Water Infrastructure Megaprojects

There are three things a state can do about water at national scale: make it, move it, or store it. Each is a distinct megaproject family with a distinct binding constraint, and the constraint is almost never the one the engineering literature optimises. Established Making it — desalination — is a purchase of electricity. Moving it is a purchase of sovereign consent plus lift energy. Storing it is a question of where the water goes when it is not wanted, and the cheapest answer is underground, which turns a civil-engineering problem into a property-rights problem.

Established The measured spine of this brief is a 2026 PRISMA systematic review in Processes synthesising 45 studies published 2015–2026, retrieved in full; the GRACE and GRACE-FO satellite gravimetry record for groundwater; and the operational literature of China's South-to-North Water Diversion, which is the only inter-basin transfer of continental ambition that actually exists. The single most useful sentence in that record is the review's own: the levelized cost of desalinated water remains high “due to the high energy consumption”. Almost every desalination figure in public circulation has been separated from that clause, and a cost without its energy assumption is not a cost.

Frontier This brief was written under a research pass whose retrieval was severely constrained: agency and multilateral sources — the World Bank, USGS, FAO, IRENA, the National Academies, the Bureau of Reclamation — were unreachable, as were the Copernicus journals where much of the hydrology is published. Several figures the commission asked for therefore do not appear: the numeric thermodynamic minimum for seawater separation, South-to-North cost and displacement, the California State Water Project and Central Arizona Project accounts, and global managed-aquifer-recharge volumes. Each is named below with the instrument that would close it, and no number is printed in its place. The central argument does not need those constants. It needs the structure, and the structure is verified.

Established Membrane desalination is a 3–5 kWh/m³ process, and everything about its economics follows from that number and the price of a kilowatt-hour. The 2026 Processes review gives reverse osmosis a specific energy consumption of 3–5 kWh/m³, with a stated average of 3 to 4 kWh/m³ for RO specifically. Established Thermal technologies reach 15 kWh/m³; multi-stage flash is given as consuming up to 12 kWh/m³. Established The levelized cost of desalinated water is reported as USD 0.5–2.0/m³, with an alternative band of USD 0.6–1.5/m³ in the same paper. Established The review states the causal chain explicitly: cost “remains high (USD 0.5–2.0/m³) due to the high energy consumption of up to 15 kWh/m³ in thermal processes.” Established Note what the two bands do together: a roughly fivefold spread in energy intensity tracks a roughly fourfold spread in cost. That is not a coincidence, it is the dependency, and it is stated inside the source rather than inferred by this brief. Established The same review reports desalination serving more than 300 million people by 2025.

Established The thermodynamic floor is real, old, and settled in structure — and this brief will not print its numerical value, because that value could not be verified in this pass. The second-law literature establishes that the minimum separation work rises with feed salinity and with recovery ratio, and that it remains relatively stable up to roughly 80% recovery before increasing sharply (Cerci, Cengel & Wood 1999, abstract read). Established The modern framing is the same: exergy-based analysis and minimum separation work “define the limits of desalination” (Taktour et al. 2026). Frontier The specific least-work figure for standard seawater at 50% recovery — and therefore the multiple by which a best-in-class plant exceeds it — is unverified here. The instruments that would close it are named and were not obtained: Gordon & Hui (2016) and Qian et al. (2013), both in Desalination, both confirmed to exist and neither read. Established What survives without the constant is the shape, and the shape carries the argument. The thermodynamic penalty hides in the recovery ratio, not in the membrane. High-recovery designs are attractive precisely because they shrink brine volume, and they buy that reduction at a steeply nonlinear energy price above about 80%. Frontier The brine problem and the energy problem are therefore coupled with opposite signs: pushing on one worsens the other, and this brief could find no published joint optimisation of the two.

Established Brine is 1.5 litres for every litre of product, and it is a chemistry problem as much as a salinity one. The review gives the ratio as 1.5 L of saline solution for every litre of freshwater. Established Brine salinity across the 45-study synthesis runs 40–75 g/L, with maxima above 75 g/L recorded in the Arabian Gulf, and measured plume salinities of 45–70 g/L. Established Contaminant loads exceed drinking-water guidance by multiples: boron at 1.8 mg/L against a WHO limit of 0.5 mg/L, a factor of 3.6; bromate at 25 µg/L against a WHO limit of 10 µg/L, a factor of 2.5. Frontier Boron matters twice over, because it is phytotoxic well below the concentrations that trouble human drinking water, which makes it a specific and under-discussed obstacle to irrigating with desalinated water.

Established The most decision-relevant numbers in this whole subject are two salinity deltas and a dilution ratio. Posidonia seagrass is damaged where salinity exceeds +1 g/L above ambient; benthic communities show a significant decrease above +2 g/L above ambient; and the dilution required to stay under those thresholds is 20:1 to 50:1. Established Plume geometry is current-limited rather than volume-limited: hypersaline plumes extend 100–300 m in slow currents but under 50 m on the Chilean coast where currents are strong — a sixfold difference driven by hydrodynamics, not by plant design. Frontier The operational conclusion is that brine impact is a near-field siting question, and that the scarce input is not capital but suitable coastline. Frontier One caveat is load-bearing and this brief states it rather than burying it: all of the threshold and plume figures above come from a single synthesis. Primary Mediterranean Posidonia monitoring was not obtained in this pass, and a threshold quoted from a review is a claim about the review.

