective way to reduce the overall human water consumption. restoration scenarios were considered when calculating ACW. On the other hand, the annual water consumption from human 6 3 3.3.4. Cropland to be deducted activities (ETcon) was estimated as 1378.9 10 m (Table 2) in this In 2014, the water consumption per hectare of cropland in study. The average annual water usage recorded by the annual Bortala and Jinghe watershed has reached 4204 m3. Under the “core bulletin of water resources in Bortala is 1550 106 m3, which is restoration” scenario (lake size restored to 522 km2), about 171.1 106 m3 higher than our estimation. Generally, water 76,189 ha cropland should be reduced to close the gap between consumed by human activities should be less than amount of water ACW and human water consumption according to eq. (14) and eq. use, considering part of water will be infiltrated into ground or (15). Under the “complete restoration” scenario (lake size restored return to environment. In this case, this gap is reasonable. As a to 800 km2), about 141,342 ha cropped land should be fallowed to summary, validated by observation and compared with previous fully restore the environment and biodiversity in Ebinur Lake basin. studies and official records, the approaches adopted in this study to What will happen if the cropland were returned to water in this estimate ACW and ETcon are reliable and the accuracy of the ACW study region? There was a total of 225,970 ha cropped land in and ETcon estimation are satisfactory. Bortala and Jinghe watersheds by the end of 2014. If 76,189 ha Our study also suggested a trade-off method of returning (z33.72%) land was fallowed to make the lake area reaches cropland to water to balance the water demand between human 522 km2, there will be 149,781 ha cropland left, which is still activities and environment restoration. Is it a feasible way? slightly larger than the cropped area in this region in 2000. Ac- Considering improving IE failed on water saving at basin scale cording to the statistical report on economic and social develop- (Grafton et al., 2018) and reducing ET through mulching, zero tillage ment of Bortala in 2014, the grain production was about 0.90 and regulated deficit irrigation could lead to yield loss (Ward and million tons and the production per capita was 1521 kg, which was Pulido-Velazquez, 2008), reducing cropland area is an alternative

Table 2 Water consumption from human activities during period 2009 to 2014 (unit: 106 m3).

Items Qm Qb ETagr_con ETwin_con ETurb_con Sum (Controllable ET, ETcon) 2009 2.1 10.3 1008.0 316.9 40.0 1377.3 2010 3.1 10.4 996.9 311.0 45.0 1366.4 2011 3.8 10.5 899.9 298.4 43.7 1256.3 2012 5.8 10.5 1240.3 283.5 38.0 1578.1 2013 7.4 10.6 977.1 232.2 43.3 1270.6 2014 9.5 10.6 1123.5 240.7 40.3 1424.6

Avg.2009-2014 5.3 10.5 1040.9 280.5 41.7 1378.9 fraction 0.38% 0.76% 75.49% 20.34% 3.03% 100.00% 738 H. Zeng et al. / Journal of Cleaner Production 217 (2019) 732e741 promising way to restore the damaged endorheic lake. In reality, the land fallow policy has been carried out by Chinese governments in Beijing and Hehei, which provide subsidies to fallow land especially at groundwater over-exploited regions (Wang et al., 2016). The study of Yan et al. (2015) also suggested to fallow agricultural land in Hai Basin for water saving.

