Groundwater is falling in 71% of the world's monitored aquifers. The cut is already being made — on farms first, and on the smallest users hardest.
The Cut Has Already Been Made
Groundwater is not vanishing everywhere at once — it is being reassigned
Groundwater levels are falling in 71% of the world's monitored aquifer systems, according to an analysis of 170,000 wells across 1,693 aquifer systems published in Nature in January 2024 [1]. ✓ Established In 36% of those systems the water table falls by more than 10 centimetres a year, and in 12% by more than half a metre [1]. The scarcity is real. What is less understood is that the adjustment has already begun, and that in most basins it is being made by default rather than by decision.
Water scarcity is usually narrated as a coming conflict between states. The evidence points somewhere less cinematic and considerably more consequential. Agriculture accounts for about 72% of global freshwater withdrawals [3]. ✓ Established Groundwater supplies close to half the water used for irrigation, and unsustainable groundwater use is embedded in food that crosses borders every day [28]. When an aquifer is drawn down faster than it recharges, the water does not disappear from the economy in a single event. It is progressively withdrawn from the lowest-value, least-protected use and delivered to the highest-value, best-defended one. That process has a name. It is reallocation, and it is already the governing fact of water policy in the Ogallala, the North China Plain, the Punjab, the Central Valley and the Iranian plateau.
The reason reallocation happens by default is physical. A falling water table raises the cost of lifting water through deeper wells, larger pumps and more electricity per cubic metre. That cost is a charge on every user in the basin, and it is not felt equally. A municipal utility passes it to ratepayers. An industrial user absorbs it inside a product whose value per cubic metre is orders of magnitude higher than a field of forage. A smallholder with a shallow well and a five-horsepower pump can do neither. Long before an aquifer is empty it has become unavailable to the users with the least capital. The hydrology is uniform across the basin. The distribution of the loss is not.
The Nature analysis found that declines had accelerated over the past four decades in 30% of the world's regional aquifers, with the sharpest drawdowns beneath cultivated land in dry regions [1]. ✓ Established Ninety per cent of the aquifers showing accelerating decline lie in areas that grew drier over the same period [2]. In regional aquifers under Iran, India, Spain, the United States, Saudi Arabia, Morocco and Mexico, median rates of decline exceed one metre a year [1]. The Food and Agriculture Organization reports that renewable freshwater availability per person fell a further 7% in the decade to 2025 [3]. ✓ Established None of this is a forecast. It is a measurement of a transfer already under way.
The finding comes from the largest assessment of its kind, covering 170,000 monitoring wells across 1,693 aquifer systems in countries representing roughly 75% of global groundwater withdrawals [1]. Declines exceed 10 centimetres a year in 36% of systems and half a metre a year in 12% [1]. The same dataset shows reversals in 16% of aquifers with long records, which establishes that depletion responds to policy rather than being fixed by climate alone [1].
The macroeconomic stakes are now quantified rather than asserted. The Global Commission on the Economics of Water, reporting in October 2024, concluded that an unmanaged water crisis could place more than half of global food production at risk by 2050 and reduce median global GDP by about 8%, with low-income countries losing 10% to 15% [4]. ◈ Strong Evidence The World Bank had earlier put regional losses at up to 6% of GDP by 2050 in the worst-affected areas, naming northern Africa, the Middle East, India and China [5]. The same analysis contains the finding that matters most for policy. Where governments improve efficiency and move as little as a quarter of water to higher-value uses, projected losses fall sharply and in some regions vanish [5].
That sentence is the entire argument for treating this as an allocation problem rather than a supply problem. The volume of water available to a basin is largely fixed by climate and geology. The value extracted from that volume is not. Agriculture generates a small fraction of the economic output per cubic metre that industry and services generate, which is why every scarcity episode ends with water moving out of farming. The open questions are not whether the transfer happens but how fast, who is compensated, which crops and which communities absorb the contraction, and whether the receiving cities are ever required to pay the full social cost of what they take.
This report follows the transfer through its stages. It begins with three aquifers running the same arithmetic at different speeds, moves to the pricing rules that made depletion rational for the individual pumper, traces the way depletion is exported inside food, tests the two technologies most often nominated as substitutes, identifies who actually bears the cut, and assesses the instruments a handful of jurisdictions have used to govern the process rather than let it govern them. The evidence supports one conclusion above all others. Reallocation is not avoidable. Ungoverned reallocation is.
