Fuel life-cycle assessment: where the calculation actually becomes contestable

A scientist analyzing a fuel sample in a laboratory with an industrial refinery in the background at dusk

The same batch of used cooking oil, two carbon-intensity certificates, and values that do not overlap. The gap is not a calculation error: it comes from three methodological decisions taken before the first gram is counted, each defensible on its own, and whose combined effect is enough to push two results apart by a factor close to two. Here is where those decisions sit, and how to spot them on a certificate.

What is being compared, and in which unit

Megajoules compare only the energy actually carried through to the engine. The chain then breaks down into two unequal halves. From well to tank, one adds up the production of the feedstock, its transport, its conversion, distribution and fuelling. From tank to wake, only one thing happens: the carbon it contains comes out as carbon dioxide, and that portion is not negotiable. It is not, however, strictly identical from one fuel to another: a paraffinic product, richer in hydrogen than a kerosene loaded with aromatics, releases slightly less carbon dioxide for the same energy delivered. The gap stays narrow compared with those that follow, but it is a reminder that this portion is calculated from composition rather than decreed. Everything that is debated lies upstream. Added to this is a convention for equivalence between gases, whose chosen time horizon makes pathways where methane leaks heavier or lighter.

The carbon footprint of fuel is measured in energy units, not volume or weight. The biggest differences come from upstream processes, not the combustion itself.

An Airbus A320neo burning 2,500 kg of sustainable aviation fuel (SAF) at 30°C will emit 7,200 kg of CO2 if the SAF is from waste oils, but 8,500 kg if it’s from palm oil – a 1,300 kg difference from the same fuel type.

Well-to-wake chain and position of the boundariesFive stages aligned left to right: cultivation and collection, upstream transport, conversion and refining, distribution and fuelling, in-flight combustion. An upper brace spans all five stages and reads well to wake; a lower brace stops at the fourth stage and reads well to tank. Two notes flag the usual exclusions.well to wakeCultivationcollectionTransportupstreamConversionrefiningDistributionfuellingCombustionin flightwell to tankbiogenic CO2 countedas zero by conventionnon-CO2 effects (contrails, NOx): out of scope
The debate takes place entirely upstream of the aircraft. Combustion releases the carbon with no room for interpretation; it is the upstream that varies. The usual boundaries also exclude capital goods.

The boundaries: what is decided not to count

Three exclusions shape almost every framework. Capital goods first: the plant, its reactors, agricultural equipment, ships, excluded on the grounds that they are amortised over considerable volumes. The argument holds for a refinery running for decades, much less for a pilot unit that will produce little.

The second is the heaviest exclusion for aviation: non-carbon-dioxide effects are out of scope. Contrails, nitrogen oxides, particulates, none of that enters a fuel’s carbon intensity, even though these terms weigh heavily in the forcing attributed to air transport. The paradox is complete when a fuel low in aromatics and sulphur reduces the emission of non-volatile particulate matter, and therefore the number of ice-nucleating particles available to form a contrail: this gain appears nowhere on the certificate.

The third is an accounting convention: biogenic carbon dioxide is counted as zero at combustion, in exchange for the uptake counted at cultivation. The symmetry is clean for a dedicated crop; it becomes asymmetric for a waste, for which neither the upstream uptake nor the combustion is counted. This is consistent with the rule, it is not a measurement.

Co-product allocation, the first lever

No production makes kerosene alone. Hydrotreating of oils and fats also yields naphtha, renewable diesel, propane and tail gases; Fischer-Tropsch synthesis produces a range of cuts to be fractionated; an alcohol-to-jet pathway leaves heavy fractions. All the process burdens, and those inherited from upstream, must be allocated among these outputs.

Four co-product allocation methods, and what each one shifts
Method Allocation key What it serves well Its weakness
Energy Energy content of the outputs Reproducible, independent of markets Treats a tail gas burned on site as a valuable product
Mass Mass of the outputs Auditable, data always available Burdens heavy cuts regardless of their use
Economic value Revenue from the outputs Reflects the industrial intent of the site Follows prices: a batch changes intensity from one quarter to the next
System expansion Credit for emissions avoided at the displaced product Comes closer to the real effect on the system Requires designating a displaced product; can yield a negative intensity

These four methods are not ranked by increasing rigour: they are four readings of the same plant. The energy method is the most reproducible, which explains its regulatory favour. The economic method fits industrial intent better but makes a batch’s intensity dependent on the quarter’s price ratio. System expansion comes closest to causal reasoning and is the most dangerous: it requires designating a displaced product, and can push the intensity below zero without anything physical having changed.

The status of the feedstock, the second lever

Feedstock classification precedes every calculation. The answer does not change a single line of the process, but it decides what share of the upstream burdens the fuel must carry.

