Haute Lumière

Commerce · VII.06 · MMXXVI · daylight

La Bourse  /  Volume VII  /  Nº VII.06

Watershed as the Unit of Account

Volume VII — Planetary and Cosmic

---

THE PLATE

---

THE LETTER

You have probably been handed a sustainability report with a water number in it, and you have probably noticed that you could not check it. Not because anyone was dishonest. Because the number had no boundary. Litres withdrawn, by a company, in a year, across nineteen countries, is a sum of things that happened in places that have nothing to do with each other. It cannot be reconciled against anything, because there is nothing it has to add up to.

This chapter is about the one boundary in environmental economics that does not have that problem. A watershed — a catchment, a basin, whatever your country calls it — is the area of land from which water drains to a single point. Draw that line and something unusual becomes true: the accounts must close. Every drop that falls inside the line leaves as streamflow past the gauge, leaves as vapour, or is still there in soil, snow, aquifer or reservoir. There is no fourth destination. Water cannot be offshored, restated, or booked to a subsidiary in another jurisdiction.

That is a rarer property than it sounds, and it is the whole argument. An accounting boundary that closes gives you a residual, and a residual is what turns a report into a set of books. If your precipitation, your evapotranspiration, your outflow and your change in storage do not reconcile, the gap is not a rounding difference. It is either an error you have to find or a withdrawal somebody has not declared.

Volume VII opens on the largest frames — planetary boundaries as a balance sheet in VII.01, the bioregion as an economic unit in VII.05. This chapter takes the narrower and harder claim. Of all the boundaries a bioregion might be drawn on — watershed, ecotone, language, foodshed, soil order — exactly one of them comes with an accounting identity already attached. That one is worth using for the books, whatever you use for the culture.

What follows is four basins that keep real accounts, the arithmetic of a real payment-for-watershed-services scheme with all three of its terms, and the one term that kills most such schemes and almost nobody prints.

— The Editors

---

DISCOVERY

What is already working

Start with the fact that this is not a proposal. Basins are being governed by number, some of them for seventy years, and the record is better than the literature's tone suggests.

The Rhine, and a commission with no power to compel anyone. The International Commission for the Protection of the Rhine has coordinated Switzerland, France, Germany, Luxembourg and the Netherlands since 11 July 1950 and its legal basis — the Bern Convention of 1963 — renewed as the Convention on the Protection of the Rhine in 1999 — gives it almost no enforcement authority. It sets programmes, it measures, and it publishes. After the Sandoz warehouse fire at Schweizerhalle on 1 November 1986 sent firefighting water carrying up to 30 tonnes of pesticides into the river, killing organisms for hundreds of kilometres and halting drinking-water abstraction along some 900 kilometres to the Netherlands, the riparian states adopted the Rhine Action Programme in 1987 with a number and a date: halve the discharges of 40 named dangerous chemicals within ten years, and make the river clean enough for salmon to return.

Now read the scorecard the commission published about itself. Heavy metal discharges fell 72 to 95 percent by 1996 and phosphorus inputs 65 percent. Nitrogen fell 26 percent against a 50 percent target, and the ICPR's own activity report calls that "by far not achieved." By 1997 just 36 of 76 assessed substances met their targets and the rest — mercury, cadmium, copper, zinc, lindane, diuron, the PCB group — did not, and the commission printed the list. The salmon did return: considered lost in the Rhine in 1958 9,586 adults were recorded in the system between 1990 and 2020, with more than 28 percent of potential salmon habitat reconnected, and 71 fish species are present today.

That "by far not achieved" is the most valuable sentence in seventy years of basin governance, and it is why the rest of the scorecard is worth believing. A commission that can only count turned out to be enough, because the counting was shared, annual, public, and honest about its own misses.

And the Rhine has the only large watershed payment with a price printed in a treaty. Under the 1976 Chlorides Convention, France undertook to cut chloride discharges — largely from the Alsace potash mines — by at least 60 kilograms a second, with a first stage of 20 kg/s. The capital cost of 132 million French francs was apportioned in the treaty text: the Netherlands 34 percent, Germany 30, Switzerland 6. The abatement happened upstream and the downstream beneficiaries paid 70 percent of the capital, with the party furthest downstream paying the largest single share. It is the cleanest upstream–downstream bargain ever written down, and it has a second half worth as much as the first: a further 60 kg/s stage was rejected by the Rhine Ministers in 1988 as too costly. The deal held exactly as long as marginal abatement cost stayed under downstream willingness to pay, which is the arithmetic of the fifth section below, written as a treaty.