4.2. Driving forces of lake shrinkage

It is also worthwhile to investigate the driving forces of lake Fig. 7. Mann-Kendall test of mean annual temperature at Jinghe and Wenquan shrinkage, which will provide guidance on preventing lake meteorological stations (P ¼ 0.05). shrinkage. In general, climate change and human activities are two major driving forces that cause lake shrinkage. which is an important source of water recharge for Ebinur Lake. Wang et al. (2014b) estimated that the glaciers in Ebinur Lake basin 4.2.1. Climate driving forces decreased by 14.7% during the past 40 years, which corresponded Precipitation and temperature are two climate variables that to a roughly 20.5% change of water volume in Ebinur Lake. With the have prominent impact on lake area. In this study, the nonpara- increased precipitation and temperature (or increased melting metric Mann-Kendall trend test (Mann, 1945) was used to assess glacier water), the water volume of Ebinur Lake is expected to in- the changes in trend of precipitation and temperature (Fig. 6 and crease. However, the Ebinur Lake has experienced severe area Fig. 7). The trend variation of target variable is indicated by the UF shrinkage since 2003. Therefore, other factors rather than climate curve. If the UF value is greater than 0, the sequence has an upward change may play a stronger role in driving the depletion of Ebinur (or positive) trend; and if the UF value is less than 0, a downward Lake. In the following section, the impact from human activities on (or negative) trend is presented. The standard normal test statistic lake shrinkage is discussed. (Gocic and Trajkovic, 2013) was used to judge the significance of the change in trend. If the absolute value of UF exceeds the 95% significance level (U0.05 ¼ ± 1.96), the upward or downward trend is 4.2.2. Human driving forces fi fi signi cant. If UF and UB have intersection within the signi cance 4.2.2.1. Expansion of cropland and wind-break forest. intervals, the point of intersection indicates where the abrupt Agriculture accounts for major water withdrawals from streams, change of trend occurred. lakes and groundwater (Rulli et al., 2013). The rapid expansion of Fig. 6 showed the trend assessment of annual precipitation at agricultural has resulted in increased water consumption of the Jinghe and Wenquan meteorological stations. An increasing trend basin. Fig. 8 illustrated the changes of cropped land area in Bortala was observed for precipitation at Jinghe station (Fig. 6a) starting and Jinghe watershed from 1990 to 2014. Table 3 listed the statistics fi from 1982 (positive UF curve), while the signi cant positive trend of area and ET for cropland and artificial vegetation in Bortala and e was observed during 2002 2005. For Wenquan station (Fig. 6b), an Jinghe watershed at five years (1990, 2000, 2010, 2013 and 2014) increasing trend of precipitation has been observed since 1990, based on the analysis of Landsat and HJ satellite imagery. fi > while the positive trend was not signi cant ( 1.96) until 2002. In 1990, 2000, 2010, 2013 and 2014, the cropped land in the Fig. 7 showed the trend assessment of annual temperature at Bortala and Jinghe watershed was estimated to be 91.1, 143.7, 195.5, Jinghe and Wenquan meteorological stations. At Jinghe station 218.7, and 226.0 103 ha, respectively (Table 3). Table 3 also (Fig. 7a), the temperature presented an increasing trend since 1990, showed the cropped land increased by 52.6 103 ha from 1990 to fi andthetrendissigni cant after 2002. Furthermore, a shift of domi- 2000 and by 51.8 103 ha from 2000 to 2010. The expansion of nate trend from negative to positive was observed in 2000. At Wen- agricultural land also resulted in the enlargement of farmland quan station (Fig. 7b), an increasing trend of temperature has been windbreak, which protects crop from frequent strong wind. The fi observed since 1997, and the trend became signi cant after 2007. A remote sensing imagery indicated that the area with planted shiftof dominate trend from negative to positivewas observed in 2001. vegetation was 69.1 103 ha in 1990, 85.5 103 ha in 2000, In sum, both precipitation (Fig. 6) and temperature (Fig. 7) presented an upward trend in Ebinur Lake basin since 2000s, which agrees well with the findings from Wang et al. (2014a). The positive trend of precipitation in the 1990s corresponded well to the rapid expansion of Ebinur Lake from the end of 1990s to the early 2000s. The rise of temperature may accelerate the melting of glacier,

Fig. 6. Mann-Kendall test of annual precipitation at Jinghe and Wenquan meteoro- Fig. 8. Changes of cropped land area in Bortala and Jinghe watershed from 1990 to logical stations (P ¼ 0.05). 2014. H. Zeng et al. / Journal of Cleaner Production 217 (2019) 732e741 739

Table 3 Cropped land, artiﬁcial vegetation area and water consumption change from 1990 to 2014 at Bortala and Jinghe watershed.

