<!-- ARTICLE_BODY_START -->
The two percentages describe different achievements. In an author manuscript posted on 6 March 2026, Tharun Sankar and colleagues reported an 11.6 percent reduction in observed contrail distance per flight distance within the observation region for 1,232 flights eligible for an avoidance intervention, compared with controls. For the 112 flights that actually followed the avoidance plans, the reported reduction was 62 percent. The trial integrated forecasts into airline flight planning; its randomisation assigned city pairs to treatment or control in successive periods. The larger number is a result for a selected operational subset, not the reduction achieved across all eligible flights. The trial’s March 2026 author manuscript
This gap between an available option and an executed flight is central to contrail mitigation. A forecast must identify an atmospheric opportunity. A planner must find a feasible alternative. The aircraft must be able to fly it. Observations must establish what happened, and a climate calculation must assess whether the change was beneficial after accounting for additional fuel and any replacement contrails.
None of those steps is captured by the phrase “fly around the clouds.” The cloud of interest may not exist when the route is planned. It may be invisible in the instrument used to verify the flight. It can move and spread after the aircraft has departed. And reducing the amount of visible white line is not necessarily proportional to reducing its warming effect.
The attraction is nevertheless substantial: operational changes could address a short-lived climate influence using aircraft that already exist. The challenge is to make that opportunity measurable without treating every successful avoidance as a complete climate account. The relevant evidence now includes field observations, flight trials and increasingly detailed models, but those forms of evidence answer different questions.
Before there is a cloud to avoid
Contrails are ice clouds associated with aircraft exhaust in sufficiently cold conditions. The FAA explains that their formation and persistence depend on both the atmosphere and the exhaust, including fuel and engine characteristics. Some disappear relatively quickly; others persist and spread into contrail cirrus. A line that ends abruptly can reflect a change in the surrounding conditions rather than an aircraft deliberately changing what it emits. FAA’s contrail explanation
The formation calculation is often expressed through the Schmidt–Appleman criterion. In Ulrich Schumann’s documented contrail model, this asks whether mixing hot, moist exhaust with cold ambient air produces local saturation with respect to liquid water, allowing condensation and subsequent freezing. Ambient temperature and pressure matter, but so do the water and heat associated with the exhaust and the aircraft’s propulsion efficiency. It is not simply a rule that every aircraft above a single altitude makes a trail. Schumann’s contrail model and formation criterion
Two questions must then be separated: can the initial ice cloud form, and can it persist? Relative humidity with respect to ice compares ambient water vapour with the equilibrium amount over ice at the local temperature. An ice-supersaturated environment can support continued growth of ice. That condition is not identical to saturation over liquid water. Using a humidity percentage without identifying its reference can therefore obscure the physical question being asked. Ice supersaturation and its meteorological representation
A deliberately simplified example shows the importance of the distinction. Suppose an air parcel contains 110 units of water vapour where the equilibrium amount over ice is 100 units. Its relative humidity with respect to ice is 110 percent. If an estimate instead represents that parcel as containing 99 units, it falls on the other side of the persistence threshold. These are invented amounts, not measurements at a flight level, and the example does not model the time required for crystals to grow or disappear.
The threshold does not turn the atmosphere into neat horizontal shelves. A flight samples a path through space and time. Conditions at one point do not establish conditions several minutes later along that path, nor conditions in a nearby layer. A single reported humidity value can be useful while still being insufficient to describe the region a proposed trajectory would cross.
This is why a contrail-aware route needs meteorological information at the scale relevant to the decision. A forecast that captures the broad weather system can still place a thin favourable layer incorrectly. Conversely, a local discrepancy need not invalidate every possible avoidance route. The useful test concerns the paths being compared, not whether a map is entirely right or entirely wrong.
Following the ice after the aeroplane has gone
Schumann’s model follows contrail segments as they move with the atmosphere and evolve, rather than leaving them attached to the original flight track. It represents processes including mixing, growth and loss of ice, and eventual disappearance through sublimation or precipitation. This lifecycle perspective is essential: a newly formed line and its later spreading cloud are stages of the same atmospheric disturbance, not independent emissions to be counted twice.