New York City, and the plant it did not build. Under the United States Surface Water Treatment Rule a large surface supply must be filtered unless it can demonstrate that its source water is protected well enough not to need it. In 1997 New York City signed a Memorandum of Agreement with New York State, the Environmental Protection Agency, environmental organisations and the watershed communities themselves, committing roughly $1.5 billion to protecting the Catskill and Delaware watersheds — land acquisition, septic replacement, farm planning, stream corridors — in exchange for a Filtration Avoidance Determination. The determination has been renewed ever since. The city has acquired or protected over 130,000 acres. Most of the world knows this story as a ratio; we are going to check the ratio in the next movement, and it holds up better than its critics expect.

The Murray–Darling, and the only basin on earth that bought water back at scale. Australia's Water Act 2007 and the Basin Plan of 2012 did something no other jurisdiction has managed: they set a legally binding cap on how much water may be taken from a basin, quantified the reduction required — 2,750 gigalitres a year of long-term average annual yield, later reduced by 70 under the Northern Basin Review and offset by 605 through supply measures, leaving a standing target of 2,075 — and then went into the market and bought entitlements from willing sellers at a published price.

And it very nearly finished. As at 30 June 2026 the recovery register shows 2,073.1 gigalitres a year contracted against that 2,075, with the department putting what remains at under one percent of the task. A basin spanning several states, with a century of entitlements already issued over it, has bought back more than two thousand gigalitres a year of its own diversions from willing sellers — and it can tell you to three decimal places which method delivered each part of it. Whatever else is contested, and much is, the volumes are public and the argument is conducted in units. That is more than any other large basin can say.

One honest note on those decimals, because the chapter is about books that close: subtracting the register from the target leaves 1.9 gigalitres a year, while the department states 8.6. Recovery targets are set catchment by catchment, so a surplus in one valley does not close a shortfall in another and the basin-level subtraction understates what is left. That is the likely reason and this chapter has not verified it — which is exactly what a residual is for: it is printed, not reconciled away.

And the accounting standard that already exists. Australia's Water Accounting Standards Board did the thing this chapter is arguing for and did it in the language of finance: general-purpose water accounting reports, with a statement of water assets and water liabilities, a statement of changes in them, and an assurance regime modelled on financial audit. The Bureau of Meteorology has published a National Water Account against it. The balance sheet is not a metaphor somebody in this book invented. It is a published standard with an auditor's opinion attached.

And the quiet one, which is the one to copy. New York's Watershed Agricultural Council is a farmer-led body that holds the relationship with the watershed's farms and delivers whole-farm plans to them. The city does not contract with four hundred farms. It contracts with one organisation that contracts with four hundred farms. That single structural choice is worth more than any hydrology in this chapter, and the arithmetic in the next movement will show exactly how much.

Four basins, four continents, one pattern: in every case somebody agreed what the units were before they argued about who got them. Ostrom's design principles, which govern Volume VI, put clearly defined boundaries first for exactly this reason. A watershed hands you the boundary for free. It is the only economic frontier drawn by gravity, and gravity does not lobby.

---

THE ARITHMETIC

What works, what does not, and where the line sits

First, the identity. This is the whole of the chapter in one line.

        P  =  ET  +  Q  +  ΔS  +  ε

  P    precipitation onto the basin           the only credit
  ET   evapotranspiration                     the loss to the sky
  Q    streamflow past the gauge              the transfer downstream
  ΔS   change in storage                      soil, snow, aquifer, reservoir
  ε    the residual                           what you have not found yet

That is double entry. Not by analogy — by structure. P is the credit, the rest are debits, and they must sum. A national economic account has no such constraint; you can revise GDP and nothing anywhere fails to balance. Here, if ε is not near zero, somebody is abstracting water nobody has recorded, or your gauge is wrong, or your storage estimate is. The residual is the instrument. It is a suspense account with hydrology in it, and like every suspense account it is where the interesting things are found.

Second, the worked basin. Take New York City's Catskill and Delaware watershed: 1,000,000 acres, which is 404,686 hectares or 4,047 square kilometres. The city draws about 900 million gallons a day from it — ninety percent of a supply the Department of Environmental Protection puts at more than a billion gallons a day. That draw, spread over the basin, is a depth of 307 millimetres a year.

Now compute the ET term rather than assuming it. Zhang, Dawes and Walker (2001) fitted a two-parameter relation to more than 250 catchments worldwide that gives mean annual evapotranspiration from rainfall, potential evapotranspiration, and one coefficient for vegetation — 2.0 for forest, 0.5 for grass. At 1,200 millimetres of rainfall and 700 of potential evapotranspiration:

  cover     ET (mm/yr)   Q (mm/yr)
  ---------------------------------
  forest        670         530
  pasture       516         684
  ---------------------------------
  difference   +154        −154

Here is the cut, and it is the one nobody puts in the brochure. The forested watershed yields 154 millimetres a year less water than the same land under pasture. Over 4,047 square kilometres that is 451.9 million gallons a day of water that does not arrive — and the city's entire draw is 900. Keeping the Catskills forested costs New York City about half as much water as New York City drinks.

