The Murray-Darling Basin missed a legislated 450-gigalitre water recovery deadline at about 6 percent delivered, and in July 2026 14 of the 39 projects offsetting a further 605 gigalitres were withdrawn

What is happening?

The Murray-Darling Basin Plan, made in 2012 under the federal Water Act 2007, commits Australia to taking water back from irrigation and returning it to the river system. Two targets sit inside it. The first, called Bridging the Gap, began as 2,750 GL/y to be recovered over seven years and was reduced to about 2,075 GL/y by amendments in 2017 and 2018. The second, added in 2018, is a further 450 GL/y of environmental water, and it carried a legislated deadline of 2024-06-30.

That second deadline passed with about 27 GL recovered, roughly 6 percent of the target. Two years before it fell due, on 2022-07-31, the figure stood at 2 GL, or 0.4 percent. An independent panel review had already reported in March 2020 that the deadline was unrealistic.

Nothing in the statute attaches a consequence to missing it. The federal law obliges the minister to take all reasonable steps to increase the water available for the environment by 450 GL/y, and it names no penalty and no fallback. Before the 2024 deadline arrived, the Water Amendment (Restoring Our Rivers) Act 2023 had already moved it to 2027-12-31, a change of law rather than a change of outcome. Recovery has run faster since: 169.8 GL by 2025-09-30, being 38 percent of the target, 246.3 GL by 2026-03-31, and about 380 GL by 2026-06-18, being about 85 percent. The government has stated an aim of 400 GL by December 2026.

The other target lost ground in the opposite direction in July 2026. Of the Bridging the Gap total, 605 GL/y was never meant to be recovered as water at all. In 2017 the basin governments agreed to offset that volume with 39 supply measure projects — changes to river structures and operating rules intended to deliver an equivalent environmental outcome using less water. On 2026-07-06 the Murray-Darling Basin Authority reported that 14 of those projects had been withdrawn because they could not be completed by the December 2026 deadline. Thirteen were complete and ten remained on track. The withdrawn projects represented up to 355 GL of equivalent benefit, and the exact shortfall is not settled because the reconciliation process has not finished.

Whose problem is this?

The body that must deliver the water is answerable to the parliament that can move the date, and in 2023 that is what happened. The people at the downstream end of the system are not a party to either decision.

Where does this problem end?

The boundary here separates a volume of water from an ecological result.

This document is about whether a legislated volume of water arrived by a legislated date, and it treats river condition as the reason that volume was legislated rather than as the thing being measured. A recovery target can be met in full while a river keeps declining, and it can be missed in a year when rainfall makes the river look healthy.

What is the state now, and what should it be?

Now

Should be — this problem is unusual in that the target state is written down in full. Federal law names the volume, 450 GL/y of additional water available for the environment, and it names the date, 2027-12-31. The Bridging the Gap component names a second volume of about 2,075 GL/y, of which 605 GL/y is to be delivered not as water but as an equivalent environmental outcome from the 39 supply measure projects. What the law does not name is what follows if either arrives late, and the record shows that the answer so far has been an amendment.

How big is it?

About 2 million people live in the Murray-Darling Basin, and about 1.4 million more live outside it while drawing their water from it, chiefly in Adelaide. Both figures come from a single secondary source citing 2010 material, and they are 16 years old as of 2026-08-15. No later official count was opened in this round, because every automated request to the basin authority pages returned an access refusal. Search summaries seen during research suggested a higher current basin population of about 2.3 to 2.4 million, and that figure could not be confirmed against any source that opened, so it is not used here.

The count measures dependence rather than harm, and the difference matters for how the number is read.

Basin residents include the irrigation communities whose water is the source of the recovery, so a person inside this count may be made better off by the shortfall rather than worse. A person counted here is not necessarily harmed today, and a river reach that is failing does not appear in this count at all.

The physical scale is easier to state and rests on the same kind of secondary source. About 11,500 GL is extracted from the basin each year against an annual extractable volume of about 13,000 GL, and about 95 percent of what is extracted goes to irrigation. Basin irrigation is described as producing about 40 percent of Australian agricultural output. Against those quantities, the two recovery targets amount to roughly 2,075 GL/y and 450 GL/y, and the second of them stood at about 85 percent delivered on 2026-06-18.

Under what conditions does it arise?

1. The duty is written as effort, not as result. The federal obligation is to take all reasonable steps toward the 450 GL/y. It names no penalty, no automatic consequence, and no party with standing to insist on the volume itself. An obligation to take all reasonable steps is discharged by the steps, not by the result.

2. The deadline can be moved by the body that set it. When the 2024 date approached with about 6 percent delivered, the answer was the 2023 amending Act, which moved the date to 2027-12-31. The same parliament that wrote the deadline can move it, and moving it costs less than meeting it.

