Climate change challenges efficiency of inter-basin water transfers in alleviating water stress

Inter-basin water transfer (IBT) is widely used to mitigate water shortage at the cost of compromising water availability in water-exporting regions. Yet, we do not know how efficient are the IBTs in alleviating inter-regional water stress in a changing climate and water supply-demand context. From a socio-hydrological perspective, we here quantify the efficiency of more than 200 IBTs across the United States by a Stress Relief Index that measures the impact of water redistribution on the overall water stress level. Based on the assumption that an IBT-induced increase and reduction in water availability would respectively constitute a positive and negative impact on regional water security, we show that 29% of the IBTs could be considered socially inefficient by 2010 as they shift water stress from water-receiving to water-exporting and downstream regions. Future stress escalations induced by growing population, declining runoff, and increasing demands for energy production and irrigation will alter IBT efficiency disproportionately. The inefficient IBTs would amount to 32% and 35% by the end of the 21st century under the scenarios of representative concentration pathway (RCP) 4.5 and RCP8.5, with 7 ∼ 16 IBTs reaching a tipping point that their role in the water system could switch from alleviating to aggravating the overall water stress. Our results indicate that the evolving climatic and socioeconomic status can largely affect transfer efficiency, highlighting the need of basin-level adaptation strategies for sustainable use of the IBTs.


Introduction
Inter-basin water transfers (IBTs) have played a critical role in securing freshwater supply in many water deficient regions (Shiklomanov 2000, Yevjevich 2001, Liu and Zheng 2002. IBTs are designed to alleviate regional water stress by diverting surface water from a 'water rich' area to highly stressed regions, and potentially reduce the detrimental impacts caused by unsustainable local water uses (e.g. groundwater overdraft) (de Graaf et al 2019). Worldwide, 500 billion out of the estimated 42 trillion m 3 (1.2%) of renewable water resources are redistributed by IBTs annually (Shiklomanov 2000), such as the South-to-North water transfer project in China and the Snowy Mountains Scheme in Australia (Ghassemi andWhite 2007, Zhang et al 2020). In the United States, 22 billion m 3 of water is transferred each year via over 200 IBTs, with the annual transfer magnitude ranging from 500 m 3 to 4 billion m 3 per IBT (Petsch 1985, Mooty andJeffcoat 1986). The transferred water only accounts for 1% of surface freshwater resources (2.1 trillion m 3 yr −1 ) and 6% of surface water demands (356 billion m 3 yr −1 ) in the U.S. (Duan et al 2018), but it is an important source of freshwater for arid or densely populated areas. One example is southern California, where over 80% of the surface freshwater resources comes from the 8 billion m 3 yr −1 of water transferred from the San Joaquin River and Colorado River (Ashoori et al 2015, Duan et al 2019, supporting a population of 13 million in the Los Angeles Metropolitan Area and over nine million acres of farmland production for both local consumption and global trade (Hoekstra andChapagain 2011, Mount andHanak 2016).
The adequacy of surface freshwater resources at the country or regional scales can be interpreted by the water stress index (WS) defined as the ratio of off-stream water demand (water withdrawal or water use) to water availability (Wada et al 2011, Pedro-Monzonís et al 2015. The level of water stress is usually considered high when the ratio exceeds 0.4 due to the concern for sufficient instream water uses for navigation, hydropower generation, and ecological and environmental demands (Vörösmarty et al 2000, Oki et al 2001, Richey et al 2015. The aqueducts of IBTs are artificially integrated into the river networks, and disrupt the upstream-downstream flow passage due to water withdrawals in one basin and return flows in another. Such artificial modification to the natural hydrological system alters water balance in hydrologically connected drainage basins and influence downstream environments, which subsequently reduces water stress in the water-receiving basin at the cost of compromising water availability in the water-exporting basin.
Individual IBTs across the globe have been well studied from various perspectives of hydrology (Long et al 2020), water management (Gohari et al 2013, Barnett et al 2015, economics (Zhao et al 2015), and ecosystem conservation (Davies et al 1992, Grant et al 2012, Vargas et al 2020. However, two considerable gaps remain in our understanding of the IBTs' role in water stress alleviation. First, most of the previous studies have focused on the impact of individual IBTs, while the aggregate hydrological impacts of multiple IBTs have been rarely quantified at high spatial resolutions due to a lack of standardized and systematic data collection at national or continental scales (Emanuel et al 2015). For example, Long et al (2020) reported that water transferred through the South-to-North water transfer project accounted for 40% of groundwater storage recovery in Beijing; Zhao et al (2015) argued that water transfer cannot mitigate water stress in China, but the results were based on water stress evaluated coarsely at the provincial level (30 provinces) and the hydrological connections among regions were not considered. Second, little is known about the potential variations in the IBTs' role under the compound influence of dynamic hydroclimatic condition and water supply-demand context. For example, Emanuel et al (2015) examined the effectiveness of water transfer projects in the United States by comparing the transfer magnitudes to streamflow in water-exporting and receiving regions without considering the spatiotemporal differences in water demand. Although the physical characteristics of drainage basins (e.g. topography, soil properties) can be assumed to remain static within the time scale of interest to water managers, pre-transfer water availability and water stress is constantly changing driven by environmental and anthropogenic stressors such as the changes in climate, population, water use efficiency, and economic and energy structures (Vörösmarty et al 2000, Maupin et al 2014. The nonlinear response of downstream water availability to the combined impacts of water transfer, climate change, and water-use behaviors in upstream areas could be highly complex for large basins. It is important to understand whether the IBT infrastructure could fulfill its role efficiently in a changing environment. Here we present a dynamic water stress model and a set of efficiency metrics based on a high-resolution geography of water transfer, water use and availability, analyzing the role of IBTs in stress alleviation in historical and future contexts. We compiled historical records of 228 IBTs that transferred water across the 2009 8-digit hydrologic unit code (HUC-8) watersheds in the conterminous United States (CONUS), and projections of watershed runoff, population, water withdrawal, and consumptive water uses under future scenarios of climate change and adaptation in the 21st century. We consider explicitly how the topology of river networks and off-stream water uses determine the character of IBTs in sustainable water supply. We hypothesize that the large-scale climate change and associated adaptation strategies would alter regional water availability and water demand from both hydrological and socioeconomic aspects, potentially resulting in spatially diverse changes in IBT efficiency in water stress alleviation.