Established Desalination is not globally distributed; it is a regional industry with a handful of very exposed seas. The review reports regional brine discharge of more than 35 Mm³/day in the Arabian Gulf, more than 55 Mm³/day across MENA by 2025, 2–3 Mm³/day for Algeria and 1–2 Mm³/day for Morocco. Established Coastal salinity in the Arabian Gulf has risen 2–3 g/L. Frontier Read against the +1 and +2 g/L ecological thresholds, that is the near-field effect accumulating to basin scale in the world's most desalination-dense and most weakly flushed water body. Speculative If the thresholds transfer, the Gulf has already crossed the ecological carrying capacity that the same review's own numbers imply — and the optimistic case requires Gulf ecosystems to be more salinity-tolerant than Mediterranean seagrass systems, which is untested in the retrieved literature. Frontier The global brine total reported in the same review is subject to a suspected misattribution serious enough to have its own entry in section 14; it is not reproduced here.

Established Installed capacity and delivered production differ by roughly an order of magnitude, and this brief reports production because that is what it could verify. Global desalinated water production, as represented in global hydrological models, was 3.7 km³/yr in 2005 — about 10 Mm³/day — projected to 5.1–8.3 km³/yr for 2011–2040 and 23.4–52.7 km³/yr for 2041–2070 (Hanasaki et al. 2016). Frontier Global installed capacity, plant count and country shares are not stated in this brief. The field's inventory of record is Jones, Qadir, van Vliet, Smakhtin & Kang (2019) in Science of the Total Environment, whose existence and standing are confirmed — OpenAlex records 1,935 citations — and whose text could not be obtained, its abstract absent from the metadata record. Established Desalination at very large unit scale has its own diffusion literature treating it as an adoption and policy phenomenon rather than a technological one (Feitelson & Jones 2014), which is the right frame: the plants are not the hard part.

Established Satellite gravimetry has made groundwater depletion measurable, and the honest version of those measurements carries error bars that popular reporting drops. For Turkiye over April 2002 to March 2026, terrestrial water storage anomaly is −0.68 ± 0.14 cm/yr, groundwater storage anomaly −0.57 ± 0.16 cm/yr, and volumetric groundwater loss 4.52 ± 0.93 km³/yr across 789,748 km², for a cumulative net loss of about 108 km³. Established That is roughly ±21% relative uncertainty on a national depletion rate, and it is the honest precision of the instrument. Established Over the GRACE era 2003–2016, the Northwestern Sahara Aquifer System lost above 30 km³ and the Nubian Sandstone Aquifer System around 50 km³ — and the attribution differs, the first to “both climate…and water abstraction” and the second to “only…water abstraction”. Established That distinction is the difference between overdrawing a rechargeable resource and mining a fossil one. Nubian is being mined. Established The instrument is not uniformly a depletion story: Papua Island over 2005–2024 shows TWS at +0.43 cm/yr and groundwater storage at +0.38 cm/yr.

Established GRACE measures total mass; the groundwater compartment is a modelled partition, and the ground-truth correlation is weaker than the confidence of most claims made from it. Over the High Plains Aquifer (450,000 km²), seasonal GRACE TWS against summed storage components gives R = 0.82, but GRACE-derived against measured groundwater storage gives R = 0.58. Frontier The resolution problem is being attacked from both ends: machine-learning downscaling from 0.5° to 0.1° over Egypt reaches R up to 0.91 and RMSE up to 1.43 and resolves accelerating depletion in the west and centre of the country, while Level-1B regression is producing spatially refined global trend fields explicitly aimed at “identification of regions where human land and water use alter local freshwater availability”. Established Groundwater is now a first-class term in global hydrology rather than a residual: adding an explicit groundwater scheme to a global river-routing model improves simulated discharge at about 70% of gauging stations and brings simulated storage into better agreement with GRACE.

Established Long-distance transfer at continental ambition has exactly one existence proof, and its published record is operational rather than financial. China's South-to-North Water Diversion meets an ecological replenishment demand for northern rivers and lakes of about 2.04 billion m³ per year. Established Middle Route real-time control holds canal water level within a 0.25 m target interval and meets user demand “five times more rapidly” than the baseline scheme. Established Winter is a first-order constraint, not a footnote: well-water pumping for ice melting adds 1.1 million m³ to total pumped volume even under optimised operation. Established Algal growth in the main canal concentrates “along curves”, which is a maintenance liability specific to very long open conveyance. Frontier The project's total cost, its design transfer capacity and the number of people displaced are not stated in this brief. The sources that hold them — the World Bank, the Chinese Ministry of Water Resources, and the Webber, Rogers and Crow-Miller academic literature — were unreachable in this pass. Frontier The same applies to the two American built cases the commission asked for: California's State Water Project and the Central Arizona Project appear here without figures, because GAO reporting on CAP repayment, Bureau of Reclamation project histories and the National Academies' Colorado River Basin assessment could not be retrieved. Naming the gap is more useful than filling it from memory.