Consider a purely geometrical example. If a cloud segment moves with a steady horizontal wind of 20 metres per second for thirty minutes, it travels 36 kilometres. That calculation assumes constant wind and ignores deformation; it is not a prediction of an actual contrail. It explains why a later satellite image should not be compared with the aircraft’s earlier ground track as though both were drawn on the same stationary sheet.
The cloud can also change shape while it moves. A useful mental picture is a ribbon carried by a flow whose speed and direction vary across the ribbon, not smoke fixed behind a moving vehicle. That picture is an explanatory analogy, not a substitute for a microphysical model. It highlights why a forecast needs the atmospheric history after formation as well as the conditions encountered by the aircraft.
Particle number is another quantity that cannot be substituted directly for cloud effect. As a mathematical illustration, take a fixed mass of ice divided into identical spheres. Reducing their number by a factor of eight while holding total mass constant doubles each sphere’s radius. Their combined geometrical cross-sectional area becomes half the original area: eight times fewer particles, each with four times the area. Real contrail crystals are not necessarily identical spheres, and their ice mass is not fixed as they exchange water with the air. The calculation merely demonstrates that particle number, mass and projected area are different quantities.
An intervention that changes the initial particles can therefore alter later evolution without producing an identical percentage change in every cloud property. A measurement of fewer crystals is meaningful evidence. It is not, by itself, a measurement of the same percentage reduction in lifetime, radiative forcing or global temperature response.
The practical implication is a chain of observations and assumptions. Initial exhaust particles influence the young contrail; the environment influences its subsequent growth and survival; its evolving properties influence radiation. An uncertainty near the beginning can propagate through the chain, but a precise measurement of the beginning does not eliminate uncertainty farther along it.
The white line has two radiative effects
Contrail ice can scatter incoming sunlight back towards space and reduce the outgoing terrestrial radiation that escapes. These effects have opposing signs in the planetary energy balance. A coupled ICON–CoCiP modelling study published in 2025 explicitly treats interactions between contrails and the surrounding atmosphere, including water uptake and radiation. Its approach also illustrates why background cloud conditions and feedbacks matter when estimating the additional effect of contrail ice. Schumann and Seifert’s coupled-model study
At night there is no incoming sunlight for a contrail to reflect at that location, but the cloud’s subsequent lifetime may extend into daylight. During the day, both radiative effects need consideration. A rule based only on the clock at departure would miss the location and timing of the later cloud. The objective is not a categorical dislike of white lines; it is a reduction in the relevant net climate effect.
An invented energy-budget calculation makes the distinction tangible. Suppose one cloud produces a positive contribution of 12 units from its longwave effect and a negative contribution of 8 units from reflected sunlight over a specified interval. The net is positive 4 units. Removing the cloud removes both contributions, not just the positive 12. A second cloud with contributions of positive 12 and negative 14 would have a negative net of 2 under the same accounting convention. Those arbitrary numbers are not representative contrail values.
There is a dimensional distinction as well. Radiative forcing describes a change in energy flow, commonly expressed per unit area. Integrating an appropriate forcing over area and time gives an energy quantity. For example, a hypothetical uniform 1 watt per square metre over 1 square kilometre for 1 hour corresponds to 3.6 billion joules. This calculation does not establish a surface temperature change. It only converts power, area and duration into energy.
A claim about contrail length therefore leaves several questions unanswered. How wide did the cloud become? How long did it persist? What was the net radiative effect during that lifetime? What background conditions would have existed without it? Two equal lengths of initially visible trail need not answer those questions in the same way.
This is one reason to resist describing a flight’s entire climate effect with a photograph. The photograph can document a cloud. The climate inference requires additional information about properties, time and context. Good observations are indispensable, but they do not remove the need to specify what quantity is being inferred from them.
