Unstabilised approach: the criterion exists, the decision stays human

Approche non stabilisée : le critère existe, la décision reste humaine

Two hundred feet above the glidepath, fifteen knots too fast, the flaps still in mid position, and the thousand-foot gate coming up: everyone on board knows what the manual asks for. In most cases the aircraft continues anyway. This is neither ignorance nor indiscipline; it is the way the decision gets made, under an automation that keeps the approach looking calm right up to the last moment.

A stabilisation criterion is not a description, it is a lock

The criterion does not describe a nice approach. It manufactures a decision. Without it, the question ‘will it work?’ stays open until touchdown, and an open question always closes on the side of continuing: at every instant, the remaining gap looks smaller than it did a moment before, and the effort already invested argues for pressing on. The criterion cuts that continuum by setting a discrete test at a height known in advance, typically one thousand feet above the aerodrome in instrument flight conditions and five hundred feet visually. The exact value belongs to the operations manual, and many operators raise that gate for certain runways, certain approach types or certain surface conditions.

What it tests fits in a short and conjunctive list: flight path held, laterally and vertically; landing configuration set; speed within a narrow window around approach speed; rate of descent contained, on the order of a thousand feet per minute beyond which a specific manoeuvre is required; thrust stabilised, hence out of idle; checklists complete; briefing done. All of these, not most of them. It is that conjunction that gives the criterion its value: a single missing item is enough, and there is no need to demonstrate that the approach is dangerous in order to discontinue it. The criterion reverses the burden of proof, and that is roughly all it delivers.

The consequence is widely misunderstood. The criterion says nothing about the safety of the approach in progress. An approach found unstabilised at the gate can be perfectly recoverable, and most are: that is precisely what wears the rule down. It does not claim to predict the accident, it sets the point beyond which one stops betting.

The stabilisation criterion is like a safety net – it doesn’t guarantee the approach is safe, but it forces a decision point. If any one condition isn’t met at the gate height, the approach must be abandoned. It’s not about predicting danger, just about making a clear choice before things get too close to the runway.

On a typical Airbus A320 approaching runway 27L at Paris Charles de Gaulle, the stabilisation gate is set at 1,000 feet. At this point, if the aircraft is 200 feet high, 10 knots fast, and the flaps are still at 2 degrees (instead of the required 3), the approach must be discontinued. Even if the crew thinks they can catch up, the rule requires a go-around.

Deux profils d’approche de part et d’autre du stabilisation gateSide view. A steady profile descends in a straight line to the runway threshold. A second profile, drawn as a dashed line, stays clearly above it, steepens, and only joins the first after the dashed vertical marking the stabilisation gate. Two points mark the remaining height gap at that instant.heightdistance coveredthresholdgate de stabilisationlate catch-upprofile held
At the gate, the two profiles are still apart. The dashed track converges, but it only joins the held profile after the gate: at the moment of the check, the remaining margin for catching up is far smaller than the trend suggests.

Why the lock trips anyway

Flight data monitoring programmes regularly produce the same shape of result: approaches flagged as unstabilised are numerous, and only a small minority of them are followed by a go-around. The explanation is not moral, it is structural, and it comes down to three mechanisms.

The first is a cost asymmetry. Continuing requires no action, no call, no report: it is the option that chooses itself when nobody does anything. Going around requires a manoeuvre, a reconfiguration, negotiation with air traffic control, fuel, time, sometimes an explanation on return. One branch is free in the short term and invisible, the other is immediately visible. The criterion exists to reverse that flaw, but it only reverses the text: the surrounding system of costs stays exactly where it was until the operator deals with it.

The second is a proximity effect. The closer the goal, the more evidence a crew demands of itself before giving it up. At thirty miles, a mere reservation is enough to ask for a delay. At three hundred feet, near certainty is needed. This is not a character trait, it is the ordinary working of commitment to a task already under way, and it is corrected by a decision prepared in advance, never by will in the moment.

The third is the weight of the call. In most manuals, the criterion is checked by the monitoring pilot. But making the call means contradicting the pilot flying about what they are doing, at the moment they are busiest. A crew that has not agreed before the descent who says what, and what that call triggers, discovers on final that speaking has a cost.

The system makes continuing the easy choice – no extra work, no extra communication. Going around feels like admitting failure, even when the rules say it’s the right thing to do. The closer you get to the runway, the harder it is to walk away, even when the approach isn’t stable.

