Skip to main content
← Getting The Data

The Volunteer Network

Two flight-tracking systems watch the same sky. One is commercial — Flightradar24, FlightAware, ADS-B Exchange since its sale — and it is the one every screenshot in this case came from. The other is a volunteer network of private receivers, owned by nobody, feeding open databases under an open licence.

They see the same radio signal. They do not publish the same thing, they do not answer to the same people, and one of them keeps an archive the other has no obligation to keep.

This page is about the second system: how it collects, how the networks share with each other, what it will show you that the paid trackers will not, where its archives live — and the sweep this investigation built on top of it to stop asking "was this aircraft here" and start asking "what was here".

How a position gets from an aircraft to this page​

Nothing in this chain is owned by an aviation company. That is the whole point of it.

StepWhat happensWho runs it
1An aircraft's transponder broadcasts its identity, position and altitude on 1090 MHz, twice a second, unencrypted, to nobody in particular. This is ADS-B Out, mandatory in nearly all controlled US airspace since 1 January 2020.The aircraft
2A software-defined radio dongle on somebody's roof — a $25 RTL-SDR and a quarter-wave antenna — receives it.A hobbyist
3dump1090 or its modern successor readsb decodes the raw pulses into aircraft records. Both are open source.Open-source maintainers
4A feeder client forwards those records to one or more aggregators. Most feeders send to several at once.The same hobbyist
5The aggregator merges thousands of receivers, runs MLAT — multilateration, timing the same signal at four or more stations to locate aircraft that broadcast no position of their own — and serves a live map.adsb.lol, airplanes.live, adsb.fi, ADS-B Exchange
6At the end of the UTC day the aggregator writes each aircraft's whole track to a trace_full_<hex>.json file and keeps it.The aggregator
7That archive is mirrored off-site, in adsb.lol's case to public GitHub Releases under the Open Database Licence.adsb.lol
8We download it, filter it and publish what it says.This investigation

Step 7 is the step that matters most to this case, and it exists only because of step 3. The decoders are open source and the data is open-licensed, so nothing stops a third party from holding a complete copy. A commercial tracker's history is a product; a volunteer network's history is a public file. When a record disappears from the first, the second is where you go looking — and three of the four routes this investigation uses to recover erased flight data exist because of that one licensing decision.

Why there is more than one network, and what happened in January 2023​

For years the unfiltered-tracking world was essentially one site. ADS-B Exchange was built by volunteers on the explicit promise that it would never suppress an aircraft — not a billionaire's, not a government's, not a police department's.

In January 2023 it was sold to JETNET, a business-aviation data company backed by the private equity firm Silversmith Capital Partners, for a figure reported at around $20 million. (JETNET press release, Aviation International News, RTL-SDR.com)

The volunteers who had built and fed it had not been consulted and were not paid. A large number of them pulled their receivers and pointed them somewhere else, and the somewhere else had to be built in a hurry. adsb.fi, theairtraffic.com and — the one that grew the largest — airplanes.live all date from that moment. adsb.lol was already running and absorbed feeders too.

Three consequences follow, and every one of them shapes what this investigation can and cannot do:

  • There are now several independent archives of the same sky. adsb.lol and airplanes.live are not mirrors of each other. They are separate networks of separate receivers, so an aircraft present in both was heard twice, by two different sets of antennas, and that is the strongest form of confirmation a trace can have. It is also why this repo pulls both and publishes the disagreements.
  • ADS-B Exchange's own deep history went behind a paywall. Its globe_history endpoint returns 403 without a subscription. What survives free is a monthly sample — one complete day, the first of each month, back to 2016 — and that sample is the only free route into 2022 at all.
  • Coverage before roughly November 2023 is thin. The replacement networks were new. An aircraft missing from a 2022 or early-2023 query may simply have been over a part of the country where the new network had no receivers yet. That is a fact about the network, not about the aircraft, and it is the single easiest mistake to make with this data.

The part the paid trackers will not do​

This is the practical reason an investigation like this one belongs on the open track, and it is not a small effect.

The FAA runs two programmes that let an aircraft owner disappear from public view:

  • LADD — Limiting Aircraft Data Displayed. The owner asks the FAA to suppress the tail from the FAA's own data feed, which is the feed the large commercial trackers license and blend with their receiver data. LADD does not touch the radio broadcast at all: the aircraft keeps transmitting its real identity in the clear. (FAA, NBAA)
  • PIA — Privacy ICAO Address. A stronger measure: the aircraft is issued a temporary ICAO address not linked to its tail number and broadcasts that instead. This one does change what goes out over the air, so it defeats the volunteer networks as well — an aircraft on PIA appears as a real track belonging to nobody.