Why a small subset attracts so much attention
Roger Teoh and colleagues’ 2024 global modelling study used flight trajectories, meteorological reanalysis and engine-emission estimates to simulate contrails for 2019–2021. In its 2019 calculation, approximately 2 percent of flights accounted for 80 percent of annual contrail energy forcing. The model was especially sensitive to humidity corrections and particle-emission assumptions. The concentration is a result of that historical simulation, not a fixed property of every airline, year or route. Teoh and colleagues’ global study
Concentration creates an attractive possibility: large benefits without moving every aircraft. It also creates a prediction problem. Knowing after the event which flights had large effects is not equivalent to identifying them beforehand. The more strongly a strategy depends on a small high-impact subset, the more its outcome can depend on whether that subset was correctly selected.
Imagine a fleet of one thousand flights in a hypothetical day. Twenty account for most of the chosen contrail-impact measure. A targeting system that modifies twenty flights has not necessarily captured those twenty. It might select ten genuinely important cases and ten low-impact ones. Alternatively, it might identify the important cases but find feasible alternatives for only half. The fraction of flights changed does not reveal the fraction of impact avoided.
A retrospective ranking is also vulnerable to selection optimism. If the largest effects are identified using a detailed reconstruction of weather that was unavailable before departure, an operational planner cannot simply reproduce the ranking. The experiment must preserve the distinction between the information available when a choice was made and the information used later to evaluate it.
That distinction need not undermine targeted avoidance. It identifies what must improve: selection, feasibility and verification. A programme might make progress by finding slightly fewer of the highest-impact flights more reliably, rather than by drawing an ever larger set of uncertain regions on the planning screen. Whether that trade is worthwhile depends on the consequences of missed opportunities and unnecessary deviations.
An informative report would therefore show the candidate population, the selected flights, the alternatives generated, the routes released and the routes flown. Each transition can explain a different limitation. Reporting only the last successful subset can demonstrate possibility while leaving the scale of the implemented benefit unclear.
A forecast field is not an instruction to the crew
The CocipGrid model described by Zebediah Engberg and colleagues in January 2025 turns trajectory-based contrail calculations into four-dimensional fields suitable for flight planning. It evaluates small potential contrail segments across position, altitude and time, with outputs differentiated by aircraft–engine groups. The resulting fields can be combined with other planning costs. They describe predicted consequences of flying through the atmosphere, not a standalone authority to change a flight. The CocipGrid model paper
A planner needs a comparison between complete feasible alternatives. A lower predicted contrail effect along one segment may be outweighed by a change elsewhere. The aircraft may encounter different winds, fly a different distance or create a new contrail after leaving the region originally targeted. A useful calculation follows the proposed trajectory rather than treating each coloured grid cell as an independent decision.
This can be demonstrated with another invented example. Route A has a climate score of 100 units under a specified assessment method: 60 assigned to contrails and 40 to the other included contribution. Route B reduces the contrail component to 20 but raises the other component to 42. The total falls from 100 to 62, a 38 percent reduction, while the contrail component alone falls by two-thirds. Both percentages are correct for their respective quantities; neither should be reported as the other.
Now suppose a third route removes the original contrail segment but generates a different one with an assessed contribution of 50. Calling the first segment “avoided” does not establish that the whole flight improved by 60 units. The replacement contribution belongs in the comparison. A route optimisation that checks only the region it deliberately bypassed could miss its own unintended outcome.
All of these scores are illustrative. Real assessments require physically meaningful quantities and a stated comparison method. The point is structural: the calculation needs a boundary around the whole change, not just the part that looks beneficial.
The feasible alternatives also have to satisfy the operational constraints that govern ordinary flight planning. The climate objective belongs inside that process. A general-interest article cannot turn a forecast into a flight-level recommendation, and a promising climate score cannot certify the safety or availability of a trajectory.