That debit appears in no watershed-protection business case ever written, and it is not small, and it is not an argument against the programme. The city is buying quality with quantity, deliberately and correctly, because it has quantity to spare and no appetite for a filtration plant. The point is that only a closed account shows you the trade. Single-entry books show the credit and lose the debit, which is how a scheme in a basin with no spare quantity ends up paying people to deepen the scarcity it was convened to relieve. The global synthesis is unambiguous: Farley, Jobbágy and Jackson (2005) analysed 26 catchment data sets and 504 observations and found annual runoff fell by 44 percent (±3) where grassland was afforested and 31 percent (±2) where shrubland was — 75 percent (±10) with eucalypts. Trees are not a water supply. They are a water quality strategy with a water quantity price, and the price is printable.

Note what that does to the figure above. The Zhang model puts the forest's runoff 22.5 percent below pasture; the measured catchments put it at 44 percent. The two routes share no assumptions — one is a water-balance model fitted to rainfall and potential evapotranspiration, the other a synthesis of paired catchment experiments — and they agree in sign and in order of magnitude, with the model coming in smaller by a factor of 1.95. The cut above is the conservative end of the evidence, not the generous end.

Third, the famous ratio, checked. The ratio entered the literature through Chichilnisky and Heal in Nature in 1998 and has circulated ever since as roughly $1–1.5 billion of watershed protection against $6–8 billion for a filtration plant — almost always quoted without its denominator. So check it against the only plant New York actually built. The Croton Water Filtration Plant entered service in 2015 at a final cost of $3.2 billion against an original projection of $800 million — a 4.0× overrun — for 320 million gallons a day of capacity. That is $10.0 million of capital per million gallons a day, and it is not an estimate: it is a receipt. At that realised price, filtering the Catskill/Delaware supply would have cost $9.0 billion at 900 mgd and $12 billion at a 1,200 mgd design. The city's own later figure is "more than $10 billion", reached by a completely different route.

The famous ratio was not inflated. Checked against a receipt, it was low.

Annualise it. Nine billion dollars at a 4 percent real discount rate over 40 years carries a capital charge of $454.7 million a year; add the city's own stated operating figure of at least $100 million and the avoided cost is $554.7 million a year, or $1.69 per thousand gallons delivered. Against the programme, the ratio depends entirely on the denominator nobody states:

  $1.5bn annualised over 10 years   $150.0m/yr   $0.46/kgal    3.7 : 1
  $1.5bn annualised over 27 years   $55.6m/yr    $0.17/kgal   10.0 : 1

Say which one you mean. A single ratio with a hidden denominator is how a real finding becomes folklore.

Fourth, the honest negative, and it is severe. Can New York demonstrate that the land it protected caused the water quality it enjoys? No. Bosch and Hewlett's 1982 review of catchment experiments established the threshold the field still uses: a change in forest cover below about 20 percent of catchment area cannot normally be detected in streamflow at all. New York has protected 130,000 of 1,000,000 acres — 13 percent. The most successful watershed payment programme on earth has protected less land than would be required to measure its own effect. Andréassian (2004) and Bruijnzeel (2004) reach the same place from the tropics: attribution weakens as basins get larger, and at basin scale it is mostly gone. Pattanayak, Wunder and Ferraro (2010) went looking for payment schemes with credible counterfactual evaluation and found very few; Naeem and colleagues, writing in Science in 2015 found that most such programmes lacked an adequate scientific basis for the service they claimed to buy.

So most watershed payments cannot demonstrate the service they buy. That is true, it is the strongest objection in the field, and the design in the next two movements does not depend on solving it.

Fifth, the third term. A payment for watershed services has three terms, and published business cases print two. Write the break-even parcel size:

        A*  =  t / (b − c)

  t   fixed transaction cost per contract per year
  b   downstream benefit per hectare per year
  c   upstream opportunity cost per hectare per year

In the Catskills, with 15 percent of the basin genuinely at risk of conversion, b is $9,138 per hectare per year — the avoided $554.7 million spread over 60,703 hectares. Against a forgone land rent of $250 and a transaction cost of $1,200 a contract, A\* is 0.135 hectares. Any parcel bigger than a seventh of a hectare pays for its own paperwork. The third term is invisible here, which is precisely why the famous case taught the world a lesson the world could not use.

Now a basin shaped like most basins. A utility serving 100,000 people at 400 litres a head a day moves 14.6 million cubic metres a year. At an incremental treatment saving of $0.30 per thousand gallons it avoids $1.16 million a year; across 10,000 at-risk hectares that is $115.71 per hectare per year. At a forgone rent of $250, b − c is negative and no parcel size clears at all. At a forgone rent of $80, b − c is $35.71 and A\* is 33.6 hectares — and where the median holding is 2 hectares, transaction cost runs at $600 per hectare per year against a net benefit of $35.71. The scheme is short by a factor of 17, and it fails on paperwork, not on ecology.