On a Boeing 737 approaching runway 09 at London Heathrow, a crew might be 150 feet high and 8 knots fast at the 500-foot gate. The monitoring pilot calls “unstabilised” but the pilot flying hesitates – they’re only 30 seconds from touchdown. They continue, catching up just before landing. Flight data later shows they were actually 100 feet high and 12 knots fast at the gate, but the crew didn’t want to “waste” a go-around.

The burden of proof a crew sets itself before going aroundA solid curve rises slowly then very steeply as the aircraft nears touchdown: this is the amount of evidence a crew demands of itself before discontinuing. A dashed horizontal line represents the constant evidence supplied by the stabilisation criterion. Beyond the crossing point, marked by a vertical line, the criterion alone no longer suffices to carry the decision.evidence required to give upproximity to touchdowntipping pointevidence supplied by the criterion
The criterion supplies constant evidence; the inner threshold, meanwhile, rises with proximity to the goal. Past the crossing point, the text alone no longer carries the decision: only a rule stated before the descent remains stronger than commitment in the moment.

The trend trap. At the gate, the gap is shrinking: speed is coming down, the glidepath is getting closer, everything is heading the right way. But the criterion bears on a state, not on a derivative. An approach that is converging is an approach that has not yet converged, and that is exactly the situation where the margin for catching up – available drag, speed margin and remaining distance – has already been largely spent. The call is made on what the instruments show at the stated height, not on what they will show ten seconds later.

What automation changes about the whole matter

A pilot trained on a modern transport aircraft rarely sees an unstable approach as a turbulent one. It is more nearly the opposite: the flight path is held by the autopilot, thrust is managed, the display shows a profile closing in. Nothing is stirring in the cockpit, and the absence of commotion is taken as a sign of normality. The disorder that used to raise an alarm has disappeared; the gap has not.

Managed modes do not remove the energy excess, they redirect it. An aircraft too high on a computed profile corrects by descending harder, and speed benefits from that; an aircraft too fast on a held glidepath corrects by holding idle longer, and the energy stays stored as speed. In both cases the display stays orderly: an active mode, a target announced as eventually captured, no alarm. What is missing from the screen is the sum, the total energy against the remaining distance, and no standard instrument shows it directly.

A perverse form of learning is added on top. When automation catches up, and it often does, the crew registers that late convergence works. Each recovered approach nudges the personal threshold of what is acceptable, with no event ever contradicting it. The day a tailwind on final, a heavier weight or a delayed gear extension changes the equation, it is the shifted threshold that serves as the reference.

Two ways of organising the decision to discontinue, and what each one lets through
Point of comparison Single gate at a fixed height Gate, then re-evaluation through to touchdown
What it guarantees A decision is made at least once, at an instant known to both pilots The aircraft’s state stays checked against a limit all the way to landing
What it lets through Any degradation appearing after the gate Little in theory, a lot in practice if the exit criteria are not written down
Workload imposed on the crew Low and concentrated in a few seconds High and continuous; requires short calls and an unambiguous division of roles
Condition for it to hold up A height suited to the conditions and the approach type, not a single value Explicit exit parameters after the gate: speed, thrust, alignment, alarms
Typical defect observed ‘Stabilised’ called, then monitoring relaxes Too many calls, hence trivialised and eventually ignored

Behind the parameters, a single variable: energy

The criterion’s five or six parameters are not independent. They are partial measurements of a single quantity: the aircraft’s total energy against what it has left to dissipate before the threshold. Height is positional energy, speed is kinetic energy, and configuration is the only tap available for bleeding it off, each notch of slats, flaps or gear extension consuming a share of the remaining capacity.

Two pilots in a cockpit during a landing approach at dusk, with an airplane visible outside the windshield.
Human judgment remains critical in managing unstable approaches.

That is where the most ordinary and most misunderstood manoeuvre comes from: delaying configuration. Extending late is borrowing. Distance and time are gained, noise and fuel burn are pushed back, and the bill comes due later, in the shape of a deceleration to be carried out in a shorter space. As long as the debt is repaid before the gate, everything is in order. When the late extension is demanded by air traffic control, imposed by an accepted shortcut, or by a tailwind that was not anticipated, the debt is taken on without the deadline having been calculated.

A point of method follows from this. Speed and height are exchangeable: an excess of one converts into the other, but the conversion costs distance. An aircraft in excess on both counts has no exchange left to make, it can only dissipate, and its capacity to dissipate is bounded by the configuration already extended. It is that case, high and fast together, that makes catching up impossible, not either one taken alone. The question to ask on the last turn is therefore not height alone, but what margin will be left to descend and slow down at the same time.