Feedstock status and the share of upstream burdens allocatedA staircase rising from left to right. On the horizontal axis, four statuses for the incoming feedstock: waste, residue, co-product, main product. On the vertical axis, the share of upstream burdens allocated, from none to majority. The step rises by one level at each change of status.share of upstream burdens allocatednonelowsharedmajoritywasteresidueco-productmain productfeedstock status
The classification of the feedstock weighs more heavily than any process optimisation. Moving from co-product to waste wipes out the upstream agricultural burdens without changing a single valve. The jump happens in steps, not continuously.

Classified as waste, a feedstock arrives at the gate carrying only the burden of its collection and transport: cultivation, fertiliser, irrigation, harvesting, all of it disappears. No yield gain offers an equivalent effect. The battle is therefore fought over classification far more than over engineering, and the incentive to reclassify virgin oil as used oil grows exactly with the intensity gap between the two.

What holds this classification together is not the calculation but traceability: a mass balance documented at every link, audits at the collector’s premises, consistency between declared volumes and the population actually served. A framework that does not go down to the collector certifies nothing but a declaration. What remains, almost always absent, is the counterfactual: a used oil that used to go into animal feed does not become available without consequence, and residues left in the field maintain a carbon stock that removal depletes.

Land use, the third lever

Direct land change can be calculated: the carbon stock of soil and biomass before and after cultivation begins is compared, and the difference is amortised over a conventional period set by the framework. The difficulty is not the calculation but the reference state. What was the plot before: permanent grassland, fallow, degraded forest, land already cultivated? The result swings with the answer, and the answer is a matter of records more than measurement.

Indirect land change, by contrast, cannot be measured. It models what happens elsewhere when a piece of land stops supplying the food market: what land is brought into cultivation in compensation, in which country, on what carbon stock. This requires an economic equilibrium model, elasticities, an assumption about the yield response and a reference scenario. Two equally well-documented models give different orders of magnitude. Hence the so-called low-risk schemes, which seek to demonstrate additionality – intermediate crops, yield gains beyond trend, abandoned land brought back into cultivation – and which still rest on a counterfactual. The key point to retain: a framework that does not include this term is not saying it is zero, it is saying it does not count it.

Where the factor of two appears

Three methodological choices and eight defensible resultsA tree starts from a single point on the left and splits three times: first by the choice of boundaries, then by the allocation method, and finally by the treatment of land use. The eight endpoints range from the lowest to the highest result, and a brace on the right marks a gap of roughly double.lowest≈ ×2highestthe same batchboundariesallocationland use
Three binary decisions are enough to produce eight equally defensible figures. None of them is fraudulent: each is justified within its own framework. It is their combination that opens the range.

Three binary decisions taken in cascade produce eight results, all defensible, for the same batch out of the same plant. For an agricultural pathway, the gap between the low and high bound can approach a factor of two. For a waste pathway it narrows, since the agricultural upstream has vanished from the calculation – but that is precisely where the status of the feedstock carries the whole weight, and where fraudulent reclassification becomes most profitable.

The trap. Lining up in a single table intensities coming from different frameworks, or from different editions of the same framework, and drawing a supplier ranking from it. The gaps there measure the method, not the performance: a pathway can look better simply because its certificate excludes indirect land-use change. A ranking only makes sense at identical framework, version and boundaries.

Synthetic fuels shift the problem

Electrolysis fuels transfer all burdens to two inputs. Electricity first. Depending on whether one uses the grid’s average mix, the marginal mix, or a power purchase agreement backed by an identified plant, three very different intensities emerge for the same facility. Two requirements underpin the credibility of the third option: temporal correlation, producing when the plant produces rather than relying on an annual average, and additionality, a plant built for this use rather than diverted from another.

Carbon next. Captured from a fixed source, it is cheap but raises a question of ownership: does the reduction belong to the industrial operator who captured it, or to the fuel that uses it? The no-double-counting rule is a negotiated convention, not a physical property. Captured from the air, it removes the ambiguity but costs far more energy, which feeds back into the first term. Since every conversion has an efficiency, the intensity of a synthetic fuel is essentially that of its electricity divided by the efficiency of the chain.

What an operator can actually verify

Operators check that they have been handed a readable and consistent document, and keep what is needed to justify it. Two declaration regimes coexist, and they must be told apart at a glance.

Two engineers in hard hats and safety vests review a document in a control room with multiple monitors.
Data-driven decisions are critical for accurate fuel life-cycle assessments.
Default value and calculated value: two declaration regimes
Criterion Framework default value Calculated and verified value
Origin of the figure Scale published by pathway and by feedstock Real site data, link by link
Level adopted Deliberately conservative As accurate as the data allow
Burden of proof The feedstock’s membership of the category Every item, every flow, every assumption
Exposure to method choices Fixed by the framework Full: boundaries, allocation, land
Records to keep Certificate and feedstock traceability Report, source data and verification

Two reading traps close in even before these questions of method. The first is the declaration base: a certificate often covers the pure synthetic component, whereas what goes into the tank is a blend with fossil-derived kerosene. Carrying the pure component’s intensity over onto the fuelled volume credits the operator with a reduction it has not bought; the conversion must go through the proportion actually blended in, certified batch by batch. The second is the displayed reduction percentage: it is calculated against a fossil reference value set by the framework, and two frameworks do not use the same one. At identical intensity, changing the comparator changes the percentage – two reductions expressed as percentages can only be compared if the reference is the same too.