Three Basins, One Arithmetic
The Ogallala, the North China Plain and the Punjab solve the same equation at different speeds
The Ogallala fell 1.52 feet across southwest Kansas between January 2024 and January 2025, a steeper drop than the 1.43 feet of the previous year [6]. ✓ Established Punjab extracts groundwater at 156% of its annual recharge [8]. The North China Plain lost water at 3.35 gigatonnes a year before a state intervention of extraordinary scale began to reverse it [9]. Three basins, three political systems, one equation.
The Ogallala, or High Plains aquifer, underlies eight American states and supplies roughly 30% of the groundwater used for irrigation in the United States, supporting close to a fifth of national agricultural output [7]. ✓ Established Kansas Geological Survey measurements taken in January 2025 recorded a decline of 1.52 feet across southwest Kansas over the preceding year and 1.34 feet in the northwest, the latter nearly three times the 0.47 feet lost a year earlier [6]. In parts of the aquifer, water levels have fallen more than 200 feet since large-scale irrigation began [7]. Federal analysis suggests up to 40% of the aquifer may no longer support irrigated agriculture within decades [7].
The consequence is not desert. It is dryland farming, which is a different and much poorer business. Irrigated maize in western Kansas yields a multiple of what the same field produces on rainfall alone, and the fixed costs of the farm in land, machinery and debt do not fall in proportion. The transition therefore removes income before it removes acreage. Communities register the change first in the tax base and the school roll, then in the land price. The aquifer never announces its exhaustion. It raises the pumping lift by a foot a year until the marginal quarter-section stops paying, and the decision to stop irrigating is taken one field at a time by people who would not describe themselves as parties to a reallocation.
The North China Plain ran the same equation on a larger scale. Downscaled satellite gravimetry puts the groundwater depletion rate at 2.43 centimetres a year, equivalent to 3.35 gigatonnes annually [9]. ◈ Strong Evidence Total water storage across the region fell by roughly 49 cubic kilometres between 2005 and 2019 [10]. Drawdown funnels opened beneath Beijing, Tianjin and Hebei, and the land above them began to compact. What distinguishes the case is what followed. China did not reprice the water. It moved the water, and then it moved the accounting.
The Punjab case is the clearest demonstration that depletion is a policy output rather than a natural misfortune. India's Central Ground Water Board recorded the state's annual extraction at 26.27 billion cubic metres in 2025, a stage of extraction of 156.36%, the highest of any Indian state [8]. ✓ Established One hundred and eleven of the state's 153 assessment blocks are classified as over-exploited [8]. Tarn Taran extracts at 202.49% of recharge, Ludhiana at 194.65%, Fatehgarh Sahib at 182.13% [8]. In central Punjab the water table falls by roughly one metre a year, and producing a kilogram of paddy consumes between 3,500 and 4,000 litres of groundwater [22].
The 2025 assessment by the Central Ground Water Board puts Punjab's extraction at 26.27 billion cubic metres against a sustainable limit 56% lower, with 111 of 153 blocks over-exploited [8]. ✓ Established The figure has improved from 163.8% in 2021-22 to 152.22% in 2025-26 as canal deliveries were expanded, and 95 dark-zone blocks recorded gains [8]. Improvement from an extraction rate above 150% is progress toward a smaller deficit, not toward balance.
China's answer to the same arithmetic was to build. The first phases of the South-to-North Water Diversion had delivered 81.33 billion cubic metres by June 2025, supplying 185 million people across 45 cities and releasing 11.8 billion cubic metres into more than 50 northern rivers [27]. ✓ Established Beijing's plain water table has risen more than 13 metres over a decade and the city's over-extraction zones have been eliminated [27]. Groundwater across the wider plain has risen at about 0.7 metres a year since 2020, passing 2005 levels by 2024 [10]. ◈ Strong Evidence It is the largest documented aquifer recovery on record.
The qualification matters as much as the achievement. The recovery replaces roughly half the storage lost between 2005 and 2019 [10], and it was purchased with an inter-basin transfer whose construction cost is estimated in the tens of billions of dollars, alongside the resettlement of populations along its route [27]. ⚖ Contested What China demonstrates is that aquifer depletion is reversible at a price. What it does not demonstrate is that the price is generally payable. Very few basins have a wetter neighbour within reach, a state able to finance a transfer of that magnitude, and the administrative capacity to enforce pumping restrictions across three provinces at once.
