Methodology

The page that keeps us honest. For every module: where the data comes from, what we do to it, where it fails, and how often it moves.

Three commitments hold across the site. We aggregate rather than accuse: open AIS alone cannot attribute intent, so no module names a vessel in a sensitive matter. We link rather than copy: every figure carries its source. And we would rather publish a smaller, honest number than a larger, fragile one.

Incident database

Sources: NGA ASAM archive (public domain, closed June 2024); UKMTO advisories, MICA Center and IMB PRC reporting for the current period, curated by hand.
Method: Each event is typed by a keyword hierarchy (hijack > mine > jamming > seizure > strike > boarding), geocoded from the reported position, and zoned by bounding box. Vessel details are carried only when publicly reported by the source.
Limits & bias: ASAM understates robbery in under-reported regions and stopped updating in June 2024. Keyword typing misclassifies edge cases, roughly 3 to 5 percent on manual review. Curated 2026 entries cover the major reported events, not every incident; each carries its source link and positions are often approximate.
Frequency: Archive: fixed. Curated period: weekly, with the brief.

Strait transit monitor

Sources: IMF PortWatch daily chokepoint transit calls (satellite AIS via the UN Global Platform); live feeds from Fintraffic Digitraffic (Baltic, CC BY 4.0) and aisstream.io (world straits).
Method: Daily transit calls are aggregated into complete ISO weeks per chokepoint. Each week is compared to the same chokepoint's 2024-25 baseline; the rerouting index divides weekly Cape of Good Hope roundings by Suez transits.
Limits & bias: Transit counting misses transponder-dark passages by design. PortWatch data lands with a few days' lag and its methodology can be revised upstream; baselines inherit whatever bias the 2024-25 reference period carries.
Frequency: Weekly, with the Monday refresh; the live strait panels update in seconds.

AIS integrity watch

Sources: aisstream.io live position stream (volunteer shore receivers), processed in memory; Natural Earth 50m land polygons for the on-land test.
Method: One bounded listening session per week (~12 minutes) covers all six zones at once. Five heuristics run on the raw stream: jamming clusters, impossible jumps, speed anomalies, dark gaps above a duration threshold, and positions solidly on land. Only zone-level counts are written.
Limits & bias: Sampling makes the counts a floor, not a census — comparable week to week, biased low wherever receiver coverage is thin. Heuristics measure the signal environment, not intent: a dark gap can be a failure, not a choice. Thresholds trade sensitivity for false positives and are documented in the repository. No vessel is ever identified, which is a deliberate cap on what this module can claim.
Frequency: Weekly sampled session, with the Monday refresh.

War risk zones atlas

Sources: Public circulars and industry notices (JWC listed areas, industry HRA notices).
Method: Each area is stored as an approximate perimeter with a validity interval and its source circular, then rendered over the incident record of the selected month.
Limits & bias: Perimeters are reconstructions for visualisation, not the legal texts. The circulars themselves are the only authoritative reference; ours may lag or simplify.
Frequency: On publication of relevant circulars.

Weekly risk brief

Sources: Everything above, plus cited public reporting.
Method: Around three hundred words and one chart, written by a human, with the chart following the observatory's figure conventions: the title states the conclusion, the source sits below.
Limits & bias: An editorial product: selective by design. What it omits is a judgement, and the judgement can be wrong.
Frequency: Weekly, Mondays.

Corrections: if a figure or classification is wrong, write and it will be corrected with a note. Errors are part of the method; hiding them is not.