How long container ships wait for a berth at ten US gateways, against how long they then spend working cargo. Anchoring is inferred from physics, not from the crew-set status field: a ship at anchor swings through a wide arc, a berthed ship is held by mooring lines. Derived from the TT-001 archive, so a threshold change is a re-run.
Hours waiting at anchor before a berth — RIGHT-CENSORED AND THIN. 44 of 206 calls anchored but had not berthed when the window closed, and those are the LONGEST waits, so the bias grows exactly when congestion is worst — in a crisis this would sag rather than spike. Only 92 of 1,241 observed anchor hours (7%) are verified as queueing. Usable once the archive exceeds the waits it measures.
· from 1970-01-01, open-ended
Hours alongside berth by gateway — WINDOW-CLIPPED. Berth durations are floored by the length of the archive, not measured: 34% of berth hours sit in stops reading the full archive length, where a genuine container call runs 12-48h. Inflates the figure and so deflates the waiting rate. Usable once the archive is long relative to a call.
· from 1970-01-01, open-ended
Average waiting rate by gateway — THREE OPEN BLOCKERS. (1) Right-censoring deletes the longest waits and the bias grows with congestion. (2) Not reconciled against the World Bank CPPI annex, whose published berth-share converts arithmetically to this same rate. (3) The fleet-mix check is unrun: waiting falls with ship size while service time rises, so this could measure fleet composition rather than port performance. See berthing/PHASES.md.
· from 1970-01-01, open-ended
Their pages stay up and their data is not served. A correction is worth publishing.
Source: Own AIS receivers, via the TT-001 archive. The full method — what is measured, what is inferred, and what can mislead — is published as TT-009 method.
describe_series, query_series, check_agreementtrimtab://method/TT-009