The commercial trackers honour LADD. The volunteer networks do not, because they take no FAA feed and are under no obligation to. That is the founding principle ADS-B Exchange was built on and the one its successors kept.

The size of that gap is measurable in this investigation's own data. Sweeping 10 September 2025 — the day of the assassination — the open archive put 141 notable aircraft on the ground inside the two event circles. Most of them were flagged LADD in the aircraft database: business jets at Provo and Salt Lake City whose owners had asked to be left off the public displays, sitting there in the open archive with tail number, type and operator attached.

Any research that starts at a commercial tracker starts with those aircraft already removed — and it will never know which ones were taken out, because the removal is silent. This is the strongest single argument for running the open pass first, and it is why this section's charter says the free pass always runs first.

Where the archives actually are​

Four free routes, all in use, none needing an account or a payment. The detail — endpoints, gotchas, retrieval dates and the control tests — is on Open Flight Data.

RouteWhat it isReachShape
globe.airplanes.live/globe_historyThe airplanes.live network's own daily archive~Nov 2023 → today, throughoutOne aircraft, one day
adsb.lol/globe_historyThe adsb.lol network's daily archive2023 → today, with two dead bandsOne aircraft, one day
github.com/adsblol/globe_history_YYYYadsb.lol's entire database, mirrored to GitHub Releases, ODbL, one ~2–5 GB tar per UTC day2023 → today, including the dead bandsThe whole sky, one day
samples.adsbexchange.com/tracesADS-B Exchange's free monthly sample1st of each month, back to 2016One aircraft, one sampled day

The third row is different in kind from the other three, and that difference is what the rest of this page is about. The first, second and fourth all answer "where was this aircraft" and need a hex code before you can ask. The GitHub mirror holds every aircraft the network heard that day — roughly 74,000 to 95,000 airframes — so it can answer a question none of the others can.

The geographic sweep​

Everything this investigation had pulled from the open archives before 26 August 2026 was a per-tail question: was SU-BTT at Provo on 10 September? That question can only ever confirm or fail to confirm an aircraft somebody had already named. It holds about 9% of the aircraft-days the speaking-event windows need, and it is structurally incapable of finding an aircraft nobody thought to look for. A recent hunt for overlaps at airports near the events came back empty for exactly that reason: coverage, not effort, was the ceiling.

The sweep turns the question inside out:

What was within 50 miles of this city on this day — whoever it turns out to be?

Code: apis/public_open_source/code/geo_sweep.py, with the target builder in code/lib/targets.py. No dependencies, Python 3.

How it works. Each sourced US speaking event in tpusa_events.csv becomes a circle: the US Census centroid of the event city, 50 miles in radius, which contains every airport within 50 miles of that city by construction. Each event is widened to ±1 day so an arrival the evening before and a departure the morning after are both inside the window — a UTC day boundary falls in the middle of the American evening, so this is not optional. That produces 187 event-days across 278 UTC dates. For each date the whole day's tar is streamed and never stored: curl pipes it into Python, each aircraft is decompressed in memory, and a byte pre-filter discards the ~92% that were never in the right degree square before any JSON is parsed.

What lands on disk. A CSV row for every aircraft that entered any circle — that is the coverage record, and it is the part that makes an absence mean anything. The full track is kept only for aircraft that were on the ground inside an event circle and were foreign-registered, unregistered, military or PIA, plus any tail this repo already tracks. Every day gets a _sweep.meta.json recording the release tag, the asset URLs, the HTTP result, the byte count and the UTC time of the pull. Nothing overwrites a previous sweep — a re-sweep lands beside the first, because the diff between two pulls of one URL on two dates is how this repo proves something stopped being retrievable.

The control cities are swept on the same days, in the same run, by the same code. Des Moines, Chattanooga, Spokane, Albuquerque, Syracuse and Shreveport have no connection to this case. "Foreign-registered jets were within 50 miles of the event" means nothing until you know how many were within 50 miles of Des Moines that day. This investigation has already had to retract one published finding for skipping a control test, and the control is built into the sweep rather than bolted on afterwards so it cannot be skipped again.

