In 2023, federal money paid for 64 conservation projects across the four Upper Division States — just under $16 million, for an estimated 37,810 acre-feet of water those farms and cities agreed not to consume. Everyone did what they were paid to do. Nobody can tell you where that water went.
That is not a scandal. It is a mechanism. And if you are a commissioner, a watershed council, or a landowner being asked to trust the next round of conservation credits or a corporate water offset, the mechanism is the part worth understanding.
the leak, in plain language
Here is the whole story in three moves.
- Someone pays a farmer to use less. A field goes fallow, or a hay crop gets one cutting instead of three. The water that would have been consumed by the crop stays in the ditch, then the creek, then the river.
- The water re-enters the stream. Under Western water law, water left in the river is just river water. It carries no tag that says "paid for."
- The next diverter takes it. Downstream, a headgate opens under a decree that has been valid for a century. The conserved acre-foot becomes someone else's irrigated acre. Legally. Correctly.
The farmer did the hard thing. The downstream user followed their decree. The payor spent real money. And the volume everyone meant to save may never reach the reservoir it was meant to protect.
That is what happened with the 2023 System Conservation Pilot Program. The program estimated how much each project did not consume, and it paid for that estimate. It had no way to track whether the water reached Lake Powell, and Upper Basin water managers have said as much for years. The Colorado River District's general manager put it bluntly: it "inherently just flows downstream."
If you want the farmer's side of this arrangement, what the checks pay and what they do not, read getting paid not to farm. This post stays on what happens to the water after the check clears.
why water accounting exists
Accounting is the industry's answer to the leak, and it is the right answer to that question.
In October 2024 the Upper Division States, acting through the Upper Colorado River Commission, approved a provisional accounting framework with the Bureau of Reclamation covering conservation activities in 2025 and 2026. It is running now — the commission approved its first Water Year 2025 accounting report in April 2026. Strip the acronyms and it does three things:
| step | what it asks | how they answer it |
|---|---|---|
| measure the saving | how much did the project not consume? | satellite evapotranspiration (eeMETRIC), plus the methods refined in the 2023 and 2024 pilot |
| follow the water | how much of that saving reaches an Upper Basin reservoir? | estimate losses between the field and the reservoir, including intervening points of diversion |
| book the balance | whose account does the stored volume sit in? | provisional accounting that can later be converted to credit under a further agreement |
That middle row is the fix. Somebody is finally asking whether the acre-foot arrived, not only whether it was left behind. Location matters too: storing conserved water high in the basin, close to where it was saved, gives the people who saved it more say over when and how it is released.
Some irrigators had already worked this out locally. In the Dolores Water Conservancy District, farmers who took system conservation payments arranged to hold the saved water in nearby McPhee Reservoir as carryover for the next dry year. Same program. No leak. Because there was a bucket, and a bookkeeper, close to the field.
Tracking conserved consumptive use across state lines, through reservoirs, past a thousand headgates is difficult, expensive work, and the people doing it are trying to make a paid fallow mean something. Water accounting is how you turn a leak into a protected volume.
a protected volume is still an allocation
Here is where the ledger stops being able to help you.
Say the accounting works perfectly. Every conserved acre-foot is measured at the field, tracked to the reservoir, and booked to an account. You have stopped the leak. What do you actually hold?
You hold a volume that moved from one column to another. The river did not get bigger. The snowpack did not melt slower. The meadow above the diversion did not hold more of the spring pulse into August. You protected a claim on last year's factory output. You did not touch the factory.
A water credit is an accounting entry. A watershed is a factory. Accounting turns a leak into a protected volume; it does not turn a protected volume into a wetter basin.
The Colorado River is the clearest example only because it is one of the most closely accounted rivers on Earth. How the credits get used, and who they protect, is a story about the rules that expire in 2026, and it lives in its own post: what happens to the Colorado River after 2026. This post is about the mechanism, and the mechanism is the same on any river.
| question | the ledger can answer | only the land can answer |
|---|---|---|
| how much did the farm not consume? | yes | — |
| did the saving reach storage? | yes, with loss estimates | — |
| whose account holds it? | yes | — |
| will next spring's melt arrive slower and later? | — | yes: canopy, shade, soil, meadow storage |
| will a wet year carry into a dry one? | — | yes: floodplains, beaver, riparian sponge, recharge |
| is there more water in the basin? | no | only if the places that make it are in better condition |
The left column is allocation. It is necessary, and it is what every credit, offset, and certificate on the market today is built to do. The right column is production. Nothing on the ledger reaches it.
what a water offset is, and what it buys
A water offset is the corporate version of the same idea: a company pays for conservation somewhere in a basin to balance the water it consumes somewhere else. A conservation credit is the public-agency version: a state or district earns a booked volume for conservation it funded.
Both inherit the leak. If the conserved volume can be diverted again before it reaches storage, the offset bought a receipt for water that may already be in someone else's ditch. That is exactly why offset programs now lean on registries, tracking, and storage accounts. They are plugging the same hole the Upper Basin is plugging.
And both inherit the ceiling. A well-tracked offset is an honest allocation. It is still an allocation. If the company's actual exposure is a basin that produces less water each decade, no quantity of receipts changes what the basin produces.
A water offset is a payment for someone else's conserved volume. If that volume can be recaptured before it reaches storage, the offset bought a receipt, not water. If it cannot be recaptured, the offset bought an allocation, not a wetter basin.
the farmers are not the leak
Credit accounting also sits on top of rights that were settled late or not at all. Any ledger that counts conserved water has to decide what it does with water that was never allowed to be used in the first place — see tribal water settlements and watershed protection.
It is worth saying plainly, because the language of credits and leaks invites the wrong villain.
The irrigator who fallowed a field for anywhere from $150 to more than $600 an acre-foot gave up a season of production, a hay contract, sometimes a hired hand. The downstream diverter who took the water did what a decree entitles them to do; they may not have known the water was "conserved" at all. The commission that paid for it was moving fast under a federal deadline with the tools it had.
The leak was in the design, not in anyone's conduct. Accounting is the design fix. The people on the ground were never the problem, and any credit program that treats them as one will not have participants for long.
track the volume. fund the factory.
So what do you do with this, if you sit on a council, a commission, or a deed?
Keep the ledger. If you are paying for conservation, you should know whether it arrived. Provisional accounting is the correct response to a real leak, and the cousins building it deserve the benefit of the doubt.
Then put something beside it that the ledger cannot do: fund the condition of the land that produces the acre-foot in the first place. Headwater meadows that hold snowmelt. Soils on irrigated ground that keep more of what falls on them. Riparian corridors that spread a flood into groundwater instead of sending it past you in June. The people who already pay for shortage, through hauling, fallowing, new pipes, and treatment, are paying after the fact. This is paying before.
That is what ensurance is for. A certificate of ensurance is tied to one named place and funds its present condition: a specific meadow, a specific ranch's riparian reach, a specific soil-building program. It is not a fungible acre-foot. It is not a water credit, and it does not claim to offset anyone's use. It can sit beside your conservation account without pretending to be one. The volumes are small today. The instrument is live, and the distinction is the point.
How the landscape actually makes water, and why it is the cheapest storage you will ever fund, is in the water cycle: broken, and how to put it back.
If you are a government or a watershed collaborative weighing a credit program, this is the conversation: which named places in your basin produce the water your ledger counts, and what would it take to hold their condition?
see how ensurance works for governments →
Read next: who buys a water credit, the people already paying for shortage, and what they are actually shopping for.