Established The most uncomfortable result in the subject is an evaluation of transfer as built, and it runs against the case for building it. A peer-reviewed assessment of inter-basin transfer finds it “not only fails to improve water consumption…but also further restricts residents' water use due to the high water price caused by this project” (Sheng, Tang & Webber 2020). Established This is an ex-post evaluation of an operating scheme, not a modelling objection raised in advance. Frontier Transfers are also not hydrologically neutral: simulation finds climatic influence intensity “positively related to the quantity of water to be transferred”, which means a sufficiently large transfer is a regional climate intervention in its own right. Established And the reason continental schemes do not get built was named by the profession itself at the height of the debate: “Institutional constraints are strong and nationalistic reactions to proposals…have contributed to delays in development” (Laycock 1971, JAWRA). That is a contemporaneous diagnosis from the flagship American water resources journal, not a modern reinterpretation.

Frontier The live shift is from treating brine as a disposal cost to treating it as an ore body, and it has just become computationally routine. A 2026 published protocol implements an end-to-end, plant-resolved techno-economic model quantifying “the recoverable energy and mineral value of desalination reject brine”, integrating five energy-recovery pathways and five mineral-recovery pathways across 300 plants with verification layers, and running in 1–2 minutes. Frontier The mainstream review literature now carries the same proposal, closing with “diffuser optimization, integration of renewable energies, and brine recovery through the extraction of strategic minerals”. Speculative What is missing is the answer: no verified figure exists in this pass for the fraction of brine mineral value recoverable at positive margin, and the whole proposition turns on whether further concentrating an already-concentrated stream pays for itself.

Frontier Outfall engineering, not membrane chemistry, is where the measurable environmental lever sits. Optimisation of discharge design finds that pre-dilution with treated wastewater reduces cost while pre-concentration increases it — a clean statement that the cheapest brine mitigation available is co-location with a municipal wastewater stream. Frontier Frontier Numerical assessment of coastal brine discharge has been carried far enough to report salinity-rise footprints as areas in km², which is the right output format for a siting decision.

Frontier Pressure-retarded osmosis runs the salinity gradient backwards and has reached commercial pilot scale on precisely the two waste streams the outfall result identifies. A 460 m³/day system operates on “concentrated seawater and treated sewage water”. Speculative If it scaled, the two binding terms — energy cost and brine disposal — would partially cancel each other, because the disposal event becomes a generation event. Handwave The scale gap is the whole story and it is enormous: 460 m³/day against production plants operating at 100,000 to 1,000,000 m³/day, three to four orders of magnitude, with no published route across it.

Frontier Two genuinely different separation mechanisms are being pushed as alternatives to pressure-driven membranes. Battery-materials capacitive deionization is offered as a path to “higher capacity and lower energy consumption” than membrane separation — a charge-based rather than pressure-based mechanism, historically strong at brackish concentrations and weak at seawater salinity. Frontier Thermally enhanced reverse osmosis is a live route to reduced specific energy, published in Desalination in 2024; this brief confirms the paper exists and did not obtain its results. Frontier Frontier Forward osmosis has a comparative energy-modelling literature against RO, likewise confirmed and unread.

Frontier The field is moving from thermodynamic idealisation to fleet-level empiricism, which is a healthier place to be. Machine learning is being applied to predicting specific energy consumption across real seawater RO plants, and specific-energy decomposition for RO has its own methodological literature. Frontier Speculative Both are named instruments for the measurement that matters most and does not yet exist in public: the distribution of actual plant energy consumption against the theoretical floor. Until that distribution is published, claims that the fleet is near the limit are claims about the best plant, not the fleet.

Frontier Green hydrogen is about to become a major new desalination demand driver in exactly the water-scarce regions already desalinating. Life-cycle assessment of water sourcing for hydrogen production via seawater desalination, with and without brine treatment, across the UAE, Spain and Australia, finds region-specific sustainability outcomes. Speculative The interesting implication is a competition nobody has priced: hydrogen electrolysis and municipal supply bidding for the same desalinated cubic metre and the same cheap electron.

Frontier Brine valorisation has crossed from concept into laboratory results with numbers attached. Reject brine used to pretreat spruce sawdust into an activated-carbon substitute increases adsorption capacity four times against untreated material. Established The crispest negative result in the area is older and worth keeping: coupling desalination to solar saltworks for zero liquid discharge works chemically but is defeated because “truck transport costs are prohibitive”, which localises the entire idea to plants adjacent to existing saltworks. Frontier Product-water safety, including genotoxicity assessment of desalinated drinking water, is a downstream question receiving far less attention than intake and brine; the paper exists and this brief did not read it.

Established Here is the ordered chain from where desalination is now to desalinated water cheap enough to irrigate with, which is the target state that would change the world rather than a city. Irrigation water is bought at roughly one to two orders of magnitude below the verified USD 0.5–2.0/m³ band, so this is a genuine workback to something currently considered impossible.

Frontier Link 1 — establish the true floor and the fleet's distance from it. Measure least separation work in kWh/m³ at stated salinity, temperature and recovery, against the measured distribution of plant specific energy consumption. Established Scientific, but nearly closed: the thermodynamics is settled and only the fleet data is missing. Frontier Link 2 — drive specific energy down the remaining margin via energy-recovery devices, thermally enhanced RO and capacitive deionization. Industrial and incremental. This is the link the engineering community works on, and it is the link with the least remaining headroom.

