A cell in the landing distance table provides a value for a given mass, altitude, and temperature. This value has never been measured as such: it comes from a family of test points flown at other masses, by other temperatures, transmitted by telemetry, corrected by a model, wrapped by a negotiated method, then penalized twice before being printed. Knowing where each margin was applied changes the reading of the table — and especially the reading of a discrepancy observed in a row.
A test point does not produce a number
Raw data On the test card, the line is summarized in one sentence: landing, landing flaps, target mass, maximum manual braking, dry runway. Yet what the test produces is not a distance. It is a package of sampled time series: altitude-pressure, indicated speed, total temperature, attitudes and angular velocities, accelerations on the three axes, control and surface positions, engine speeds and flows, braking pressures, aircraft position. The distance does not appear anywhere in the raw data. It will be calculated later, by integrating a position between two instants that someone will have had to define — and defining the touchdown instant is no longer measuring, it is agreeing.
The test engineer therefore does not read a result, he reads a trajectory, and first judges whether the point is usable: stabilization held to the threshold, calm enough atmosphere, technique conforming to that which will be written in the manual. A poorly flown point is not corrected, it is rejected, and it consumes flight time — the rarest resource of a campaign. This sorting is done in the telemetry room while the aircraft is still in the air: it is the only time when you can ask it to try again.
Les données brutes d’un essai ne donnent pas directement la distance d’atterrissage. Elles contiennent des mesures brutes comme la vitesse, la température et la position, mais la distance réelle est calculée plus tard en combinant ces données. Un point d’essai mal exécuté est jeté, pas corrigé, car le temps de vol est trop précieux.
Un Airbus A320 effectue un essai d’atterrissage avec une masse de 65 tonnes, à une altitude de 1 000 pieds et une température de 25°C. Les données brutes montrent une vitesse de 140 nœuds au toucher, mais la distance réelle n’est pas mesurée directement. Elle sera calculée plus tard en intégrant la position de l’avion entre le seuil de piste et l’arrêt complet, en tenant compte des corrections de modèle et des marges de sécurité.
Calibrate before measuring
Calibration No quantity enters the chain without its own calibration curve, plotted on the bench, linked to standards whose lineage is known, and accompanied by a declared uncertainty. A pressure sensor is not linear, it drifts with temperature, it ages; a strain gauge has a factor that depends on the bonding. Calibration does not eliminate these defects, it describes them: after correction, a residual error remains whose limit is known. It is this limit, and not the measurement itself, that will travel to the end of the chain.
Then comes what electronics add. Quantization transforms a continuous signal into a finite step; the anti-aliasing filter, essential, introduces a group delay different depending on the channels. Above all, the timestamp must be common to all parameters: a derivative calculated from two desynchronized channels produces a physically false value while remaining perfectly plausible. Synchronization is not a refinement, it is a condition of validity.
There remains a correction of another nature: it does not correct the instrument but the aircraft. A static port measures a local field deformed by the airframe, and this deformation depends on the angle of attack, hence on the speed and configuration. The position error is measured in flight — trailing cone, passage in front of a reference station, reference aircraft. This is not a workshop adjustment but a certified datum, on which all speeds displayed to the crew depend.
The right reflex. Before interpreting a discrepancy between two displayed values, ask yourself which one is already corrected. A calibrated speed, an indicated speed, and a speed from an inertial or satellite source do not tell the same story, and the disagreement between them is often the correction itself, not a failure.
Reducing means bringing each point to a reference condition
Reduction The point was flown at a mass that decreases continuously, at an altitude-pressure that was not round, at a temperature that deviated from the standard atmosphere, with a non-zero wind component and an engine whose actual thrust was never measured in flight. The manual, on the other hand, must provide values for all masses, all altitudes, and all temperatures in a domain. Between the two, there is reduction.
Its principle is to move from dimensional quantities to groupings that absorb the conditions of the day: mass relative to the pressure ratio, engine speed and speed relative to the square root of the temperature ratio, aerodynamic coefficients rather than forces. Expressed in this way, points obtained on different days line up along the same curve. The wind component, runway slope, and everything else are similarly corrected, and all are brought back to a reference atmosphere.
Thrust is the least intuitive link: no one directly measures the thrust of an engine installed in flight. It is calculated by an engine model, calibrated on the bench and then cross-checked in flight by independent methods. The model thus becomes a piece of the measurement chain, on the same level as a sensor, with its own validity and limits.
The trap. A reduction law is only valid where it has been confronted with measurements. Extended beyond — mass never flown, altitude-temperature pair never encountered, intermediate deflection never demonstrated — it continues to produce a smooth and credible number, without any warning sign. The difficulty is that an extrapolation is not visible in the result: it is only visible in the table of flown points, which the final number reader never has in front of them. Bad surprises rarely come from a faulty sensor; they come from a validity domain crossed without anyone having to decide.
From dispersion to guaranteed figure
Dispersion A test point is not flown once. It is repeated, and the repetitions do not overlap: the piloting technique varies from one repetition to another, the atmosphere is never the same twice, the ground effect depends on the exact height at the threshold crossing, the condition of the brakes and tires evolves throughout the day. This dispersion is not a measurement defect, it is the real behavior of the aircraft.
