PortInsight Anchorage Index

Latest: September 2026

In September 2026, vessels spent longest at anchorage in Los Angeles (median 20.8 h) and shortest in Yokohama (13.7 h).

Los Angeles: Container ships bound for San Pedro Bay are held offshore by a vessel queuing system and are not counted, so this is anchorage time for tankers and bulk carriers rather than a container queue.

Yokohama: Yokohama shares Tokyo Bay with Tokyo, Kawasaki and Chiba, so a ship waiting at the bay entrance may be bound for any of them, and the waiting areas north and east of the measured area are left out.

Median by ship type

The headline figure covers every commercial vessel that anchored. Split by the type each one broadcasts, the same port can read very differently — a type is shown only where at least 20 stays were completed, so a dash means too few to publish rather than none.

Methodology

What is measured

Measured from public AIS data: completed stays at anchor inside the port area by commercial vessels (IMO number, 100 m and longer), by the month the vessel left the anchorage. Bunkering and lay-up anchorages are excluded from the headline and bunkering is reported separately. Stays are tracked for at most 72 hours; longer stays count as 72 hours, so P90 values at the cap are shown as "≥72 h". The figure covers all time at anchorage, not only the queue for a berth, so it is a different measure from the berth-queue figures published elsewhere and can read higher or lower than them.

Why other sources show different numbers

Port congestion figures from different sources measure different things, so ours can read higher or lower than theirs without either being wrong. A berth-queue figure counts only the wait for a berth, often for container ships alone; ours also counts time at anchor waiting for orders, fuel, crew or a tide, which makes it higher. An average over every arrival includes the ships that went straight to a berth; ours is measured only over the ships that anchored. Each port page publishes the share of calls that anchored at all, and that share is what a median should be read against.

Our figure reads lower where ships wait outside the area we monitor. The wait happens, we do not see it, and what remains inside the area is shorter. Where we know this is the case the port is described below, and where it would make the headline mislead, the port is withheld rather than published.

Where the monitored area stops short

Singapore: Waiting grounds beyond Singapore’s own waters — the anchorages off the Malaysian east coast and in the Malacca Strait — are observed separately and reported beside the headline, never in it, so that Singapore stays comparable with the other ports.

Antwerp: The waiting anchorages for the Scheldt ports — Westhinder and Oostdyck off the Belgian coast, and Steenbank off the Oosterschelde — lie outside the monitored area, which covers the river and the port. Most stays measured inside it are short stops in the dock area, so the figure understates the time ships actually wait. Those anchorages are shared with Ghent and Zeebrugge and cannot simply be added to Antwerp, so they are observed separately from October 2026, and Antwerp is withheld from the index until they have been watched for a full month.

Los Angeles and Long Beach have run a vessel queuing system since 2022: container ships bound for San Pedro Bay hold position far offshore and are given a berth slot before they approach, rather than queueing at anchor. That wait happens outside the measured area, so this figure covers the ships that do anchor in the bay — mostly tankers and bulk carriers. Read it as anchorage time for that traffic, not as the container queue and not as a measure of whether the port is backed up.

Yokohama shares Tokyo Bay with Tokyo, Kawasaki, Chiba and Yokosuka, and ships waiting at the entrance to the bay may be bound for any of them. AIS destination fields are too inconsistent to tell them apart: of the commercial ships at anchor in the measured area, only about one in ten names a Tokyo Bay port at all. The measured area therefore reaches north far enough to include Kawasaki, while waiting areas further north towards Tokyo and east towards Chiba are excluded, because counting them would mean counting other ports’ traffic. The figure is best read as anchorage time in the Yokohama approaches rather than a clean count of Yokohama arrivals.

When a month is published

A port’s month is published only when it was measured on all but at most two days, from at least 100 completed stays, with no more than 15% of candidate stays dropped for missing vessel data. A month that falls short is shown as insufficient rather than published with a weaker figure, and an index month needs at least 5 ports that pass.

The series starts in September 2026, when PortInsight moved to a new AIS data provider. Earlier months are not comparable and are not published.

Changes to the measurement

• 21 Aug 2026: AIS data moved to a new provider. Stays measured before this were fragmented — Singapore read about 2 hours where the same waits now read about 17 — so earlier months are not published.
• 19 Sept 2026: Hamburg’s monitored area was extended to the Elbe Reede and German Bight anchorages. Before that only short stops on the river were seen, so Hamburg is published from October 2026.
• 20 Sept 2026: Singapore’s waiting grounds outside its own waters began to be observed. They are reported separately and never enter the headline.
• 1 Oct 2026: “Vessels at anchor” was found to be counting ships at a berth. It is now the average number at anchor at any one time, worked out from total time at anchor, so it cannot contradict the median.
• 6 Oct 2026: Antwerp’s monitored area was found to leave out the Scheldt waiting anchorages: a probe counted 36 anchored commercial vessels in them, at least 15 bound for Antwerp. Antwerp is withheld until those anchorages have been observed for a full month. Waiting grounds outside any port’s area are now counted with the same anchoring rule as the headline, so Singapore’s separate figure is not comparable across this date.