The type of operation then shifts where control is centred. A network operator that fuels most of its departures at a handful of hubs negotiates a physical supply and mainly checks consistency between volumes delivered, volumes uplifted and volumes declared. A dispersed operator – business aviation, aerial work, a highly branched regional network – most often buys through book-and-claim, unable to fuel wherever it lands: its risk is not the measurement but the registry and the uniqueness of the retirement. The same certificate is therefore not audited the same way depending on the network’s structure.

Next comes the chain of custody. Mass balance follows physical accounting along a real chain. Book-and-claim detaches the environmental attribute from the molecule: the operator buys a reduction produced elsewhere and puts locally available fuel into its tanks. This is not a trick, it is what allows purchasing where fuelling is not possible, but it only holds if a registry retires the attribute exactly once. Without an auditable registry, the same tonne gets resold.

The right reflex. On any certificate received, locate six elements before taking the figure at face value: the framework and its version, the declared boundary, the allocation method, whether or not indirect land-use change is included, whether the value is a default or a calculated one, and the nature of the chain of custody. If any one of the six is missing, request it on receipt of the batch, not at the time of the audit.

What this changes before an audit or a qualification

Four things must be locatable without hesitation: the emissions declaration procedure and the framework it applies, the monitoring plan and the measurement points it imposes, the person who commits the operator by signing the certificate, and the retention period for the records. These are the four points a verifier opens first, never the calculation itself.

The grey area does not close for all that: non-carbon-dioxide effects remain out of scope even though they weigh heavily, indirect land-use change remains a matter of modelling, and nothing in physics prevents the same attribute from being claimed twice if registries do not communicate. These gaps are discussed openly on the FL330 technical forums.

For anyone preparing a qualification or an interview, what is expected is not to quote a figure: it is to be able to say which allocation method the operator applies, whether indirect land-use change enters the calculation, and what distinguishes a physically delivered batch from a purchased attribute. These three statements are worth more than a recited number.

One final, obvious point that becomes costly when forgotten: the applicable regulatory text and the operator’s approved documentation prevail over this article, including when the imposed framework produces a less flattering figure than a supplier’s.

Questions fréquentes

Why can two certificates covering the same fuel display different intensities?

Because a carbon intensity is not a measurement but the result of a calculation framed by a reference standard. The choice of boundaries, the co-product allocation method and the treatment of land use are each set differently from one framework to another, and each of these choices shifts the result. Two figures can therefore only be compared if they come from the same framework, the same version and the same boundaries.

What is excluded from a well-to-wake assessment?

Three exclusions are almost universal. Capital goods, meaning the plant, agricultural equipment and ships, excluded on the grounds of their amortisation over large volumes. Then non-carbon-dioxide effects: contrails, nitrogen oxides and particulates. Finally, by convention, biogenic carbon dioxide emitted at combustion, deemed offset by the uptake counted at cultivation.

Why does the status of the feedstock weigh so heavily on the result?

Because it decides what share of the upstream burdens the fuel must carry. Classified as waste, a feedstock brings only its collection and transport: cultivation, fertiliser, irrigation and harvesting disappear from the calculation. No process gain offers a comparable effect, which explains why most controls focus on the traceability of the feedstock rather than on the plant.

Is indirect land-use change a measured figure?

No, it is a modelling result. It describes what happens elsewhere when a piece of land stops supplying the food market, which requires an economic equilibrium model, price elasticities, an assumption about the yield response and a reference scenario. Two equally well-documented models produce different orders of magnitude. A framework that does not include this term is not assuming it is zero: it is choosing not to count it.

Is a book-and-claim purchase equivalent to a physically delivered batch?

For declaration purposes, it all depends on the framework and the registry. Book-and-claim detaches the environmental attribute from the molecule and allows the purchase of a reduction produced where fuelling is not possible, which is legitimate. The condition is that an auditable registry retires the attribute exactly once; without that, nothing physically prevents it from being claimed twice.

Is a synthetic fuel automatically less carbon-intensive?

No. For an electrolysis-and-synthesis pathway, the product’s intensity is essentially that of its electricity divided by the efficiency of the whole chain, since every conversion costs energy. Depending on whether the grid’s average mix, the marginal mix or a power purchase agreement backed by an identified plant is used, with or without temporal correlation and additionality, the same facility can display very different results.

Commentaires

Leave a Reply

Découvrez l’écosystème FL330

FL330, c’est toute l’aviation professionnelle — explorez nos autres espaces.