The Price That Was Never Charged
Depletion is not a market failure — it is the predictable output of the pricing rules
Punjab has spent Rs 1.25 lakh crore on free agricultural electricity since 1997 and budgeted a further Rs 10,000 crore for 2025-26 alone [22]. ✓ Established The subsidy does not buy water. It buys the energy to lift water that is already free at the point of extraction, which converts an exhaustible common resource into a flow whose marginal cost to the user is approximately zero.
Groundwater in most jurisdictions is not sold. It is captured. Rights attach to the land above the aquifer, extraction is bounded by the capacity of the pump rather than by the recharge of the formation, and the cost the farmer faces is the cost of lifting, not the cost of the resource. Under those rules depletion is not deviant behaviour. It is the rational response of every individual pumper to a rule set that rewards speed. The farmer who conserves does not bank the saved water. It flows to the neighbour who did not. Economists have described this structure for seventy years. Very few water codes have been rewritten to reflect it.
Energy pricing then removes what little friction remained. Free or near-free agricultural electricity operates in Punjab, Andhra Pradesh, Karnataka and Tamil Nadu, and subsidised tariffs apply in most other Indian states [22]. ✓ Established Punjab's annual subsidy bill rose from Rs 604.57 crore in 1997-98 to Rs 10,000 crore budgeted for 2025-26 [22]. The policy is not irrational from the standpoint of the state that adopted it, since it delivered national grain self-sufficiency and is now electorally immovable. Its hydrological effect is nonetheless precise. It sets the marginal cost of the last cubic metre lifted at zero at exactly the moment when the correct price should be rising with depth.
United Nations indicator 6.4.1 measures value added in dollars per cubic metre of water used by sector, and the gap between agriculture and the industrial and service sectors is consistently an order of magnitude or more [3]. Agriculture takes about 72% of withdrawals globally [3]. ◈ Strong Evidence That asymmetry explains why every sustained scarcity episode ends with water leaving farming, and why the World Bank found that shifting as little as 25% of water to higher-value uses would eliminate most projected GDP losses from scarcity [5]. Value per cubic metre is a poor guide to what a society should grow. It is an excellent predictor of what a market will take.
The value-per-cubic-metre comparison is analytically powerful and politically radioactive, for a defensible reason. Food is not valued at the margin the way semiconductors are. A basin that reallocates water to its highest-value uses maximises measured output and can still end up importing every calorie it consumes, converting a domestic water risk into a foreign supply risk. Chile is the cautionary case. Its 1981 Water Code created perpetual, tradable and largely unconditional rights, and by the 2010s roughly 1% of users held about 80% of them, with export agriculture in the Petorca valley expanding while residents were rationed [26]. ⚖ Contested The market allocated efficiently by its own metric and produced an outcome most Chileans rejected.
Chile's 2022 reform, Law 21.435, is instructive precisely because of what it could not do. It declared access to drinking water and sanitation an essential human right, established that human consumption and household subsistence take priority when rights are granted or restricted, and introduced time limits on new concessions [26]. ✓ Established It did not redistribute the rights already issued. A water code can change the rules for future allocations far more easily than it can reclaim allocations already capitalised into land values, farm debt and pension portfolios. Every jurisdiction now attempting reallocation is discovering the same asymmetry.
We are rapidly changing the global hydrological cycle, due to climate change and ecosystem degradation. This is threatening human wellbeing, the global economy and the resilience of societies in the face of rising shocks.
— Johan Rockström, Co-Chair, Global Commission on the Economics of Water, October 2024Pricing reform fails for a reason that is rarely stated plainly. Volumetric water charges are a transfer from farmers to the state, and the farmers can see it while the beneficiaries, meaning future users, downstream towns and the aquifer itself, cannot organise. Metering is expensive, politically visible and, in basins with millions of wells, administratively close to impossible. The instruments that have actually worked in practice avoid the price signal altogether and operate on quantity instead, through allocations, caps and buybacks. That is not an argument against pricing. It is an observation that quantity instruments are easier to enforce against a constituency that will resist either one.
The deeper problem is that the cost of depletion is not borne by the person who causes it, and never has been. A well drilled 40 metres deeper by a farmer with capital lowers the water table beneath a neighbour who has none. The external cost is real and measurable, first in the additional energy the neighbour must buy and eventually in the abandonment of the neighbour's well, and it appears nowhere on the balance sheet of the transaction that produced it. Depletion is therefore not evidence that water markets have failed. In most basins there is no market. There is a race.
