Envisioning a sustainable agricultural water future across spatial scales

Sustainable agricultural water systems are critical to ensure prosperous agricultural production, secure water resources, and support healthy ecosystems that sustain livelihoods and well-being. Many growing regions are using water unsustainably, leading to groundwater and streamflow depletion and polluted water bodies. Often, this is driven by global consumer demands, with environmental and social impacts occurring in regions far from where the crop is ultimately consumed. This letter defines sustainable agricultural water limits, both for quantity and quality, tying them to the impacts of agricultural water use, such as impacts on ecosystems, economies, human health, and other farmers. Imposing these limits will have a range of both positive and negative impacts on agricultural production, food prices, ecosystems, and health. Pathways forward exist and are proposed based on existing studies, showing the gains that can be made from the farm to global scale to ensure sustainable water systems while sustaining agricultural production.


Introduction
Ensuring sustainable agriculture and water systems is imperative in the face of climate change, growing populations, and changing diets. Due to their tight interconnections, these systems cannot be managed separately, as limited water supply and impaired water quality harm agricultural production, and agricultural production contributes to degraded water supplies. Irrigated agriculture accounts for 44% of global crop production (Alexandratos and Bruinsma 2012), with 62% sourced from surface waters and 38% from groundwater (Mekonnen and Hoekstra 2010). Agricultural water use has increased in recent decades (Wada et al 2012), including groundwater use in regions already experiencing groundwater depletion (Wada et al 2012, Dalin et al 2017 and those that have not traditionally required irrigation (Jame and Bowling 2020). However, irrigation benefits agriculture as it increases yields and buffers against climate variability (Troy et al 2015, Li andTroy 2018). It is therefore critical we ensure sustainable water-food systems, including both the quantity and the quality of the water used by and discharged from agriculture.
Many studies examine water stress from a quantity perspective (Wada et al 2011, Mayer et al 2012, Rushforth and Ruddell 2016, Herbert and Döll 2019, but water quality limits on availability are often overlooked and it is more difficult to define hard limits for sustainable water management, leading the World Health Organization to call water quality 'the invisible water crisis' (Damania et al 2019). In most high-income countries, agriculture has replaced point sources from industry and wastewater as the largest source of contamination (Mateo-Sagasta et al 2017). Agricultural runoff is typically high in nitrogen and phosphorus (Carpenter et al 1998, Howarth et al 2002. It can lead to significant saline loading in rivers (Nauman et al 2019) and aquifers (Pulido-Bosch et al 2018), which can decrease crop yields (Zörb et al 2019). Given the importance of agriculture as a polluter, ensuring clean water requires agriculture to implement best practices.
The linkages between agriculture, water use, and environmental impacts occur across spatial scales. Global food preferences, population, and political decisions, such as ethanol use of corn, drive agricultural production in areas distant from the consumption of that product. That agricultural production is then dependent on the local water resources and may drive local environmental degradation. Farm-level decisions, such as crop mix, fertilizer and pesticide inputs, drainage and irrigation water management, can lead to downstream impacts in areas distant from the agricultural production, like the Gulf of Mexico's hypoxia impacting fisheries off the Louisiana coast (Zimmerman and Nance 2001).
Multiple definitions of sustainable water management exist (United Nations 1992, Mays 2006, many of which contain qualitative, subjective, and normative language, making it difficult to implement (Larsen andGujer 1997, Russo et al 2014). One exception is the quantification of sustainable limits using the Ecological Limits of Hydrologic Alteration (ELOHA) framework, in which quantified hydrologic alteration is linked to quantified ecological impacts to determine limits on water use (Poff et al 2010). Extending the ELOHA approach to include water quality as an additional hydrologic alteration, and non-ecological impacts associated with hydrologic alteration, has the potential to define sustainable water limits that account for the three pillars of sustainability: economic, social, and environmental.
We define agricultural water management to include agricultural practices that impact water resource quantity and quality. This letter first reviews the external drivers of agricultural water use. It then extends the ELOHA framework to define sustainable water limits. It examines the consequences of imposing sustainable water limits on agriculture, freshwater ecosystems, and other connected systems, concluding with pathways forward. This letter contributes to the literature in two ways: (1) expanding an existing framework for determining sustainable water limits to explicitly include water quality and non-ecological water sustainability targets, and (2) establishing connections across spatial scales, from the local to global, in water sustainability and its drivers, including consumer food demands and agricultural production.