Items Cropped land Artiﬁcial vegetation

1990 2000 2010 2013 2014 1990 2000 2010 2013 2014

Area (103ha) 91.1 143.7 195.5 218.7 226.0 69.1 85.5 104.8 109.0 110.0 ET consumption per ha (m3/ha) 4238 4238 4388 3836 4204 3860 3860 4177 3825 3984 ET of sum (106m3) 386.1 609.0 857.9 838.9 950.1 266.7 330.0 437.7 416.9 438.2

104.8 103 ha in 2010, and 110.0 103 ha in 2014, with an incre- production can be increased by expanding cropland in this region. ment of 40.9 103 ha from 1990 to 2014 (Table 3). In the down- For example, in the 1990s, the cropland was expanded greatly due stream of Bortala and Jinghe basin, the cropland also showed a to the incentives given by local government through farmland significant increasing trend. The rapid expansion of irrigation reclamation. In addition, profitable cropland leasing and crop agriculture around Ebinur Lake is threatening the health and sus- production drive farmers to enlarge their land. Input-output survey tainability of its ecosystem. from 61 cotton farmers and 85 maize farmers revealed that the net Assuming the water consumption of crop and manmade vege- earnings from land leasing was 4200 to 4500 CNY/ha in 2013, and tation in 1990 and 2000 remains the same as the average of that this value increased to 4800 CNY/ha in 2014. The net earnings from from 2009 to 2014, the ETcon from cropped land were estimated as cotton and maize planting was 12,729 CNY/ha and 5158.5 CNY/ha in 386.1, 609.0, 857.9, 838.9, 950.1 106 m3 in 1990, 2000, 2010, 2013, 2013, respectively. Stimulated by land leasing and crop planting 2014 respectively, and the ETcon from manmade vegetation were income, the cropped land reached a maximum area of 266.7, 330.0, 437.7, 416.9, 438.2 106 m3 for the five years (Table 3). 225.97 103 ha in 2014. From the late 1990s to early 2000s (Fig. 6), Overall, the ETcon of cropped land and artificial vegetation increased Xinjiang Uygur Autonomous received abundant precipitation, by 564.0 106 m3 and 171.5 106 m3 respectively from 1990 to which led to environmental and ecosystem restoration and accel- 2014 with a total increase of water consumption of 735.5 106 m3. erated the arable land reclamation as well. After 2006, the If left unattended, the Ebinur Lake will suffer from dramatic ecosystem was imbalanced as the decrease of precipitation. By shrinkage as the Aral Sea in Central Asia with current water con- 2013, the water area of Ebinur Lake had reduced to its historical sumption rate. minimum mainly caused by human activities. The decrease of runoff difference between upstream hydrological station and downstream outlet also indicated a rapid increase 4.2.2.2. Dams construction. The diversion of surface inflow can also of water consumption from irrigation agriculture. From 1998 to cause a rapid decrease in lake size (Williams, 2002), such as the 2014, an average of 68.10% runoff was lost between Shankou station case of Lake Urmia in Iran (Ouria and Sevinc, 2016). Ebinur Lake is and the outlet at Jinghe basin, and the loss reached to 92.14% in suffering from this issue due to the over-construction of dams in 2014. Similarly, an average of 47.42% runoff was lost between Bole the basin. In China, reservoir are classified into four categories ac- station and the outlet at Bortala basin, and the loss reached to cording to the reservoir storage capacity (SL262-200, China), 74.18% in 2013. The left panel in Fig. 9 showed the accumulated namely, large reservoir Ⅰ (reservoir storage capacity > 1 109 m3), runoff anomalies at Bole station in Bortala basin (red line) and large reservoir Ⅱ (0.1 109 m3 < reservoir storage capac- Shankou station in Jinghe basin (blue line). A sharp increase of ity 1 109 m3), medium reservoir (0.001 109 m3 <¼ reservoir accumulated runoff anomalies were respectively observed after storage capacity < 0.1 109 m3) and small reservoir (0 < reservoir 2004 at Shankou station and after 2009 at Bole station. The right storage capacity < 0.0001 109 m3). By 2013, at least 30 reservoirs panel in Fig. 9 showed the ratio of runoff loss at the Bortala (red have been built in Kuitun, Bortala and Jinghe basins. Six medium to line) and Jinghe (blue line) basins. An increasing trend of runoff loss large sized reservoirs were built before 1977 in Kuitun basin, was observed for both Jinghe and Bortala River. By comparing the including one large reservoir Ⅱ (Liugou in 1957 with storage ca- slopes of regression line of the two basins, it is found that the runoff pacity of 0.102 109 m3) and five medium reservoirs (Chepaizi in loss at Jinghe (1.9) is larger than that of Bortala (1.34). This indicated 1952, Kuitun in 1957, Huanggouyiku in 1957, Huanggouerku in the agricultural water consumption in Jinghe basin was more 1970, and Quangou in 1977). These dams retained flow from Kuitun intense than that in Bortala basin. to Ebinur Lake since 1977. In Bortala, there are one medium The over-expansion of cropland in study region is due to both reservoir (Wuyi in 1968) and four medium reservoirs (Etuokesaier fi political incentives and economic pro t. Xinjiang Uygur Autono- in 1987, Konghaquan in 1994, Xiatianji in 1997 and Halatuluke in mous is the largest province in China and is endowed with abun- 2014). The total reservoir storage in Bortala was about dant arable land. If water availability is not a constraint, crop 0.195 109 m3 in 2015, while the total reservoirs storage in Kuitun was 0.303 109 m3 in 2009 (Deng et al., 2012). In other words, nearly 0.4 109 m3 water has been stored in reservoirs in the upstream of Ebinur Lake. This is one of the major reasons that caused the shrinkage of Ebinur Lake. Lake Urmia in Iran disappeared due to damming for irrigation and hydropower, which takes nearly 90% of its water inflow (Stone, 2015). If the reservoir construction within Ebinur basin is not constrained and well-regulated in the future, the tragedy of Urmia Lake may reoccur in Ebinur Lake. As a summary, compared with climate change, the human activities would have stronger impact on the Ebinur Lake shrinkage. Fig. 9. The accumulated runoff anomaly (left panel) at Bole station in Bortala basin (red line) and Shankou station in Jinghe basin (blue line), and the ratio of runoff loss 4.3. Implications (right panel), ratio of runoff loss ¼ (runoff at upstream station - runoff at downstream outlet)/runoff at upstream station, at the Bortala (red line) and Jinghe (blue line) basins. In arid and semi-arid region, the shrinkage of endorheic lake 740 H. Zeng et al. / Journal of Cleaner Production 217 (2019) 732e741 proposes a big challenge for the sustainable development of adopted in this research can be applied to the restoration of other ecosystem. A lot of evidences confirmed that overexpansion of inland lakes in arid regions as well. irrigation agriculture was the major reason for lake shrinkage. Take The approach proposed by this study is simple and operational. Aral Sea as an example, the irrigation area expanded from 500 It is driven by remote sensing technology and the data are readily million ha in 1965 to 790 million ha in 2000, which remarkably available; thus, the method can be widely applied in a timely reduced the discharge (Micklin, 2007). The dramatic shrinkage of fashion. However, there are still some limitation of this study and Lake Urmia was caused by the dam construction and intensive can be improved in the future. The relationship between water size irrigation (AghaKouchak et al., 2015). of inland lake