Depletion Travels in Food
Virtual water moves aquifer drawdown into someone else's trade balance
Approximately 11% of the non-renewable groundwater used for irrigation worldwide is embedded in internationally traded food, and two-thirds of that total is exported by three countries alone, namely Pakistan, the United States and India [13]. ◈ Strong Evidence Around 43% of global virtual water exports originate in regions that are already water-scarce [14].
Every tonne of wheat carries the water that grew it. The accounting convention is called virtual water, and it converts a local hydrological deficit into an entry in a trade statistic where no regulator is looking for it. The concept is not a metaphor. Research on groundwater depletion embedded in international food trade found that roughly 11% of the non-renewable groundwater abstracted for irrigation leaves the country inside crops, with Pakistan, the United States and India accounting for about two-thirds of the flow [13]. ◈ Strong Evidence The importing country records a food purchase. The exporting country records an aquifer decline that no trade agreement mentions.
The geography of the flow unsettles the standard efficiency argument. A 2025 analysis of international crop trade found that about 43% of global virtual water exports come from water-scarce regions [14]. Trade does mitigate aggregate water stress, and moving crops from wet regions to dry ones is one of the largest water-saving mechanisms in the world economy. The aggregate figure nonetheless conceals a substantial counter-flow in which the water-poor subsidise the water-rich. India is the clearest instance. It is the world's largest rice exporter with close to 40% of global trade, and that surplus is produced disproportionately in Punjab and Haryana, the two states with the deepest groundwater deficits [8].
Nothing in this arrangement is accidental, and it is not principally driven by foreign demand. It is driven by domestic policy. Guaranteed procurement prices for rice and wheat, free electricity for pumping, and a canal system that cannot deliver in the summer combine to make groundwater-intensive rice the rational crop for a Punjabi farmer even where the state's own hydrology says otherwise [22]. ✓ Established The export is the residual. Reallocating water away from that crop is not a negotiation with the world market. It is a negotiation with a procurement policy, an electricity tariff and a rural credit system that were all built to encourage exactly what is now being discouraged.
The Arizona case shows the same logic operating across a border rather than within one. Saudi Arabia banned domestic cultivation of alfalfa and other green forage crops in 2018 after exhausting its own fossil aquifers, and the Saudi-owned producer Fondomonte moved the production to the Ranegras Plain in western Arizona, where groundwater in rural basins is essentially unregulated [23]. ✓ Established The company leased about 3,088 acres of state farmland for roughly 83,000 dollars a year and in a single year pumped 31,196 acre-feet, equal to 81% of all groundwater withdrawn from the basin [23]. The hay is shipped to the Middle East. The drawdown stays in La Paz County.
A cubic metre of groundwater sold abroad inside a crop leaves no claim behind. The importing economy gains a cheaper calorie, the exporting farm gains a season of revenue, and the aquifer records a permanent debit for which no party to the transaction is liable. This is the only major resource flow in the world economy with no depletion accounting attached. Oil has reserve statements, fisheries have quotas, forests have certification, and non-renewable groundwater has a customs code for wheat.
Arizona's response arrived nine years after the operation began. The State Land Department cancelled one lease and declined to renew three others, and the state has begun considering extraction limits in rural basins that have never had them [23]. The delay is the point. A basin without a monitored extraction cap has no mechanism to distinguish a use that adds a few hundred thousand dollars of lease revenue from a use that removes the water supply of every other landowner. Reallocation happened. It simply happened through a lease negotiation rather than a water policy, and the parties whose water was reallocated were not present.
The corrective is not trade restriction, which would raise food prices in the importing countries least able to bear them and would not change the pumping incentive in the exporting basin. The corrective is domestic. Where the extraction cap binds at the well, the crop mix adjusts and the trade pattern adjusts with it, without any need for the trade system to know what a virtual water flow is. Every jurisdiction that has meaningfully reduced groundwater depletion did so by constraining abstraction at home. None did it by regulating what left the port.
The Substitutes That Do Not Substitute
Desalination and drip irrigation are real technologies with the same disqualifying limit
Desalinated water costs between 0.50 and 2 euros per cubic metre delivered for irrigation, against 0.05 to 0.35 euros for conventional surface irrigation water [20]. ✓ Established Drip irrigation reliably raises the value produced per unit of water and, at basin scale, frequently raises consumption rather than lowering it [21]. ⚖ Contested Both technologies work. Neither substitutes for a cap.