Established Link 3 — collapse the energy price term. This binds. The measurement is a levelized-cost decomposition stating how many USD/m³ are attributable to electricity at a stated USD/MWh. The source that gives the cost band states energy as the cause in the same sentence, so this is not an inference. Frontier Link 4 — solve intermittency if the cheap energy is solar. The measurement is membrane life and specific energy as functions of duty cycle. RO prefers steady state, photovoltaics do not supply it, and cycling membranes is damaging. No verified load-following penalty was obtained in this pass.

Frontier Link 5 — solve coast-to-field conveyance and lift. This binds, and it is routinely left outside the boundary of quoted costs. The measurement is USD/m³ per 100 km of conveyance per 100 m of lift. Desalinated water arrives at sea level on a coastline; the aquifers it is supposed to substitute for are inland and elevated. Frontier Link 6 — dispose of agricultural-scale brine within ecological limits. The measurement is salinity delta at the plume edge against the +1 g/L and +2 g/L thresholds, and dilution achieved against the 20:1–50:1 requirement. Established At a 1.5:1 brine ratio, irrigation-scale volumes make this a supply constraint on suitable coastline, with strong currents and a co-located wastewater stream as the scarce inputs.

Frontier Link 7 — manage boron and soil salinity in irrigated ground, measured against the 0.5 mg/L guideline and against phytotoxicity thresholds that are lower still. Under-discussed and scientific. Established Link 8 — establish water rights and pricing that let agriculture actually buy the water. Plausibly the true binding link, and the one built transfer scheme that was evaluated ex-post failed its version of this test by raising the price to end users.

Established The adjudication: links 3, 5 and 8 bind — energy price, conveyance, and institutional pricing. Membrane science is link 2, and link 2 is nearly exhausted. A programme that buys membrane research to make water cheap is optimising the one term with almost no headroom left, and this misallocation is the single most consequential claim this brief makes.

Established Three of this subject's dependencies are on retrieval rather than on research, which is an unusual and slightly embarrassing thing to have to say about a mature field. The global inventory of desalination capacity and brine production exists, is cited nearly two thousand times, and was not obtainable here. Frontier Until it is read, the field's headline capacity numbers propagate through reviews without anyone checking them against the primary — which is exactly how the error documented in section 14 appears to have happened.

Frontier The numeric thermodynamic minimum is a dependency on two specific papers, not on new science. Gordon & Hui (2016) and Qian et al. (2013), both in Desalination, are the named instruments. Nobody needs to run an experiment; somebody needs to open the journal. Speculative The same is true of managed aquifer recharge volumes: the IAH Commission on Managed Aquifer Recharge global inventory and the published global MAR assessments exist and were rate-limited out of this pass, which is why no recharge volume appears anywhere in this brief.

Frontier The genuine research dependencies are four. First, the load-following penalty for photovoltaic-driven RO — membrane life against duty cycle — without which solar desalination at continental scale cannot be costed. Speculative Second, the joint optimisation of recovery ratio against brine ecological cost, which nobody appears to have performed despite both curves being published. Frontier Third, primary field measurement of brine effects rather than thresholds inherited from review to review. Fourth, a global site inventory identifying coastlines that combine strong currents with an existing treated-wastewater outfall, which is a desk study on existing data rather than a research programme.

Speculative And one dependency runs the other way, on a subject nobody has costed at all: pairing pressure-retarded osmosis energy recovery with brine-disposal siting. The pilot runs on brine plus treated sewage; the outfall optimisation says brine plus treated sewage is the cheapest disposal route. The combined system has not been analysed, which means the case for it is currently made only by juxtaposition — including here.

Established The first required experiment is not an experiment. It is a library retrieval, and it should be done before anything else in this brief is treated as settled. Obtain Jones et al. (2019) and read its two headline quantities. Speculative If its desalinated-water and brine figures stand in the ratio the 2026 review's own brine-to-product number implies, then a widely read systematic review has mislabelled one as the other, and every downstream citation of it is understating global brine by about 1.5 times.

Frontier Measure the fleet, not the champion. Publish the distribution of specific energy consumption across operating seawater RO plants, with feed salinity, recovery ratio and energy-recovery-device configuration attached to each point, and plot it against the least-work curve. Frontier The machine-learning fleet study is the instrument. The output that matters is the shape of the tail, not the minimum.

Frontier Run the duty-cycle experiment that solar desalination depends on. Operate identical RO trains at constant load and at photovoltaic-shaped load, and measure membrane life, cleaning frequency and specific energy as functions of cycling depth over at least one membrane replacement interval. Speculative This is a two-year test on existing equipment and it would settle whether solar desalination needs storage, overbuild, or neither.

Frontier Do the joint recovery-and-brine optimisation. For a defined coastline, compute total system cost including ecologically-constrained disposal as a function of recovery ratio, using the near-80% energy nonlinearity on one side and the +1 g/L and +2 g/L thresholds with the 20:1–50:1 dilution requirement on the other. Speculative The result is a recovery ratio that is optimal for the site rather than for the plant, and no such curve appears to have been published.