Humidity errors have a weather pattern
A November 2025 study by Oliver Driver, Marc Stettler and Edward Gryspeerdt compared ice-supersaturation information around North Atlantic extratropical depressions. It found that modelling performance varied with the weather system’s structure, rather than behaving like one uniform error everywhere. Adjusting humidity could change detection performance, but did not resolve all the differences between regions within those systems. Driver and colleagues’ primary analysis
This matters because a correction can improve one property while leaving another unresolved. Matching the overall frequency of ice-supersaturated conditions does not prove that the conditions occur at the right places. A map can have the correct total coloured area and still put the important boundary on the wrong side of a flight path.
Consider two simplified forecasts for a narrow layer. One predicts the layer in exactly the right position but makes it too extensive. The other predicts the correct extent but shifts the entire layer. Both can produce the same count of incorrect grid cells. Yet they can imply different choices for a particular route. Verification should be related to the intended use, rather than assuming that a single map score describes all operational consequences.
Thomas Dean and colleagues’ July 2025 author manuscript examined forecast stability and trajectory optimisation using forecasts and a reanalysis-based evaluation. Its experiments found that useful avoidance could remain possible despite poor pointwise agreement, because relevant spatial errors were relatively small. The reported benefit was assessed against reanalysis; it was not a measured reduction from a fleet of rerouted flights. That distinction is essential when interpreting the encouraging modelling result. The forecast-stability author manuscript
An operational strategy consequently needs room for uncertainty without pretending that the entire sky must be known perfectly. A very narrow avoidance opportunity may disappear if a forecast boundary shifts modestly. A broader alternative may be more robust but incur a larger operational cost. Comparing those alternatives is a decision problem tied to the specific flight, not a contest to produce the most visually detailed weather map.
The assessment should also record forecast age. A plan developed many hours before departure and a tactical update shortly before reaching a region use different information. Evaluating the earlier plan with the later forecast without saying so would exaggerate what the planning system knew when it acted.
Fuel creates a second, longer accounting problem
Burning extra conventional jet fuel produces extra carbon dioxide even when a contrail is avoided. IATA’s stated CO2 Connect methodology uses a combustion factor of 3.16 kilograms of carbon dioxide per kilogram of jet kerosene. That is a combustion conversion, not a complete lifecycle assessment and not a conversion from contrail cloudiness to carbon dioxide. IATA’s methodological explanation
For an explicitly hypothetical flight burning 20,000 kilograms of fuel, an additional 0.5 percent would be 100 kilograms. Applying the stated factor gives 316 kilograms of additional combustion carbon dioxide. These numbers are chosen for arithmetic, not as a typical flight or an observed avoidance penalty. The extra carbon dioxide cannot be dismissed because the percentage sounds small; its significance depends on the benefit achieved and the comparison framework.
Fuel records also require a clear definition. Fuel loaded before departure is not automatically fuel burned on the flight. A proposed plan’s estimate is not the same quantity as measured consumption. A comparison can be affected by aircraft type, distance, operating conditions and other differences between groups. Those distinctions matter especially when the expected penalty is small relative to normal variation.
In the March 2026 trial, the aircraft-adjusted fuel-per-distance difference for the 112 flights flown according to avoidance plans was 0.02 percent, with a reported 95 percent interval from −1.17 to +1.16 percent. This is not proof of exactly zero penalty. The manuscript’s larger eligible group had a different statistical result; the fuel statement here concerns the flown-as-planned subgroup. The trial’s fuel results, Table 3
There are two separate questions behind that result. What fuel difference can be resolved in the trial data? And what fuel difference would be expected if the intervention were applied elsewhere under different conditions? A small, uncertain estimate in one setting does not answer the second question for every fleet. Equally, the possibility of additional fuel does not establish that every avoidance is counterproductive.
A sensible report can preserve both facts: the measured comparison did not resolve a significant increase for that subgroup, and fuel remains part of the evaluation. Turning either statement into a universal claim would discard useful information.
A better intensity is not always less fuel
Normalisation can help compare aircraft and routes, but it changes the question being answered. Consider two completely hypothetical flight plans. The baseline covers 1,000 kilometres and burns 5,000 kilograms of fuel, or 5 kilograms per kilometre. An alternative covers 1,100 kilometres and burns 5,280 kilograms, or 4.8 kilograms per kilometre. Its fuel intensity is 4 percent lower, while its total fuel consumption is 5.6 percent higher.