and the annual inflow was explored by statistic Our ultimate goal is to find the sweet spot or a good balance regression in this study, the accuracy of which is greatly affected by between the lake restoration and agricultural development. the length and accuracy of historical data. Although Bradley curve Although enhancing IE may help to save water at farm scale, it does was identified as the optimal fitting function in this study, its R2 not help at basin scale (Grafton et al., 2018). Reducing ET is the was below 0.65. This indicates the variation of annual inflow is not another way to save water (Wu et al., 2014). However, reducing ET very well captured (considering R2 > 0.8 for very good case) by through mulching, zero tillage, and regulated deficit irrigation Bradley function, which would lead to overestimation or underes- could lead to yield loss (Perry, 2011; Yan et al., 2015), thus reducing timation of actual annual inflow. Therefore, the physical model, for the area of cropped land is an alternative to reduce ET from up- example, the Bayesian networks, is recommended and will be stream and increase the inflow of inland lake. Our study suggested explored in our future study for annul inflow estimation. Secondly, terminating the agricultural reclamation and reducing the irrigated only six years of annual precipitation and ET data were available cropland to help restore the Ebinur Lake. In the real world, reducing and used to estimate the ACW of Ebinur Lake, which may cause irrigated cropland and returning water from cropland to inland lake uncertainty in ACW calculation, the long-term historical data has been adopted by government, such as Bortala and Turpan. should be employed in future. Considering the central Asia has been suffered from lake shrinkage for a long time (Bai et al., 2011), returning irrigated land to water Acknowledgements would be a new way to restore the damaged lake. The remote sensing technology may play an important role in The authors acknowledge the financial support from the Na- monitoring the area of endorheic lake, the area of cropland and tional Key Research and Development Program of China estimating ET and precipitation. Abundant remote sensing precip- (2016YFA0600304; 2016YFC0503401), the Strategic Priority itation and ET products have been generated, such as TRMM Research Program of Chinese Academy of Sciences (XDA19030204), (Huffman et al., 2007) and GMP precipitation products (Hou et al., the National Natural Science Foundation of China (41561144013), 2014), ET estimations from ETWatch (Wu et al., 2012), SEBAL the International Partnership Program of Chinese Academy of Sci- (Bastiaanssen et al.,1998a,1998b) and SEBS (Su et al., 2002) models. ences (121311KYSB20170004), and the Director Youth Fund from As the advance of technology and the improvement of the accuracy, Institute of Remote Sensing and Digital Earth, Chinese Academy of remote sensing products are all very helpful in decision support of Sciences (No.Y5SJ1700CX). We also acknowledge the Bozhou Water environment restoration. Other emerging technology, such as the Conservancy Bureau of Xinjiang Uygur Autonomous Region for machine learning and cloud computing may provide efficient al- providing the annual inflow of Ebinur Lake and the annual hydro- gorithms and novel platforms for information extracting and logical runoff from Jinghe and Bortala hydrological station. We knowledge sharing and may benefit the tradeoff between envi- would also like to thank China Meteorological Data Sharing Ser- ronment restoration and human water consumption in the future. vices System for providing the daily precipitation, temperature and annual precipitation raster. 