Drivers of water use in agriculture
Water sustainability, or the lack thereof, arises in response to various external pressures. These pressures occur at a range of scales from global to local and are most tightly linked to decisions and actions at the local farm scale, which then impact agricultural water management.
Global demand for agricultural output. Global agricultural markets, including prices, impact local decisions made by agricultural producers. Globally, increasing population increases demand for agricultural production; increasing income increases demand for meat and other livestock-derived products (Milford et al 2019, Komarek et al 2021, which are more water intensive (Hoekstra and Mekonnen 2012). Unsustainable irrigation demands at the local scale are commonly associated with global export market demand, often driven by cotton, sugar cane, fruit, and vegetable export (Rosa et al 2019, Richter andHo 2022). Increasing domestic and international demand for pistachios and almonds has been accompanied by an increase in California acreage of both trees Horwath 2014, 2016), which have significant water demands (Fulton et al 2019). Biofuels were seen as a climate change mitigation strategy, but they have environmental consequences, both in terms of expanding agricultural land use (Delucchi 2010) and removing land from food production (Rulli et al 2016).
Climate change is projected to alter the hydrologic cycle through changes in timing, mean, variability, and extremes (Douville et al 2021). In some cases, this may result in changes in water stress, with more projected areas with increased water stress than less (Gosling and Arnell 2016).
Governance and water rights. Political and hydrologic boundaries rarely align, leading to varying water governance within a watershed. For example, the Saskatchewan River basin is in two countries, three provinces, and First Nations land, with provinces and the federal government having different management responsibilities. This fragmentation and overlapping of political boundaries and responsibilities inhibits effective governance (Gober and Wheater 2014), which is critical for reaching sustainability goals.
Water rights can reduce flexibility in water management, impeding sustainability. In parts of the western US, prior appropriation rights are predicated on the principle of 'use it or lose it' , encouraging farmers to use more water than needed, while the newer regulated riparian system used in many eastern states may not protect groundwater resources during drought (Jame and Bowling 2020). In Australia, the Murray-Darling basin allows for trading of water rights and entitlements, providing economic gains (Grafton et al 2012) with the potential for environmental flow gains (Docker and Robinson 2014). This points to both the importance and potential for including governance and water rights into water sustainability frameworks.
Risk management practices. There are built-in incentive structures to overuse water and other inputs in crop production. Crop insurance may require water application, leading to increased irrigation (Deryugina and Konar 2017). Fertilizers provide an opportunity for higher yields, but overapplication may occur due to input uncertainty (Paulson and Babcock 2010), risk aversion, or reapplication due to weather (O'Connell and Osmond 2022). However, best management practices can lead to reductions in nitrogen application without yield reductions (Good and Beatty 2011 and references within), indicating that there are opportunities for improvement.
Water infrastructure built to supply or drain agricultural water can contribute to sustainability issues. Subsurface drainage of agricultural land can lead to flashier streamflow (Sloan et al 2016), more linear baseflow recessions (Schilling and Helmers 2008), increased phosphorus transport (King et al 2015), and increased nitrate-nitrogen transport (Kladivko et al 2001). Farmers who have invested in irrigation equipment to supply water may be hesitant to decrease or eliminate irrigation water use, as there is no economic benefit to them. Reservoirs were constructed throughout the US to supply water for irrigation with a well-established literature about how dams impact streamflow, suspended sediment, and stream temperature (e.g. Sabo et al 2010).
These drivers will present challenges to implementing sustainable water limits, defined in the following section, and will be impacted by any implemented water limits as discussed in section 4.

Sustainable agricultural water management and its indicators
In 1987, the Brundtland Report defined sustainability as 'meeting the needs of the present without compromising the ability of future generations to meet their own needs' (World Commission on Environment and Development 1987). Water sustainability balances societal water needs with those of the environment, which include sustaining dynamicallystable water levels, particularly groundwater, and water quality to ensure future generations can meet their water needs.