There is no contradiction. The alternative uses less fuel for each unit of distance but travels farther. Neither number alone decides whether the climate trade is worthwhile. The total additional consumption belongs in the comparison of the two complete flights; the intensity may be useful for understanding aircraft performance or adjusting comparisons across different routes. Reporting the intensity change as a total fuel saving would be a category error.
The same care is needed when aggregating a programme. Suppose, again hypothetically, a small subgroup of long flights experiences a modest percentage increase while many short flights are unchanged. Averaging the percentage changes flight by flight does not necessarily give the percentage change in total fuel. The long flights contribute more kilograms to the sum. A fleet result should identify whether it weights flights equally, weights by fuel, or uses another stated method.
A practical evidence package can retain both views. It can report the actual mass of fuel used within the study boundary and a normalised comparison designed to account for differences between groups. Those results should have clearly different labels. If one cannot be measured reliably, the report should say so rather than allowing the other to stand in for it.
These distinctions are particularly important when comparing avoidance techniques. A lateral detour, a change in vertical profile and a fuel-composition intervention need not affect distance, consumption and exhaust particles in the same proportions. A common outcome measure is necessary before their reported percentages can be ranked. Otherwise an apparently superior method may merely have been evaluated with a different denominator.
Comparing a short-lived cloud with accumulated carbon
David Lee and colleagues’ assessment of aviation forcing, published online in 2020 and in a 2021 journal volume, distinguishes radiative forcing from effective radiative forcing and examines different ways to compare aviation’s climate influences. Its discussion explains why equivalence between short-lived effects and carbon dioxide depends on the chosen metric and time horizon. A single multiplier is not an immutable physical property of every flight. Lee and colleagues’ primary assessment
This is not merely a choice of units. A short-lived cloud and an emission that contributes to long-lasting atmospheric carbon have different time profiles. A comparison can ask about accumulated radiative influence over a defined period, temperature at a particular future time, or another specified objective. Those questions need not assign identical relative weight to the same intervention.
An illustrative decision calculation can stay honest by using a single declared score rather than pretending to convert everything into physical tonnes. Suppose a candidate deviation would avoid 60 score units if the targeted warming contrail would otherwise form, and would add 5 units through its fuel consequence. If the relevant probability of the avoidable event were 0.7, the simplified expected benefit would be 0.7 times 60 minus 5, or 37 units. This assumes no replacement contrail and no other consequence.
If that probability were only 0.05, the same calculation would give negative 2 units. The example does not estimate actual probabilities or recommend a threshold. It shows why both confidence and the magnitude of the benefit matter. An action can look attractive when a forecast is certain and unattractive when the same event is unlikely.
The simplification would fail if the deviation sometimes produced a new warming cloud, if the fuel estimate changed with the route, or if the score mixed incompatible horizons. Those are not small details to append after an optimisation. They determine what is being optimised.
A robust claim should therefore say which climate measure was used, which consequences were included and how the result changes under reasonable alternative assumptions. A positive result under one clearly identified metric is evidence under that metric. It should not be relabelled as an identical reduction in all forms of climate impact.
Seeing a trail, and seeing the right trail
A study published on 30 October 2026 by Thymen Woldhuis and colleagues provides unusually direct evidence about formation. The researchers matched IAGOS aircraft carrying atmospheric instruments with high-resolution Sentinel-2 and Landsat imagery. Their dataset contained 543 aircraft observations. Accounting for measurement uncertainty, the Schmidt–Appleman criterion explained 98.3 percent of observations using onboard meteorological measurements, compared with 92.1 percent using ERA5 data. The authors also emphasise limits on conclusions about later dispersion. This is a test of formation and early evolution, not 98.3 percent accuracy for forecasting a flight’s climate effect. Woldhuis and colleagues’ primary study
That distinction is especially valuable because a disagreement has several possible origins. The atmospheric input may be wrong. The physical model may omit a relevant process. The observation may fail to detect an existing trail or incorrectly identify another cloud. Without separating those possibilities, a verification exercise can blame the wrong part of the system.