5. Conclusions References Facing the dilemmas that improving IE does not help water saving at basin scale and reducing ET through mulching, zero tillage AghaKouchak, A., Norouzi, H., Madani, K., Mirchi, A., Azarderakhsh, M., Nazemi, A., and regulated deficit irrigation will lead to crop yield loss, this study Nasrollahi, N., Farahmand, A., Mehran, A., Hasanzadeh, E., 2015. Aral Sea syn- drome desiccates Lake Urmia: call for action. J. Great Lake. Res. 41 (1), 307e311. proposed an alternative approach to achieve the balance between Bai, J., Chen, X., Li, J., Yang, L., Fang, H., 2011. Changes in the area of inland lakes in lake restoration and human water consumption by returning arid regions of central Asia during the past 30 years. Environ. Monit. Assess. 178 agricultural land to water. The methodology includes four steps: (1) (1), 247e256. Bai, J., Chen, X., Yang, L., Fang, H., 2012. Monitoring variations of inland lakes in the identifying the Qtar of endorheic lake at different restoration sce- arid region of Central Asia. Front. Earth Sci. 6 (2), 147e156. narios; (2) calculating ACW in the upstream based on Qtar; (3) Bastiaanssen, W.G.M., Menenti, M., Feddes, R.A., Holtslag, A.A.M., 1998a. A remote setting ET deduction goal based on water consumption balance; sensing surface energy balance algorithm for land (SEBAL). 1. Formulation. J. Hydrol. 212e213, 198e212. and (4) estimating crop area to be deducted to achieve the ET Bastiaanssen, W.G.M., Pelgrum, H., Wang, J., Ma, Y., Moreno, J.F., Roerink, G.J., van deduction goal. The ET deducted in this way is the real water saving der Wal, T., 1998b. A remote sensing surface energy balance algorithm for land rather than the palliative one from enhancing IE. The proposed (SEBAL).: Part 2: Validation. J. Hydrol. 212e213, 213e229. Bao, A., Mu, G., Zhang, Y., Feng, X., Chang, C., Yin, X., 2006. Estimation of the rational method was implemented in Ebinur Lake of China with two water area for controlling wind erosion in the dried-up basin of the Ebinur Lake 2 2 restoration scenarios (lake area restored to 522 km and 800 km ). and its effect detection. ChSBu 51 (1), 68e74. The results revealed that the annual inflow should be increased to Coe, M.T., Foley, J.A., 2001. Human and natural impacts on the water resources of the e 602.6 106 m3 and 876.6 106 m3 to make lake area restored to Lake Chad basin. JGRD 106 (D4), 3349 3356. 2 2 Deng, M., Li, X., Zheng, Y., Zhang, Y., Hao, F., Liang, L., 2012. Water resource future 522 km and 800 km respectively. At the current stage allocation for industry development and urbanization in Golden Triangle Zone. (2009e2014), the average annual human water consumption in the Kuitun River Basin. ALG 35 (4), 527e536 [In Chinese]. upstream is 1378.9 106 m3, and it is to be reduced by Foley, J.A., DeFries, R., Asner, G.P., Barford, C., Bonan, G., Carpenter, S.R., Chapin, F.S., Coe, M.T., Daily, G.C., Gibbs, H.K., Helkowski, J.H., Holloway, T., Howard, E.A., 6 3 6 3 320.3 10 m and 594.2 10 m respectively to achieve the two Kucharik, C.J., Monfreda, C., Patz, J.A., Prentice, I.C., Ramankutty, N., Snyder, P.K., restoration goals. The overexpansion of cropland and artificial for- 2005. Global consequences of land use. Sciences 309 (5734), 570. est were found to be the main contributors to the shrinkage of Gao, G., Shen, Q., Zhang, Y., Pan, N., Ma, Y., Jiang, X., Fu, B., 2018. Determining spatio- temporal variations of ecological water consumption by natural oases for sus- Ebinur Lake. Finally, the feasible solution for ecosystem restoration tainable water resources allocation in a hyper-arid endorheic basin. J. Clean. of Lake Ebinur was provided for both scenarios. The methodology Prod. 185, 1e13. H. Zeng et al. / Journal of Cleaner Production 217 (2019) 732e741 741