Sustainable agricultural water management does not mean returning to a pristine water system, but rather that the impacts of the alteration are within acceptable limits. Every resource allocation has tradeoffs, and in practice, water sustainability is the politically accepted state of water use that can continue indefinitely. It is what society is willing to accept in terms of environmental impacts from the current water demands that sustain a region's economy. This section defines a framework for determining limits on hydrologic alteration to prevent unacceptable environmental impacts as defined by stakeholders.

Defining sustainable agricultural water limits
To determine sustainable water limits, we suggest expanding the ELOHA framework (Poff et al 2010) to include water quality and non-ecological targets. ELOHA defines the bounds of hydrologic alteration based on the acceptable consequences, determined by stakeholders, thereby incorporating the idea that sustainability is a socio-political construct that requires compromise. A similar approach is often used in defining water quality indices, where different water quality metrics are given locallyrelevant weighting factors determined through expert input using the DELPHI structured communication technique (Sutadian et al 2016). In the US, the Clean Water Act establishes the total maximum daily load (TMDL) framework which determines load reductions needed to improve water bodies with water quality impairment to regulatory limits, involving stakeholders to varying extents to allocate allowable loads between various point and non-point sources. Figure 1 lays out the steps of the framework for defining sustainable agricultural water limits. First the management area should be defined. Then the systems that have or would be impacted by hydrologic alteration should be identified, such as freshwater ecosystems, human health, or water infrastructure. If possible, the relationships between the hydrologic alteration and these systems' impact-such as decreased low flows and salmon productivityshould be quantified. This relationship and the acceptable impact established by stakeholders then determine the sustainable limit. An example of acceptable consequences may be moderate species decline but not population or community collapse (Poff et al 2010). If the relationship cannot be determined between the alteration metric and the impact, presumptive standards should be used for groundwater extraction (Gleeson and Richter 2018), streamflow alteration (Richter et al 2012), and the more stringent of federal (i.e. the US Clean Water Act), local, or WHO water quality regulations. Figure 2 lays out two hypothetical examples of hydrologic alteration-impact relationships that could be used to determine sustainable water limits. The top row illustrates the causal chain of groundwater pumping for irrigation leading to declining groundwater levels (a), which lead to declining baseflow and therefore streamflow, particularly during low flow months (b). Figure 2(c) sketches the potential relationship between low flow depletion and species abundance, with a sustainable target of low flows chosen before there is a species collapse. This sustainable target would be tied to groundwater levels and lead to limits on groundwater pumping. Both relationships-declining water table impacts on baseflow and baseflow depletion impacts on species abundance-are areas of active research (e.g. Palmer and Ruhi 2019, Yarnell et al 2022), and modeling frameworks exist and continue to be developed that could estimate these relationships (Horne et al 2019).
The bottom row lays out a water quality scenario of fertilizer application in agricultural areas with subsurface drainage. Fertilizer application leads to higher nutrient inputs to streamflow (d) and (e), with targets tied to the relationship between nitrate loadings  There can be multiple impacts of interest. Likely water quantity metrics include mean, low, and high streamflow, streamflow variability, and groundwater levels. In some cases, water quality indices, which are weighted averages of different water quality metrics, may be preferable to use as a single metric to quantify water quality impairment and its impacts. Water quality metrics include stream temperature, nutrients (e.g. nitrate and soluble reactive phosphorus), electrical conductivity, biologic oxygen demand, and turbidity (Terrado et al 2010, Hurley et al 2012, Misaghi et al 2017. These metrics are frequently monitored, are critical to freshwater ecosystem health and can serve as indicators of contamination sources that may be difficult to monitor. Alteration-impact relationships vary spatially, with different watersheds having different concerns. The expanded ELOHA framework should be applied at the local watershed scale, but some of the targets may come from larger regional scales, such as maintaining aquifer levels. Ensuring sustainability across spatial scales is necessary to avoid disproportionate impacts outside of the local watershed. In some regions, water systems may require time to recover, particularly in those regions with significant groundwater depletion (Basso et al 2013, Steiner et al 2021) or legacy pollution (Basu et al 2022). A balance will have to be struck, as current users should not be unduly penalized for past actions.