Suppose, in an invented observation experiment, one instrument detects half of one hundred real trails and another detects all of fifty real trails. Both report fifty detections. The equal counts do not establish equal underlying formation. Detection probability and sampling determine what the counts mean. An avoidance study must understand whether its measurement process treats the compared groups consistently.
A missing line also needs interpretation. “No contrail was detected by this method during this observation window” is an observation statement. “No contrail of any kind existed at any time” is much stronger. Neither should automatically be substituted for “the flight caused no contrail warming.” The last claim would require an assessment of the cloud’s relevant lifecycle and the method’s limits.
Attribution adds another challenge. In busy airspace, several flights can pass through similar conditions near each other in time. An observed trail must be associated with the correct aircraft if the analysis is to evaluate a specific intervention. Wind displacement, timing and the geometry of observation all matter. A plausible visual match is not the same thing as a validated assignment.
The best verification systems make these uncertainties inspectable. They retain observation windows, association criteria and excluded cases, rather than presenting a final percentage without the chain that produced it. This makes a result easier to reproduce and helps identify whether a disagreement calls for a better forecast, a better detector or a better physical model.
The flight not taken is part of the experiment
The hardest measurement is absent by design. If a flight successfully avoids a warming contrail, the original route is not flown by that aircraft at that moment. The cloud that would have formed there is a counterfactual. Observing clear air behind the rerouted aircraft does not by itself prove that the original route would have produced a persistent warming trail.
This is why a comparison group has value. In a properly designed experiment, it can estimate outcomes under the alternative policy without requiring the same aircraft to fly both routes simultaneously. But the unit of assignment, adherence and analysis must be understood. A comparison of all eligible flights addresses a different question from a comparison restricted to the flights that followed the proposed intervention.
Consider a hypothetical programme offering a new plan to one hundred flights. Only twenty use it. Suppose those twenty are the flights with the easiest operational opportunities, not a random selection of all one hundred. Their results can establish what happened in the selected circumstances. They cannot automatically establish what would happen if all one hundred were forced into the same policy.
The reasons for non-adherence are themselves informative. An unavailable route, a conflicting operational requirement, a late change in conditions and a decision not to use a planning option are different events. Grouping them all as “the forecast failed” would be as misleading as ignoring them because a subset succeeded.
A second trap is the baseline plan. If an analyst compares an avoidance route with an old plan generated before major weather information arrived, the apparent difference may include ordinary replanning rather than the avoidance intervention alone. A credible counterfactual should be as comparable as possible in the information and operational constraints it uses, while acknowledging that it remains a modelled alternative.
These considerations do not require an impossible standard of proof. They require a claim whose scope matches the design. A trial can show operational feasibility, a reduction in observable trails or a modelled climate benefit. Each is useful, and each needs its own explanation of what was compared.
Airspace is shared, not a collection of isolated optimisations
EUROCONTROL’s May 2025 account of MUAC’s work describes both operational trials and simulations. It reports that controllers in a 2023 simulation preferred planned avoidance over tactical changes, and that blocking flight-level bands affected capacity, becoming impractical at high traffic loads in that simulation. These findings concern the tested operational concepts; they are not a universal prohibition on contrail avoidance. MUAC’s research and operational account
The systems issue is straightforward. An alternative that is feasible for one aircraft in isolation may conflict with alternatives proposed for others. If many independent planners all choose the same attractive region, the combined set of plans may not be available. The problem resembles allocating a shared resource, although aircraft operations have their own non-negotiable safety requirements and professional procedures.
A hypothetical scheduling example makes the difference visible without offering flight instructions. Suppose two feasible alternatives are available to each of ten planned movements, but one alternative has only enough capacity for three under the relevant operating constraints. A calculation that assigns all ten to that option has not found ten feasible improvements. It has found ten individually attractive requests that need a system-level allocation.