Speculative Build the siting map. Intersect coastal current regimes strong enough to hold plumes under 50 m with the locations of treated-wastewater outfalls of adequate volume, and with distance and lift to demand centres. This is a desk study on existing datasets, it is the obvious global optimum for where desalination should go, and as far as this brief can establish nobody has produced it.

Frontier Two field programmes are needed to replace inherited thresholds with measurements: longitudinal Posidonia monitoring at operating Mediterranean outfalls, and a Gulf-specific salinity-tolerance study. Speculative The entire optimistic case for continued Gulf expansion rests on an untested assumption that Gulf benthic communities tolerate more than Mediterranean ones.

Established The plant itself is a solved artefact and the engineering difficulty has migrated outward to the interfaces. Energy-recovery devices — pressure exchangers returning the energy in the reject stream to the feed — are what took reverse osmosis to 3–5 kWh/m³, and their accounting has a dedicated thermoeconomic literature, confirmed to exist and not read here. Frontier Diffuser design at the outfall is the named mitigation lever in the review literature, and the optimisation result is unambiguous about direction: dilute before discharge, do not concentrate.

Established Very long conveyance is an active-control problem, not a passive one. The South-to-North Middle Route holds level within 0.25 m under real-time control and responds to demand five times faster than the baseline scheme, which is a statement about instrumentation and control theory rather than about concrete. Established Cold-climate operation adds a first-order term: ice-melting well-water pumping raises pumped volume by 1.1 million m³ even optimised, which is the constraint every high-latitude transfer proposal inherits and almost none of the classical proposals costed. Established Established Biofouling has a geometry: algal growth concentrates along canal curves, so maintenance is not uniform along the route.

Frontier Subsurface storage is the cheapest storage available and its engineering is mostly about the well and the water, not the reservoir. Aquifer storage has near-zero evaporative loss and near-zero capital cost per cubic metre relative to surface reservoirs, which makes it the natural buffer for both intermittent solar desalination and seasonal transfer. Speculative Speculative The engineering conditions are recoverable fraction, clogging management at the injection interface, and geochemical compatibility between injected and native water. Frontier No recharge volumes appear in this brief because the global inventories were unreachable, and a brief that quoted them from memory would be worth less than one that says so.

Frontier Desalination is being pulled into the energy system rather than merely supplied by it. Pressure-retarded osmosis treats the salinity gradient as a generator; capacitive deionization borrows battery materials; and real-options analysis is being used to value photovoltaic desalination as an option under price uncertainty rather than as a discounted cash flow. Speculative Each is a case of the water industry adopting the energy sector's physics and its finance at the same time, which is what a derived commodity looks like from the inside. See Energy Storage Revolutions and Planetary-Scale Energy Systems for the supply side this subject is a derivative of.

Frontier Hydrogen is the adjacent technology most likely to change desalination's demand curve within a decade. Electrolysis needs high-purity water and the best solar and wind resources sit in exactly the water-scarce regions that already desalinate, so the two industries will compete for the same electron and the same cubic metre. Speculative Nobody has priced that competition. See Hydrogen Economies.

Frontier Brine mineral recovery makes this subject adjacent to the resource economy rather than to utilities. A reject stream containing recoverable magnesium, bromine and lithium is a mining feedstock that has already been pumped and pressurised, and the 300-plant techno-economic protocol is the instrument that will decide whether that is a real industry or an argument. Speculative On present evidence it is an argument with a good model attached. See Resource Economies and Industrial Ecology.

Established Two adjacencies are institutional rather than technical. Conveyance belongs to Continental Irrigation Systems, which carries the transfer schemes and the reasons almost none of them were built; and the delivery record of very large public works belongs to Megaproject Governance and Historical Megaprojects. Frontier Downstream, cheap desalinated water is the input assumption behind Desert Greening and much of Future Agriculture, which is why its price deserves more scrutiny than it usually gets.

Established The institutional requirement that binds hardest is the least glamorous: an enforceable rule about what may be discharged, at what salinity delta, at what distance from what. The ecological thresholds are published and specific — +1 g/L for Posidonia, +2 g/L for benthic communities, 20:1 to 50:1 dilution. Frontier A regulator that writes those numbers into a discharge consent turns siting into an engineering optimisation with a defined objective. A regulator that does not turns the near-field into an externality, and the Arabian Gulf's measured 2–3 g/L coastal salinity rise is what that looks like at basin scale after several decades.

Frontier The second requirement is an accounting system for groundwater, and it is prior to any allocation rule. Nobody can allocate a stock nobody measures. Satellite gravimetry now supplies national-scale numbers with stated uncertainty — 4.52 ± 0.93 km³/yr for one country over a 789,748 km² domain — but the correlation between satellite-derived and well-measured groundwater is R = 0.58, which means the satellite disciplines a national budget and cannot adjudicate a farm-level dispute. Speculative A workable regime needs both: gravimetry as the audit, metered wells as the ledger, and a legal definition of a right held over a subsurface volume rather than over a pumping act.

