Why Does a Drone Float Before Landing? Ground Effect and Fountain Flow
What the research says about the cushion of air near the floor, why a quadcopter is four jets rather than one rotor, and what that means when you land your own build.
Bring a small drone down toward a pad and, a few centimetres up, it seems to stop. You are still lowering the throttle, yet the drone hangs there, sometimes drifting a little, before it finally settles. Pilots call it “floaty”. The usual name for the effect is ground effect, and the usual explanation is a cushion of air under the props. That explanation is right as far as it goes. The part most explainers skip is that a quadcopter is not one rotor. It is four jets that meet under the drone, and that changes what happens in the last few centimetres.
This article follows the Roverwing visual explanation, then separates measured results from the simplified animation and considers what they mean for a drone build.
1. Far from the floor, the air just goes down
A multirotor flies by throwing air downward. Each propeller pulls air from above and pushes it into a column under the drone, and the reaction holds the drone up. Far from any surface, that column carries on downward without interference. The whole system is steady: the same throttle holds the same hover.
2. Near the floor, the air spreads sideways
Bring the drone close to the ground and the picture changes. A study of rotorcraft outwash in the NASA technical reports server (Tanner and colleagues, 2015) describes how the downwash must rapidly spread out near the ground: a mostly vertical stream turns into an outwash that runs sideways along the floor as a thin, fast layer called a wall jet. In the same paper, the thrust boost is described as significant when a rotor is about two rotor radii, one rotor diameter, or less above the ground, and a rotor at three radii or more is generally treated as out of ground effect.
That one-diameter figure is where “within about one prop width” in the video comes from. It is a rule of thumb for rotors in general, not a fixed height for every drone. A drone with 10 cm props feels it much lower than one with 30 cm props, because the distance scales with the rotor.
3. The cushion: pressure builds, the same throttle lifts more
What does the blocked, redirected air do to the props? Gao and colleagues (2019) measured it on a Crazyflie 2.0, a palm-sized quadcopter with a rotor radius of 23 mm. Using particle tracking, they saw that near the ground the jets under the rotors stagnate. They read that as a region of higher pressure under the rotors, with the downwash reduced. On a tethered Crazyflie, net lift rose as the drone came closer to the floor at every throttle level they tested. So the same throttle lifts more near the floor, which is the “float”.
The same paper measured a larger drone, an AscTec Hummingbird with 9.8 cm prop radius. It needed up to 9.63 % less throttle to hover 5 cm above the ground than far from it, and the effect began to show in the throttle below about 0.4 m. Those numbers belong to that drone. They are examples of the scale, not constants: the height where it starts depends on how big the props are, and the size of the gain depends on the drone.
The classic theory for this is Cheeseman and Bennett’s work on helicopter rotors from 1955, which treats the ground with a mirror-image trick and predicts thrust rising as the rotor nears the floor. Gao and colleagues point out that this model treats the rotor as a single point source and has been shown to underpredict the ground effect on small quadcopters. That is the first hint that a multirotor is not just a scaled-down helicopter.
4. The twist: four jets meet under the belly
A quadcopter has four rotors, each throwing its own jet. Near the floor, each jet spreads sideways as a wall jet, and now the wall jets from neighbouring rotors run into each other under the middle of the drone. Where they collide, they cannot keep going along the floor, so they turn upward.
Dekker, Baars, Scarano, Tuinstra and Ragni (2023, published in the Cambridge journal Flow) measured this with side-by-side rotors. The two wall jets collide along a line between the rotors and turn up as a fountain. At a height of two rotor radii the fountain rises above the rotor disks and is bent back into the rotors. That re-ingested flow is turbulent and switches back and forth between the rotors, which gives unsteady blade loading and thrust fluctuations. This study used two rotors, not four, so “four jets meet” is the natural extension of what they measured rather than something they tested on a full quadcopter.
A computational study of a whole quadcopter (Tanabe and colleagues, 2021) points the same way. In their simulation the power needed for the same thrust falls as the drone nears the ground, rises locally at about one rotor radius because air recirculates into the rotors, and then falls sharply again closer still. They also found the outwash from the gaps between the rotors is stronger and reaches much higher.
