The approach was stabilized, the height call came, the pitch attitude was set as it had been a thousand times before — and the wheels do not touch. The runway markings slide by, the aiming point passes under the nose, and the hand hesitates between holding it and letting go. Yet the technique is not at fault: in the final seconds, the aircraft no longer answers the same equations as on final approach, and the few knots the approach had tolerated suddenly have nothing left to dissipate them.
When an airplane gets very close to the ground, the air flowing around its wings behaves differently. The ground blocks some of the usual air movements, which reduces drag and increases lift. This makes the plane float more easily, which can be tricky to control during landing.
A Boeing 737 with a 35-meter wingspan flying at 150 knots approaches a runway. At 35 meters above ground (one wingspan), ground effect is just starting. At 3.5 meters (one-tenth of a wingspan), induced drag drops by 50%, making the plane float longer and requiring careful control to avoid overshooting the runway.
What the ground changes is not the air under the wing
The image of a cushion of compressed air between the wing and the runway is convenient, and wrong. What proximity to the ground changes is the field of velocities induced by the wing’s vortex system. A lifting wing leaves behind two wingtip vortices whose circulation imposes, in the wing’s vicinity, a downward velocity component: downwash. But the ground is an impermeable surface that the flow cannot cross. It behaves like a mirror: the flow organizes itself as if a second vortex system, symmetrical and of opposite sense, existed beneath the runway, and the effect of this image system is to cancel part of the downwash where the wing is working.
Everything else follows from this. Less downwash means a local relative wind that is less deflected downward: at the same pitch attitude and flight path, the effective angle of attack of the airfoils increases, so lift increases. And the lift vector, tilted less rearward, leaves a smaller induced-drag component. The wing behaves as if its aspect ratio had grown — the classic formulation in terms of effective aspect ratio — and induced drag, which varies as the square of the lift coefficient relative to that aspect ratio, collapses precisely in the regime where it dominated: high angle of attack, low speed, full landing configuration.
The scale of the phenomenon is the wingspan, not the height
The governing parameter is not the radio-altimeter height but the ratio of wing height to wingspan. This is what makes transpositions from one aircraft to another misleading: at the same height, the larger the wingspan, the smaller this ratio and the more firmly established the phenomenon — a large-span aircraft is already deeply into it where a modest-span type is only just beginning to feel it. The same height call therefore does not denote the same aerodynamic state from one type to another, and a flare technique imported from a smaller aircraft is triggered at the wrong moment. The effect becomes measurable below a height of roughly one wingspan, and the essential part plays out in the last tenth of that — that is, during the flare, and the flare alone.
| Wing height relative to wingspan | Induced-drag reduction | What the crew observes |
|---|---|---|
| Beyond one wingspan | Negligible | Nothing: performance is that established outside ground effect |
| About half a wingspan | Small, on the order of ten percent | Nothing discernible in manual flight |
| About a quarter of a wingspan | On the order of a few tens of percent | The aircraft holds level flight at lower thrust than expected |
| About a tenth of a wingspan (wheels close to the runway) | On the order of half | Near-zero deceleration, floating, reluctance to descend |
Drag fades away just when it is counted on
An aircraft’s deceleration in flight is the ratio of the difference between thrust and drag to its mass. At the landing reference speed, the aircraft flies at a high lift coefficient: the induced share of drag is then comparable to the parasite share, sometimes greater. It is exactly this share that ground effect cuts away. Just when the crew expects aerodynamics to consume the excess energy, the aerodynamic drag available shrinks to the parasite share, which is nearly insensitive to angle of attack.
Added to this is what thrust continues to supply. An engine at flight idle does not deliver zero force, and the devices that actually decelerate the aircraft — spoilers, reversers, brakes — are all conditioned on touchdown. During the flare, the only braking is aerodynamic, and that is precisely what proximity to the ground has just weakened. The aircraft does not float merely from excess lift: it floats because nothing is slowing it down anymore.
The consequence is geometric, not sensory. A weak deceleration, sustained for a few seconds at a speed on the order of the reference speed, translates into a distance covered above the runway that runs into the hundreds of feet. That distance is deducted from the remaining runway length, before the first braking means has even begun to act.
The tail sees a less-deflected flow: the nose-down moment
The horizontal stabilizer works in the wing’s wake, and therefore in its downwash. When this decreases, the flow reaching the tail is less inclined downward: its local angle of attack shifts in the positive sense, the downward-directed load it was producing diminishes, and the resulting moment on the aircraft is nose-down. Entering ground effect therefore means receiving, at the same time, extra lift and a moment that lowers the nose — two contradictory sensations that share a single origin.
In practice, holding the flare attitude requires more nose-up control input than the same attitude change would require at altitude, and the trim’s equilibrium state ceases to represent what the hand must actually supply. The magnitude of the phenomenon depends on the architecture: a high tail, less immersed in the wake, does not react the way a low tail placed squarely in it does. This is a type-specific characteristic, described in the manufacturer’s documentation, and not a universal rule transposable from one fleet to another.
