What is actually being asked
The three documents are not three topics. They are one review seen from three angles, and each hands work to the next.
The Carbon Leakage Review, led by Frank Jotzo and published in February, closed with a recommendation to introduce a border carbon adjustment for cement and clinker first, to consider lime, hydrogen and ammonia, steel and iron, and glass later, to refuse export rebates outright, and — the part that matters here — to remove concessional trade-exposed baselines for any commodity once a border adjustment covers it. Its recommendations were handed to the Safeguard review.
The Climate Change Authority was given a deliberately narrow brief: the post-2030 baseline decline rate, and the extent to which the reforms are actually driving abatement at the facilities themselves. Its preliminary arithmetic puts a decline rate of about 4.5 per cent a year against the bottom of Australia’s 2035 target range and about 7 per cent against the top — against an indicative 3.285 per cent currently written into the rules. It says explicitly that the coverage threshold and international units are outside its scope.
The department’s consultation paper is the wide one. It carries the decline rate, credit restrictions, trade-exposed concessions, the border adjustment, banking and borrowing, coal baselines, road-versus-rail coverage, and the coverage threshold. It opens by stating that the review “is not intended to make fundamental reforms” and that changes will generally take effect from 1 July 2030.
Within that, the coverage threshold gets a short section and a clean table: drop from 100,000 tonnes to 75,000 and 48 more facilities come in, carrying 4.1 Mt of newly covered emissions; drop to 50,000 and it is 108 facilities and 7.8 Mt. The biggest increases, the paper notes, are in metal ore mining and oil and gas extraction.
Then, forty lines later, in a passage about how baselines would be set for newly covered facilities, comes this:
“Under either approach, the minimum baseline — currently 100,000 tonnes — would fall to be consistent with the new coverage threshold.”
Safeguard Mechanism Review Consultation Paper, August 2026, section 3.5.1That sentence is not costed anywhere in the paper. It is the largest number in the section.
The floor
Two provisions of the Safeguard Rules work against each other, and the interaction is not a drafting accident — it is deliberate, and it has consequences the review does not appear to have priced.
The obligation has a floor. A facility’s baseline is set to 100,000 tonnes if the number calculated under the ordinary rules comes out below 100,000.
National Greenhouse and Energy Reporting (Safeguard Mechanism) Rule 2015, s10(1)The reward does not. Credits are issued against the baseline that would have been calculated “if subsection 10(1) had not been enacted” — that is, against the unfloored number.
Same instrument, s56(4)The logic is sound. Without the floor, a shrinking facility’s baseline would chase its emissions down to nothing and the administrative cost of covering it would exceed the abatement. Without the asymmetry, a facility parked on the floor could bank credits for abatement it was already going to make. So the floor caps the obligation and the asymmetry cancels the reward.
What that produces, in practice, is a growing group of facilities for which the scheme has switched off in both directions. They face no declining constraint, because their baseline has stopped falling. And they earn nothing for abating, because between the floored baseline they must not exceed and the unfloored one credits are measured against, there is a band where abatement produces nothing sellable.
The register shows how fast that group is growing.
The trace is the modal case, not a named facility: a baseline that reaches exactly 100,000 tonnes in 2024–25 and is then held there. Forty-three facilities are on precisely that number today, so they all project to precisely 72,838 tonnes in 2029–30. Everything between the two lines is abatement that moves a facility further into compliance and produces nothing sellable.
Source: National Greenhouse and Energy Reporting (Safeguard Mechanism) Rule 2015, s10(1) and s56(4); emissions reduction contribution schedule 0.951 (2023–24) declining 4.9 points a year to 0.657 (2029–30). Analysis by Heliovulcan.
The two panels share a time axis but not a scale — a count and a tonnage are different measures and never belong on one pair of axes. The lower panel starts in 2025–26 because the held-back tonnage cannot be computed for the filed years: for a facility already on the floor, the register publishes the floored number and not the one behind it. The projection holds production and emissions intensity fixed. It does not model the separate convergence of facility-specific intensity toward the industry average, which would push more facilities onto the floor, not fewer — so 64 is a conservative count.
Show the numbers
| Year | Facilities on the floor | Baseline held above the path | Basis |
|---|---|---|---|
| 2023–24 | 33 | — | Filed |
| 2024–25 | 43 | — | Filed |
| 2025–26 | 46 | 0.266 Mt | Projected |
| 2026–27 | 53 | 0.537 Mt | Projected |
| 2027–28 | 55 | 0.833 Mt | Projected |
| 2028–29 | 60 | 1.153 Mt | Projected |
| 2029–30 | 64 | 1.509 Mt | Projected |
Source: Clean Energy Regulator published Safeguard baselines and emissions, FY2023–24 and FY2024–25 (227 facilities with a baseline in the later year). Analysis by Heliovulcan.