5. So is the ground always a help?
No, and this is the honest limit of the whole topic. Dekker and colleagues cite He and Leang (2020), who measured two rotors spaced two radii apart at two radii above the ground. They saw a sudden thrust drop of about 8 % below free hover, which they linked to fountain flow re-entering the rotors. So near the ground thrust is not always gained, and at some heights it can be lost.
Get Valekar Developer’s stories in your inbox
Join Medium for free to get updates from this writer.
That is why the video says thrust “flickers, even drops”. The cushion tends to add lift, the fountain can take some of it away or make it unsteady, and which one wins depends on the height, the spacing of the rotors and the drone.
6. Does it make a drone wobble?
The video uses the word “wobble” for the thrust fluctuation, and it is worth being clear about what is and is not shown. No source in this research measured a quadcopter wobbling on its landing approach caused by the fountain. What the sources do show is unsteady thrust in the fountain, recirculation into the rotors, and, in one landing study, sideways drift. Shi and colleagues (2019, the “Neural Lander” paper) flew a 1.47 kg Intel Aero quadcopter and found that a standard controller struggled to reach the ground and drifted sideways while landing in ground effect. A controller that learned the near-ground forces cut the horizontal drift by up to 90 %. The authors describe ground effect as both helpful and disruptive to stability. They do not say the fountain caused the drift. The link between a fluctuating thrust under the drone and a drone that skates or wobbles is a reasonable inference, not a measured result.
7. What to do with this on your own build
None of the sources tested a landing technique, so treat this section as practical reasoning rather than a result.
- Plan the landing rather than chasing the float. The studies do not establish a universal descent technique. Use your aircraft’s documented landing procedure and controller settings; do not treat the animation as a reason to rush the final descent.
- Expect the zone to scale with the props. Larger props start feeling the floor higher up. Do not copy a height from a different drone.
- Check your flight controller’s landing behaviour. Some controllers are tuned with ground effect in mind and slow down near the ground. The research above says the effect is real, so a landing that wobbles may be worth a tuning look rather than a pilot fix.
- Do not trust a single number. The 9.63 % throttle reduction, the 0.4 m height and the 8 % thrust drop each come from one specific drone or test rig.
What the video shows, and what it does not
The video is a 3D simulation built in Blender. The amber air particles, the red thrust arrows and the fountain column are illustrations drawn to show the mechanism, not recorded airflow. The facts behind them come from the studies listed below.
Sources
- Tanner, Overmeyer, Jenkins, Yao and Bartram, Experimental Investigation of Rotorcraft Outwash in Ground Effect, AHS 71st Annual Forum, 2015: https://ntrs.nasa.gov/api/citations/20160006428/downloads/20160006428.pdf
- Gao, Di Franco, Carter, Quinn and Bezzo, Exploiting Ground and Ceiling Effects on Autonomous UAV Motion Planning, ICUAS 2019: https://par.nsf.gov/servlets/purl/10188556
- Dekker, Baars, Scarano, Tuinstra and Ragni, Unsteady flow behaviour of multi-rotors in ground proximity, Flow, Cambridge University Press, 2023 (CC BY 4.0): https://www.cambridge.org/core/product/392CEF3D620FC660ADE5DDA407FA7BD9/core-reader
- Tanabe, Sugawara, Sunada, Yonezawa and Tokutake, Quadrotor Drone Hovering in Ground Effect, Journal of Robotics and Mechatronics 33(2), 2021: https://www.jstage.jst.go.jp/article/jrobomech/33/2/33_339/_article
- Shi and colleagues, Neural Lander: Stable Drone Landing Control using Learned Dynamics, ICRA 2019: https://arxiv.org/abs/1811.08027
Next in the series
Part 1 covered why a drone that descends straight down can wobble and fall (vortex ring state). Part 3 asks why bigger propellers use less power for the same lift.
The narration in the video is an AI clone of the creator’s voice. The visuals are original 3D simulation.
