How high do flying ants actually go?

Every summer or two a swarm somewhere gets thick enough to appear on weather radar. It happened over southern England in 2025, over Kent and Sussex in 2020, and again on 8 July 2026 during the third heatwave of that year, when Fox Weather ran the story as bands of what looked like light drizzle falling on ground that was bone dry.

The mechanism is not complicated. A winged ant is roughly the size of a raindrop and bounces the beam back in roughly the same way, so a dense enough flight reads as weather until somebody checks it against the gauges.

What none of those articles ever get round to is the obvious follow-up. If the swarm is being painted by a radar built to look at rain clouds, how far up is it?

A weather radar does not look sideways along the ground. It sweeps a beam tilted slightly upward, so the further from the dish you get, the higher the beam is sitting, and the earth curving away underneath adds to that. By the time you are far enough out to be looking at a different county, the beam is well up into the air.

That gives you a floor and nothing else. The radar tells you the ants reached the height of the beam. Whether they stopped there or carried on it cannot say, because there is no second beam above them to check.

It also tells you nothing about which ants. A radar return is a return. It does not sort a harvester ant from a black garden ant, and as it turns out that distinction matters more than I had assumed.

The gap got filled last year, and I had not registered it until I went looking properly.

Giannetti, Schifani and Grasso, publishing in Current Zoology in 2025, 3D-printed a small cone-shaped adhesive trap, mounted it on a commercial drone, and flew it through nuptial flights in Italy. The design detail that makes it work is unglamorous: they found the trap caught insects during horizontal passes but hardly anything while the drone was climbing or hovering, so they restricted the glue to the inner faces of the cone. That way a capture can only happen during level flight, which means every ant on the trap has a known altitude rather than an altitude smeared across the whole ascent.

Over 12 sampling days and 264 flights they caught 24 males and 22 females. Roughly one flight in six brought something back.

The captures, by species:

Then a scatter of bycatch: two Lasius emarginatus queens at 45 to 50 metres, a Camponotus male at 25, four Tetramorium males at 15, 15, 20 and 120, and a single Temnothorax male at 115.

So the whole spread runs from 15 metres to 120. Most of it sits between 60 and 110.

That is a good deal lower than the picture I had been carrying around. Not hedge height, plainly, but well inside the range of a tall building rather than somewhere up among the swifts.

The finding I did not see coming is what predicts the altitude.

They fitted a linear mixed model with species identity as a random factor, testing wing surface area and wind speed against capture height. Wing area came out significant and negative: bigger wings, lower flight. The small species were the ones turning up at the top of the range, and the heavyweights were near the bottom. Camponotus at 25 metres, Colobopsis at 30 to 35, and the tiny Tetramorium and Temnothorax males providing the two highest captures in the whole dataset.

It is not the intuition you would start with. A bigger insect with more wing sounds like it ought to be the one that gets up there.

It is also not the first time the result has appeared. Helms, Godfrey, Ames and Bridge got to the same place in 2016 by a completely different route. They fitted altitude loggers to Purple Martins nesting in Oklahoma and then went through what the birds were bringing back to their chicks. The martins themselves ranged up to 1,889 metres and their feeding trips topped out at 922, and ants turned up throughout the lower atmosphere to a maximum of around 160 metres. Across a 63 mg span of dry weight, maximum flight altitude fell by 60% from the lightest species to the heaviest.

Two studies sharing no equipment, no continent and no species list, pointing the same way. That is worth a good deal more than either on its own.

The Oklahoma work also turned up the exception that makes the rule readable. The cornfield ant, Lasius neoniger, is light, so it should have flown high, and instead it never appeared above about 22 metres. The reason is that it is a female-calling species: its queens mate on the ground near their own nest rather than in a swarm overhead, so they have no business being up there in the first place. Every other species in the analysis belongs to a genus that does mate on the wing.

So size predicts altitude, but only among the ants that are using the sky to meet in. Which is less a caveat than a clue about what the height is actually for.

I had this partly wrong, and it is worth saying so rather than quietly editing it out.

The version of this post I published before reading the paper argued that wind is what caps the climb: flights cut off above about 6.3 m/s at ground level in Hart and colleagues' 2018 UK dataset, wind gets stronger with height, so I reasoned the ceiling was the point where a small insect can no longer hold station. It read well. The drone data does not support it. Wind speed was not a significant predictor of flight altitude in their model, and size was.

The two findings are not in direct contradiction, and I want to be careful about the distinction. Hart's threshold is about whether a flight launches at all, measured at ground level. Giannetti's result is about how high the ants go once they are already up. A colony can perfectly well refuse to fly on a windy afternoon while wind has no bearing on the altitude it picks on a calm one. Those are different questions and the two datasets answer different ones.

