Drones over the fence: what Remote ID reveals near US airports

What this report is, in one minute

Most drones flown in the United States are required to broadcast a Remote ID signal, a kind of digital licence plate that says where the drone is, how high it is, and where its pilot is standing. The requirement tracks FAA registration, so it applies to drones weighing 250 grams or more and to any drone flown commercially under Part 107, whatever its weight.⁴ Dronetag operates one of the world's largest networks of receivers that listen for those signals. When a receiver picks one up, we know a compliant drone was in the air at that place and moment.

We took every Remote ID flight our network saw close to a US airport, lined those flights up against the airport's exact coordinates, and asked two simple questions. How near did the drone get, and how high did it climb. Then we compared the heights to the rules the Federal Aviation Administration sets for that airspace. The airports here sit inside controlled airspace, the tightly managed zones around runways where a drone is not supposed to fly at all without prior permission, and where the permitted height is often just a few dozen feet or nothing at all.

The lines a drone is not meant to cross

To call a flight an incursion, you need a rule to measure it against. US drone flight runs mainly under the FAA's Part 107, and three limits matter for everything that follows.

Two practical points follow from this. First, the airports in this study are commercial and military fields that sit in controlled airspace, so any flight in these perimeters is one that should have been coordinated with the FAA in advance. Second, the single hardest and most defensible line to test is the 400 ft ceiling, because it applies everywhere and needs no assumptions about which grid square a drone was in. When a drone's height above its own take-off point passes 400 ft this close to a runway, it has crossed a line that almost no routine authorisation would have opened.

Schematic only. The grid value at any given point is set per square on the FAA UAS Facility Map for that airport and is not to scale here.

What the network actually saw

Between April 2025 and June 2026, receivers on the Dronetag network logged 897 distinct Remote ID drone flights that came within a mile and a quarter of a US airport reference point. Those flights touched 48 airports spread across 24 states, and were picked up by 73 different receivers. A further set of aviation detections on the UAT band, the signals crewed aircraft broadcast, was identified and removed before any of the figures below, so what remains describes drones and only drones. Importantly, most detections were caught by receivers already deployed in the vicinity for other reasons rather than by sensors mounted at the airports themselves, which is why we treat them as secondary, opportunistic detections rather than a fixed monitoring census.

Where in the country

Every airport the network caught activity near, plotted across the country. Each circle marks an airport, sized by how many Remote ID flights we recorded within a mile and a quarter of it. The clustering on the west coast reflects where our receivers are dense rather than where the airspace is busiest, so read the map as a picture of coverage, not a ranking of risk.

Where the flights clustered

The table below ranks the airports by how many Remote ID flights the network caught nearby. Read the concentration as a map of where our receivers are dense, not as a league table of the worst-behaved airspace in the country. California dominates because network coverage there is strong, not because it is uniquely troubled.

Grid breach counts flights that exceeded the pre-approved ceiling for their exact FAA grid cell. Above 400 ft is the subset that also passed the blanket Part 107 ceiling. Both are measured among flights that carried an above-take-off height, so a dash means no height was recorded at that airport rather than a clean record. Closest approach and max height use above-take-off readings.

How high were they really flying

This is where the report earns its title. For the 496 flights with a trustworthy height above their take-off point inside the perimeter, we plotted how high each one climbed and set it against the ceilings. The distribution is not reassuring. Most flights sit in the band a lawful drone might occupy, yet a long tail pushes well past where any routine approval would reach.

A word of honesty about the number itself. Height comes from the drone's own broadcast, and the most reliable form of it is height above the take-off point, which is what every altitude figure in this report uses. We deliberately set aside readings expressed relative to the ellipsoid or as pressure altitude, because those are not measured from the ground and would inflate or distort a direct comparison to an above-ground ceiling. That choice removes roughly a hundred flights from the altitude analysis and makes the exceedance figures a floor rather than a ceiling.

