Aviation's 2.5%

The four figures above describe the same activity and cannot be compared with one another: the first counts tonnes, the second watts per square metre, the third degrees, the fourth take-offs. Almost every discussion about aviation and the climate runs aground exactly there, in the move from one to the next. The simplest one comes first.

In 2024 the world emitted about 43.2 billion tonnes of CO₂ counting fossil fuels, cement and land use — four times the 1940 figure, and a record. Aviation, on the same scale, is the line that looks as if it were resting on the axis.

So far, the reading that closes the argument. But the same line, divided by the total instead of set beside it, tells the other half: aviation's share went from 0.32% in 1940 to 2.47% in 2019, while almost every other sector lost relative weight over the same stretch.

A small share does not mean a sector standing still. It means a small sector inside an enormous total — one that grows faster than the total does.

The denominator is worth looking at closely, because it is the first source of confusion: here the total includes land use. Against fossil fuels alone, the same 2019 share becomes 2.8%. And if you look at the perimeter of the numerator, the same question has at least three answers: commercial passenger flights alone, as counted by the OECD, come to 863 Mt in 2019, while the Lee and Bergero series, which also covers freight, charter and general aviation, counts 1,037 Mt. Neither is wrong: they count different things. Those same three answers turn up, with the same air of contradiction, in any public argument on the subject.

The obvious question, in front of the previous chart, is why the share rises precisely while aircraft get more efficient. The answer is a product of two factors moving in opposite directions, and one of them moves harder.

Between 1990 and 2019 the CO₂ per passenger-kilometre fell from 357 to 157 grammes: −56%, a real technical achievement, delivered by engines, aerodynamics and above all fuller aircraft. Over the same period passenger-kilometres went from 1,964 to 8,597 billion: ×4.4. The product, which is the emissions, nearly doubles: ×1.9.

This is why the two usual camps can both cite true figures and reach opposite conclusions. «Every flight pollutes half as much as thirty years ago» and «flying pollutes twice as much as thirty years ago» are both correct statements about two different quantities. The second is the one the atmosphere sees.

Here the accounting changes in kind, and this is where aviation parts company with other sectors. An aircraft does not emit carbon dioxide alone: it emits nitrogen oxides at altitude, water vapour, and particles around which, in cold saturated air, trails form that can turn into cirrus and last for hours. The standard way of adding up effects that different is effective radiative forcing, in milliwatts per square metre.

In the 2018 balance CO₂ is worth 34% of the total: 34.3 of the 101 mW/m² net. Contrail cirrus alone is worth more than half. Aviation's total is about 3.5% of anthropogenic forcing — against the 2.5% of the tonnes alone — and Klöwer and colleagues, applying the same accounting to temperature, attribute to flying about 4% of the observed human-caused warming, on the order of 0.04 °C. Aviation's cumulative emissions since 1940 are 33 Gt, roughly 2% of the world's cumulative total.

Two warnings, which matter more than the figures. First: 2.5% and 3.5% neither add up nor replace one another — they are percentages of two different quantities, and swapping them produces numbers that sound stronger and mean nothing. Second: the non-CO₂ effects carry far wider uncertainty than CO₂ — the published interval for the total runs from 55 to 145 mW/m², at 5-95% — and they do not persist in the same way. A tonne of CO₂ acts for centuries; a contrail is gone by morning. Comparing them requires choosing a time horizon, and the choice changes the answer: that is why the European rules on monitoring non-CO₂ effects prescribe publishing several horizons — 20, 50 and 100 years — rather than a single number. At 20 years contrails dominate the comparison; at 100 CO₂ takes it back. Neither horizon is «the right one»: it is a choice, and it has to be stated.

The ECAC area — 44 states, the Europe of air traffic control rather than the Europe of the Union — is the perimeter where the question can be asked flight by flight instead of through aggregate estimates. In 2025 it passed its 2019 levels for the first time.

The return to pre-pandemic levels happened in the number of flights, not in the emissions. Comparing 2025 with 2019, EUROCONTROL measures the same traffic and more CO₂: sectors have grown longer, turboprops have been replaced by regional jets, long-haul has come back.

In Europe the same activity weighs more — about 4.4% of total emissions — and not because Europe flies worse: it is the rest of the European economy that has decarbonised while flying has not, so the same slice takes up more room in a smaller pie. The share is set to keep rising for the same reason: the sector has fewer levers than the others, and its main ones — sustainable fuels, fleet renewal — are slow. In EUROCONTROL's plan for 2050, operational measures — routes, profiles, traffic management — account for 10% of the effort. Ten per cent of an enormous effort is still a great deal.

If the technological levers are slow and demand keeps growing, what is left to look at in the short run is how far each flight sits from a reference built on the same journey. On this, in Europe, there are two estimates: they use different references and measure different things, and the difference between them is not a hidden margin.

The first is EUROCONTROL's. In the Performance Review Report 2025 the Performance Review Commission estimates the air-traffic-management «benefit pool» empirically at about 9%, comparing each flight with the tenth percentile of comparable flights — that is, with what somebody else, on that same sector and in that same aircraft type, actually managed to do. It is a measure of plausibly recoverable margin, because its reference is a flight that existed.

The second is this site's, and it means something else: it compares each flight with an ideal flight — same aircraft type, great-circle route, optimal altitude and speed profile, the same real wind — and finds, across 1,833,127 ECAC flights over 197 days, 12.1% more CO₂, of which 7.5 points of route and 4.6 of profile. The two figures are not in contradiction: the first measures what could be recovered, the second how far a theoretical limit sits from every real flight, because a real flight has to respect separation, route structure, closed airspace and arrival queues, among many other operating conditions. That list explains why the limit is out of reach; it is not a breakdown of the 12.1%, which the model does not attribute to any of those causes in particular.

In one line: the 9% measures what could realistically be recovered by comparing real flights with one another; the 12.1% measures the distance from a physical limit no flight can reach. It follows that the difference between them is not 3 points of inefficiency the institutions are ignoring: it is the part of the gap that exists because the sky is full of other aircraft. Subtracting one figure from the other produces no quantity that means anything.

The value of the gap is not saying how much is wasted. It is saying where it accumulates: in which phase of the flight, on which routes, with what regularity.

«Where» should be taken literally. A high figure on the flights that touch an airport describes those flights — the shape of the airspace, the procedures, the queues waiting for them — and not the conduct of the airport, which decides almost none of those queues. The findings read that way throughout, and the method says where it is soft.

• It does not say flying is the main problem. 2.5% is still 2.5%: electricity, industry and heating remain orders of magnitude larger.
• It does not say 3.5% replaces 2.5%. They are percentages of different quantities, with different uncertainties, and each should be quoted with its metric attached.
• It does not estimate how much CO₂ could be avoided. The gap this site measures is the distance from a theoretical physical limit, not an operational recovery margin. For that, the published reference is EUROCONTROL's 9%.
• It does not cover air freight separately, nor compare flying with other modes for the same journey: both are legitimate questions that need different data from these.
• It is not a primary source. Every figure here except this site's own comes from the publications listed below; where sources diverge, the divergence is flagged rather than averaged away. The full list, with the date each was verified, is published as context-sources.json.

Every figure on this page that is not this site's own was verified against its source on 2026-08-20, and is re-checked at each release. Unlike everything else here, these numbers cannot be recomputed from the published data: they come from other people's publications, and the machine cannot tell when they go out of date.