Gerrits, A.M.J., Savenije, H.H.G., Veling, E.J.M., Pfister, L., 2009. Analytical derivation agriculture: agronomic constraints and hydrological realities. Agric. Water of the Budyko curve based on rainfall characteristics and a simple evaporation Manag. 96 (11), 1517e1524. model. WRR 45 (4). Poppenborg, P., Koellner, T., 2014. A Bayesian network approach to model farmers’ Gocic, M., Trajkovic, S., 2013. Analysis of changes in meteorological variables using crop choice using socio-psychological measurements of expected benefits of Mann-Kendall and Sen’s slope estimator statistical tests in Serbia. GPC 100, ecosystem services. Environ. Model. Softw 57, 227e234. 172e182. Rulli, M.C., Saviori, A., D’Odorico, P., 2013. Global land and water grabbing. Proc. Grafton, R.Q., Williams, J., Perry, C.J., Molle, F., Ringler, C., Steduto, P., Udall, B., Natl. Acad. Sci. Unit. States Am. 110 (3), 892. Wheeler, S.A., Wang, Y., Garrick, D., Allen, R.G., 2018. The paradox of irrigation Steduto, P., Hsiao, T.C., Fereres, E., 2007. On the conservative behavior of biomass efficiency. Sciences 361 (6404), 748. water productivity. Irrigat. Sci. 25 (3), 189e207. Hashizume, M., Ueda, K., Nishiwaki, Y., Michikawa, T., Onozuka, D., 2010. Health Stone, R., 2015. Saving Iran’s great salt lake. Sciences 349 (6252), 1044. effects of Asian dust events: a review of the literature. Nihon Eiseigaku Zasshi Su, H.C., 2006. Change in Ebinur Lake area and its impact on eco-environment. 65 (3), 413e421. J. Glaciol. Geocryol. 28, 941e949. Hellegers, P.J.G.J., Soppe, R., Perry, C.J., Bastiaanssen, W.G.M., 2010. Remote sensing Su, Z., 2002. The Surface Energy Balance System (SEBS) for estimation of turbulent and economic indicators for supporting water resources management de- heat fluxes. Hydrol. Earth Syst. Sci. 6 (1), 85e100. cisions. Water Resour. Manag. 24 (11), 2419e2436. Wang, L., Li, Z., Wang, F., Edwards, R., 2014a. Glacier shrinkage in the Ebinur lake Howell, T.A., 1990. Grain, dry matter yield relationships for winter wheat and grain basin, Tien Shan, China, during the past 40 years. JGlac 60 (220), 245e254. sorghum-southern high plains. Agron. J. 82, 914e918. Wang, X., Li, X., Xin, L., 2014b. Impact of the shrinking winter wheat sown area on Hou, A.Y., Kakar, R.K., Neeck, S., Azarbarzin, A.A., Kummerow, C.D., Kojima, M., agricultural water consumption in the Hebei Plain. J. Geogr. Sci. 24 (2), Oki, R., Nakamura, K., Iguchi, T., 2014. The global precipitation measurement 313e330. mission. Bull. Am. Meteorol. Soc. 95, 701e722. Wang, X., Li, X., Xin, L., Tan, M., Li, S., Wang, R., 2016. Ecological compensation for Huffman, G.J., Bolvin, D.T., Nelkin, E.J., Wolff, D.B., Adler, R.F., Gu, G., Hong, Y., winter wheat abandonment in groundwater over-exploited areas in the North Bowman, K.P., Stocker, E.F., 2007. The TRMM multisatellite precipitation analysis China Plain. J. Geogr. Sci. 26 (10), 1463e1476. (TMPA): quasi-global, multiyear, combined-sensor precipitation estimates at Ward, F.A., Pulido-Velazquez, M., 2008. Water conservation in irrigation can in- fine scales. J. Hydrometeorol. 8, 38e55. crease water use. Proc. Natl. Acad. Sci. 105 (47), 18215e18220. Kendy, E., Zhang, Y., Liu, C., Wang, J., Steenhuis, T., 2004. Groundwater recharge Williams, W.D., 2002. Environmental threats to salt lakes and the likely status of from irrigated cropland in the north China plain: case study of Luancheng inland saline ecosystems in 2025. Environ. Conserv. 29 (2), 154e167. County, Hebei province, 1949e2000. HyPr 18 (12), 2289e2302. Wu, B., Jiang, L., Yan, N., Perry, C., Zeng, H., 2014. Basin-wide evapotranspiration Liu, H., Yin, Y., Piao, S., Zhao, F., Engels, M., Ciais, P., 2013. Disappearing lakes in management: concept and practical application in Hai Basin, China. Agric. semiarid northern China: drivers and environmental impact. Environ. Sci. Water Manag. 