IBTs
The location and transfer magnitudes of IBTs in 1973-1982 were collected from survey questionnaires in the eastern (Mooty and Jeffcoat 1986) and western (Petsch 1985) United States. Among the 256 IBTs originally identified in the reports, 23 were excluded due to the lack of transfer magnitudes and another five were excluded because they were associated with transboundary river basins (Emanuel et al 2015). The remaining 228 IBTs include 122 projects in the east and 106 in the west. Although dated, this database is the most comprehensive national-scale data of IBTs with detailed flow volumes Dzombak 2017, 2019), and has been widely used for national water resource assessments (Emanuel et al 2015, Brown et

Historical population and water use
Historical data of population and water uses were obtained from the water census reports compiled by the U.S. Geological Survey (Solley et al 1998, Maupin et al 2014. The water use datasets were rescaled from counties (3109 counties) to HUC-8 watersheds based on weighted areal averages. Water withdrawal and consumption were linearly interpolated within each five-year reporting interval to generate a continuous time series of water demand to be compared to the variations in water availability (Duan et al 2018).
2.1.3. Reference scenarios of water demand and supply in the 21st century Two reference scenarios of future water demand and supply were used to investigate the response of IBT efficiency to climate change and socioeconomic adaptation. An intermediate water stress (IS) scenario and a high water stress (HS) scenario were created based on the trajectory of a series of climatic (e.g. precipitation, temperature, radiation) and socioeconomic (e.g. population, energy structure, water use efficiency) factors to represent a future with and without climate change mitigation strategies, respectively (Duan et al 2019). The IS scenario was driven by climate change under the representative concentration pathway (RCP) 4.5 scenario, intermediate population growth, and power generation complying with the Clean Power Plan (CPP). Meanwhile, the HS scenario was driven by climate change under the RCP8.5 scenario, fast population growth, and power generation without the CPP. The CPP is a U.S. Environmental Protection Agency (EPA) program issued under the Clean Air Act (42 U.S.C. §7401) that aims at a reducing carbon emission from fossil fuel plants and extending tax credits for renewable energy. Energy structure and energy consumption was projected based on the Annual Energy Outlook provided by the Energy Information Administration of U.S. Department of Energy. Climate projections from 19 Global Climate Models (GCMs) of the fifth phase of the Coupled Model Inter-comparison Project (CMIP5) (http://maca.northwestknowledge.net/) was collected to enable a robust quantification of the major uncertainties from model structure. Future population were obtained from the Integrated Climate and Land-Use Scenarios v1.3 datasets compiled by the EPA (www.epa.gov/iclus/iclus-downloads). Water demand and consumption of surface freshwater was simulated by extrapolating the past trends with consideration of future demographic, socio-economic, and climatic disturbance on thermoelectric (NETL 2011), irrigation (Döll 2002), and domestic water uses. More details can be found in Duan et al (2019) and Brown et al (2013).