Examples of applying sustainable agricultural water limits
One way to assess water sustainability is through water stress indicators, which have been the subject of numerous studies from the local (e.g. Mayer et al 2012, Rushforth and Ruddell 2016) to global scales (e.g. Wada et al 2011, Herbert andDöll 2019). Figure 3 maps the surface water stress index (figure 3(a)) and groundwater stress index (figure 3(b)) over the contiguous US when only water quantity is considered as is typically done (see Supplementary Information for methods and data). Similar water stress maps have been constructed in many other studies, but the quantification of water quality impairment is generally thrown into the category of economic water scarcity with assessments done at the country scale. Figure 3(c) demonstrates the impact of incorporating quality into the quantification of physical water scarcity, using a simple method that assumes water supply from impaired water bodies is unavailable for use (Supplementary Information). Including water quality increases water stress in 20% of the HUC12 watersheds (figure 3(c)) compared to only considering water quantity (figure 3(a)), with 5% more watersheds nationwide experiencing severely water stressed conditions (WSI > 0.6).
National scale using presumptive standards. We can examine how water stress would be reduced if sustainable limits were imposed for agriculture. Figure 3(d) presents the remaining water stress if a presumptive standard is applied across the US. The presumptive standard here is that agricultural withdrawals are limited to 40% of supply and agricultural pollution is reduced to keep water quality levels at or below the levels set by the EPA for impaired water bodies. The decrease in water stress between figures 3(d) and (c) demonstrates the result of imposing these standards. The water stress index is reduced in 12% of all watersheds in the country and these are the most extremely impacted watersheds in the country. The average water stress index reduces from 2.3 to 0.8, and the percentage of severely water stressed watersheds decreases from 14.3% to 8.7%. Water stress remains in metropolitan regions like Boston, where it results from municipal use. This example shows the potential for implementing sustainable water limits in agriculture and the impacts they would have in reducing surface water stress.
Locally defined targets at the watershed scale. Where possible, locally defined targets should be developed tailored to the local conditions. The ELOHA framework was tested over catchments in the Potomac River basin to evaluate if flow alterationecological response relationships can be developed (Buchanan et al 2013). They found that alterations in the duration and frequency of high and low events and in flashiness led to responses by stream macroinvertebrates, demonstrating that the response relationships can be quantified. To implement, stakeholders would need to determine the degree of invertebrate decline that was acceptable, and then the alterationimpact relationship they developed would be used to set sustainable limits. In the nearby Upper Tennessee basin, McManamay et al (2013) implemented the ELOHA framework, also finding that multiple hydrologic variables must be included to determine ecological response as a target variable for setting limits. In addition to hydrologic flow variables, they found other alterations could be important, including stream temperature, highlighting the importance for including water quality indicators with water quantity when setting sustainable water targets.
Alteration-impact relationships do not have to be tied to ecological indicators. In Kansas, a local management area set a target of 20% reduction in irrigation water use to extend the usable life of the High Plains Aquifer. This target was determined by the agricultural producers in the region, and they reduced agricultural water use by 31% during the study period

What does an ideal sustainable agricultural water world look like?
For water quantity, sustainable water management would result in dynamically stable aquifer levels and streamflow statistics. For water quality, this would result in water suitable for the intended use. For both, the allowable alteration should be driven by human and ecological needs, which vary by hydrological and ecological regimes. Sustainable water management will provide stability for water users, with the potential to implement water savings during wet years to augment dry years. At the regional watershed level, the same benefits will accrue: dynamically-stable, clean water with certainty for water users because limits imposed at the local level are designed to meet both local and regional sustainability constraints. This results in water sharing between users and the environment that ensures water sustainability across spatial scales.