The key to a stable approach is managing the aircraft’s total energy – how high and how fast it is, and how much it can slow down before the runway. If you’re both high and fast at the same time, you’re in trouble because you can’t trade one for the other. The stabilisation gate forces you to check this before it’s too late to do anything about it.

On an Airbus A330 approaching runway 32L at Dubai on a hot day (35°C), the aircraft is 300 feet high and 20 knots fast at the 1,000-foot gate. The crew thinks they can catch up, but with the high temperature reducing performance, they can’t slow down fast enough. The aircraft touches down 1,500 feet down the runway, well past the aim point, and the crew has to use maximum braking to stop in time.

Energy state shown on two axes, height and speedTwo crossed axes: speed deviation on the horizontal, height deviation on the vertical. A central dashed rectangle represents the acceptable domain. A dashed diagonal through the origin represents states of equal total energy, where a speed excess is paid for in height. A series of points coming from the upper right quadrant, both high and fast, converges by an arrow towards that rectangle but only reaches it after the largest dot, which marks the aircraft’s state at the stabilisation gate.speed excessheight excessexcess energyat the gateacceptable domainconstant energyinsufficient energy
High and fast at the same time, there is no exchange left, only dissipation. The large dot marks the state at the gate: still outside the domain, it only enters it afterwards, and at the cost of a configuration already fully extended.
What each parameter really measures, and what automation shows of it
Parameter checked Quantity actually being monitored Typical late cause What the display can hide
Speed Kinetic energy still to be dissipated Residual tailwind below the cloud base, wind shear close to terrain Managed thrust holds the target as long as drag is enough, without signalling that it has run out of margin
Rate of descent Gap between the path being flown and the path required High intercept, track shortened by air traffic control The vertical mode announces a successful capture even though the required slope exceeds what is sustainable
Configuration Remaining capacity to dissipate energy Delayed extension, speed held to an imposed gate Nothing shows how much margin has already been spent
Thrust Balance reached between drag and energy supplied Prolonged idle late in the descent A stable idle looks like stabilised thrust
Lateral flight path Time left to finish everything else Late alignment, short base leg Automation aligns cleanly, but sometimes after the gate

The right reflex: call the value, not the verdict. At the gate height, a call stating ‘140, one thousand five hundred feet per minute, flaps not out yet’ forces both pilots to compare a state against a written limit. A call that says ‘stabilised’ or ‘not stabilised’ asks the speaker to carry a judgement alone, and a judgement can be negotiated. A number cannot.

The gate is a moment, the approach is a duration

Most manuals have the criterion checked once. Deviations, for their part, do not wait for that moment: a gust under the lee of terrain, wind shear, a late alignment correction, speed held to a short gate, delayed gear extension, a slip to catch up on the glidepath, a runway change accepted at the last moment. It is not rare for an approach that ends badly to have complied at the gate and stopped complying afterwards, in the last seconds when no one was watching the parameters any more because the gate had already been passed.

A serious manual therefore writes two distinct things: the conditions to be met at the gate, and the conditions whose loss, after the gate, requires discontinuing. The second set is shorter than the first, often limited to speed, alignment, thrust and the absence of an alarm, but it must exist in writing. Failing that, the only criterion applicable below the gate reverts to each person’s own judgement, which is exactly what the whole device was meant to avoid.

One case deserves to be added, rarely addressed: the landing actually discontinued, decided after the thrust reduction, or even after a first touch. This is no longer the same manoeuvre nor the same performance. The aircraft is in landing configuration, at low speed, possibly with the ground spoilers deployed and the autopilot disconnected. The climb requirements that cover this case in certification are not those covering the approach go-around: they relate to the more draggy landing configuration and to a lower speed. Knowing at what point one’s own operator considers the landing to be committed to, if that has been written down, is a matter of preparation, not improvisation.

The stabilisation gate is just one checkpoint, but problems can happen before or after it. A good manual should have rules for the whole approach, not just at the gate. And if you decide to go around after you’ve already started landing, it’s a different (and harder) manoeuvre than a normal go-around.

On a Boeing 777 approaching runway 25L at Frankfurt, the crew passes the 1,000-foot gate with everything stable. But then ATC asks them to extend their downwind leg to allow another aircraft to land first. This forces them to fly a much longer final approach, and by the time they’re ready to land, they’re too high and too fast. They try to catch up but run out of time and space, touching down 1,200 feet past the aim point. The approach was stable at the gate, but became unstable later.