Against the FAA grid, cell by cell

The 400 ft ceiling is the line that applies everywhere, which is what makes it hard to argue with, but near an airport it is also the most generous line in the sky. The FAA divides the airspace around every towered field into a grid of cells roughly a quarter of a square mile each, and gives every cell its own maximum altitude for near-automatic authorisation, stepping from 400 ft all the way down to zero. To see how these flights truly sit, we took the actual FAA UAS Facility Map, found the exact cell beneath each flight's highest point, and read that cell's own ceiling.⁵

The result is stark. Of the 496 flights with a height inside the perimeter, 476, or 96%, were above the altitude their own cell pre-approves. The reason becomes obvious once you see the map. 475 of them were flying in cells rated at zero feet, the surface areas wrapped around the runways where the grid pre-approves no drone flight at all and any operation would need slow manual coordination that is rarely granted.

This is not the story of one careless group. Broken down by class of operation, 97% of professional and enterprise flights and 95% of consumer flights were above their cell ceiling. The pattern is close to universal, which is the point. Whatever the operator or the airframe, flying a drone this close to these runways means flying above the altitude the airspace pre-clears for that location.

The fleet behind the signals

A Remote ID broadcast carries a serial number rather than a plain-language name. Dronetag's own recognition service does the translation, resolving that serial into a specific make and model, which lets us describe the fleet near these airports down to the exact aircraft. The picture is heavily consumer and prosumer, dominated by a single manufacturer, and broadcast overwhelmingly over Wi-Fi.

Among identified aircraft, roughly 83% were DJI, led by heavier survey and inspection platforms such as the M350 RTK alongside popular compact models like the Mini 4 Pro. That range invites an obvious question. Is this mostly hobbyists straying too close, or working operators going about their business?

Professional or consumer?

Sorting every resolved model into the class of operation it is built for gives a clearer answer than manufacturer alone. We use three groups. Professional and enterprise covers platforms designed for commercial, survey, agricultural, public-safety or defence work, the Matrice and Enterprise lines, Agras sprayers, Skydio, Fotokite and the like. Consumer and prosumer covers the camera and FPV drones aimed at hobbyists and content creators. A third group sets aside the standalone Remote ID broadcast modules, since those are fitted to a third-party airframe we cannot identify from the signal.

Setting aside the modules and the unidentified, the split between resolved airframes is close to even, around 51% professional and enterprise against 49% consumer and prosumer. This is not a picture of stray hobbyists alone. Capable working platforms are near these airports in roughly equal number, which cuts both ways for interpretation. Professional operators are far more likely to hold an authorisation, so their presence is not in itself a red flag. At the same time it means a large share of this traffic is heavier, faster and more capable than a toy drone.

Model by model

Because our recognition service resolves the serial rather than a category, we can go all the way to the individual model. It identified 47 distinct drone models across the sample. The chart shows the twenty most common, and the full breakdown follows beneath it.

Around seventy detections carried a Dronetag Beacon or Mini transmitter fitted to a third-party airframe. The Beacon is one of our best-selling products and is widely used by pilots and manufacturers to meet the Remote ID requirement, so these represent ordinary operator flights and are counted as such. A further 74 flights are logged here as unidentified. For these the network captured the drone's Remote ID location broadcasts, enough to place it near the airport, but not the separate identification message that carries the serial number. With no serial there is nothing for our recognition service to resolve into a make and model, and for most of these the broadcast transport went unrecorded as well. These partial detections cluster at a handful of sites rather than spreading evenly, which points to local receiver coverage rather than anything about the drones themselves.

Mostly daylight, not always

Activity follows the pattern you would expect from camera and mapping work, concentrated through the daytime and into the afternoon in local time. A smaller group, around forty flights, fell in the late-night and pre-dawn window when additional night-operation requirements apply and when a drone is hardest for anyone on the ground to see.

How the analysis was built

The pipeline is deliberately simple so that each step can be checked. Five datasets feed it.

• Airport reference points. Coordinates for US commercial primary airports and active army airfields, used as the fixed points every distance is measured from.
• Perimeter detections. Remote ID flights whose broadcast placed them within a 1.2-mile radius of an airport reference point, drawn from the detection network's database.
• Per-flight telemetry summary. For each flight, the closest point of approach to the airport and the highest point reached inside the 1.2-mile circle, with the altitude type recorded for every reading.
• Enriched flight records. Broadcast transport and aircraft type for each operation, with make and model resolved from the Remote ID serial by Dronetag's recognition service, used for the fleet analysis.
• FAA UAS Facility Map. The FAA's published grid of per-cell altitude ceilings, matched to each flight's location for the cell-by-cell comparison.