Nineteen of the 43 currently on the floor already emit less than their floored baseline — one of them under 300 tonnes against a 100,000-tonne allowance. They are inert: no obligation, no credit, no reason to invest, and no reason to leave.
And the cohort is not a random sample of the register. Of the 64 facilities that reach the floor by 2029–30, 20 are metal ore mines, 12 are coal mines and 7 are oil and gas facilities. Metal ore mining is the sector that stands out: it is 23 per cent of the register and 31 per cent of the floor cohort. Coal runs the other way — 30 per cent of the register, 19 per cent of the cohort — because coal mines are, on average, larger emitters and stay above the floor. What the floor catches is the small end of a big industry: the satellite pit, the ageing operation, the plant in care and maintenance.
Only the two movers are labelled, in percentage points. The rest sit within four points of their register share and are there for context; hover, focus or the table below gives every value. The eight sectors shown cover 56 of the 64 facilities.
Show the numbers
| Sector | Of the register | Of the floor cohort | Change, points |
|---|---|---|---|
| Metal ore mining | 23.3% (53) | 31.2% (20) | +7.9 |
| Coal mining | 30.0% (68) | 18.8% (12) | −11.2 |
| Oil and gas extraction | 12.8% (29) | 10.9% (7) | −1.9 |
| Non-ferrous metal manufacturing | 6.6% (15) | 7.8% (5) | +1.2 |
| Road freight transport | 1.3% (3) | 4.7% (3) | +3.4 |
| Electricity generation | 2.6% (6) | 4.7% (3) | +2.1 |
| Cement, lime and plaster | 3.5% (8) | 3.1% (2) | −0.4 |
| Pipeline and other transport | 1.3% (3) | 3.1% (2) | +1.8 |
Source: Clean Energy Regulator published Safeguard baselines and emissions, FY2024–25, classified by the register’s own industry field. Analysis by Heliovulcan. Percentages are of 227 facilities and of the 64 projected onto the floor in FY2029–30.
Why the threshold question is a floor question
Because the minimum baseline moves with the coverage threshold, the choice the department is consulting on is not only about admissions. It is also about whether the 64 start declining again.
The two bars are the same unit and not the same kind of thing — that is the point of the chart. The teal bar is baseline withdrawn: every tonne of it is an abatement obligation that lands the year the change takes effect. The orange bar is emissions brought inside the scheme, of which only the part above a newly issued baseline is ever an obligation — on the department’s own Option 2 for new entrants, close to none of it in the first year. Read that way, the smaller bar is doing more work than the larger one.
| Coverage threshold | Facilities on the floor in FY2029–30 | Baseline held above the declining path | New facilities admitted | Newly covered emissions |
|---|---|---|---|---|
| 100,000 t — current | 64 | 1.51 Mt | — | — |
| 75,000 t | 44 | 0.11 Mt | 48 | 4.1 Mt |
| 50,000 t | 0 | 0.00 Mt | 108 | 7.8 Mt |
| 25,000 t | 0 | 0.00 Mt | 291 | 14.2 Mt |
Source: Floor columns — Heliovulcan reconstruction from the published FY2024–25 Safeguard register, projected to FY2029–30. Admissions columns — Safeguard Mechanism Review Consultation Paper, August 2026, table 3.3. The two right-hand columns are the department’s; the two left-hand ones are not published anywhere. “Baseline held above the declining path” is the sum, over floor-bound facilities, of the difference between the statutory minimum and the projected unfloored baseline.
Read the rows against each other. Moving the threshold from 100,000 to 75,000 tonnes admits 48 facilities carrying 4.1 Mt of emissions — but those emissions are not abatement. A newly covered facility gets a baseline, and only the gap between its emissions and that baseline is an obligation. Under the department’s own Option 2, where new entrants start near the top of the decline curve, the obligation in the first year is close to nothing.
The floor column is different in kind. It is 1.4 Mt of baseline that stops being granted, to facilities that are already regulated, already reporting, and already inside the register. It requires no new production variables, no new administrative machinery and no lead time. It is the cheapest tonne in the section, and it is not in the table.
Three things have to be said about that 1.4 Mt, because they cut in different directions.
It is a lower bound, not an estimate. For the 43 facilities already parked on the floor, the published baseline of 100,000 is already the floored number — the true calculated baseline behind it is lower, by an amount the register does not disclose. Projecting 100,000 forward therefore overstates their unfloored 2029 baseline, which understates the held-back tonnage. The real figure is larger.