Desalination is no longer exotic. More than 22,000 plants operate worldwide with a combined capacity of around 135 million cubic metres a day, and the best seawater reverse osmosis installations have pushed specific energy consumption to roughly 1.8 kilowatt-hours per cubic metre, close to double the thermodynamic minimum [20]. ✓ Established Large plants now deliver drinking water at 0.40 to 0.80 dollars per cubic metre, and Dubai has contracted supply at about 0.31 dollars [20]. For municipal supply in a coastal city with cheap electricity, desalination is a solved problem. For agriculture it is not, and the reason is arithmetic rather than engineering.
The European Commission puts the delivered cost of desalinated irrigation water at between 0.50 and 2 euros per cubic metre, against 0.05 to 0.35 euros for conventional surface water and 0.15 to 0.60 euros for reclaimed wastewater [20]. ✓ Established A crop consuming several thousand cubic metres per hectare per season cannot absorb a tenfold increase in water cost unless its output price is extraordinary. Greenhouse tomatoes and soft fruit can. Alfalfa, rice, cotton, maize and wheat, the crops responsible for the overwhelming majority of the water being depleted, cannot at any plausible price. Desalination therefore relieves cities and coastal high-value horticulture. It does not relieve the aquifers, because the aquifers are being emptied by the crops it cannot serve.
Desalination produces water at sea level. The aquifers being depleted are frequently hundreds of kilometres inland and hundreds of metres higher, including the Ogallala, the North China Plain, the Punjab and the Iranian plateau. Pumping and conveyance costs scale with distance and elevation and are often larger than the desalination cost itself. The technology is least available exactly where the deficit is largest.
Two further constraints deserve to be stated rather than assumed away. Global desalination capacity now produces more than 150 million cubic metres of brine a day, a discharge stream larger than the freshwater output and one whose disposal is manageable at an open coast and problematic almost everywhere else [20]. The energy requirement, while much reduced, also remains material at scale, since replacing a meaningful fraction of the water currently drawn from the world's stressed aquifers would demand generating capacity comparable to that of a mid-sized industrial economy. Neither point disqualifies desalination. Both bound the fraction of the problem it can address.
The second candidate substitute fails in a more interesting way. Drip and micro-irrigation reduce the volume applied to a field, and the intuition that this saves water at basin scale is wrong often enough that hydrologists have given the failure a name. In a flood-irrigated system a substantial share of the water applied returns to the aquifer or the river as recharge and is recovered downstream. Efficient irrigation removes that return flow, so withdrawals fall while consumptive use, meaning the water actually removed from the basin, stays flat or rises [21]. ◈ Strong Evidence Farmers then deploy the water freed on paper to expand the irrigated area or to switch to thirstier, higher-value crops, and the basin ends up drier.
The International Water Management Institute's synthesis of the evidence concluded that higher irrigation efficiency rarely reduces water consumption, because reduced return flows and expanded irrigated area offset the field-level saving [21]. ◈ Strong Evidence Israel's transition to drip irrigation raised yields per unit of water without reducing total agricultural consumption, as farmers continued to use their full allocations more productively [21]. The policy implication is narrow and firm. Efficiency subsidies delivered without a binding cap on abstraction subsidise expansion, not conservation.
This matters because irrigation modernisation is the single most popular water policy in the world. It is capital expenditure rather than restriction, it is attractive to every farm constituency, and it produces a measurable field-level saving that can be reported to a parliament. It is also, absent a cap, the most reliable way to accelerate depletion while appearing to address it. The remedy is not to stop funding drip systems. It is to attach the subsidy to a reduction in the licensed abstraction volume, so that the field-level gain is realised as basin-level saving rather than as expansion.
Taken together, the two substitutes define the boundary of the technical solution space. Desalination can supply coastal cities and a narrow band of high-value crops, and can therefore free some agricultural water by removing municipal demand from the same source. Efficiency can raise output per cubic metre substantially, which makes a given cut less painful to absorb. Neither creates water in a continental interior, and neither by itself reduces abstraction. Every documented case of a stabilised or recovering aquifer involves a quantity constraint. The technologies determine how expensive that constraint is to live with. They do not remove the need for it.
Who Actually Loses
The cut lands on wells, on ground level, and on the users with the least standing
More than 5,000 domestic wells in California's Central Valley are projected to run completely dry by 2040, with a further 4,000 partially dewatered, as the state's groundwater law takes full effect [12]. ◈ Strong Evidence Approximately 56,000 square kilometres of Iran, some 3.5% of the national territory, is subsiding because of aquifer depletion [18]. The cut is not distributed by hydrology. It is distributed by well depth, capital and legal standing.