At the national and global scales, this approach should meet water sustainability goals in a manner that ensures food needs are met, human health is protected, and ecosystems are preserved. Studies show this is a plausible goal. Improving crop water productivity globally has the potential to increase agricultural production while saving water (Brauman

Consequences of water (un)sustainability
Implementing sustainable water limits may adversely affect agricultural production. Constraining irrigated production of maize, soybean and winter wheat in the US to sustainable water limits would decrease yields by 20%, 6%, and 25%, respectively (King et al 2015) and raise the shadow price of water. Similarly, fertilizer limits on farmers could reduce yields, driving up food costs. In the short run, food prices will increase (9% reported by Haqiqi et al (2023)), motivating local production expansion, redistributing agricultural production along with the associated jobs and environmental costs to more water-favorable regions (Graham et al 2021), and increasing irrigation in traditionally rainfed regions of the eastern US (Jame and Bowling 2020). These externalities can also be passed to related regions through virtual water trade that allows water-stressed regions to import food, thereby avoiding agricultural water use locally. If the virtual water comes from other waterstressed regions, imposing sustainable water limits in one location simply transports unsustainable water use to other regions, not unlike the 'leakage effect' widely reported in carbon policy studies (Babiker 2005, Jakob 2021, Grubb et al 2022. The indirect leakage effect is generally more moderate than the direct effect due to adaptations throughout the system such as substitution of food consumption, adjustment of agricultural technology, and changes in trade networks (Calzadilla et al 2010, Haqiqi et al 2023, Liu et al 2017. The nature of the effect, however, varies depending on factor endowment and comparative advantages of the trade partners, as well as the substitutability between domestic and foreign goods and environmental relevance (Boulay et al 2013, Hoekstra 2017, Hogeboom 2020. In addition to the spatial spillover, the externalities can spread to other agricultural sectors and even beyond agriculture, with the purpose of increasing the net return for a unit of water used. As the more stringent sustainability standard raises the cost of water, competition for water tends to reallocate resources to the uses with higher value to offset the higher cost. The definition of water productivity can differ across scales and stakeholders. In some cases, there may be a switch to less waterintensive crops or a change from low-value to highvalue crops (Júdez et al 2011). This was seen after a recent policy change in Kansas to promote groundwater conservation, where farmers were able to reduce water use by 31% through improvements in irrigation efficiency and an expansion in the acreage of crops that use less water (Deines et al 2019), indicating that implementing water sustainability limits can be done without a collapse of agriculture. There is potential to innovate in the rainfed agricultural sector to increase water productivity without sacrificing yields (Jägermeyr 2020) or to expand irrigation to supplement rainfed production where water is available (Rosa et al 2020). However, innovations in irrigation efficiency have not always improved water availability (Perry 2007, Perry et al 2009, Grafton et al 2018. Irrigation efficiency improvements can have unintended environmental consequences, such as reducing return flows to streams, leading to depletion from the current baseline and reducing groundwater recharge (Kendy andBredehoeft 2006, Ward andPulido-Velazquez 2008). For those no longer able to irrigate or fertilize, land values may decrease (Torell et al 1990), leading to inequities. These inequities may occur at the national scale as well, as countries importing irrigation intensive commodities are at higher risk of being affected by sustainable practices (Dalin et al 2017). In some hot spots, prioritizing water sustainability and food production over other needs may lead to water shortages for electricity, industry, and domestic use (de Vos et al 2021). These negative impacts highlight the importance of taking a larger systems view of sustainability, including other connected sectors like trade and energy.
Implementing sustainable agricultural water limits may conflict with other sustainability goals. Limiting irrigation transfers the pressure of ensuring food security to rainfed production, requiring larger cropped areas to account for lower and more variable yields in rainfed production which may then raise concerns about land use change and pollution. Haqiqi et al (2023) found that imposing a sustainable groundwater policy in the US led to US irrigated acreage decreasing by 11.6 million hectares, but global cropland increasing by 29.5 million hectares, effectively exporting the environmental costs of production. Imposing nitrate conservation measures in the Mississippi River basin shifts production outside of the basin (Liu et al 2022a). In 2008, food prices in Brazil increased significantly, with rural poor farmers benefiting and urban populations having increased poverty (Ferreira et al 2013). This bifurcation in impact based on net food producers and consumers has been seen in multiple countries with net increases in poverty with rising food prices (Ivanic and Martin 2008).
Just as implementing sustainable water limits can lead to increased food production costs, so can inaction and continuing unsustainable water use. First, there will be increased agricultural production costs as water supplies dwindle, including increased marginal pumping costs (Kanazawa 1992). In India, studies have shown that yields, cropped area, and cropping intensity have declined due to declining groundwater levels (Zaveri et al 2016, Bhattarai et al 2021, Jain et al 2021, and poverty increases with declines in groundwater (Sekhri 2014). Then, there will be external costs to human health and the environment, such as nitrate pollution possibly increasing gastric cancer risk (Picetti et al 2022) and pesticide use leading to increases in medical disability (Lai 2017).