Established The third requirement is a pricing regime, and the evidence here is genuinely against the intuitive case. The one large transfer scheme that has been evaluated ex-post was found not merely to fail to improve consumption but to restrict residents' use through the price the project itself created. Established And the diagnosis of why continental transfers do not get built has been stable for over half a century: institutional constraints and nationalist reaction, published in 1971, in the profession's own flagship journal, at the height of the debate. Frontier A brief that treats the institutional obstacle as softer than the engineering one has the ordering exactly backwards.

Speculative The fourth requirement is a binding cap on total withdrawal, established before new supply arrives. Every historical scheme assumed transferred or manufactured water would substitute for over-abstraction. The recurring outturn is that new supply is absorbed by expanded demand, which is the Jevons pattern applied to water. Frontier Without an enforced ceiling on abstraction in the receiving basin, a desalination plant and a transfer canal both function as subsidies to expansion rather than as remedies for depletion — and the depletion continues underneath the new supply.

Established The distributional finding in this subject is that a water megaproject can make water more expensive for the people it was built to serve. That is not a hypothetical: it is the measured outcome reported for inter-basin transfer, where the scheme “further restricts residents' water use due to the high water price caused by this project”. Frontier Large capital works are financed, and the financing appears in the tariff. Any brief that reports a megaproject's delivered volume without its delivered price has reported half of it.

Established Displacement does not end at commissioning. A 2024 study of a South-to-North resettlement village found six factors needing renovation first, among them winter insulation and water quality — a decade after a water-supply megaproject moved those households. Frontier The detail is worth holding onto precisely because it is small: the people relocated to build a water project had a water problem.

Frontier Brine distributes benefit and harm across different populations. Product water goes to a city; the plume stays in a bay, and the ecological thresholds that govern it are on the order of one gram per litre. Speculative In an enclosed sea shared by several states, one country's discharge is another country's ambient salinity, and there is no obvious instrument for that. The Arabian Gulf is the working example.

Established Aquifer mining is an intergenerational transfer with a measured magnitude. The Nubian Sandstone system lost about 50 km³ over 2003–2016, attributed to abstraction alone rather than to climate — a fossil stock being converted into current consumption. Frontier Two further ethical questions are open rather than answerable here: boron in irrigation water at concentrations phytotoxic below the human guideline, and the genotoxicity and health-risk profile of desalinated drinking water, which has a literature this brief could confirm exists and could not read.

Speculative If desalination is an energy price with a plant wrapped around it, then cheap firm electricity converts water from a geographic endowment into a manufactured commodity — and that is one of the larger changes a civilisation can undergo. Coastal deserts stop being marginal. Rivers stop being destinies. Frontier The transformation is real but bounded by geometry: manufactured water appears at sea level on a coastline, and the lift and conveyance to anywhere else is a separate megaproject that historically does not get built.

Frontier The strategic fact of the last half century is that continental transfer was not defeated, it was outflanked. Desalination made the coastal half of the demand disappear without requiring a treaty with a neighbour, leaving transfer schemes with only the hard inland residual and the same institutional blockers named in 1971. Speculative Nobody appears to have formally priced this outflanking, which is a striking hole given how much twentieth-century water planning it silently invalidated.

Frontier The countervailing civilizational risk is that manufactured supply licenses continued mining of the stock. Groundwater depletion is measurable at national scale and the attribution is sometimes unambiguous — abstraction alone, no climate term. Speculative A civilisation that can manufacture water at the coast and continues to mine fossil aquifers inland has bought itself an option, not a solution, and the option expires basin by basin.

Frontier The quieter civilizational variable is storage, because it decides how much of the above is usable. Surface reservoirs evaporate and occupy valleys; aquifers do neither, which makes subsurface storage the natural buffer for both intermittent solar desalination and seasonal transfer, and the reason the recharge question keeps reappearing under different names. Speculative A society that can store water underground and account for it honestly can run an intermittent supply system; one that cannot must either overbuild generation or keep the plants on firm power. Handwave That is a legal-instrument problem dressed as a hydrogeological one, and it is the least discussed load-bearing element in the whole subject.

What these horizons track is not membrane performance, which is nearly exhausted as a variable, but the price of firm electricity, the enforceability of discharge limits, and whether anyone ever builds coast-to-interior conveyance. Read them as conditional on the energy system rather than on the water industry: if power prices fall as far as the more aggressive projections in the energy briefs assume, the 25-year row arrives early and the 50-year row becomes reachable; if they do not, none of the rows below move at all, however good the membranes get.

• 10 yr: Frontier Fleet-level specific-energy distributions published, closing the gap between the champion plant and the industry. Brine mineral recovery either demonstrates positive margin at one commercial plant or quietly stops being discussed. Green hydrogen becomes a named line item in desalination demand forecasts in the Gulf, Spain and Australia. Discharge consents in at least one jurisdiction written against explicit salinity-delta thresholds rather than against dilution ratios.
• 25 yr: Speculative Global desalinated production reaching the upper part of the modelled 23.4–52.7 km³/yr band projected for 2041–2070, concentrated in the same regions and the same enclosed seas. Managed aquifer recharge established as the default buffer for intermittent solar desalination, if and only if the water-rights question over subsurface storage is settled. Pressure-retarded osmosis either crosses two orders of magnitude from its 460 m³/day pilot or is retired.
• 50 yr: Speculative Desalinated water at irrigation prices somewhere in the world, contingent on electricity at a price that makes the energy term small and on a conveyance corridor nobody has yet financed. Alternatively, and on current evidence more likely, agriculture reallocates rather than buys: the cheapest new water remains the water someone else stops using.
• 100 / 250+ yr: Handwave Continental solar desalination complexes as standing infrastructure, with brine chemistry treated as an ore stream and the disposal problem solved by consuming the salt rather than moving it. Atmospheric water harvesting at scale in arid interiors, which on the thermodynamic argument sketched in section 14 requires either a sorbent breakthrough nobody has demonstrated or energy so cheap that the comparison stops mattering. Both are coherent; neither is close.