A falling water table takes the shallowest well first, and the shallowest well almost always belongs to a household rather than to an agricultural operation. California recorded about 700 dry-well reports over the two years to February 2025, with more than 200 monitored wells standing at all-time lows, concentrated in the San Joaquin Valley [12]. ✓ Established Research on the state's groundwater plans projects more than 5,000 domestic wells completely dry and about 4,000 partially dewatered by 2040 [12]. These are households in rural, low-income communities that cannot spread the cost of a new borehole across a ratepayer base, and whose water was taken by a pump they have no standing to challenge.
The agricultural cut in the same basin is larger but slower and better compensated. Irrigated farmland in the San Joaquin Valley is expected to contract by up to 900,000 acres, about 20% of the valley's irrigated land, as the Sustainable Groundwater Management Act is implemented, with a plausible range of 500,000 to 750,000 acres fallowed by 2040 and an annual economic impact estimated at around 7 billion dollars [11]. ◈ Strong Evidence Roughly 10,000 acres came out of production in the Eastern Turlock subbasin in a single recent year [11]. The land retires. The question the law does not answer is which land, and who holds it.
Colorado has run that experiment for fifty years and published the results. In the Arkansas Valley more than 100,000 acres of irrigated farmland have been permanently dried after their water rights were sold to Front Range cities [16]. ✓ Established Crowley County's irrigated acreage fell from more than 50,000 acres in the 1970s to a few thousand by 2024, a decline of over 90% [16]. A 2026 Colorado State University estimate puts the annual economic loss at 1,400 to 1,600 dollars for every acre taken out of production [16]. What remains is not farmland held in reserve. It is blown sand that state highway crews remove from the roads with front-end loaders.
A water right sold to a city is sold once. The purchase price compensates the individual seller in full and compensates the community not at all. The farm supplier, the equipment dealer, the school district and the tax base receive nothing and lose everything the acre would have generated in perpetuity. Colorado has now legislated revegetation obligations on buyers, which addresses the dust. It does not address the arithmetic.
Land subsidence is where depletion becomes irreversible in the strict sense. Remote sensing analysis finds roughly 56,000 square kilometres of Iran subsiding, about 3.5% of the country, with 3,000 square kilometres sinking faster than 10 centimetres a year and some locations exceeding 35 centimetres [18]. ✓ Established Parts of Tehran drop by up to 30 centimetres annually [18]. Mexico City reaches 500 millimetres a year in places, progressively damaging the metro system and the water network itself [18]. Northern Jakarta has fallen more than four metres since the 1970s [18]. Compaction destroys aquifer storage capacity permanently, so the basin loses not only its water but its ability ever to hold that much again.
Iran shows what happens when the losses compound without an allocation mechanism. The country extracts roughly 63.8 billion cubic metres of groundwater a year against natural replenishment of about 45 billion [19]. ✓ Established By late 2025 the reservoirs supplying Tehran had fallen to 12% of capacity and those serving Mashhad to around 3%, and the president raised the possibility of rationing followed by partial evacuation of the capital [19]. Water shortages and the accompanying power cuts drew demonstrations that began among students and spread to truck drivers, bakers, farmers and pensioners [19]. ⚖ Contested The proximate trigger was drought. The structural cause was four decades of subsidised extraction with no binding cap.
The pattern across these cases is consistent enough to state as a rule. The first loss falls on domestic wells, which are shallow and unrepresented. The second falls on the smallholder, who cannot finance a deeper bore. The third falls on the rural community whose water rights were sold individually and whose economic base disappears collectively. The fourth falls on the physical basin itself through compaction, at which point the loss becomes permanent. Large irrigators and municipal utilities are affected last and compensated best. No basin in the world has run this sequence in the opposite direction.
The Instruments Being Tried
Markets, caps, buybacks and transfers, ranked by what the evidence shows each delivers
The southern Murray-Darling Basin carries about 31.9 billion Australian dollars of water entitlements on issue, with 771 million dollars of turnover in 2024-25 [17]. ✓ Established A Kansas conservation district cut water use by more than 20% and halved its depletion rate without reducing farm income [15]. ◈ Strong Evidence The instruments differ sharply in what they can be shown to have achieved.
Four families of instrument are in use, and they are not interchangeable. Markets reallocate a fixed volume between users. Caps and allocations fix the volume. Buybacks retire volume permanently by purchasing it. Transfers import volume from another basin. Only the last changes the physical quantity available, and only the middle two change the total abstracted. A market without a cap reallocates a volume that is still being drawn down, which is why sequencing matters more than instrument choice. The cap has to come first, or the market simply prices a deficit.