What the sweep found on the day itself​

10 September 2025. The archive held 74,405 aircraft that UTC day. 1,113 of them entered the 50-mile circle around Orem. Working purely from geography — no tail numbers, no hex codes, no prior list — the sweep recovered every aircraft this investigation already knew about:

AircraftTypeOn the groundNearest fieldUTC window
SU-BTTFalcon 7XyesKPVU, 0.8 mi13:07:53 – 13:27:34
SU-BNDGulfstream IVyesKPVU, 0.8 mi16:05:35 – 20:29:38
N102DZGulfstream VyesKSLC, 1.1 mi15:00:51 – 22:18:44
N888KGChallenger 300yesKPVU, 0.2 mi19:06:07 – 21:07:54
N872RAHawker 400yesKPVU, 0.2 mi22:25:35 – 23:32:34
N560TWCitation ExcelyesKPVU, 0.7 mi15:54:12 – 16:36:42
N40JDPremier IyesKPVU, 0.7 mi20:18:12 – 21:22:18
N59906NavajoyesKPVU, 0.1 mi15:08:45 – 18:08:35
N1098LGlobal Expressno — airborneKPVU, 8.4 mi14:58:21 – 19:02:25

This is a validation result, not a new discovery, and it should be read as one. Every row above was already on this site, pulled tail-by-tail from the two daily archives on 24 August 2026. The sweep reached the same aircraft by a completely different road, and the timings match the previously published figures to the second — SU-BTT's first ground fix at 13:07:53 UTC, SU-BND's at 16:05:35 UTC, matching the September 10 flight timeline exactly.

What is new is the denominator. The old method could say "SU-BND was at Provo". The sweep can say SU-BND was at Provo and so were 1,112 other aircraft, and here is every one of them, which is the only basis on which anybody can argue about whether two Egyptian jets on a Utah ramp is remarkable.

Three things in that 1,113 could not have been found by asking about known tails at all:

  • Aircraft with no registration in the database, appearing only as a raw hex address.
  • Aircraft broadcasting a non-ICAO address — the ~-prefixed records that are TIS-B or ADS-R rebroadcasts rather than a real airframe address.
  • The LADD-flagged business jets described above, which a commercial tracker would have silently removed.

What this system cannot tell you​

None of these limits is a technicality, and none of them may be dropped when a result from this sweep is quoted anywhere else on this site.

  • A trace proves presence. It never proves purpose, and it never proves occupancy. The sweep can put an airframe on a ramp. It cannot put a person on it, and no public record establishes who was aboard any aircraft in this case, or who tasked any of them.
  • A circle is not an airport and a position is not a landing. Every row carries the distance in miles to the nearest field. Reading a nearest-field label as a destination is exactly the mistake that produced the Provo-versus-Dugway correction on this site.
  • An empty day is a coverage fact, not an absence. A swept day with no hits means the archive was asked and held nothing. Transponder off, outside receiver coverage, and a wrong claimed date all look identical from here — and a volunteer network's coverage is wherever volunteers happen to live.
  • Ground coverage is the weakest coverage there is. An aircraft on a ramp is transmitting from ground level behind buildings. Some fields are heard well, some are barely heard at all, and the difference is antenna geometry rather than anything about the aircraft.
  • 2022 cannot be swept. There is no daily archive on GitHub before 2023. The following-planes claim starts in 2022, and for that year the only free source is the monthly sample — one day in thirty, which can place an aircraft somewhere but can never support a rate.
  • PIA defeats it. An aircraft on a Privacy ICAO Address is in the archive as a real track with no identity attached. It is not hidden; it is anonymous, which is different and harder.
  • 31 event rows are not swept at all. Two name only a month rather than a date, five have no usable city, and 28 of the 2022 rows fall below the archive floor. They are listed by name in data/geo_sweep/targets.json rather than quietly dropped.

Do it yourself​

Every command below is free, needs no account and no key, and can be run by any reader with a terminal. That is the test this section exists to pass.

# One aircraft, one day, straight from a volunteer network
curl -s https://globe.airplanes.live/globe_history/2025/09/10/traces/d3/trace_full_0101d3.json | gunzip | head -c 400

# The same day from the second, independent network
curl -s https://adsb.lol/globe_history/2025/09/10/traces/d3/trace_full_0101d3.json | gunzip | head -c 400

# The whole sky for one UTC day, streamed and filtered, nothing stored
TAG=v2025.09.10-planes-readsb-prod-0
curl -sL https://github.com/adsblol/globe_history_2025/releases/download/$TAG/$TAG.tar.a{a,b} \
  | tar -xf - --include '*trace_full_0101d3.json' -O | gunzip

# This investigation's sweep
cd site/docs/Planes/following/apis/public_open_source/code
python3 geo_sweep.py --plan                    # what it would sweep, in priority order
python3 geo_sweep.py --run --date 2025-09-10   # one day
python3 geo_sweep.py --run --jobs 8            # the whole target set, resumable
python3 geo_sweep.py --report                  # event circles vs control circles

One gotcha worth stating plainly, because it costs an afternoon: the files inside the GitHub tarballs are gzip-compressed despite ending in .json. Pipe them through gunzip or you will conclude the archive is corrupt.