Which CO2 the ETS Will Bill Your Incinerator For

Bolted to a waste-to-energy stack, a few platforms up, sits a heated cabinet about the size of a wardrobe. Inside are the analysers of the continuous emissions monitor. Its sample line runs hot, held above the flue-gas dew point so nothing condenses before the gas reaches the cells, and it has trended carbon dioxide every few seconds for years. Nobody paid for that CO2. It was a permit figure: measured, logged, reported, filed. An environmental line item with no invoice behind it.

Not for much longer. On 17 July 2026 the European Commission proposed folding waste incineration into the EU Emissions Trading System, and the same CO2 stream that cabinet already watches picks up a price.

Look closely and one detail sets how steep that price runs. The meter on the stack sees every molecule of carbon dioxide leaving the furnace. The scheme charges for only some of it. Carbon that arrived as plastic, synthetic fibre, or rubber counts as fossil, and each tonne of it costs an allowance; carbon that arrived as food, cardboard, paper, or garden trimmings counts as biogenic, and it surrenders at nothing. So the genuinely new instrument on the plant isn't a CO2 analyser. You already run one. It's whatever can tell a verifier, and hold up when that number is challenged, how much of this hour's CO2 came from a shampoo bottle rather than the apple core packed beside it in the sack.

The invoice only reads half the stack

Only the fossil share carries a cost. That one rule reorganises what's worth measuring on a plant that, until now, treated all its CO2 as a single reported quantity. A waste boiler's total carbon dioxide is comparatively tractable. Where it came from is the hard part, and it's the part the allowance account turns on.

The burn itself is a separate discipline, the per-shift fight to hold combustion steady on a fuel that shows up in a bin lorry, which I've walked through in keeping combustion efficient on a waste-to-energy boiler. This piece begins one step downstream of that. Not how completely the carbon oxidises, but what fraction of it was fossil to begin with.

That fraction is neither small nor steady. The fossil carbon in mixed municipal waste rides mostly in the plastics, and the plastic load swings with the bag: heavier after a supermarket delivery day, lighter once the spring garden collections come in. A plant that pins one number to the whole year is estimating the exact quantity it now pays against. Worse, the direction of the estimate isn't neutral. Guess high to stay safe and you hand over money for carbon the scheme was always going to wave through. And guess low, and a correction finds you later.

Turning a plant's own meters into a carbon figure an auditor will accept is ground I've covered before, for product-level footprints. The ETS narrows that question to a single point. Not a cradle-to-gate number for a customer, but one verified, fossil-only CO2 total the scheme bills against. The obligation phases in from 2031 and reaches the full 100% of verified emissions in 2034, with temporary opt-outs for member states running equivalent measures until the end of 2035, as the International Carbon Action Partnership reads the proposal. Coverage begins at installations burning more than 3 tonnes an hour, which is nearly every municipal plant. Six years of lead time sounds generous. It shrinks fast once you count what has to be instrumented, validated, and signed off before the first surrender falls due.

The proposal softens the blow in places without changing the measurement job. Waste-to-energy feeding a district-heating network can qualify for free allocation on part of its output, and auction revenue is meant to help local authorities move waste up the hierarchy toward recycling. Neither of those lets you off measuring. Free allocation is set against verified emissions; it lowers what you pay, not what you have to prove. You still need the fossil-CO2 number, produced the same defensible way, before any allowance or opt-out is netted off it.

Four ways to name the fraction

Four methods are in real use to put a number on the biogenic-versus-fossil split, and they don't cost the same, update at the same rate, or persuade a verifier with equal ease.

Standard factors. The rulebook lets you take a default biogenic fraction, apply it to your waste tonnage, and calculate the fossil CO2 from there. No new hardware. No sampling. You report a weight off the weighbridge, multiply, and the year's number is done. It's the cheapest route and the bluntest one: a national or default figure that knows nothing about what your particular catchment set out at the kerb this month.

Composition sampling. Crews pull graded samples of the incoming waste, hand-sort them into material classes, and send fractions off for lab work. Done carefully, it's a genuine look at the feed. But at any affordable frequency, it's a few snapshots a year of a stream that turns over by the truckload, and the sampling bias (what a few hundred kilograms grabbed off the tipping floor really stand for) is hard to bound. Change a commercial waste contract and last quarter's sort no longer describes the fuel.