Australia operates the most developed water market in the world and illustrates both properties. The southern Murray-Darling Basin's major entitlements were valued at about 31.9 billion Australian dollars in 2024-25, up 3%, with turnover of roughly 771 million dollars, a 28% increase on the prior year [17]. ✓ Established The market moves water to its highest-value use within a season at a speed no administrative allocation can match. It also has no capacity whatsoever to reduce total extraction, which is why the Commonwealth has had to buy entitlements back, taking 51.2 gigalitres through open tender and 69.6 gigalitres from portfolio purchases by September 2025, with the programme limit raised to 300 gigalitres that November [25].
| Risk | Severity | Assessment |
|---|---|---|
| Domestic wells fail before agricultural cuts bind | The shallowest wells fail first and their owners have no standing in the allocation process. California projects more than 5,000 dry domestic wells by 2040 under an active groundwater law [12]. | |
| Aquifer storage destroyed by compaction | Subsidence removes storage capacity permanently. Around 56,000 square kilometres of Iran is subsiding and parts of Tehran drop up to 30 centimetres a year [18]. | |
| Efficiency subsidies that raise consumption | Irrigation modernisation without a binding abstraction cap raises basin-scale consumptive use, converting a conservation budget into an expansion subsidy [21]. | |
| Rural economic base collapses with the water sale | Individually rational entitlement sales aggregate into community failure. Crowley County lost more than 90% of its irrigated acreage and the businesses that depended on it [16]. | |
| Reliance on transfers most basins cannot finance | The North China Plain recovery required the largest inter-basin transfer ever built, and roughly half the lost storage remains unreplaced [10] [27]. |
The quantity instruments carry the strongest evidence base and the least glamour. Kansas created Local Enhanced Management Areas that allow irrigators in a district to impose a binding allocation on themselves by vote. In Sheridan County District 6, farmers agreed to a five-year allocation cutting water use by about 20% [15]. ✓ Established Water level declines slowed from an average of 1.5 feet a year between 2008 and 2013 to 0.68 feet between 2013 and 2017, more than halving the depletion rate [15]. Net farm income did not fall. Producers raised sorghum output by 335% and cut maize by 23%, and better scheduling, soil moisture monitoring and seeding rates absorbed most of the reduction [15].
The Kansas result is the most encouraging finding in this file, and it needs to be stated with its limits. It is a self-imposed cap in a district with relatively homogeneous producers, a shared understanding of the resource, and an initial level of use inefficient enough that a 20% reduction could be absorbed without threatening farm viability. It does not establish that a 50% reduction is painless, and it does not transfer automatically to a basin with millions of unmetered wells and no district structure. What it does establish is that the first tranche of reallocation costs far less than the political debate assumes, provided the constraint is credible, quantified and enforced against everyone at once.
The Case For Managed Reallocation
Water leaves agriculture in every sustained scarcity episode. The only variable under policy control is whether the process is governed, compensated and sequenced.
Agriculture takes about 72% of withdrawals and generates a fraction of the output per cubic metre available in industry and services [3].
The World Bank found that moving 25% of water to higher-value uses eliminates most of the projected GDP damage from scarcity [5].
Sheridan 6 cut water use by more than 20% and halved its depletion rate without reducing net farm income [15].
The Case Against Reallocation by Market
Value per cubic metre ranks a household well far below an export orchard. Distribution is not a rounding error in this calculation.
Crowley County lost more than 90% of its irrigated acreage through voluntary transactions that compensated sellers and no one else [16].
Under Chile's 1981 code roughly 1% of users came to hold about 80% of water rights, and the 2022 reform could not redistribute them [26].
A basin that reallocates to the highest bidder can maximise measured output while converting a water risk into an import dependency.
Colorado needed legislation in 2026 to compel buyers merely to revegetate the land they had dried decades earlier [16].
Transfers occupy a category of their own because they change the volume rather than its owner. China's South-to-North diversion had delivered 81.33 billion cubic metres by mid-2025 to 185 million people in 45 cities [27], and the North China Plain water table has risen at about 0.7 metres a year since 2020, passing 2005 levels by 2024 [10]. ◈ Strong Evidence It is a genuine reversal at continental scale. It also required a wetter basin within reach, a construction budget in the tens of billions of dollars, mass resettlement along the route, and enforcement capacity across multiple provinces. Roughly half the storage lost between 2005 and 2019 remains unreplaced [10].