• Depends on This subject sits downstream of the energy system almost entirely. Cheap firm power is what decides whether desalinated water is a municipal product or an agricultural one, which makes Planetary-Scale Energy Systems, Energy Storage Revolutions and the fission and fusion briefs the real upstream. It also depends on conveyance, which is a separate discipline with a separate failure history — see Continental Irrigation Systems — and on the delivery record of very large public works in Megaproject Governance.
• Requires (not on this map) Three conditions this subject waits on that no experiment will produce. Cheap firm electricity is the thesis: the levelized cost of desalinated water is dominated by the energy term, its own source says so in the sentence that gives the number, and no plausible membrane improvement moves the cost as far as a halving of the power price would. Enforceable brine discharge regulation is what converts published ecological thresholds — +1 g/L, +2 g/L, 20:1 to 50:1 dilution — into a siting constraint that engineers optimise against instead of an externality that accumulates, as it has in the Arabian Gulf. And aquifer withdrawal accounting and allocation is prior to everything about storage: managed recharge is the cheapest buffer available and is unusable at scale until a right can be held over a volume of stored subsurface water, with satellite gravimetry as the audit and metered wells as the ledger.
• Enables What it enables is more consequential than what it costs. Manufactured water at agricultural prices is the load-bearing assumption underneath Desert Greening and large parts of Future Agriculture, and it is the reason those briefs should be read with this one's price argument in hand. Coastal urban growth beyond local hydrology — see Sustainable Megacities — is already running on it. Electrolytic hydrogen in arid regions needs it as a precondition, not as a convenience.
• Adjacent Adjacent rather than dependent: Ocean Engineering for outfalls, intakes and the near-field hydrodynamics that decide brine impact; Resource Economies and Industrial Ecology for brine as a mineral feedstock rather than a waste stream; Arctic Engineering for the cold-climate conveyance problem that every high-latitude transfer proposal inherits and rarely costs; and Historical Megaprojects for the abandoned continental schemes whose failure mode — institutional rather than technical — is the one this subject keeps rediscovering. The adjacency worth watching is pressure-retarded osmosis, which sits on the boundary between this brief and the energy briefs and would, if it scaled, move brine disposal from the cost column to the revenue column.

Frontier Start with an apparent error in a 2026 systematic review, because it is the most concrete thing in this brief and it is unresolved. Belhadj, Rezzoug & Benmoussa (Processes 14(15):2391) report global brine production as 95 Mm³/day in 2015, rising to over 110 Mm³/day in 2025. Using only figures internal to that same paper, that is almost certainly the desalinated water figure rather than the brine figure. The review gives a brine-to-product ratio of 1.5 litres of brine per litre of freshwater; at that ratio a global brine total of 95 would imply freshwater production of about 63 Mm³/day. But the same paper reports MENA brine discharge alone exceeding 55 Mm³/day. MENA at 55 of a global 95 would make one region 58% of world brine, which is implausible and inconsistent with the paper's own treatment of MENA as one region among several. Read the other way — 95 as desalinated water, implying global brine of about 141 — MENA's 55 is about 39%, which matches the known regional concentration. Speculative This brief reports the arithmetic and stops there. The primary inventory that would settle it, Jones et al. (2019), was blocked, so this is a suspected misattribution and not a confirmed one. If the primary turns out to carry a water figure near 95 and a brine figure near 141, then a 2026 review has swapped the labels and every citation downstream of it understates global brine by about 1.5 times. Until someone opens that paper, the honest position is that the number is in doubt and should not be repeated in either direction.

Established “Desalination now costs fifty cents a cubic metre.” That figure is real and it is the bottom of a verified USD 0.5–2.0/m³ range, quoted by a source that names energy consumption as the cause in the same sentence. A desalination cost without four attached assumptions — the electricity price, the technology (3–5 kWh/m³ for RO against up to 15 for thermal), the feed salinity, and whether conveyance and brine disposal sit inside the boundary — is not a cost, it is a quotation. The rule this brief follows is that no desalination cost appears without its energy assumption attached, and the reader should apply the same rule to every other source.

Frontier “Reverse osmosis is already at the thermodynamic limit, so there is nothing left to get.” Half true, and used to shut down inquiry. What is verified is that minimum separation work stays relatively stable up to about 80% recovery and then rises sharply, which locates the thermodynamic penalty in the recovery ratio rather than in the membrane. High-recovery designs are pursued because they shrink brine volume, and they pay a steeply nonlinear energy price for it. Frontier So the two problems are coupled with opposite signs, and the interesting remaining engineering is the joint optimisation, which nobody has published.