This is what kills most watershed payment schemes and it is almost never printed. Ferraro (2008) named the mechanism — transaction costs and asymmetric information are largely fixed per contract, so they fall as contract size rises — and the fix falls straight out of the arithmetic. Aggregate the counterparty. One contract with a 3,000-hectare water-user association carries 40 cents a hectare a year of transaction cost and clears by 89×. It replaces 1,500 contracts with one, and it replaces 1,124,250 pairwise relationships with one. That is what New York's Watershed Agricultural Council is, and it is the cheapest thing in this chapter.

Sixth, the two governance arithmetics, side by side. The Colorado River Compact of 1922 apportioned 7.5 million acre-feet a year to each of the Upper and Lower Basins; the 1944 treaty added 1.5 for Mexico. Total obligated: 16.5 maf a year.

What the negotiators were looking at is now checkable, and it is worse than the usual telling. In the Meko tree-ring reconstruction of the Upper Colorado, the wettest twenty-five-year window in the entire 1,244-year record begins in 1905, at 16.83 maf a year — and Reclamation's own gauged natural flow for water years 1906 to 1922 — the record actually on the table — averages 18.03. They did not merely pick a wet period. They picked the wet period, and the measured record in front of them was wetter still. Against what the river carries:

  reconstructed, AD 762–2005    14.65 maf    gap 1.85 maf   11% of the promise
  measured, WY 1906–2024        14.63 maf    gap 1.87 maf   11% of the promise
  driest 25 years, from AD 1130 12.62 maf    gap 3.88 maf   24% of the promise
  measured, WY 2000–2024        12.40 maf    gap 4.10 maf   25% of the promise

The 2000s run 15.2 percent below the measured long-run mean. System storage capacity is 58.48 maf; at the start of water year 2026 it held 21.8 maf — 37 percent of it.

The Compact allocated a volume. A volume cannot close against a variable flow, so the gap became litigation, shortage declarations and emergency negotiation. Now rewrite the identical deal as shares of measured flow — 45.4545 percent to each basin, 9.0909 to Mexico — and at 12.40 maf each basin receives 5.636 instead of 7.5. The 25 percent haircut arrives by arithmetic, in the year the water does not, and nobody has to be sued. Australia writes entitlements this way: a Murray–Darling entitlement is a share of an announced allocation, not a promise of megalitres, which is why that basin's crisis was a crisis of volume and not additionally a crisis of law.

And the Murray–Darling supplies the second arithmetic, which is about how you buy water back. Recovery ran on two tracks: buying entitlements from willing sellers, and subsidising irrigation efficiency so that farms could give back the water they no longer lost. The recovery register says which one worked, as at 30 June 2026:

  purchase, open and limited tender   1,254.1 GL/yr   60.5%
  infrastructure                        631.4 GL/yr   30.5%
  state, gifted and other               187.6 GL/yr    9.0%
  ------------------------------------------------------------
  total recovered                     2,073.1 GL/yr

Against funding of A$2,832 million committed to purchase and A$3,120 million to gap-bridging infrastructure: the subsidy route was given more money and returned half as much water. The Productivity Commission put the unit costs at A$2,808 per megalitre for purchase against A$8,126 for infrastructure, in 2022 dollars per megalitre of long-term average annual yield — "almost three times more expensive," in its own words, at 2.89×. That comparison is the Commission's, computed from Wheeler and colleagues' evidence to it; the National Audit Office examined process and value for money inside individual procurements and published no cross-method comparison at all, and attributing this ratio to it would be a citation nobody checks.

The single hardest exhibit needs no comparison between policies. Take one irrigation network, one recovery method, four rounds, thirteen years:

  PIIOP round 1   2009   A$ 68.17m    5,700 ML   A$11,960 /ML
  PIIOP round 2   2011   A$221.80m   36,960 ML   A$ 6,001 /ML
  PIIOP round 3   2015   A$144.53m   13,663 ML   A$10,578 /ML
  OFEP            2022   A$126.48m    5,472 ML   A$23,114 /ML

Said precisely, because the loose version of this is wrong: the volume does not fall in a line — it peaks in round two and the last round recovers about what the first did. What moves in one direction is the price. The cheapest round to the last is 3.85×, and the largest round was also the cheapest. One policy running out of road, in its own numbers, published by the department that ran it.

And Grafton and colleagues, in Science in 2018 gave the structural reason the gap is worse than any of those accounts show. An efficiency subsidy buys a reduction in a farm's losses, and a farm's losses are the return flow that was somebody else's supply. The farm's books improve. The basin's do not. Only a closed basin account can see that, and a farm-level account structurally cannot.