The Colorado River is where all four instruments will be tested against one another in public. The agreements governing Lake Powell and Lake Mead expire at the end of 2026, and the Interior Department has issued operating guidelines for 2027 and 2028 alongside a ten-year decision framework, with average reductions of roughly 1.5 million acre-feet contemplated under most conditions [24]. ✓ Established Agriculture accounts for about 80% of consumptive use in the basin [24]. Arithmetic of that shape has only one solution set. The negotiation is not about whether farming loses water. It is about which farms, on what schedule, at whose expense, and with what compensation.
What the Evidence Tells Us
Reallocation is happening either way — the open question is whether it is governed
Groundwater declines have reversed in 16% of the aquifer systems with long records, and those reversals follow policy changes, managed recharge and surface water substitution rather than favourable weather [1]. ✓ Established The distinction that matters is not between basins that must reallocate and basins that need not. It is between reallocation that is decided and reallocation that is merely suffered.
Six findings survive close reading of the evidence. First, the deficit is real, measured and widespread, with 71% of monitored aquifer systems falling and declines accelerating in 30% of regional aquifers over four decades [1]. ✓ Established Second, it is not principally a shortage of technology or of hydrological knowledge. It is a shortage of enforceable limits on abstraction in systems whose pricing rules reward speed. Third, it is reversible, and the reversal is documented in 16% of aquifers with long records and demonstrated at scale in the North China Plain [1] [10].
Fourth, the cut lands in a fixed order, moving from domestic wells to smallholders to the rural economic base and finally to the physical storage capacity of the basin through compaction. Fifth, the technologies most often proposed as alternatives do not remove the need for a cap, since desalination is priced out of the crops that are draining the aquifers [20] and irrigation efficiency without a binding limit reliably increases basin-scale consumption [21]. ◈ Strong Evidence Sixth, and least discussed, the first tranche of reduction is far cheaper than the political argument suggests, as Sheridan 6 cut use by more than 20% and halved its depletion rate without reducing net farm income [15].
Set against those findings, the standard framing of water scarcity as an approaching conflict between states is not so much wrong as mistimed. Interstate water disputes are real and will intensify. But the transfer that will determine most people's exposure over the next twenty years is happening inside borders, between sectors, through instruments as unremarkable as an electricity tariff, a procurement price, a state land lease and a well permit. It is not being negotiated at a summit. In most basins it is not being negotiated at all. It is the residual of decisions taken for other reasons decades ago.
No policy can prevent water leaving agriculture in a depleting basin. Policy determines the order in which users are cut, the speed at which the cut arrives, the compensation attached to it, and whether the physical storage of the aquifer survives the transition. Those four variables account for almost the entire difference between the Kansas outcome and the Iranian one. The hydrology of the two cases is not that different.
That reframing changes what counts as a water policy. Metering and a licensed abstraction volume are water policy. So is the electricity tariff for agricultural pumping, the procurement price for rice, the terms of a state land lease, the definition of a beneficial use in a hundred-year-old water code, and the rule determining whether a city that buys a farm's water must also restore the land it dried. Ministries of water control almost none of these instruments. That is a substantial part of why the deficit persists in jurisdictions that have understood it for decades.
This study shows that humans can turn things around with deliberate, concentrated efforts. Groundwater depletion is not inevitable.
— Scott Jasechko, Bren School of Environmental Science and Management, University of California Santa Barbara, January 2024The uncomfortable corollary is that the countries with the deepest deficits are frequently those least able to run a compensated transition. Punjab's extraction rate has improved from 163.8% to 152.22% over four years through canal expansion and crop change [8], which is real progress and still leaves the state taking half again as much as it recharges. Iran, at roughly 63.8 against 45 billion cubic metres, has no comparable slack and no fiscal room to buy entitlements back [19]. ⚖ Contested Where a state cannot compensate the losers, reallocation still occurs. It occurs through well failure, migration and protest rather than through a purchase agreement.
The final observation is the one the data supports most strongly and the discourse accommodates least. There is no version of the next two decades in which agriculture retains its current share of water in the depleting basins. The figure of 72% is not a floor [3]. It is the starting point of a contraction already under way in Kansas, California, the Punjab, La Paz County and the Arkansas Valley, and it will reach every basin where extraction exceeds recharge. The reports that matter will not be about whether the reallocation happened. They will be about who was in the room when it was decided, and whether anyone was required to pay for what they took.