Established “Brine is just salty water and the ocean is enormous.” The ocean is enormous; the plume is 100–300 m in slow currents and the ecological thresholds are +1 g/L and +2 g/L above ambient. Basin-scale dilution arithmetic is the wrong calculation because the impact is near-field and hydrodynamic, and a plume does not average itself over an ocean before it reaches a seagrass meadow. Established And where flushing is poor the near-field effect does accumulate to basin scale: coastal salinity in the Arabian Gulf has risen 2–3 g/L. Established “Brine is only a salinity problem” is separately wrong: boron runs at 3.6 times and bromate at 2.5 times the WHO guideline values.

Established “Global desalination is about ninety-five million cubic metres a day.” Three different quantities circulate under one heading and are constantly interchanged: installed nameplate capacity, actual delivered production, and brine output. Modelled global production was 3.7 km³/yr in 2005, which is about 10 Mm³/day — roughly an order of magnitude below the capacity figure usually quoted for the same era, because fleets run well below nameplate. Frontier Unit systems compound it, since km³/yr and Mm³/day differ by a factor of 365. Every desalination volume needs a label and a single unit system; this is the most likely place for a large arithmetic error to enter any treatment of the subject, including this one.

Established “Desalination is a drop-in substitute for depleted groundwater.” Desalinated water arrives at a coastline at sea level. The aquifers being depleted — Nubian at about 50 km³ over 2003–2016, attributed to abstraction alone — are frequently inland and elevated. The substitution requires the conveyance megaproject, and conveyance is the part that historically does not get built. Frontier “GRACE shows aquifer X losing Y cubic kilometres a year”, quoted with no error bar, is the matching failure on the measurement side: the published relative uncertainty on a national groundwater loss rate is about ±21%, and the correlation between satellite-derived and measured groundwater storage is R = 0.58. Established “Groundwater is universally in decline” is also not what the instrument says: Papua Island shows TWS at +0.43 cm/yr and groundwater storage at +0.38 cm/yr. The global picture is heterogeneous and selecting depleting basins overstates the aggregate.

Frontier “Zero liquid discharge solves brine.” It relocates the problem into solids handling and energy. The crispest retrieved result on a ZLD variant — coupling desalination with solar saltworks — found the binding constraint was that truck transport costs are prohibitive. Established The chemistry worked; the logistics did not. That is the characteristic shape of failure in this field, and it is worth noticing how rarely it is a chemistry failure.

Handwave Atmospheric water harvesting at continental scale. No source on AWH was retrieved in this pass and this brief prints no AWH number, which is itself the most useful thing it can say about the claims in circulation. The chain that would have to hold is stated so a reader can test any proposal against it: the minimum work to condense water from air rises steeply as relative humidity falls, so in humid coastal air AWH competes with desalination on energy, and in the arid interiors where it would actually be valuable it is far worse. Speculative The defensible niche is small-scale, off-grid and humid. Any continental-scale AWH claim should be met with two demands: the assumed relative humidity, and the sorbent regeneration energy in kWh/m³. A proposal that supplies neither is not making a physical claim.

Handwave Iceberg towing. Likewise, no source retrieved and no numbers printed. The structure of the test is simple and unforgiving: tow energy over thousands of kilometres, melt loss in transit through warm water, and a delivered product that is fresh water at a coastline — which is exactly what seawater RO already supplies at USD 0.5–2.0/m³. Speculative An iceberg scheme does not have to be feasible, it has to beat that band at the same delivery point, and its advocates rarely state the comparison in those terms.

Established The strongest position in this whole subject is that water scarcity is an allocation and pricing problem rather than a supply problem, and it has direct empirical support from the largest transfer ever built. The ex-post evaluation finds transfer “not only fails to improve water consumption…but also further restricts residents' water use due to the high water price caused by this project”. Frontier Agriculture is the overwhelming majority of freshwater withdrawals and the least price-exposed user, so the marginal cubic metre in almost every stressed basin is cheaper to reallocate than to manufacture or import. Speculative What would have to be true for the megaproject case to survive this: that the price effect is transitional rather than structural, or that the no-project counterfactual was worse in ways a consumption metric does not capture. Both are arguable and neither has been demonstrated. Frontier This brief's terminal position is a declared tie on a narrower point: manufactured supply is genuinely the right answer for coastal cities with cheap power and good flushing, and genuinely the wrong answer for inland irrigated agriculture, and most public argument fails by generalising one of those cases to the other.

Frontier One last correction, and it is about how citations break rather than about water. The research pass behind this brief recorded its own bibliographic instrument returning a result set about organ cryopreservation in response to a query on reverse-osmosis energy — a silent substitution with no error and no low-confidence signal, discarded on inspection. Three other independent passes in the same project caught comparable substitutions, including a publisher URL serving a different article than the digital object identifier requested. Speculative This matters here because the subjects in this brief are governed by assessment reports assembled at speed from machine-retrieved literature, and a silent substitution in that pipeline is undetectable downstream. The suspected 95 Mm³/day mislabelling at the top of this section is what that failure mode looks like once it has reached print.