3. The cheapest recovery route was closed by policy. The 2017 and 2018 amendments attached a socio-economic neutrality condition to voluntary buyback, which restricted purchase in the water market and pushed recovery toward infrastructure efficiency works and supply measures. That choice makes each gigalitre slower and more expensive to obtain, and it converts a purchase into a construction programme.

4. A large share of the target was converted from water into projects. The 605 GL/y offset does not exist as water anywhere. It exists as 39 engineering and rule-change projects, and a project can be abandoned in a way that purchased water cannot. That is what happened to 14 of them on 2026-07-06, and the reconciliation of what was lost is still running.

5. The resource itself is shrinking while the target is fixed. The 2025 evaluation projects flows falling by 20 to 30 percent by mid-century, and by up to 50 percent in some rivers. A fixed volume target set against a falling total means the same number of gigalitres represents a larger share of the river each year it goes unmet.

What has been tried?

Two directions were pursued at once — buy the water back, and avoid buying it by engineering an equivalent outcome. The first was constrained by policy and the second has now lost more than a third of its projects.

What was found?

Why is it still unsolved?

Enforcement absent — the volume and the date are both written into federal law, and the obligation attached to them is to try.

A deadline that can be moved by the body it binds is a plan, not a deadline. The 450 GL/y target arrived in 2018 with a date of 2024-06-30 and reached that date at about 6 percent. The response was not a consequence but an amendment, and the amendment was passed before the original date fell due. An observer watching only the statute book would see a target that has never been missed, because the date is rewritten faster than it can be broken.

The second reason is that the duty is drafted so that failure is not a defined event. An obligation to take all reasonable steps is satisfied by conduct, and conduct can always be described as continuing. There is no volume at which the minister is in breach, no party named as able to enforce the volume, and no penalty for arriving after the date. What the record shows instead is an acceleration once the deadline was extended, from 2 GL in 2022 to about 380 GL by mid-2026, which demonstrates that the delivery was possible and that the original schedule was not what produced it.

The third reason is that a large part of the target was made of things that can be cancelled. Water bought and registered cannot be unbought, but a project that has not been built can simply be withdrawn, and 14 of them were. The 605 GL/y offset was agreed in 2017 as an equivalence — the same environmental outcome for less water — and equivalence is a claim about engineering that stays untested until the engineering is finished. When 14 of the 39 projects went away in July 2026, what went away was not a volume of water anyone had ever held. The authority put the loss at up to 355 GL and said the reconciliation was not complete, which means the size of the reopened gap is currently unknown to everyone including the body that administers the plan.

The fourth reason keeps the pattern stable. The Royal Commission found in 2019 that the diversion limits had been set unlawfully and that a whole review amounted to maladministration, and no source opened here reports a binding federal consequence following from those findings. A finding of unlawfulness that produces no remedy tells every later participant what the cost of the next shortfall will be.

What observation would mean it is solved?

Candidates — (a) the 450 GL/y is fully recovered and registered on or before 2027-12-31 without a further extension (b) the reconciliation of the withdrawn supply measure projects completes and the resulting shortfall is closed with recovered water rather than a further offset (c) the ecological indicators the recovery exists to move, such as the condition of native fish populations, improve across several consecutive years.

(a) is the cleanest test and it is a test of the calendar, not of the river. It is falsifiable on a fixed date, it cannot be argued about, and the recovery rate since 2023 makes it plausible. What it does not settle is whether the volume was set correctly, and the Royal Commission held in 2019 that the diversion limits it sits on were derived without regard to climate change. A target met is not the same as a target that was right.

(b) is the one that would show the pattern had actually broken. The 605 GL/y offset is where a water target was converted into a construction promise, and the withdrawal of 14 projects is the moment that promise partially failed. Closing that gap with water rather than with a new equivalence claim would be the first time in this record that a shortfall was answered with the resource instead of with a revision. Closing it with another offset, or with a third extension, would leave the mechanism exactly as it is.

(c) is the outcome everything else is a proxy for, and it moves for reasons the plan does not control. A volume can arrive on time and the river can still be in poor condition, because the volume was only ever a proxy for the condition. Fish population indicators respond to drought and flood on their own schedule, and the 2025 evaluation projects flows falling by 20 to 30 percent by mid-century regardless of what is recovered. Read alone, a good year would credit the plan with the weather. The three have to be read together, and (b) is the one that would say whether the mechanism has changed.

What is it connected to?