The costs of unsustainable agricultural water use also extend to downstream water bodies, ecosystems, and users. Wastewater treatment plants have higher removal costs due to agricultural inputs like nutrients and pesticides (Dodds et al 2009, EPA 2015, Price and Heberling 2018. The Chesapeake Bay has excess nutrient pollution due to agriculture in the Susquehanna River (Hagy et al 2004), with an estimated $1.6 billion budgeted for restoration efforts in 2022 (OMB 2022). Meeting restoration targets for the Gulf of Mexico's hypoxia zone, largely the result of agricultural pollution, is estimated to cost $2.7 billion annually.
Implementing sustainable limits will have negative impacts, but these impacts can be mitigated through policy levers as part of a planned process. Continued unsustainable water use may lead to water limitations regardless, as nature will eventually do it for us. If that occurs, we will have lost the flexibility provided by a managed system, such as preserving groundwater to ameliorate water stress during droughts. A managed process to achieve sustainable water will be far less damaging than leaving it to nature, which would be a 'chaotic disallocation' of water (Perry 2019).

What are the pathways forward?
Sustainable agricultural water limits need to be implemented in many agricultural regions globally to sustain long-term agricultural productivity, downstream users, and ecosystems. Water use limits should be negotiated by stakeholders using the expanded ELOHA framework, which both accounts for the local nature of water issues and allows for water targets that are informed by impacts of interest across spatial scales. A possible impediment to implementing this framework is establishing the alteration-impact relationship, which can be dataintensive (Williams 2018) or require modeling. In addition, these alteration-impact relationships can involve multiple causal variables, such as decreasing streamflow leading to higher stream temperatures thereby decreasing fish populations. Stakeholder engagement is critical to determine thresholds (Stein et al 2017, Mussehl et al 2022, which can also present challenges implementing an ELOHA framework. However, case studies have shown that ELOHA is feasible (Zhang et al 2016, Stein et al 2017, and presumptive standards can be used when either data or institutional capacity is lacking. The TMDL process established by the US's Clean Water Act is an existing example of presumptive standards to define water quality loading limits to different point and non-point sources.
While implementing ELOHA can provide sustainable water limits, policies are needed to ensure sustainable agricultural water use. Ensuring water quality may require revising regulation of non-point source pollution, as the current system relies on voluntary adoption of conservation practices. In the Midwestern US, this could involve regulating tile drain outflow before it joins the surface channel. Implementing ELOHA with multiple objectives/limits, such as both sustaining low flows and water quality, will likely require multiple policies. Because water is so interconnected, there could be a desire to achieve too many objectives. Therefore, realism in possible objectives, and associated limits and policies, is necessary.
To ensure sustainable water supplies, tradable water rights for surface and ground water will likely be necessary to provide flexibility in agricultural practices and water savings, like in the Murray-Darling (Docker and Robinson 2014). There is a long history of conflict over water allocations and rights that continue today, such as in the Colorado River basin, and any policy change will involve difficult decisions and possibly inaction. However, stakeholder engagement to identify possible sustainable limits can set the stage for when action is possible. Under certain conditions, crises and related focus points can open policy windows during which policy change is more likely (Michaels et al 2006, Rose et al 2020, Liu et al 2022b. Sustainable water limits may result in the spatial redistribution of crop production, and policy mechanisms can support farmers in this transition. Current agricultural regions will likely continue for as long as climate change allows as they are sited in regions with fertile soils and knowledgeable farmers with developed supply chains, but changes in crop composition can allow for significant water savings (Davis et al 2017b). Supporting shifting crop production patterns may require changing subsidies for crops and crop insurance programs. It will require supply chain development; the Midwestern US supply chains are efficient for maize and soy production but other crops face barriers. Extension programs can provide farmers with choices that work with the climate and water availability. Policy levers, including the US Farm Bill, should be explored to encourage farmers to either shift to less management intensive approaches and crops or improve water conservation practices.
This letter lays out a framework for ensuring both sustainable water quantity and quality, tied directly to the impacts of agricultural water management. This is a practical framing: tradeoffs exist with any decision involving resources, and this is particularly true when considering the interplay between agricultural production, water needs, and the environment. These frameworks must be applied across spatial scales, as we must be sustainable from the farm level to the global.

Data availability statement
The data that support the findings of this study are openly available at the following URL/DOI: www. usgs.gov/mission-areas/water-resources/science/ water-use-united-states and waterwatch.usgs.gov.