Radiocarbon. Living matter takes up carbon-14 from the atmosphere; fossil carbon, buried for millions of years, has none of it left to detect. Measure the 14C in the flue-gas CO2 and the biogenic fraction reads straight off the physics, with no assumption about waste composition at all. This is the reference method, standardised as EN ISO 13833, and nothing else argues with it on accuracy. Its constraint is cadence. You capture a sample over hours or days, ship it to a laboratory, and wait for the count. The sampling itself is a small engineering job: CO2 drawn off the duct and fixed into an absorber over a window long enough to average the swing between a plastic-heavy morning and a quiet afternoon, then counted by accelerator mass spectrometry or liquid scintillation back at the lab.

The balance method. Rather than touch the waste or the isotopes, you solve the plant's own books. Flue-gas flow, oxygen, CO2 concentration, steam raised, feedwater, combustion air: feed those into a set of mass and energy balances, standardised as ISO 18466, and the biogenic fraction falls out. No new penetration in the stack, no sample bottles. The inputs are signals the control system already logs, so the answer can refresh continuously, every hour or better. The price of that is trust in the instruments, because the method inherits every drift and bias in them, and it still wants a radiocarbon campaign now and then to prove it's honest.

The feed won't hold still, so the number shouldn't either

A figure set once a year and a figure that moves every hour are different kinds of asset when the feed behaves like municipal waste. The same heterogeneity that makes combustion a per-shift fight drags the fossil fraction around under any annual average. A standard factor freezes it. So does once-a-year sampling, at a value the plant then departs from within the week.

Why should cadence matter for a bill you settle annually? Because a continuously tracked fraction, integrated across every real tonne actually burned, lands closer to the truth than one snapshot scaled up to a year. Say your plastic content ran high in the same months your throughput ran high. An annual average measured in a quiet week understates what you truly emitted, and a verifier is within their rights to say so. The balance method, refreshing each hour from the control system, weights the fraction by real operation. Radiocarbon nails the truth for its sampling window and no other. Read that way, the two stop looking like competitors and start looking like halves of one instrument: continuous coverage from the balances, periodic ground truth from the isotopes.

Precision earns its keep at the allowance desk

Every tonne of fossil CO2 you can't pin down cleanly becomes a tonne you either overpay on or get corrected for. Set a conservative default high enough to feel safe, and you buy allowances against biogenic carbon that never owed anything. Money surrendered for potato peel. Set it low, and the verifier's correction, plus the extra scrutiny that trails one, costs more than the precision would have. But the blunt method isn't free. Its price just hides inside the safety margin you carry all year.

Sharpening the fossil fraction, from a padded default down to a validated, plant-specific figure, lifts nothing extra off the stack and burns no more fuel. It simply stops you paying for carbon the scheme intended to let go free. On a plant sized well past the coverage threshold, the fossil tonnage is large, and a few points of avoidable error on a large number is a line item worth an instrument of its own. The cheapest method on the purchase order can turn out to be the dearest on the annual allowance statement.

There's a quieter benefit too. A fraction you can defend with your own measurements is one you can also explain when it moves. Take on a new contract heavy in commercial plastic film, and the number climbs for a reason you can name. That traceability is worth as much at an assurance meeting as the accuracy itself, because the question that ends most of those meetings is simply where a figure came from.

What a verifier will actually sign

Accuracy alone doesn't close the loop. The number also has to be one the scheme's rules recognise and an accredited verifier will put a signature under. The EU's Monitoring and Reporting Regulation frames two ways to reach the total CO2 in the first place. You can calculate it, from activity data times a carbon content and an oxidation factor, or you can measure it directly at the stack, continuous CO2 concentration against flue-gas flow through the CEMS. For garbage, the calculation route hits the same wall everything else does here: you can't spec the carbon content of a fuel that arrives mixed in a lorry. So many plants lean toward the measurement-based total, and that wardrobe-sized cabinet ends up doing more work than the permit ever asked of it. The measurement route has metrology of its own to defend, mind. Turning a concentration into a mass needs the stack-gas flow, and that flow rests on a clean velocity profile in a wide, wet duct, so it draws the same verifier scrutiny as the analyser upstream of it.