Dynamic water stress model
We developed a dynamic water stress model to investigate regional water availability and stress levels in various contexts of climate, water uses, and water transfers. The HUC-8 level watershed water balance was simulated by a previously validated monthly eco-hydrological model-the US Department of Agriculture Forest Service (USDA-FS) Water Supply Stress Index (WaSSI) model (Sun et al 2011, Caldwell et al 2012. WaSSI was developed to capture land-cover specific hydrological processes and water balance components in the CONUS. Monthly precipitation and temperature for the period of 1961-2010 from the Parameter-elevation Relationships on Independent Slopes Model dataset (http://prism.oregonstate.edu/) were used to drive the WaSSI model for historical runoff simulation. Monthly precipitation, temperature, solar radiation, wind speed, and specific humidity derived from 19 CMIP5 GCMs were used to project runoff under future scenarios of RCP4.5 and RCP8.5.
Regional water availability was simulated by accumulating and routing streamflow through the river networks across the CONUS (Duan et al 2018). We assumed that water withdrawal, consumption, and transfer would occur uniformly in each watershed, while the return flows, the transferred water, and the residuals of the accumulated flow would be discharged simultaneously to surface water at the inlet of the next downstream watershed. A total of 18 777 upstream-downstream water connections among the HUC-8 watersheds were identified based on the geospatial attributes of streams at different hierarchical levels obtained from the National Hydrography Dataset (http://nhd.usgs.gov/data.html).
We further defined four scenarios of regional water availability to disentangle the independent impacts of each IBT and consumptive water uses on water supply and water stress at each watershed, including 'natural' , 'post-IBT' , 'post-consumption' , and 'post-consumption&IBT' streamflows. 'Natural' streamflow (TF m ) represents a natural state of maximum water availability without anthropogenic disturbances, as where LF is runoff generated from local watershed; UF i represents water flow accumulated from the ith upstream watershed. 'Post-IBT' (TF t ) and 'postconsumption' (TF c ) streamflow represents regional water availability disturbed by the water transfers and upstream consumptive water uses respectively, as where T j is the magnitude of water transferred in or out of the watershed through the jth IBT project; C i is water consumption that occurs in the ith upstream watershed. Similarly, 'post-consumption + IBT' streamflow (TF ct ) represents water availability under the combined influence of upstream water uses and IBTs, as IBT-induced variations in regional water availability with and without the influence of upstream water uses are quantified by comparing the simulations of TF ct to TF c and TF t to TF m , respectively.

Quantification of IBT efficiency
'Efficiency' is usually evaluated by weighing the benefit against the cost (Colby 1990). Based on the assumption that an IBT-induced increase and reduction in water availability would respectively constitute a positive and negative impact on regional water supply, we here suggest two sets of simple metrics to consistently evaluate IBT impact on regional water supply and its efficiency.

Extent metrics
Due to the upstream-downstream water dynamics among hydrologically connected watersheds, an IBT could impact a much larger area besides the watersheds exporting and receiving the transferred water. We used six metrics to evaluate the extents of positive and negative impacts of each individual IBT on water supply, including land area and population positively/negatively impacted by the transfer, and changes in the coverage of highly stressed area and population caused by the transfer (SI appendix, table S1 available online at stacks.iop.org/ERL/17/044050/mmedia).

Efficiency metrics
We developed two simple indices using only water availability and water demands to assess IBT efficiency: (a) 'Natural efficiency' is measured by difference between transfer-in ratio and transfer-out ratio (DIO). Transfer-in ratio (TI) is the ratio of transfer magnitude to the mean annual streamflow in the water-receiving watershed, reflecting the degree of benefit for water supply in the destination region. Transfer-out ratio (TO) is the fraction of mean annual streamflow removed by an IBT from the waterexporting watershed, reflecting the cost of water availability reduction and ecosystem disturbance in the source region. DIO is calculated as where TI and TO represents the ratio of transfer magnitude of an IBT to regional 'natural' renewable freshwater availability in the water-receiving watershed (TF m (r)) and water-exporting watershed (TF m (e)), respectively. (b) 'Social efficiency' is measured by Stress Relief Index (SRI), indicating the combined impacts of per unit transferred water on overall water stress across the areas that are in and downstream of waterexporting or receiving watersheds. The benefit/cost are measured by IBT-induced decrease/increase in regional water stress multiplying the impacted population, as where T is transfer magnitude of the IBT; P i and ∆WS i are the population and the IBT-induced change in water stress in the ith watershed. λ i is the weighing factor for the ith watershed that can be adjusted to differentiate water management priorities. Higher λ i could be implemented for regions involved with extreme water shortage, vulnerable ecosystems, or endangered aquatic species. Here λ i is set to be the constant one to facilitate a consistent comparison of climate change impact across the country. ∆WS is calculated as where WD is off-stream water demand. Both indices measure efficiency by weighing the positive impacts of an IBT on water supply against the negative impacts. Therefore, a larger value of DIO or SRI indicates a higher efficiency of an IBT in remedying water scarcity. DIO and SRI may be either positive or negative, with a negative value suggesting that the IBT is inefficient in a certain hydrological and wateruse context.