It is therefore necessary to decide what is published. A fit passing through the middle of the cloud would describe an average aircraft, which makes no sense for a datum intended to bound a departure. A penalizing value is retained: envelope of the cloud, statistical offset associated with a defined probability of exceeding, or a combination of both. The essential thing is elsewhere: the method is agreed with the authority before the campaign, and not after, when the results are known.
The word « guaranteed » then takes its precise meaning. The manufacturer does not claim that the aircraft will make this distance, but that a compliant aircraft, flown according to the described technique, will not do worse. The commitment is enforceable, and that is why it is expensive: every margin conceded to the envelope ends up in the commercial load sold over the entire commercial life of the type.
This is also what explains the caution of evolutions. When an aircraft receives a modification — wingtip device, new engine standard, additional flap deflection — the entire campaign is almost never redone. The reduction model is revised, the gap it introduces is estimated, and only a reduced set of confirmation points is flown, chosen where the model is least certain. The published data therefore remains based on an original campaign whose scope must be known: it is this, and not the revision date, that bounds the validity.
From the test pilot to the airline crew
Certification Between the test point and the table cell, the certification text imposes its own margins, and these are not measured: they are written. The landing distance is counted from a conventional height above the threshold, not from the touchdown. The takeoff distance retained is the greatest between the case with engine failure and a flat-rate increase of the all-engine case, of the order of fifteen percent. The acceleration-stop integrates a flat-rate reaction time, of the order of two seconds at the decision speed. The net trajectory is obtained by subtracting a gradient penalty from the measured trajectory, depending on the number of engines, of the order of eight tenths of a percent for a twin-engine aircraft.
Other margins come from what the demonstration forbids: maximum manual braking, without reverse thrust credit on a dry runway. The reference technique — stabilized approach, touchdown without prolonged flare, braking engaged without delay — is described in the manual and is part of the datum on the same level as the number. A published distance without its technique is not incomplete data: it is uninterpreted data.
Then comes the operation, which adds its own layer. For a jet aircraft, the departure rule does not allow the destination runway to be entirely consumed: the demonstrated landing distance must only occupy a fraction of it, of the order of six tenths on a dry surface, and this requirement is further tightened by about fifteen percent when the runway is announced wet. The reassessment made in flight upon arrival is based on data closer to reality — actual mass, announced surface condition, actually retained configuration — but it also receives a safety margin, of the order of fifteen percent. The wind itself is penalized, and asymmetrically: only part of the announced headwind is counted, and more than the full tailwind, because an estimation error does not have the same consequences in both directions.
The stacked margins, and who decides them
Margins The interest of the overall view is to separate what belongs to metrology, what belongs to an industrial choice, and what belongs to the text. The three are not discussed in the same place, and are not recovered in the same way.
| Étape | What is added | Nature | Who sets it |
|---|---|---|---|
| Sensor and acquisition | calibration uncertainty, quantisation, filtering delay | metrological, bounded | the flight test department, on a calibration plan |
| Position error correction | residual error of the airspeed model | model validated by cross-checking | the manufacturer, accepted by the authority |
| Reduction to reference conditions | validity range of the aerodynamic and engine models | model | the manufacturer |
| Envelope of the cloud | scatter of repeats, variability of technique | statistical or envelope | method agreed before the campaign |
| Certification requirements | screen height, all-engines factor, reaction time, net gradient | fixed, written | the certification text |
| Operating rules | fraction of available length, wet runway factor, wind factoring | fixed, written | the operating regulation |
| Operator policy | operations manual supplements, internal limits, aerodrome restrictions | volontaire | l’exploitant |
A reading imposes itself: the last two lines are the only ones on which an operator has a say, and they are also the only ones that a crew sees vary from one airline to another for the same aircraft. Everything that precedes is frozen at the time of certification and is only renegotiated at the cost of a new campaign.
What the table does not say
Assumptions The first verification to be done at home concerns the status of the data used. The approved flight manual carries enforceable values; the tables in the operations manual and the results of a calculation tool may be non-approved manufacturer data, intended for routine use and sometimes less penalizing. The two look very similar and are not interchangeable.

- The second concerns the incorporated assumptions. Is reverse thrust credit taken into account, and on what surface? What braking configuration is assumed, is the automatic system credited? What interpolation is allowed between two rows, and what to do beyond the last column — because a table is not extrapolated: its last line is a limit, not a trend.
- The third verification is that of the lineage. A performance table carries a revision, and this revision is linked to an aircraft standard: a configuration reference, a navigability directive applied, an equipment installed after delivery can make a page inapplicable to a specific aircraft in the fleet. This is what documentary checks catch the least, because the page remains readable and perfectly plausible — it is simply for another aircraft of the same type.
The lasting gray area remains that of contaminated runways. These distances do not come from tests covering each surface condition: they combine friction models and correspondences between the declared runway condition and an assumed braking efficiency. The data is based, but its uncertainty does not have the nature of that of a dry runway value — hence its often operational rather than approved status.
This chain deserves to be known in order by those preparing a type rating or a performance engineer position: it explains why two operators of the same aircraft do not display the same limitations, and why the right question in front of a surprising number is always « from which step does it come ». The FL330 training courses cover this reading from the operational side, and methodological questions are discussed on the FL330 forums.
This article exposes a mechanism, not a procedure: in case of divergence, the applicable text, the approved flight manual of the aircraft, and the approved documentation of the operator prevail over everything that has just been written.

Leave a Reply
You must be logged in to post a comment.