The step from 75,000 to 50,000 matters more than it looks. Because 43 facilities are parked on the identical number, they project to an identical number — 72,838 tonnes in 2029–30. A 75,000-tonne floor leaves them floored, just barely. A 50,000-tonne floor releases them entirely. The relationship between threshold and effect is a cliff, not a slope, and where the cliff sits depends on undisclosed data.
This is why the relationship between the threshold and its effect is a cliff rather than a slope. Forty-three facilities are filed at exactly 100,000 tonnes today, so they project to exactly the same number in 2029–30 — and they sit just under 75,000. A 75,000-tonne floor leaves almost all of them floored; a 50,000-tonne floor releases every one. No facility in the register projects below 50,000 tonnes. The 116 facilities above 150,000 tonnes are off the right of the chart and are nowhere near the floor.
Show the numbers
| Projected 2029–30 baseline, before the floor | Facilities |
|---|---|
| Below 65,000 t | 0 |
| 65,000–70,000 t | 3 |
| 70,000–75,000 t | 41 |
| 75,000–100,000 t | 20 |
| 100,000–150,000 t | 47 |
| 150,000 t and above | 116 |
| Total | 227 |
Source: Clean Energy Regulator published Safeguard baselines and emissions, FY2024–25, each facility’s baseline rescaled from its filed contribution factor to 0.657. Analysis by Heliovulcan. For facilities already at the statutory minimum the filed baseline is itself the floored number, so their position in this distribution is an upper bound — the true values sit further left.
None of this is an argument that lower is better. The department’s own reasoning against a 25,000-tonne threshold — 291 new facilities, 18 new industry groups, administrative cost outrunning abatement — is sound and this analysis does not touch it. The point is narrower: whatever threshold is chosen, its largest single effect in the first year will fall on facilities that are already covered, and that effect should be in the consultation table.
On-site abatement is a mining problem
Both the department and the Authority describe the same worry in almost the same words. Net emissions are falling fast — 7.4 per cent then 5.5 per cent across the two post-reform years. Gross emissions, the ones that actually come out of the stack, are falling at 1.9 and 2.4 per cent. The difference is being filled with credits. The Authority puts it plainly: material on-site abatement “appears limited to a few facilities”.
Neither paper says where. The register does.
Eight sectors are shown, not all of them; the tail below 150,000 credits is omitted and the totals below are of everything. Coal alone surrenders more than every non-resource sector on the register put together.
| Sector | Offset credits surrendered | Share of all surrender | Scheme credits surrendered |
|---|---|---|---|
| Coal mining | 5,050,594 | 47.9% | 1,351,561 |
| Oil and gas extraction | 1,941,035 | 18.4% | 646,868 |
| Metal ore mining | 1,562,804 | 14.8% | 299,698 |
| — three sectors combined | 8,554,433 | 81.1% | 2,298,127 |
| Non-ferrous metal manufacturing | 681,168 | 6.5% | 212,070 |
| Rail freight transport | 328,427 | 3.1% | 13,513 |
| Air and space transport | 218,307 | 2.1% | 0 |
| Ferrous metal manufacturing | 204,177 | 1.9% | 0 |
| All facilities | 10,549,152 | 100% | 2,582,112 |
Source: Clean Energy Regulator published Safeguard baselines and emissions, FY2024–25. Analysis by Heliovulcan. Shares are of offset credits surrendered; the largest sectors are shown, not every sector. Adding deemed surrenders brings the offset total to 10,831,682 — reconciling to the 10.8 million stated in the consultation paper, and putting the three-sector share at 79.0 per cent on that wider basis. The scheme-credit total reconciles to the paper’s 2.6 million.
Of the 24 facilities that surrendered offset credits equal to 30 per cent or more of their baseline — the threshold at which the regulator requires a written explanation of why more was not done on site — 12 are coal mines and 10 are metal ore mines. Twenty-two of twenty-four.
So the measures on the table — quantitative limits on credit use, discounting so that more than one credit is needed per tonne, vintage restrictions — are, whatever their stated framing, instruments aimed at mining. That is worth naming before the design is settled, because the assumption underneath all three is that the party holding the baseline can respond by investing on site. For a large and identifiable class of mines, it cannot.
The incentive and the owner
At an off-grid mine, on-site abatement mostly means one thing: changing what the power station burns, or putting solar and storage in front of it. That is the only lever of any scale, and at most sites the mine does not hold it.
We have gone through the Australian gold sector site by site — 88 operations carrying a Safeguard or pollutant-inventory identity, with station ownership, storage and contract expiry established from contractor project registers, state generation licences and company filings rather than assumed. None of those three facts appears in any national register.
Of the sites verified as having their own power station, 90 per cent do not own it. It belongs to a contractor — typically one of about seven fund-owned independent power producers — under a build-own-operate contract with a term measured in a decade or more. Two operations own their station outright.