---

DREAM

What becomes ordinary

In the version of this that has already happened, a basin publishes accounts, and the accounts are boring.

They arrive once a year, on a date everyone knows, and they look like accounts: a statement of water assets and water liabilities at the opening and closing of the year, a statement of the changes between them, a reconciliation, and a residual with a note explaining it. They are assured by someone independent, in the way a set of financial statements is assured, and the assurance opinion is short and occasionally qualified. Nobody finds this remarkable. The Bureau of Meteorology has been doing it for years and the only people who get excited are the ones who find something in the residual.

Every entitlement in the basin is a share, not a volume. When the year is dry, the announced allocation is lower and everyone's entitlement is worth less on the same day, in the same proportion, without anyone declaring an emergency. The lawyers are not involved because there is nothing to litigate: a percentage of a measured quantity is not a matter of opinion. Farmers hedge the allocation the way they hedge a wheat price, and there is a forward market in it, and the market is dull, which is the highest compliment a market can be paid.

The water utility carries a watershed asset on its balance sheet, and the regulator lets it earn on that asset the same return it earns on pipes and pumps — because the asset does the same work as a plant and the accounting should not punish the cheaper answer. When the utility's board reviews capital, the watershed programme and the treatment plant appear in the same table, in the same units, competing for the same money on the same test. Usually the watershed wins. Sometimes it does not, and that is a real answer rather than a defeat.

Upstream, the person who actually changes the land is a member of an association that holds one contract, and they deal with a neighbour who understands their ground rather than a form from a city they have never visited. They are paid for doing something visible and verifiable on their own land. Nobody has ever asked them to prove what their field did to the river, because nobody sensible would, and the whole design has been arranged so that no one has to.

Where the basin crosses a border, there is a commission, and the commission has almost no power. It cannot fine anyone, it cannot compel anyone, and it does not need to, because what it actually runs is a publication schedule. Once a year it prints how far each party got against the numbers they all agreed, and where a target was missed it says so in the report rather than in a footnote. The credibility of everything else it publishes rests on the sentences where it says a target was not met, and every party knows that, which is why the misses go in. The room is quieter than people expect. Nobody is being caught; everybody is being counted, and being counted turns out to be a thing organisations will submit to for seventy years if the counting is fair and the arithmetic is shared.

And the residual is watched. It is the one number in the report that anybody argues about, and when it moves, somebody goes and finds out why, and what they find is usually an unmetered abstraction, a leaking main, or a gauge that needs recalibrating. The basin found it because the books had to close. That is the entire benefit, and it is enormous.

---

DESIGN

The structure that gets there

Five moves. They are ordered, and the order matters, because each one removes a dependency the next one would otherwise inherit.

One: the boundary is the gauge, and the gauge comes first. Do not begin with a valuation study. Begin by establishing where the basin's water is measured, at what interval, by whom, to what precision, and who can see it. Every subsequent argument will be conducted in the units this instrument produces, so the instrument is the most political object in the system and should be settled while nothing is at stake. If a basin has one credible gauge and no valuation, it can govern itself. If it has a valuation and no gauge, it has a brochure.

Two: allocate shares, never volumes. Every entitlement is a percentage of a measured, announced quantity. This is the single design decision that separates the Colorado from the Murray–Darling, and its value is not efficiency — it is that shortage is distributed by arithmetic rather than by power. Write it into the instrument at the start. Converting a volumetric regime to a share-based one after over-allocation has been capitalised into land values is the hardest transaction in water law, and it is a decade of work.

Three: settle downstream at the intake, contract upstream on practice. This is the move that removes the attribution problem rather than solving it. The downstream buyer is not purchasing a hydrological outcome from a parcel; it is purchasing a reduced probability of having to build a plant, and that probability is settled against the one measurement the utility already makes every day under regulatory compulsion: raw water quality at the intake. Upstream, pay for practice — the septic system replaced, the riparian buffer standing, the plan followed — because practice is observable, inspectable, and now remotely verifiable at almost no cost. Attribution is done once, at basin scale, against the regulatory threshold. It is never done per hectare, because per hectare it cannot be done. New York's arrangement has worked this way since 1997 without anyone calling it that.

Four: aggregate the counterparty. Contract with one intermediary holding membership, not with every landholder. The arithmetic gave the magnitude: at a 3,000-hectare association, transaction cost per hectare falls from $600 to 40 cents and the scheme clears by 89×. The intermediary must be led by the people it contracts for — the Watershed Agricultural Council is farmer-governed and that is why farmers sign — and it must be paid a disclosed administration margin rather than an invisible one, so the third term stays on the page where somebody can manage it down.

Five: publish the residual, and treat it as the finding. The closing reconciliation is the product. Report ε in the same units as everything else, with a note, every year, including the years it is embarrassing. A basin that publishes a residual it cannot explain is operating a real accounting system. A basin whose residual is always zero is operating a press release.