Spatial extents of IBT impact on regional water supply
We identified 150 and 145 watersheds directly receiving and supplying water of IBTs, respectively (figure 1). The number of involved watersheds increase to 387 (water-receiving) and 356 (watersupplying) respectively when downstream influence of water transfer is considered. The extents of positively impacted land area and population of individual IBTs varied from 1.2 × 10 3 to 1.2 × 10 5 km 2 and from 1200 to 8.5 million people, whereas the negatively impacted area and population also varied widely from 0.7 × 10 3 -1.3 × 10 5 km 2 and from 1900 to 4.9 million people (figures 2(a) and (b)). We found that the negatively affected area of 112 IBTs exceeded the corresponding positively impacted area, including 58% (71 out of 122) in the east and 39% (41 out of 106) in the west. The negatively impacted population also exceeded the positively impacted population in 69 (57%) eastern IBTs and 38 (36%) western IBTs, respectively. These results suggest that nearly half of these existing IBTs compromised water supplies for either a larger area or population.
We cross-compared historical water stress under the four scenarios of regional water availability (equations (1)-(4)). Discrepancy among these scenarios suggests only a slight IBT-induced change in the coverage of highly stressed land area, either with (14.8%, 1.2 × 10 6 km 2 ) or without (14.3%, 1.1 × 10 6 km 2 ) the influence of upstream water consumption. Meanwhile, the corresponding proportion of highly stressed population was reduced from 18.5% to 17.6% and from 18.3% to 17.3% (figure 2(c)). We identified 12 most influential IBTs in Colorado, California, and New York that had relieved 2.3 million people from HS. These results confirmed the crucial role of IBTs in facilitating water supply for densely populated regions.

IBT efficiency in 1981-2010
Various hydroclimatic conditions and water uses led to diverse efficiency of the IBTs across the country in 1981-2010. On multi-decadal average, TI and TO (figure 2(d)) varied greatly from 1.4 × 10 −5 % to 1.2 × 10 4 % and from 1.9 × 10 −6 % to 65%, respectively. Most of the IBTs with large TI or TO (7 IBTs with a TI larger than 100%; 9 out of 12 IBTs with a TO larger than 10%) were located in the west due to the dry climate and large transfer magnitudes (see reference numbers and details in SI appendix table S2). There were 96 IBTs (55 in the east and 41 in the west) with TO exceeding TI, a condition suggesting an inefficient IBT. The least naturally efficient IBTs include three projects in New York (e.g. IBT#17E Delaware Aqueduct) and one in Nebraska (#15W Loup River power canal) that removed as much as 29%-58% of the streamflow from the water-exporting watersheds.
SRI evaluations show that IBTs with small transfer magnitudes can also be deemed efficient as they benefit considerable downstream populations. Several transfers (<10 Mm 3 ) located in Illinois (#68E City of Highland Distribution lines), Wyoming (#6W City of Cheyenne Stage 1 Diversion), and Nevada (#79W Alfred Merritt Smith conveyance) were ranked among the most efficient IBTs with the overall reductions in water stress exceeding 1.0 (thousand people Mm −3 ). Meanwhile, the negative SRI values indicate 67 inefficient IBTs (43 in the east and 24 in the west). The smallest SRI (−1.8 thousand people Mm −3 ) was found in the IBT transferring water from the Sierra Nevada to southern California through the Los Angeles Aqueduct (#111W). This project decreased water supply for 3 million people downstream by removing water from upland areas.
Contradictory results between natural and social efficiency (i.e. DIO × SRI < 0) were found at 55 IBTs ( figure 3(a)). Such inconsistency reflects different perspectives on the role of IBTs in local and basinlevel water systems. While DIO reflects the ratios of transfer magnitude to streamflow in the source and destination watersheds, SRI varies with the relevant population and the aggregate hydrological response to variations in climate and water uses. Over a short period of time, annual variation in SRI is mainly controlled by the temporal variability of the dry and wet spells since socioeconomic status is relatively stable. For instance, higher efficiency at the three largest IBTs in California ( figure 3(b)) coincided with the drought years of 1995-1997-2003(Robeson 2015.