145, 145e153. Technol. 47 (21), 12107e12114. Wu, B., Qian, J., Zeng, Y., 2017. Land Cover Atlas of the People’s Republic of China (1: Ma, M., Wang, X., Veroustraete, F., Dong, L., 2007. Change in area of Ebinur Lake 1 000 000). SinoMaps Press, Beijing. during the 1998e2005 period. IJRS 28 (24), 5523e5533. Wu, B., Yan, N., Xiong, J., Bastiaanssen, W.G.M., Zhu, W., Stein, A., 2012. Validation of Mann, H.B., 1945. Nonparametric tests against trend. Econometrica 13 (3), 245e259. ETWatch using field measurements at diverse landscapes: a case study in Hai McFeeters, S.K., 1996. The use of the Normalized Difference Water Index (NDWI) in Basin of China. J. Hydrol. 436e437, 67e80. the delineation of open water features. IJRS 17 (7), 1425e1432. Wu, B., Zeng, H., Yan, N., Zhang, M., 2018. Approach for estimating available Micklin, P., 2007. The Aral Sea disaster. AREPS 35 (1), 47e72. consumable water for human activities in a River basin. Water Resour. Manag. Micklin, P.P., 1988. Desiccation of the Aral Sea: a water management disaster in the 32 (7), 2353e2368. soviet union. Sciences 241 (4870), 1170. Wu, J., 2004. Characteristics and reasons of fluctuation of Lake surface of the Aibi Muhtar, A., Abdukerim, R., Halik, M.J.A.E.S., 2017. Assessing habitat suitability for lake, Xinjiang autonomous region. Mar. Geol. Quat. Geol. 24, 57e60. cervuselaphus in the Ebinur Lake National nature Reserve. Acta Ecol. Sin. 37 Xue, J., Gui, D., Lei, J., Sun, H., Zeng, F., Feng, X., 2017. A hybrid Bayesian network (11), 3919e3925. approach for trade-offs between environmental flows and agricultural water Nakayama, T., Yang, Y., Watanabe, M., Zhang, X., 2006. Simulation of groundwater using dynamic discretization. AdWR 110, 445e458. dynamics in the North China Plain by coupled hydrology and agricultural Xue, J., Gui, D., Zhao, Y., Lei, J., Zeng, F., Feng, X., Mao, D., Shareef, M., 2016. models. HyPr 20 (16), 3441e3466. A decision-making framework to model environmental flow requirements in Ouria, M., Sevinc, H.J.S., 2016. The role of dams in drying up Lake Urmia and its oasis areas using Bayesian networks. J. Hydrol. 540, 1209e1222. environmental impacts on Azerbaijani districts of Iran. Saussurea 6 (1), 54e65. Yamaguchi, N., Sakotani, A., Ichijo, T., Kenzaka, T., Tani, K., Baba, T., Nasu, M., 2012. Ouyang, Z., Zheng, H., Xiao, Y., Polasky, S., Liu, J., Xu, W., Wang, Q., Zhang, L., Xiao, Y., Break down of Asian dust particle on wet surface and their possibilities of cause Rao, E., Jiang, L., Lu, F., Wang, X., Yang, G., Gong, S., Wu, B., Zeng, Y., Yang, W., of respiratory health effects. Biol. Pharm. Bull. 35 (7), 1187e1190. Daily, G.C., 2016. Improvements in ecosystem services from investments in Yan, N., Wu, B., Perry, C., Zeng, H., 2015. Assessing potential water savings in agri- natural capital. Sciences 352 (6292), 1455. culture on the Hai Basin plain, China. Agric. Water Manag. 154, 11e19. Perry, C., 2007. Efficient irrigation; inefficient communication; flawed recommen- Zhang, F., Tiyip, T., Johnson, V.C., Kung, H.-t., Ding, J.-l., Sun, Q., Zhou, M., Kelimu, A., dations. Irrigat. Drain. 56 (4), 367e378. Nurmuhammat, I., Chan, N.W., 2014. The influence of natural and human factors Perry, C., 2011. Accounting for water use: terminology and implications for saving in the shrinking of the Ebinur Lake, Xinjiang, China, during the 1972e2013 water and increasing production. Agric. Water Manag. 98 (12), 1840e1846. period. Environ. Monit. Assess. 187 (1), 4128. Perry, C., Steduto, P., Allen, R.G., Burt, C.M., 2009. Increasing productivity in irrigated