Fills with researchAustralian water market design and the price effects of buyback, the sustainable diversion limits themselves as distinct from the recovery targets, First Nations water entitlements in the basin, drinking water security for basin towns, the 2026 review of the basin plan which is under way, and river basin allocation failures in other countries. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• How much water the 14 withdrawn projects actually cost the plan. The authority gave a figure of up to 355 GL and stated that the reconciliation has not finished. This is a procedure in progress rather than missing data, and until it completes nobody has published the size of the reopened gap.
• Which section of the Water Act carries the reasonable steps duty. Automated summaries in this round gave 85AC in one place and 86AC in another, and the disagreement was not settled. The wording of the obligation and the absence of a penalty rest on the legislation text that was opened; the section citation does not, so it is left unresolved here rather than picked.
• How the figure of about 2,075 GL/y is built. The reduction from 2,750 GL/y is attributed to a 70 GL/y northern basin adjustment and the 605 GL/y supply measure offset, and that arithmetic appears here only in search summaries. The Productivity Commission report confirms the 605 GL design and not the total.
• Whether the 99.5 percent completion claim still holds. It was announced on 2025-03-06, more than a year before 14 supply measure projects were withdrawn. No source opened here restates it afterwards, and the government dashboard that would carry an updated figure timed out on every attempt.
• What the current official basin population is. Only a 2010 estimate was opened. Every automated request to the basin authority pages describing population and significance returned an access refusal, so a 16-year-old figure is the most recent one this document can support.
• What happens if the 2027 deadline is also missed. No source opened here names a procedure. The record contains one target reduced and one deadline extended by statute, which is a history rather than a stated rule, and this document does not treat it as a prediction.
• Whether the water recovered so far has produced an ecological improvement. The 2025 evaluation reports native fish populations rated poor in 19 of 23 sub-catchments after about 2,069 GL was recovered. Nothing opened here attributes that result to the volume, to the timing of its delivery, or to anything else.
• What followed the Royal Commission findings at the federal level. No source opened here reports a binding federal response to the 2019 report, and the authority disputed its central criticisms. Whether any recommendation was implemented is absent from everything opened.

See the evidence

No Australian government source was read directly except the legislation text. The Federal Register of Legislation page carrying Part 2AA was opened and supplies the wording of the reasonable steps obligation, the 450 GL/y volume and the 2027-12-31 date; what it did not settle is the section number, which appears as 85AC in one summary and 86AC in another, so this document names both and asserts neither. Every page belonging to the Murray-Darling Basin Authority returned an access refusal to automated requests, and the departmental dashboard timed out three times, which is why four rows above are blank with the reason recorded rather than filled with a substitute link. The recovery volumes therefore reach this document through press reporting of government figures, and the two most consequential of them are corroborated where possible: the withdrawal of 14 projects appears in two independent trade sources on the same date with the same counts of 13 complete and 10 on track, and the extension of the deadline to 2027-12-31 appears in both the Productivity Commission review and the legislation text. Where sources disagree the disagreement is left visible. The reduction of Bridging the Gap to about 2,075 GL/y is carried only by search summaries, and the Productivity Commission report confirms the 605 GL supply measure design without confirming the total, so the figure is graded low above. The population figures rest on a single secondary source citing 2010 material, and a higher current figure seen in search summaries could not be confirmed against anything that opened, so it is excluded rather than averaged in. This is a Path A output, so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range 2,000,000–3,400,000 As of 2010

Sensitivity The width of this interval is definitional, not statistical. It is not a confidence interval and neither endpoint carries a margin. The low end counts residence inside the basin and the high end adds the population outside it that is supplied from the basin, and the two endpoints answer two different questions about who counts as dependent. Both endpoints come from a single secondary source citing 2010 material, so the figures are 16 years old against a status check date of 2026-08-15. Two corrections would move the number and neither could be computed from the sources opened here. Upward, search summaries during research suggested a current basin population of about 2.3 to 2.4 million, which was not confirmed against any source that opened and is therefore excluded rather than averaged in; every page of the Murray-Darling Basin Authority describing population returned an access refusal to automated requests. Downward, this counts dependence on the river system rather than exposure to the recovery shortfall, and no source opened here separates the two. The quantity the recovery target actually protects is not a population at all. It is the condition of the river, reported in the 2025 evaluation as native fish populations rated poor in 19 of 23 sub-catchments, and no count of people expresses that. Irrigation-dependent employment and downstream ecological loss are absent from both endpoints.

Regional breakdown No source opened here splits basin population by state or by sub-catchment. The basin spans Queensland, New South Wales, Victoria, South Australia and the Australian Capital Territory, and the only figures available are basin-wide totals from a single 2010 estimate. Dividing them by state population would be proportional allocation, and dependence on the basin follows the river network rather than where people live, so that division would be wrong in both directions at once. The externally supplied component is concentrated in one city, which a population-weighted split would spread across five jurisdictions.