Two things this design deliberately refuses. It does not pay for trees in a basin that is short of water — the Farley arithmetic says that purchase is negative on one side of the ledger, and the ledger is the point. And it does not fund irrigation efficiency out of a water recovery budget, because efficiency subsidies buy reductions in return flow, and return flow is downstream supply. Buy the entitlement instead. In the one basin that has run both at scale it came in at A$2,808 a megalitre against A$8,126, and it is the only one of the two that is unambiguously real in the basin's accounts.

---

DESTINY

How it holds when nobody is pushing

Three things keep a basin account alive, and all three are structural rather than motivational.

The account has a statutory date. Annual, published, assured. A report that must appear on a date survives every change of personnel; a report that appears when someone champions it dies with the champion's promotion. This is why the Rhine commission has outlasted three generations of national politics: its product is a publication schedule.

The money rides on it. The utility's watershed programme sits in the rate base and earns like an asset; the entitlement holder's share is worth what the announced allocation says. Where the account moves, someone's balance sheet moves. An account nobody is paid or charged against is a hobby.

The intermediary is older than the deal. A farmer-led council, a water-user association, a basin commission. It holds the relationships, the local knowledge and the contract history, and it persists through the periodic collapse of political attention that every basin experiences.

Now the honest part, because these fail in four named ways and every one has happened. They fail when the gauge is captured — when the party being measured owns the measurement, and the record quietly improves. They fail when the payment becomes an entitlement: after a decade, upstream recipients treat the transfer as income rather than as consideration for a practice, and withdrawing it becomes politically impossible even where the practice has stopped. They fail when the downstream buyer's alternative gets cheaper — membrane filtration costs have fallen for thirty years, and a watershed programme priced against a 1990s plant is priced against a number that is still moving. Re-run the avoided-cost arithmetic every five years and expect it to erode. And they fail when the scheme is asked to prove attribution it was never designed to prove, usually by an auditor four years in who asks which hectare produced which milligram. The defence against that is to have said so at the start, in writing, in the instrument: this is a purchase of a deferred capital decision, settled at the intake, and it was never a purchase of a per-parcel hydrological outcome.

A scheme that has written that sentence into its own documents survives the audit. One that has let the folklore stand does not.

And one ending that is not a failure at all, because it is worth knowing the difference. In October 1988 the Rhine Ministers declined the second 60 kg/s stage of the chloride reduction: the marginal abatement cost had risen above what the downstream parties would pay for it. The first stage stood, the money already spent stayed spent, and the deal stopped where the arithmetic said to stop. That is a walk-away price doing exactly the work a walk-away price is for, and a scheme that publishes one can be ended honourably instead of being abandoned quietly.

---

DELIGHT

What it feels like

There is a specific pleasure in a set of books that close, and people who have never kept books do not know about it. It is not the pleasure of a good result. It is the pleasure of a constrained system — the feeling of having nowhere to put the difference, and therefore having to find out what it is.

The field version of it is better still. Standing at a gauging station in February with cold water over your boots, watching a number you will later see in a published table, knowing that four thousand square kilometres of ridge and hemlock and old stone wall are being summarised by this one channel and this one instrument, and that the summary is honest because the landscape gave you no choice — that is one of the few moments in economics where the measurement and the thing measured are in the same place at the same time.

And then the third pleasure, the one that lasts: the year the residual will not reconcile, and somebody drives up a valley to find out why, and comes back with an unmetered abstraction nobody had recorded since 1974. Nothing in a quarterly report has ever produced that feeling. The basin told on itself, because the arithmetic left it no other option, and the arithmetic was free.

---

OPERATIONALIZE THIS

At the level of finance

The instrument: a watershed deferral facility, settled at the intake.

This is not a grant, an offset, or a philanthropic programme. It is a water utility purchasing the deferral of a capital project it would otherwise have to build, financed against the deferral, and it belongs in the capital plan next to the plant it displaces.

The structure. A ring-fenced facility inside the utility's capital programme, disbursing to a single upstream intermediary under a multi-year agreement, with an annual basin account published against it. Term should match the deferral being bought — ten years is the working default, reset every five.

The counterparty. One, and it must be governed by the landholders it contracts for. A farmer-led council, a water-user association, a catchment trust. The utility's contract is with the intermediary; the intermediary's contracts are with its members; the utility never contracts with a parcel. Its administration margin is disclosed as a separate line, because that line is the third term and a disclosed cost is a managed cost.

The settlement. Against raw water quality at the intake, measured against the regulatory threshold the utility already monitors under compulsion. Zero marginal measurement cost, an existing chain of custody, and a counterparty who is never asked to prove causation. Payment is for verified practice upstream plus a portfolio-level performance band at the intake, not for attribution.