On the biogenic split, the pecking order is clearer than it first appears. Radiocarbon is the reference, which makes it the thing everything else gets checked against; EN ISO 13833 exists precisely so that measurement is comparable from one lab and one plant to the next. The balance method is a recognised path to a continuous fraction, but an ISO 18466 result carries weight with a verifier when a radiocarbon campaign stands behind it, not in place of one. And a learned model (a soft-sensor estimating fossil CO2 from the plant's own history) is not a standalone method the rulebook knows. It earns a place only riding on top of the recognised two: the radiocarbon runs as its calibration truth, and full lineage from every input signal through to the reported tonne. Strip the lineage away and you have a confident number no one can trace. Worthless on surrender day.

The upkeep you're signing up for

None of these numbers survive on their own. A radiocarbon result is only as good as the sampling that fed it: an isokinetic probe, a representative extraction over a real operating window, sample bottles handled and labelled without a mix-up between the stack and the lab bench. Get the sampling wrong and the reference method quietly stops being a reference.

The balance method shifts that burden onto the instruments you run every day. Its answer leans on the flue-gas flow, the oxygen reading, the CO2 concentration, and the steam and feedwater figures, so a zirconia oxygen cell drifting two-tenths of a percent, or a flow element fouling in the duct, bends the computed fraction without tripping any obvious alarm. The discipline is the one good combustion crews already keep: heated sample lines, scheduled reference-gas checks, and a habit of cross-checking one instrument against another rather than trusting a single reading. All of it lands in a monitoring plan the verifier reads before they read a single emissions figure, with the calibration records and the sampling dates kept on file to stand behind the number.

Standard factors need almost no upkeep, which is at once their appeal and their weakness. Nothing to maintain, because nothing is being measured. You've traded the calibration chore for a permanent guess, and on a plant with real fossil tonnage that's an expensive trade to make quietly.

Set the four side by side against the things that actually decide the choice, and the trade-offs line up:

Which method fits your plant

The honest starting point is size, feed, and how far the fraction actually travels. A small plant with a steady, well-characterised feed and modest fossil tonnage may find a cautious standard factor is simply cheaper than the apparatus of measuring it: the overpayment is genuine but small, and the compliance is close to trivial. Raise the throughput and the variability, and that trade turns over quickly.

For a mid-to-large mass-burn plant on a swinging municipal feed, the workable shape is a pairing. Take the total CO2 from the CEMS on a measurement basis, take a continuous biogenic fraction from the balance method, and run a radiocarbon campaign on a set cadence to anchor it. The continuous number keeps you honest hour to hour and weights the fraction by real throughput; the isotope runs keep the continuous number honest in turn. That pairing is also the one a verifier signs most readily, because each half covers the other's blind spot.

A learned model belongs on top of that pairing, not in place of it: fusing the CEMS, the balance calculations, and the periodic radiocarbon anchors into one fossil-CO2 figure that updates live and carries its own audit trail. That's the work of deploying and validating that kind of model on a live plant, and it only pays once the measurements beneath it are sound. Sit a model on badly calibrated instruments and it just returns wrong answers faster, dressed in a confidence harder to catch than an obviously crude default.

So which one fits your plant? A defensible fossil-CO2 number, produced at a cadence that matches how much your feed moves, by whichever combination your verifier will sign. For most installations above the threshold, that means measured totals, a continuous fraction, and isotopes to anchor it. The default factor is where you start, not where a plant with real fossil tonnage should end.

Notes

The scope and timing figures here are the European Commission's 17 July 2026 proposal as summarised by the International Carbon Action Partnership; a proposal is not yet law, and the threshold, the phase-in, and the opt-outs can all move through the Parliament and Council before adoption. The method comparison is general engineering judgement, not a compliance ruling, and it has limits: not every plant lands in the same place. What a given national authority and accredited verifier will accept for your specific installation is theirs to state, and the recognised route wins every tie. Radiocarbon, the balance method, and default factors are described as they're used across the sector; the right mix for any one plant depends on its size, its feed, and the instrumentation already on the stack.