The consequences run in a straight line:
- Every disclosed battery at those sites sits on the contractor’s asset. Fifteen of the operations have storage; in each case it was installed by the power provider inside an existing contract, not procured by the miner.
- The carbon and the asset are on different balance sheets. The mine reports the generation emissions and holds the baseline. The contractor owns the plant, chooses the fuel, and captures the fuel saving first.
- The date the decision reopens is almost never public. Contract expiry is established for nine of the 88 sites. For the rest the answer is not “no contract expiry is coming” — it is that nobody outside the two counterparties has looked.
This is the structure a credit price signal has to travel through. It reaches the left box, where the obligation is. The decision it is meant to trigger — change the fuel, add solar, add storage — is made in the right box, by a counterparty whose contract does not price carbon and typically has years left to run.
Source: Heliovulcan site-by-site build over 88 Australian gold operations carrying a Safeguard or pollutant-inventory identity, from contractor project registers, state generation licences and company filings. None of the three facts — station ownership, storage, contract expiry — appears in any national register. Percentages exclude 22 sites recorded as not yet checked; a blank storage field means no disclosure was found, never that no battery exists.
This is a split incentive, and it is not fixed by making offsets dearer. Making offsets dearer raises the cost on the miner and leaves the investment decision where it already sits — with a contractor whose economics improve when fuel is saved but who did not sign up to a carbon obligation. The Authority’s question 8 asks how on-site abatement should be measured; question 10 asks whether stronger incentives are warranted. The prior question is who the incentive lands on, and for this class of facility the answer is published nowhere and is knowable only site by site.
The same caution applies, for a different reason, to coal. Both papers raise ending the concession that lets existing coal mines keep half of their baseline on facility-specific intensity past 2030 while everyone else converges to the industry average. Doing so would raise the liability on gassy mines — correctly, on the scheme’s own logic. But of 102 Australian coal mines in our screen, none is off-grid, and the emissions at issue are ventilation air methane and haul fleet diesel. Tightening coal baselines creates a compliance cost. It does not create a generation-side market, and any business case built on the assumption that it does will be wrong by an order of magnitude.
Where a lower threshold would land
The consultation paper says the biggest increase in facility numbers under a lower threshold is in metal ore mining and oil and gas extraction. It does not say what those facilities look like operationally, because the department is counting emissions and we are looking at power stations.
Four hundred and thirty-three Australian mines sit outside the Safeguard register today. Two hundred and forty-five of them are operating. Of those, 98 are off-grid — 49 gold and silver operations, 19 iron ore, the balance spread across mineral sands, base metals and bauxite. A further 54 sit on a weak connection.
Whether any individual one crosses a 75,000 or 50,000-tonne threshold depends on emissions we cannot compute from outside, and this analysis does not claim to identify them. What it does establish is the character of the population: a lower threshold does not mainly admit factories. It admits remote sites that make their own electricity, where the abatement lever is a physical power station with a fuel bill and a contract, and where — on the gold evidence above — the odds are that the station belongs to a third party.
For anyone whose business is selling into that transition, that is the material fact. A threshold change converts a discretionary cost decision at an off-grid power station into a regulated one. It does not change the physics or the capital cost. It changes who has to have the conversation, and when.
An independent rebuild of the baseline calculation, checked against the department’s own published figures
The floor analysis is not read off a table — no table publishes it. It comes from an independent reconstruction of the Safeguard baseline calculation from the rule text and the published register: 227 facilities for FY2024–25 and 233 for FY2023–24, each facility’s baseline rescaled by the legislated emissions reduction contribution from 0.902 to 0.657, then tested against the statutory minimum at four candidate thresholds.
The reconstruction is validated by reproducing figures the department published from its own data. Offset credits surrendered come out at 10,831,682 against the paper’s 10.8 million; scheme credits at 2,582,112 against 2.6 million; coal mining at 68 facilities and 31.8 Mt against the paper’s “around a third of covered facilities” and 31.7 Mt. Facilities surrendering credits above 30 per cent of baseline come out at 24 against the paper’s 23 — a definitional edge on deemed surrenders, and the one number we do not match exactly.
The ownership evidence is a separate build and a slower one. Station ownership, storage and contract expiry exist in no national register, so each is either established from a named public source or recorded as unchecked. Forty-four of the 88 gold sites are verified; 22 are explicitly not yet checked and are excluded from the percentages rather than assumed. A blank storage field means no disclosure was found, never that no battery exists — behind-the-meter storage has no register in Australia, and reading a blank as a zero has put a site with nine megawatts of installed battery at the top of a target list before now.