The balance-sheet treatment. This is where most of the money is, and it is an argument your finance function can win. Where the programme secures a durable interest — an easement, a covenant, a permanent land interest — it is a capitalisable asset with a defined life, not an operating expense, and it should be in the rate base earning the allowed return alongside pipes and pumps. Where it does not, it is prepaid capacity and should be amortised over the deferral it buys. Expensing a programme that defers a nine-billion-dollar plant is not conservatism. It is a category error that makes the cheaper option look like charity. Talk to your auditors early; it is a conversation about the useful economic life of an intangible, which they have every year.

The number that decides it. One figure, on the front page:

             PV of the capital deferral bought
        ---------------------------------------  >  1
             PV of the programme that buys it

At the Catskill/Delaware basin, deferring a $9.0 billion plant by ten years at a 4 percent real rate is worth $2.92 billion in present value; the programme that buys it costs $450.6 million in present value. Net present value $2.47 billion. The ratio is 6.5 : 1. Read the other way, the same arithmetic gives you the walk-away price: the programme could cost up to $360 million a year before building the plant becomes the cheaper answer. Publish that number. It is the honest ceiling on what upstream can ask for and the honest floor under what downstream should be willing to pay, and it is the sentence that turns a negotiation into a transaction.

The first ninety days.

Do them in that order. Every one of them is cheaper than the one after it, and the first two cost almost nothing and change the conversation permanently.

---

APPRECIATIVE QUESTIONS

Twelve, for a room

Discovery — what is already working

1. Where in this basin does a measurement already exist that everybody trusts — who made it trustworthy, and what did that take?
2. Think of a time two parties here disagreed about water and settled it without lawyers. What made the settlement possible, and who was in the room?
3. What does the person furthest upstream already do, unpaid, that the people furthest downstream are benefiting from right now?

Dream — what becomes possible

1. If this basin published a set of accounts every year that an auditor signed, what is the first thing we would learn that we do not currently know?
2. Imagine our entitlements were shares rather than volumes, and a dry year arrived. What conversation would we no longer be having?
3. If the residual in our accounts were the number everybody watched, who would become interesting who is currently ignored?

Design — what we build

1. Who could hold one contract on behalf of everybody upstream, and what would they need from us to be able to say yes?
2. What are we already measuring, for a regulator, that could serve as the settlement point without costing another dollar?
3. If we could only buy one thing upstream — one practice, on one kind of land — which one would we want, and how would we know we had it?

Destiny — how it holds

1. What would keep this account being published in a year when nobody in this room is still here?
2. Whose balance sheet would have to move for this to survive a change of leadership, and what would it take to move it?
3. If the payment became an entitlement that outlived the practice, who would notice first — and what would we want them to be able to do about it?

---

WORKS CITED

Andréassian, V. (2004). "Waters and Forests: From Historical Controversy to Scientific Debate." Journal of Hydrology, 291, 1–27.

Bosch, J. M. and Hewlett, J. D. (1982). "A Review of Catchment Experiments to Determine the Effect of Vegetation Changes on Water Yield and Evapotranspiration." Journal of Hydrology, 55, 3–23.

Brown, A. E., Zhang, L., McMahon, T. A., Western, A. W. and Vertessy, R. A. (2005). "A Review of Paired Catchment Studies for Determining Changes in Water Yield Resulting from Alterations in Vegetation." Journal of Hydrology, 310, 28–61.

Bruijnzeel, L. A. (2004). "Hydrological Functions of Tropical Forests: Not Seeing the Soil for the Trees?" Agriculture, Ecosystems & Environment, 104, 185–228.

Chichilnisky, G. and Heal, G. (1998). "Economic Returns from the Biosphere." Nature, 391, 629–630.

Farley, K. A., Jobbágy, E. G. and Jackson, R. B. (2005). "Effects of Afforestation on Water Yield: A Global Synthesis with Implications for Policy." Global Change Biology, 11, 1565–1576.

Ferraro, P. J. (2008). "Asymmetric Information and Contract Design for Payments for Environmental Services." Ecological Economics, 65, 810–821.

Grafton, R. Q., Williams, J., Perry, C. J., Molle, F., Ringler, C., Steduto, P., Udall, B., Wheeler, S. A., Wang, Y., Garrick, D. and Allen, R. G. (2018). "The Paradox of Irrigation Efficiency." Science, 361, 748–750.

Jackson, R. B., Jobbágy, E. G., Avissar, R., Roy, S. B., Barrett, D. J., Cook, C. W., Farley, K. A., le Maitre, D. C., McCarl, B. A. and Murray, B. C. (2005). "Trading Water for Carbon with Biological Carbon Sequestration." Science, 310, 1944–1947.

