Bergen

Method. A connectivity score across our own route network: a port reads higher when it is navigationally close to many other well-connected ports. The score is rescaled 0–100 within the snapshot, so the single most-connected port reads 100. Distances are Suez / Panama / Malacca-aware navigable sea miles.

Coverage. The route network spans the 180 largest commercial ports, so this ranks hubs within that covered network, not against every port on earth. The number is deterministic — no confidence grade is invented. Computed Jun 30, 2026.

Method. Each signal is normalised to 0–10 against an empirical cap, then blended weighting safety (detentions 0.40, casualties 0.35) above operational congestion (0.25). A port is scored only on the signals it has data for, and the weights renormalise — a missing signal is never credited as a safe 0.

Coverage. PSC and casualty data here is regional (US, UK, Canada), so most ports show only congestion and carry a low-confidence flag. Detention/casualty counts come from a country-scoped name match (≈60% of US detentions resolve); unmatched records are dropped, not force-fit.

Detention and casualty signals are screened against open port-state-control and marine-casualty records, combined with our own AIS-derived congestion. Updated Jun 23, 2026.