Future IBT efficiency under climate change 3.3.1. Response of IBT efficiency
With current transfer routes and volumes, future climatic and socioeconomic changes would enhance or suppress IBT efficiency disproportionately across the CONUS ( figure 4). Multi-model average results suggest increasing signal at over half of the IBTs, accounting for 50% (IS) and 54% (HS) by DIO and 54% (IS) and 61% (HS) by SRI. In particular, increasing SRI would cover more than 70% of the 106 western IBTs due to the significant rise in pre-transfer water stress in the regions benefited from the IBTs. Discrepancy between the changes in DIO and SRI is widely found in the upstream areas of large rivers such as the Colorado River and the Rio Grande River, suggesting higher degree of dependence on the transferred water when downstream effects of the changing environment are considered. However, extensive decrease in efficiency is also expected. The proportion of naturally inefficient IBTs would slightly decrease from 42% to 41%, yet socially inefficient IBTs would increase from 29% to 32% (IS) and 35% (HS).
Multi-model means in post-transfer water stress suggest more frequent escalations to HS at waterexporting watersheds (9 under IS and 12 under HS) than that at water-receiving watersheds (4 under IS and 8 under HS) (figure S3). Downgrade from high stress to lower stress levels can only be expected at 1-3 water exporting/receiving watersheds. Such results imply that some water-exporting regions may not be able to afford the transfers in the farther future, yet the transferred water may be insufficient to remedy the worsening water shortage at water-receiving regions.
Future efficiency response shows divergent trends among historically efficient and inefficient IBTs (figure 5). 46% (IS)-57% (HS) of the changes in DIO and 50% (IS)-63% (HS) of the changes in SRI suggest increasing signals in efficient IBTs and decreasing signals in inefficient IBTs. The IBTs are expected to shift towards a pattern of 'efficient gets more efficient and inefficient gets more inefficient' as climate change intensifies. This is probably associated with the shift of hydrological cycle under global warming that partially leads arid regions to become drier, such as the western CONUS. Particularly, 7 (IS) to 16 (HS) IBTs across the central (e.g. #62 W Montezuma Valley Irrigation Company Canal 2 in Colorado) and southeastern (e.g. #62E Cobb County Water System in Georgia) CONUS are expected to switch from efficient to inefficient (i.e. from positive to negative SRIs), while the opposite switch can only be found in 2 (IS)-4 (HS) cases. Efficiency change of these IBTs are expected to reach a tipping point that their role in the water system would switch from alleviating to aggravating the overall water stress.

Driving forces of potential efficiency change
We examined the relative roles of drivers of efficiency change from the historical period of 1981-2010 to the future period of 2070-2099. Future changes in natural efficiency are resulted from terrestrial water balance shifts induced by climate change. Across IBTs with increasing and decreasing DIO, streamflow variations at water-receiving and water-exporting watersheds are projected to dominate the changes in natural efficiency, with the relative contributions reaching 70% (IS)-78% (HS), respectively. SRI change was decomposed into the independent effects of population (including the subsequent changes in domestic water use), runoff, and thermoelectric and irrigation water uses (including the impacts of changing water demand and upstream water consumption) (figure 5). Population growth and migration are likely to be the largest driver, accounting for 41% (IS)-52% (HS) of SRI change based on the multi-model mean results. For example, large increases in SRI are projected in the southwest as population expands in water-receiving areas. However, the spatially inhomogeneous population growth could compromise IBT efficiency. The proportion of IBTs where the negatively affected population exceeds the benefited population would remain around 47% under the IS scenario, but increase to 62% under the HS scenario ( figure S4). Runoff variations can explain 36% (IS)-39% (HS) of the increases in SRI and 33% (IS)-34% (HS) decreases. Impact of runoff variation on efficiency tends to be enhanced through the aggregation of upstream flows across large river basins (e.g. the Rio Grande, Gila, and Colorado Rivers), where availability of upstream flow is identified as the most influential factor on water stress in downstream watersheds (Duan et al 2019).
The drop of water demand for energy production in water-exporting areas would be a major driver of efficiency increase when climate change adaptation measures are implemented. Its relative contribution reaches 37% at the IBTs switching from inefficient to efficient under the IS scenario (e.g. #27 W Springfield Pipeline in Missouri). Variations in thermoelectric water use can explain 13% (IS)-7% (HS) of the SRI increase and 25% (IS)-12% (HS) of the decrease in terms of the national average. The averaged contributions of irrigation water use are generally small (2%-6%) for IBTs with either increasing (3%-6%) or decreasing SRI (2%-5%). However, growing evaporative demand in water-receiving region could be a dominating driver of efficiency rise, such as the St. Mary Canal (#4 W) in Montana (64%-82%). Contributions of thermoelectric and irrigation water uses are consistently larger under the IS scenario, suggesting that adaptation practices could impact IBT efficiency extensively through curbing water withdrawal and consumptive uses.

