← all methods · TT-001 instrument page · raw markdown: ?raw=1
Grade: OBSERVED. Read that precisely: we heard the broadcast. It does not mean the
broadcast was true.
Ships broadcast their own identity, position, speed and navigational status over VHF radio.
AIS is mandatory under SOLAS for vessels over 300 gross tons on international voyages. We
record those broadcasts continuously at ten US container gateways and derive events from them.
AISstream.io (a free aggregator of terrestrial receivers) → our own recorder → a deterministic
event detector → SQLite. We hold the archive; nobody can revise it under us. Equally, nobody
else guarantees it: AISstream is a free service with no availability commitment.
| Position, speed | From the vessel's GPS. Reliable. |
| Navigational status | Set by the crew. Routinely stale or wrong. |
| Vessel type | Set once at installation. Codes 70–79 are all "cargo" — AIS alone cannot distinguish a container ship from a bulker. |
| Berthed / anchored / departed | Inferred by our detector from position, speed and dwell against terminal geofences. Our judgement, not the ship's claim. |
Thresholds: 5 min persistence on navigational status, 20 min on inferred state, stationary
below 0.3 kt, underway above 1.0 kt, 30 min minimum before a contradiction is recorded.
Terminal geofences: 19 polygons seeded empirically from 40 observed berthings.
1 · A ship we cannot hear is indistinguishable from a ship that is not there.
Terrestrial AIS is usually said to reach roughly 40–60 nautical miles. **Measured
2026-07-30, ours reaches further**: 47.6% of received positions came from beyond 60 nm,
with maxima of 150–213 nm, so the feed is not confined to a shore horizon. Coverage loss
still looks exactly like absence, and this measures PRESENCE, not completeness — we may
hear some ships at 120 nm while missing most of them.
Boxes were widened to ~200 nm reach on 2026-07-30. They had reached only 5–17 nm
(Oakland 5, LA/Long Beach 8), while the June-2022 PacMMS process asks vessels waiting for
LA/Long Beach to hold beyond 150 MILES from shore — which from a port centre can be
180–250 nm. At the old width a repeat of the 2021 queue would not merely have been
missed: the number would have FALLEN, because ships pushed offshore leave the box
entirely. Inflow rose from ~90 to ~463 positions per minute.
The widened boxes overlap, so attribution is NEAREST-CENTRE, never first-match.
Savannah's box contains Charleston's berths; under first-match every Charleston ship was
relabelled Savannah, and Charleston fell to 3 positions against Savannah's 157 before it
was caught. Beyond about 100 nm proximity is the wrong rule regardless — the boxes carry
plenty of traffic bound for Vancouver, Singapore and Yokohama — so attribution for
analysis is voyage- and destination-linked, in berthing/gateways.js.
This is the most dangerous property of the instrument.
2 · Duration measures are left-censored while the archive is young.
Only stays that complete inside the archive can be measured. Early in the record, long stays
are still in progress and therefore invisible, so measured durations are biased low and rise as
the archive lengthens. This is not congestion easing or worsening — it is an artifact.
Recorded as BIAS-001; the affected series are withdrawn until the archive exceeds the stays
it measures.
3 · There is a hole in the archive.
2026-07-26T17:20Z → 2026-07-27T12:16Z: the websocket died without closing, so the process stayed
alive while receiving nothing. A five-minute stale-feed watchdog now terminates such sockets.
The gap is permanent and disclosed — recorded as GAP-001 — and those days are withheld.
Eight US container gateways: Los Angeles–Long Beach, New York–New Jersey, Savannah, Houston,
Charleston, Norfolk, Oakland, Seattle and Tacoma. **Los Angeles is separated from Long
Beach and Seattle from Tacoma** as of 2026-07-30 — merged, LA/Long Beach reported a
waiting rate of 0.029, concealing that all the waiting was at Long Beach and none at Los
Angeles. New York and Newark remain merged, deliberately: their berths sit within about
5 nm and no clean two-cluster split is visible yet. Indian and Colombo zones are configured but have
little or no terrestrial coverage. Mid-ocean is not covered at all.
2026-07-26. There is no backfill and there never can be: this is a recording, not a query.
Every TT-001 series and the public feed are filtered to AIS ship-type 70–79 and length ≥ 150 m.
Without it these numbers describe harbour traffic, not trade: of recent completed berth events,
399 were tugs (type 52), 243 passenger craft and 135 towing vessels, against 32 cargo ships. An
unfiltered "berthings per day" is a tugboat counter.
State the limit plainly: AIS type 70–79 covers container, bulk and general cargo
indistinguishably, so the 150 m floor is a proxy for container traffic, not a measurement of it.
A large bulker calling at a container gateway is counted; a small feeder container ship is not.