But the specific thing I was doing, reaching past the ground-level threshold to explain the ceiling, was speculation, and the first measurement to bear on it went the other way. The authors do suggest wind probably matters for how far the high-flying species end up dispersing, which is a third question again.

Enough caveats that I would not treat these as the final numbers.

Four species is four species. Crematogaster, Dolichoderus, Messor and Colobopsis were sampled deliberately and everything else arrived as bycatch: two Lasius emarginatus queens, a lone Camponotus male, a scatter of Tetramorium. There are somewhere over 16,000 described ant taxa. Lasius niger, which accounts for a large share of the flights people actually stand in across northern Europe, has not been measured at all.

The samples are small throughout. Seven queens here, two there. A mean altitude from n equals 2 is a data point, not a distribution.

And a trap only finds ants where the drone flew. The two highest captures sit at 115 and 120 metres, at the top of the sampled range, which is exactly the shape you would expect if the range had been cut off rather than the ants had. The Oklahoma martins settle that one: they were taking ants at around 160 metres, comfortably above anything the drone reached. So 120 is not a ceiling, it is where the sampling stopped.

Height is close to the whole point of the exercise.

An ant that mates next to its own front door has achieved very little, because the only ants nearby are its siblings. Going up is how queens and males from different colonies end up in the same piece of sky at the same time. Hölldobler and Wilson's The Ants lays out a second benefit to everyone going at once: predator satiation, where a swarm gets through by being more ants than the local birds can eat before they fill up.

Then there is dispersal. A queen who comes down on her mother's nest is competing with her mother, and getting high enough for the wind to carry her elsewhere is what stops that happening.

There is a detail in the drone data that fits this neatly. Males and queens of the same species were caught in the same altitudinal band, despite sometimes being very different sizes. Which is what you would want if the flight is a meeting: the two sexes have to end up in the same volume of air or none of it works.

The stakes explain the effort. Mortality during the nuptial flight has been estimated at close to 99%.

There is a number now, and I was wrong to say there wasn't.

Ants are up there in numbers between roughly 15 and 160 metres, most of the traffic sits somewhere in the middle of that, smaller species go higher, wind is not obviously involved, and species that mate on the ground do not participate. That comes from about ten species across two continents, which is not many, though both methods are cheap enough and repeatable enough that this may not stay true for long. That is the genuinely encouraging part. The obstacle was never that the question was uninteresting, it was that the ants are small and the sky is large and nobody had a good way to sample it. Somebody now does.

I would still be wary of quoting a confident number for a species nobody has flown a trap through. But "nobody knows" has stopped being the right answer, and the shape of what replaced it, small ants high, big ants low, wind beside the point, is not what I would have guessed.

Giannetti, D., Schifani, E. & Grasso, D.A. (2025). An adhesive drone trap to study the flight altitude preferences of winged ants. Current Zoology 71(5): zoaf002. DOI 10.1093/cz/zoaf002. Open access. Source of every altitude figure above, the wing-area result and the wind null result.

Helms, J.A., Godfrey, A.P., Ames, T. & Bridge, E.S. (2016). Predator foraging altitudes reveal the structure of aerial insect communities. Scientific Reports 6: 28670. DOI 10.1038/srep28670. Open access. Altitude loggers on Purple Martins at Lake Texoma, Oklahoma, May to June 2014. Source of the ~160 m maximum for ants, the 60% decline in altitude from lightest to heaviest species, and the Lasius neoniger female-calling exception.

Garcia-Garin, O., Espadaler, X., Navàs, F., Castells-Morral, M., Ruiz-Sagalés, M., Abril, S., Gómez, M. & Ferrer, X. (2026). Opportunistic airport-collision samples from swifts reveal ant nuptial flight phenology. Scientific Reports 16: 17194. DOI 10.1038/s41598-026-48191-1. Open access. Ant remains in 31 of 72 swift carcasses from bird strikes at Barcelona airport, used to date flights.

Hart, A.G., Hesselberg, T., Nesbit, R. & Goodenough, A.E. (2018). The spatial distribution and environmental triggers of ant mating flights: using citizen-science data to reveal national patterns. Ecography 41(6): 877–888. DOI 10.1111/ecog.03140. Open access. Source of the 6.3 m/s ground-level threshold for whether flights happen.

Boomsma, J.J. & Leusink, A. (1981). Weather conditions during nuptial flights of four European ant species. Oecologia 50(2): 236–241. On the conditions that gate a launch.

Hölldobler, B. & Wilson, E.O. (1990). The Ants. Harvard University Press. On predator satiation and the economics of the founding bottleneck.