Discussion and summary
The value of water is not the same for communities under different water stress levels. Although the IBTs are designed to divert water from where it is to where it is needed, they inevitably 'rob Peter to pay Paul' in a water-stressed world and may even cause environmental consequences (Webber et al 2017). Our analysis highlights that IBT efficiency in relieving water stress is constrained by topography, population distribution, and upstream-downstream water connections (i.e. stream flows, water withdrawals, and return flows). These results provide a reasonable evaluation for resource management at different administrative levels by helping to understand the role of IBTs in the water systems. Upland transfers should be dealt with particular caution for their potentially amplified downstream influences. Overlooking the downstream impacts of water transfers or not evaluating their efficiency in a dynamic supply-demand context could lead to misleading guidance for water and land managers.
The results support our hypothesis that future climatic and socioeconomic changes would lead to spatially diverse changes in IBT efficiency in water stress alleviation. The projected decline in efficiency and post-transfer stress levels raise the question of the sustainability of current IBTs in a fast-changing environment. Increased dependence on transferred water and stress escalation in water-exporting watersheds will lead to a higher cost to externalize water stress, and probably a higher risk to trigger conflicts over water rights (Doremus 2011). Population growth in waterreceiving areas would be a major driver of efficiency rise, but future inhomogeneous increases in population could render 15% more IBTs obsolete as the suffered population exceeds the benefited population. Efficiency of IBTs in the arid western United States, particularly across large river basins, is more likely to be enhanced by the widespread decline in water availability Dunne 2016, Duan et al 2017). Nevertheless, the increasing proportions of inefficient IBTs reveal the vulnerability of IBT efficiency to climate change from both the aspects of water demand and supply. Discrepancy between the two future scenarios demonstrate that adaptation strategies can significantly affect IBT efficiency by altering upstream water availability and pre-transfer water stress.
Several limitations and caveats apply to our study. Besides the uncertainties involved with the multidisciplinary datasets, we did not consider the role of groundwater supply and its interactions with surface water flows. We have excluded groundwater in the stress evaluations to highlight the impact of surface water redistribution by the IBTs. There is evidence that highly stressed basins collocate with regions of relatively low groundwater recharge rates across the globe, and groundwater resources are unlikely to de-emphasize the stressed basins in the long term (Qin et al 2019). However, the benchmark of pretransfer water stress would be different when incorporating groundwater and saline water uses in the tabulation. The societal reliance on the IBTs could vary with the potential changes in groundwater management policy and the ratio of groundwater use to the total water demand, particularly in regions experiencing considerable changes in groundwater accessibility (Fang and Jawitz 2019). Also, the local moisture recycling disturbed by human activities may affect regional water stress to various extents, such as the enhanced evaporation due to irrigation practices (Kang and Eltahir 2018) and increased open water area in reservoirs and aqueducts.
This study presents an integrated simulationevaluation framework to assess the large-scale response of IBT efficiency to the nonstationary climatic and socioeconomic contexts. We have focused on demonstrating how the changing environment would challenge the IBT infrastructure and regional water management (Milly et al 2008). Further research on the interactions between IBTs and environmental stressors and compilation of an updated IBT inventory with information on the utilization of the transferred water and the operation strategies of IBT projects are warranted. Improved national survey of natural and anthropogenic water cycles, proper measures curbing water uses, and a flexible water transfer strategy coordinating with the environmental changes will be vital for achieving a sustainable use of the IBTs.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.

Code availability
All computer code used in this paper is available upon request from the authors. insights on the results interpretation. K D wrote the initial draft, and all authors contributed comments and edits to finalize the paper.