A spot check versus an extended degradation windowHorizontal timeline running from final intercept to the runway threshold. A narrow bar placed at the gate represents the single check most manuals provide for. A wide band, extending well on either side of the gate, represents the period during which deviations actually appear; a second, shorter band shows that sustained attention stops shortly after the gate, well before the threshold.single checkactual onset of deviationsattention genuinely sustainedinterceptconfigurationgateshort finalseuil
The check is a single point in time, the degradation is not. The upper band marks where deviations actually start appearing, the lower band where attention is genuinely sustained. It is the gap between the end of the second and the end of the first that decides the final seconds.

The trap of the satisfied call. ‘Stabilised’ is an instantaneous observation, not a lasting state. Called at the gate, it often closes out parameter monitoring at the exact moment external workload reclaims attention: an air traffic control instruction, preceding traffic still on the runway, a frequency change. Flight analyses show this regularly: the gate had been crossed cleanly and the degradation came afterwards.

A go-around is a manoeuvre, not a penalty

The reluctance is not only decisional, it is also technical. The go-around is among the least often performed manoeuvres of an airline pilot’s career. In the simulator, it is almost always performed at decision height, on instruments, often with an engine failure, at a weight set by the exercise. In line operation, the case that actually presents itself is different: all engines running, a light aircraft late in the flight, from an intermediate height, sometimes in good visibility, with a level-off only a few hundred feet away.

This case is the most surprising one. At low weight, the thrust excess is at its maximum: pitch attitude rises sharply, acceleration is brisk, configuration limit speeds arrive within seconds and the level-off comes before reconfiguration is finished. Added to that is the somatogravic illusion, which makes strong forward acceleration feel like a nose-up pitch and pushes the pilot to correct the wrong way. At night or in instrument conditions, this is the classic loss-of-control-after-go-around scenario, and it has nothing to do with a lack of knowledge of the procedure.

Automation adds a layer of transitions. The go-around command reconfigures the flight director, changes the auto-thrust mode, engages or disengages the autopilot depending on the type, and can behave differently depending on whether it is triggered before or after the thrust reduction at the flare. The navigation computer sometimes asks for the missed-approach path to be reselected. A crew that has never seen these sequences played out calmly discovers them at four hundred feet, while climbing, reconfiguring and communicating all at once. This is an exercise worth requesting explicitly in a periodic session, on the same footing as engine failures.

The right reflex: prepare the go-around like a flight path. At the approach briefing, state out loud the level-off altitude, the heading or published track, the first reporting point, the mode that will be active after the command, and the fuel: how many attempts remain possible before the question of diverting arises. Prepared, a go-around costs three sentences at the briefing. Improvised, it produces the highest workload of the flight.

Making the decision workable as a crew

Four levers separate operations where the go-around actually happens from those where it remains theoretical, and none of them rests on individual conviction.

  • The decision is made beforehand. Stating at the briefing the condition and its consequence, ‘if we are not in configuration at the gate, go-around’, turns a judgement call into execution. What has been decided calmly no longer has to be decided under load; all that remains is to recognise the condition.
  • The call is binding and shared. Any crew member who notices the deviation can make it, and it is not a suggestion: the pilot flying carries it out, the discussion happens on the ground. A manual that leaves what follows the call to the sole judgement of the pilot flying protects very little, since it hands the decision back to the person the situation involves most.
  • Discontinuing must be free. As long as a go-around triggers a debrief, a fuel justification or a remark on return, the arithmetic stays unfavourable. Operations that achieve a high compliance rate treat a compliant go-around as a non-event and turn their attention to the opposite case: continuing an approach that did not meet the criteria, which flight data monitoring detects without difficulty.
  • The cost must be known in advance. How many attempts the fuel on board allows, which alternate aerodrome has been chosen, what a second approach costs in time and slot: calculated at the briefing, that cost stops being a last-second argument. Uncalculated, it turns up exactly when the crew has the least capacity to assess it.

To make go-arounds happen when they should, crews need clear rules, shared responsibility, no penalties for doing the right thing, and to know the costs up front. It’s about making the right choice easy, not about blaming people for hard decisions.

On an Embraer 190 approaching runway 06 at Geneva, the crew briefs: “If we’re not in landing configuration at the 500-foot gate, we go around.” At the gate, the flaps are still at 2 degrees (instead of 3). The monitoring pilot calls “unstabilised” and the pilot flying immediately initiates the go-around. No discussion, no hesitation. The crew then files a normal report, with no additional scrutiny, because their operator treats compliant go-arounds as routine.