Jayachandran, S., de Laat, J., Lambin, E. F., Stanton, C. Y., Audy, R. and Thomas, N. E. (2017). "Cash for Carbon: A Randomized Trial of Payments for Ecosystem Services to Reduce Deforestation." Science, 357, 267–273.

Meko, D. M., Woodhouse, C. A., Baisan, C. A., Knight, T., Lukas, J. J., Hughes, M. K. and Salzer, M. W. (2007). "Medieval Drought in the Upper Colorado River Basin." Geophysical Research Letters, 34(10), L10705.

Muñoz-Piña, C., Guevara, A., Torres, J. M. and Braña, J. (2008). "Paying for the Hydrological Services of Mexico's Forests: Analysis, Negotiations and Results." Ecological Economics, 65, 725–736.

Naeem, S., Ingram, J. C., Varga, A. et al. (2015). "Get the Science Right When Paying for Nature's Services." Science, 347, 1206–1207.

Ostrom, E. (1990). Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press.

Pattanayak, S. K., Wunder, S. and Ferraro, P. J. (2010). "Show Me the Money: Do Payments Supply Environmental Services in Developing Countries?" Review of Environmental Economics and Policy, 4, 254–274.

Pires, M. (2004). "Watershed Protection for a World City: The Case of New York." Land Use Policy, 21, 161–175.

Postel, S. L. and Thompson, B. H. (2005). "Watershed Protection: Capturing the Benefits of Nature's Water Supply Services." Natural Resources Forum, 29, 98–108.

Sagoff, M. (2002). "On the Value of Natural Ecosystems: The Catskills Parable." Politics and the Life Sciences, 21, 19–25.

Salzman, J., Bennett, G., Carroll, N., Goldstein, A. and Jenkins, M. (2018). "The Global Status and Trends of Payments for Ecosystem Services." Nature Sustainability, 1, 136–144.

Vörösmarty, C. J., McIntyre, P. B., Gessner, M. O. et al. (2010). "Global Threats to Human Water Security and River Biodiversity." Nature, 467, 555–561.

Wunder, S., Engel, S. and Pagiola, S. (2008). "Taking Stock: A Comparative Analysis of Payments for Environmental Services Programs in Developed and Developing Countries." Ecological Economics, 65, 834–852.

Zhang, L., Dawes, W. R. and Walker, G. R. (2001). "Response of Mean Annual Evapotranspiration to Vegetation Changes at Catchment Scale." Water Resources Research, 37, 701–708.

Productivity Commission (2024). Murray–Darling Basin Plan: Implementation Review 2023. Commonwealth of Australia. (Table 2.1, Table 2.4, Box 3.1 and figure 2.1; unit costs at p. 92, computed from Wheeler et al., submission 104.)

Productivity Commission (2018). Murray–Darling Basin Plan: Five-year Assessment. Commonwealth of Australia.

Primary and official sources. Convention on the Protection of the Rhine (1999), the Bern Convention (1963), the Rhine Action Programme (1987), the Action Plan on Floods (1998) and the ICPR Activity Report 1999/2000 and Assessment "Rhine 2020", International Commission for the Protection of the Rhine. Convention on the Protection of the Rhine against Pollution by Chlorides, Bonn, 3 December 1976, United Nations Treaty Series Vol. 1404, I-23469 (Articles 2, 7 and 8). New York City Watershed Memorandum of Agreement (1997) and the Filtration Avoidance Determinations, New York City Department of Environmental Protection and the United States Environmental Protection Agency. Croton Water Filtration Plant cost and capacity, New York City Department of Environmental Protection. Water Act 2007 (Cth), the Basin Plan 2012 and the Water Amendment (Restoring Our Rivers) Act 2023, with the sustainable diversion limit adjustment mechanism, Murray–Darling Basin Authority. Surface water recovery under the Basin Plan, the recovery register as at 30 June 2026, Australian Department of Climate Change, Energy, the Environment and Water. Auditor-General Report No. 2 of 2020–21, Procurement of Strategic Water Entitlements, Australian National Audit Office — cited for procurement practice, not for any comparison between recovery methods, which it does not make. Australian Water Accounting Standard 1 and the National Water Account, Water Accounting Standards Board and the Bureau of Meteorology. Colorado River Compact (1922) and the Mexican Water Treaty (1944).

Note on figures. The basin water balance, the Zhang evapotranspiration terms, the Croton unit capital cost and the implied Catskill/Delaware plant, the annualised avoided cost, the two benefit–cost denominators, the break-even parcel size in both basins, the aggregation gain, the Colorado share-based counterfactual and the deferral ratio are all computed in lib/verify/VII_06.py and reproducible there, with every input printed and every assumption labelled. Bioregional boundaries are Chapter VII.05; the choice of discount rate is Chapter VII.08; the commons design principles this chapter leans on are Volume VI.