The sign reverses as soon as the aircraft leaves ground effect: downwash is restored, the tail reloads downward, and a nose-up moment appears. This is the same mechanism that, after a bounce or during a takeoff rotation, adds nose-up to an input already commanded.
Why the flare does not fly like a go-around pull-up
At altitude, a pull-up is a self-damping maneuver. Increasing angle of attack makes induced drag grow faster than lift: speed drops, the maneuver brakes itself, and several hundred feet remain available to correct a misjudgment. Speed is traded for altitude, and the trade stays reversible. Near the ground, every one of these terms changes status.
| Quantity | Pull-up at altitude | Flare in ground effect |
|---|---|---|
| Available vertical space | Several hundred feet | A few feet |
| Effect of an angle-of-attack increase on drag | Strong growth in induced drag | Growth blunted by ground effect |
| Damping of the maneuver | Natural, speed drops quickly | Weak, the aircraft holds itself up |
| Possible energy trade | Speed for altitude, reversible | None: speed converts only into distance |
| Pitch response | Consistent with the input commanded | Parasitic nose-down moment to counter |
| Penalty for excess speed | Load factor, a longer pull-up | Airborne distance, touchdown shifted downfield |
| Correcting an error | Release, reset, try again | A single attempt, then a go-around |
| Quantity to monitor | Angle of attack and load factor | Rate of descent and remaining runway length |
The decisive line is that of the energy trade. A few feet above the ground, there is no more altitude to buy: excess speed can only convert into distance covered. This is why the quality of a flare is not judged by feel but by a runway reference point, and why the only correction truly available is to go around.
Excess speed is paid for twice
First payment, energy. Kinetic energy grows as the square of speed: a modest excess over the reference speed represents a more-than-proportional excess of energy, which will have to be dissipated before the stop. This term is well known to everyone; it appears in the flight manual’s distance corrections.
Second payment, less intuitive: the share of drag counted on to dispel this excess is precisely the one that dwindles. At a given mass, flying faster requires a lower lift coefficient, hence produces lower induced drag — and ground effect further reduces this already-reduced term. The parasite share, meanwhile, increases; but since the two shares are of the same order in full landing configuration, total drag changes little when speed changes by a few knots, whereas the energy to be dissipated changes as its square. The distance covered during the flare therefore does not grow in proportion to the excess speed: it grows faster than that.
Third consequence, downstream of touchdown: it occurs farther down and at a higher ground speed, and braking distance likewise depends on the square of that speed. Yet the landing-distance assessment carried out before arrival assumes a touchdown point and a flare consistent with a reference technique. A float of a few seconds therefore consumes not one margin but two: that of the distance covered in the air, then that of braking. It should be added that the approach-speed additives provided for wind and gusts protect the final approach, not the flare: the documentation specifies the point at which this excess must have disappeared, and that sentence is rarely read with the attention it deserves.
The trap. Pushing forward to force the touchdown during a float. Reducing angle of attack this close to the ground does not merely lower lift: it also removes part of the gain brought by ground effect, and the rate of descent establishes itself in a fraction of a second over the few remaining feet. The nose gear touches first, the aircraft bounces on its oleo, and the recovery reflex adds nose-up to a nose-up moment already present: tail clearance then becomes the limiting factor. The symmetrical trap exists and is seen just as often: waiting, with no reference point set in advance, until the remaining length no longer permits a go-around.
The right reflex. Before the flare, set a ground limit — an intersection, a marking, a beacon — beyond which the aircraft will not be put down, and call it out loud. The decision becomes geographic and binary, and therefore tenable a few feet above the ground, whereas a decision based on a feel for energy is not. The rest of the technique consists of holding the attitude obtained rather than continuing to pull, and letting speed fall away instead of chasing it.
What a bounce reveals about the mechanism
A bounce is not a lack of finesse: it is a change of regime. On leaving the very last few feet, the aircraft exits the zone where ground effect was most intense; induced drag is restored, the lift obtained at a given attitude decreases, and speed has already fallen during the preceding float. Angle of attack is high, the margin is thin, and the second contact can be harder than the first even though nothing more was done.
This is what underlies the distinction, present in most documentation, between a light bounce — hold the attitude, adjust thrust, land — and a large bounce, which calls for a go-around. The exact line is specific to the type and the operator, as is the point of no return beyond which, spoilers deployed or reversers engaged, the landing must be carried through to its conclusion.
| What is observed | Mechanism at work | What it costs | Workable outcome |
|---|---|---|---|
| Prolonged float, stable attitude | Reduced drag and residual thrust: nothing decelerates | Distance covered in the air, touchdown outside the zone | Go-around at the reference point set before the flare |
| Ballooning at the start of the flare | Excess lift at a given angle of attack, speed too high | Gain in height, speed spent for nothing | Hold the attitude; go around if height keeps increasing |
| Long touchdown followed by a bounce | Exit from ground effect at high angle of attack and low speed | Harder second contact, tail clearance eroded | The type’s written criterion: land or go around |
| Touchdown on the nose gear | Angle of attack deliberately reduced at low height | Pitch oscillation, structural loading | None: this is prevented, not corrected |
The last line is the one that receives the least attention in training, because it corresponds to a mistake rather than a random hazard. Yet it is the most instructive: it shows that a flare can be destroyed not by a lack of energy, but by a deliberate action mistaken for a remedy.