Four outcomes, very unequal costsA two-by-two matrix crossing the approach’s actual state, stabilised or not, with the decision taken, continuing or going around. Three boxes carry a known and bounded cost: ordinary outcome, cost in fuel and time, expected outcome. The fourth, continuing an unstabilised approach, is marked as the only one whose cost is unbounded.stabilised approachunstabilisedcontinuinggo-aroundordinary outcomenothing to debriefunbounded costthe only one to avoidknown costfuel and timeexpected outcometo be treated as normal
Three boxes cost something measurable, the fourth does not. A go-around from a stabilised approach costs fuel and time, two quantities known in advance; continuing an unstabilised approach incurs a cost with no known ceiling.

What to go and check in your own documentation

The subject does not sit well in generalities, because what matters is written elsewhere than in generic recommendations. Six points are worth noting once and for all, ideally before they are needed: the exact gate height and its variants by condition, approach type and runway state; the list of parameters and their tolerances, in particular the speed window, which varies from one operator and one type to another; whether or not criteria apply between the gate and touchdown; who makes the call and how binding it is; what is written about a landing discontinued late; and finally what the flight data monitoring programme records as an event and what is done with it.

Two grey areas remain, and it is better to name them than to ignore them. The visual approach and the visual circling manoeuvre required around the aerodrome, first, where no glidepath is displayed: the notion of a flight path held there becomes a matter of judgement, and criteria transposed as they stand lose part of their meaning. Steep approaches, second, whose nominal parameters would fall outside the tolerances of an ordinary approach and which require their own limits, written for them. In both cases the crew needs an explicit rule, not an extrapolation. Debriefs shared among crews are often more instructive on these cases than rereading the text.

For anyone preparing a type rating or a selection, one last point is worth hearing: what is observed in the simulator is not reciting the list, it is how quickly a candidate abandons a compromised approach and how clean the flight path that follows is. A brilliant catch-up is noticed less than a go-around decided early and flown correctly, because the first skill is not repeatable and the second is. It is also the only way to bring the manoeuvre into the repertoire of moves available under load.

The applicable regulation, the operator’s approved operations manual and the type’s flight manual take precedence over everything written here; where they differ, they govern.

Questions fréquentes

At what height must the stabilisation criterion be checked?

The height is not universal: it is set by the operations manual, with a higher gate in instrument flight conditions than in visual flight. Many operators use one thousand feet above the aerodrome on instruments and five hundred feet visually, and raise that gate for certain runways, certain approach types or certain surface states. What matters to the crew is less the value itself than the fact that it is stated at the briefing and identical for both pilots.

Can the approach continue if the parameters are converging at the gate?

The criterion bears on an observed state, not on a trend. An approach that has not yet converged at the gate has already spent part of its margin for catching up: the available drag, remaining distance and time are smaller than the convergence suggests. Unless the operations manual states otherwise in writing, a favourable trend is not grounds for continuing.

Who makes the unstabilised call, and is the pilot flying bound by it?

In most manuals, checking the parameters at the gate falls to the monitoring pilot, but the call can be made by any crew member who notices the deviation. Where the system holds up, the call is binding: the pilot flying carries out the go-around and the discussion happens on the ground. This specific point is worth checking in one’s own documentation, because a manual that leaves what follows to the judgement of the pilot flying hands the decision back to the person the situation involves most.

Is an all-engine go-around simpler than one on a single engine?

It is worked less often and often more surprising. At low weight late in the flight, the thrust excess is at its maximum: pitch attitude rises sharply, acceleration is brisk, configuration limit speeds arrive within seconds and the level-off comes low. Added to that is the somatogravic illusion, which makes strong forward acceleration feel like a nose-up pitch and pushes the pilot to correct the wrong way, at night or in instrument conditions.

What happens if the decision is made after the thrust reduction, or even after a first touch?

This is no longer the same manoeuvre: the aircraft is in landing configuration, at low speed, possibly with the ground spoilers deployed and the autopilot disconnected. The climb requirements that cover this case in certification are not those of the approach go-around: they relate to the more draggy landing configuration and to a lower speed. Some operators write down a point beyond which the landing is committed to; it is worth knowing what one’s own documentation says before finding oneself there.

Does a go-around have to be justified on return?

A go-around carried out in accordance with the criteria is a normal manoeuvre, provided for by the procedures and covered by certified performance: it does not need to be treated as an incident. What deserves the attention of a safety management system is the opposite, continuing an approach that did not meet the criteria, which flight data monitoring detects without difficulty. Reporting obligations remain those set by the operator and by the regulation applicable to occurrence reports.

Commentaires

Leave a Reply

Découvrez l’écosystème FL330

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