The steps

1. Match each detection to its nearest airport by coordinate. Fifty of the fifty-one reference points matched a listed airport exactly; one did not map cleanly and its flights are noted rather than leaned on.
2. Identify and remove aviation detections. Records broadcasting on the UAT band, the transport used by crewed aircraft, carry no drone make, model or type and only aviation altitude references. All 75 were excluded before analysis, leaving 897 drone flights.
3. For every flight, take the closest horizontal approach to the airport and the maximum height reached within the 1.2-mile circle. Distances are horizontal great-circle distances and do not include altitude.
4. Keep only above-take-off height readings for any comparison to an above-ground ceiling. Discard ellipsoid and pressure-altitude readings from the altitude analysis to avoid mixing reference frames.
5. Compare the retained heights to the 400 ft Part 107 ceiling for the blanket figure, then to the FAA grid for the per-cell figure.
6. Overlay each flight's highest point onto the FAA UAS Facility Map by point-in-polygon, read the ceiling of the exact cell containing it, and compare the height to that ceiling. Where a point fell inside more than one overlapping cell we took the most permissive ceiling, so the exceedance count is conservative rather than inflated. Flights recorded at or below their take-off height were not treated as exceeding a zero-foot cell.
7. Resolve make and model from each Remote ID serial with Dronetag's recognition service for the fleet composition, and derive approximate local time from longitude for the timing view.

The honest boundaries

A finding is only as strong as the caveats it survives, so here are the ones that keep this analysis defensible. None of them soften the core result, and several of them strengthen it.

• This counts only drones that broadcast. Silent or non-compliant drones never enter the dataset. Neither do the lightest recreational craft, since drones under 250 grams flown for fun are not required to register or to broadcast Remote ID at all. Every number here is a lower bound on real activity, not a full census.
• Coverage is opportunistic. Detections come mainly from receivers already in the vicinity, so the geographic spread reflects where the network listens, not a ranking of the most-flown or riskiest airports.
• We do not claim any single flight was illegal. Some operators, particularly professionals, may have held an authorisation or waiver. What the data supports is the pattern and the scale, above all the share of flights that exceeded a ceiling almost no routine approval would open near a runway.
• Height is drone-reported. Values come from each aircraft's own broadcast. We reduce the risk of misreading by using above-take-off heights only.
• The grid comparison uses each flight's highest point. We match that single peak to its FAA cell rather than tracing the whole path cell by cell, which would need full point-by-point telemetry. The zero-foot reading also carries a specific meaning worth repeating: it marks airspace with no pre-approved altitude, not a blanket ban, so a flight there is above what the grid pre-clears rather than automatically unlawful.

Sources and notes

¹ FAA, Part 107 operating limits and the 400 ft above-ground ceiling, with waivers required to exceed it under 14 CFR §107.51(b). See faa.gov and the FAA Aeronautical Information Manual, Chapter 11.

² FAA, drones in controlled airspace around airports require prior authorisation via LAANC or FAA DroneZone before flight. See faa.gov UAS Data Exchange (LAANC).

³ FAA UAS Facility Maps set the maximum altitude for near-automatic authorisation per grid square, stepping from 0 to 400 ft. Regulatory details are current at the time of writing and should be reconfirmed against the FAA source before external publication.

⁴ FAA, Remote ID applies to any drone that must be registered. Registration is required for drones weighing 250 grams (0.55 lb) or more flown recreationally, and for every drone flown under Part 107 regardless of weight. Sub-250 gram recreational drones are exempt. See faa.gov Remote Identification of Drones and drone registration guidance.

⁵ Cell-by-cell ceilings are read from the FAA UAS Facility Map dataset, the same grid that underpins LAANC authorisations, using the map effective in July 2026. The FAA describes the maps as job aids for processing authorisation requests rather than a legal instrument, and refreshes them periodically, so the per-cell figures should be paired with the data date and read as the pre-approval guideline they are.

Scope note: all detections describe drones broadcasting Remote ID. This report does not characterise drones that do not broadcast, and Dronetag's detection scope is defined by its receiver network and the Remote ID standard.