The objective might favour the cases with the largest robust climate benefit for the least operational disruption. It might also need to account for fairness, workload and the ability to explain decisions. Those are design questions for the responsible institutions, not something a weather model can settle by producing a probability field.
The distinction between strategic planning and tactical adjustment also matters for evaluation. An option generated well before departure allows a different coordination process from a request made shortly before an aircraft reaches a sensitive region. A programme should record where the intervention entered the workflow and which decisions remained with dispatchers, crews and air traffic control.
No climate result removes the need for those responsibilities. Contrail mitigation is an additional planning objective within aviation operations, not a reason for passengers or outside observers to prescribe a different altitude to a flight.
Fuel composition offers a different intervention
Avoidance changes where an aircraft flies. Changing fuel composition can change the particles emitted along the route. Raphael Märkl and colleagues’ 2024 ECLIF3 study compared contrails produced with 100 percent HEFA-SPK sustainable aviation fuel and conventional Jet A-1 under similar atmospheric and engine conditions within one flight. It measured 56 percent fewer ice particles per mass of fuel for the HEFA-SPK comparison, alongside a 35 percent reduction in soot-particle emissions. These are measured particle results for the study conditions, not a universal 56 percent reduction in aviation warming. Märkl and colleagues’ in-flight measurements
This route to mitigation is complementary in concept, but the effects cannot simply be added. If changing fuel reduces the impact of a contrail that a route change would already have prevented, adding both benefits independently double-counts some of the same avoided effect. A combined assessment should evaluate the aircraft, fuel and trajectory together.
For example, suppose an imaginary baseline has a contrail score of 100. One intervention alone reduces it to 60, while a different intervention alone reduces it to 30. It does not follow that applying both gives a score of negative 10 by subtracting reductions of 40 and 70. The interventions may overlap, interact or affect different parts of the process. The joint outcome needs its own calculation or evidence.
Nor does a fuel described as sustainable cease to produce combustion products simply because its production pathway differs. Lifecycle carbon accounting and contrail microphysics ask related but distinct questions. The first needs information about production and supply; the second needs information about fuel composition, engines, particles and the atmosphere. One label cannot supply all those measurements.
The particle findings are valuable precisely because they isolate a physical response under studied conditions. Turning them into a broader marketing percentage would weaken rather than strengthen the evidence. The next questions concern how that response varies across engines, operating conditions and atmospheric environments, and how it affects the later cloud.
What an operational result should disclose
On 2 October 2026, SESAR’s CICONIA project reported results from Air France trials conducted between February and April that year. Its account describes nearly 4,300 flights analysed, 245 identified as high-potential cases and 130 effectively rerouted. It reports an 8 percent reduction in assessed contrail climate impact across the trial scope, distinct from the larger reduction reported for rerouted flights. These are project-reported assessment results, not direct measurements of a change in global temperature. SESAR’s dated trial account
The useful feature of such a report is not merely its headline result. It is the opportunity to follow the population through successive decisions. How many flights were screened? How many had a potentially worthwhile change? How many alternatives were operationally acceptable? How many were flown? How much of the final outcome was observed directly, and how much depended on a climate model?
Fuel should appear beside the benefit, with its own uncertainty and definition. So should the climate metric, forecast version, observation coverage and handling of replacement contrails. A claim that survives these details is more credible than one that requires them to remain out of view.
There is also a difference between evidence that a method can work and evidence about routine deployment. A trial can establish a promising result while still leaving questions about seasons, regions, fleet composition and workload. Broader testing should be designed to learn where the result changes, rather than merely repeat the most favourable conditions until the same headline appears again.
The end point is not a sky without white lines. It is an accountable reduction in aviation’s climate effect, obtained through feasible operations and evaluated against a fair alternative. Contrail mitigation deserves attention because that objective may be achievable in selected circumstances now. It deserves scrutiny because the route from fewer observed trails to less warming passes through enough physics, inference and operational choice that no single percentage can carry the whole argument.