On takeoff, the same effect in reverse
The phenomenon is symmetrical, and its takeoff version costs more. Leaving the ground at insufficient speed means taking off benefiting from lift that ground effect boosts and drag that it reduces. The flight seems established. Then the aircraft climbs and leaves the phenomenon behind: induced drag is restored, the thrust required increases, and the acceleration needed for the obstacle-clearance flight path is no longer available. The aircraft can then stop climbing, or even move closer to the ground, without any engine parameter having moved.
Added to this is the nose-up moment that appears on leaving ground effect and adds to the rotation input already commanded. This is one of the reasons the rotation rate is prescribed rather than left to judgment, and why the temptation to rotate early on a short or contaminated runway works exactly against the intended goal: the aircraft becomes airborne sooner, but in a state where it no longer accelerates.
What to go check in your own documentation
A few points are worth establishing before they are needed. Where does the initiation height you use come from: a call prescribed by the manufacturer, or a fleet habit never written down? Is the thrust reduction during the flare described as an event or as a rate, and tied to what height? Is post-bounce handling formulated with an observable criterion, or passed on orally from one pilot to another? Do approach-speed additives have an upper bound, and does the documentation explicitly state where the excess must have disappeared?

Two gray areas remain. The first is the operational definition of a long touchdown: most operators prescribe a touchdown zone without saying what to do about its second half when the runway is wet, even though that is precisely where the float is paid for twice. The second is the behavior of an automatic flare, which is not that of a manual flare: the control law has its own management of height and thrust, and the comparison of the two techniques on the same type is rarely documented on the crew side, even though it would explain many surprises when switching from one to the other.
For anyone preparing a type rating or reviewing a transport aircraft performance training module, the question posed concerns the coupling, never the percentages: ground effect simultaneously gives more lift, less drag, and a nose-down moment. Being able to state these three signs, and say which one reverses on leaving the phenomenon, is enough to handle most individual cases, from floating to early rotation.
This article describes a physical mechanism and its usual consequences; it does not stand in for any procedure or approved flying technique. In case of divergence, the applicable regulation, the flight manual, and the operator’s approved documentation prevail over what is written here.
Questions fréquentes
Is ground effect really a cushion of compressed air under the wing?
No, and the image is misleading about the consequences. What proximity to the ground changes is the field of velocities induced by the wing’s vortex system: the ground forbids any vertical velocity through it and cancels part of the downwash. The result is more lift at a given angle of attack and, above all, much less induced drag — something a simple pressure cushion would not explain.
Why does the flare require more nose-up control input than a pull-up at an equivalent altitude?
Because the horizontal tail works in the wing’s downwash, and this decreases near the ground. The flow reaching it is less inclined downward, the downward load it was producing is reduced, and this results in a nose-down moment on the aircraft. Holding the flare attitude therefore requires countering this moment, in addition to producing the attitude change itself.
From what height does ground effect start to be felt?
The relevant parameter is not height alone but its ratio to wingspan. The effect becomes measurable below a height on the order of one wingspan, and the essential part plays out in the last tenth of a wingspan — that is, during the flare. This is why the same radio-altimeter height does not place two aircraft of very different wingspans in the same aerodynamic state.
Why does a few extra knots at the threshold cost so much distance?
Because the surcharge is paid twice. The energy to be dissipated grows as the square of the speed, and at the same time a higher speed requires a lower lift coefficient, hence produces less induced drag to dissipate it — a term that ground effect reduces still further. The distance covered above the runway therefore grows faster than the excess speed, and braking then starts from a higher ground speed.
Should you push forward on the controls to put down an aircraft that is floating?
No — that is the action that turns a float into an incident. Reducing angle of attack at low height removes part of the lift gain tied to ground effect, the rate of descent establishes itself in a fraction of a second, and the nose gear touches first, with a bounce and pitch oscillation as a result. The clean way out is a go-around, decided at a runway reference point fixed before the flare.
Is takeoff affected by the same phenomenon?
Yes, and symmetrically. An early rotation leaves the ground with lift boosted and drag reduced by ground effect; as the aircraft climbs, it leaves the phenomenon behind, induced drag is restored, and the acceleration needed for obstacle clearance disappears. The nose-up moment that appears on exit also adds to the rotation input already commanded, which is why the rotation rate is prescribed.

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