Singapore Strait Holds 65% Ship-to-Ship Share While Rotterdam Climbs to 59%: The STS Subtype Reads Differently From the Atlas

The chokepoint atlas told you where dark events cluster. This is what happens when you isolate the ship-to-ship transfer signature inside those clusters. Singapore is structural. Rotterdam, Houston, and the Persian Gulf are not.

The Setup

Of the 26,370 AIS dark events the detection pipeline recorded in the last 30 days, the ship-to-ship rendezvous fingerprint — vessels resuming AIS within meters of an independent reappearance, plus speed, course, and dwell signatures consistent with a transfer — fires in roughly half of all events globally. Localize that fingerprint to the chokepoints from the atlas, and the picture shifts from "where do gaps happen" to "where do meetings happen during gaps":

• Singapore Strait (1–2°N, 103–105°E): 4,496 events, 65% STS-like, mean risk 21.6, mean gap 30 hours.
• Rotterdam approach (51–52°N, 3–5°E): 786 events, 59% STS-like, mean risk 18.1, mean gap 15.6 hours.
• Houston Ship Channel (29–30°N, 95–94°W): 397 events, 51% STS-like, mean risk 19.4, mean gap 27.9 hours.
• Persian Gulf (23–27°N, 50–58°E): 101 events, 74% STS-like, mean risk 16.4, mean gap 40.1 hours.

The Singapore figure has not moved. Comparing the same lat/lon box month-over-month: the prior 30-day window also resolved at 64–65% STS share, with identical mean stsRaw weight and identical mean risk. Singapore is not surging — it is the constant against which the others should be read.

The Chain

The dark-event pipeline ingests AIS transponder records, opens a gap clock when transmission halts, and on reappearance computes a set of independent fingerprint families: kinematic plausibility, identity continuity, prior-incident proximity, and proximity-to-other-reappearance. The last of these is the STS family. When a vessel's reappearance point coincides geographically and temporally with another vessel's — and the two had no port call in between — the risk_factors.stsRaw component fires and contributes to the composite risk score. This is a heuristic, not a confirmed handshake — but at scale, the geography of where it fires tells you where ships are quietly meeting.

The atlas already established that gap-duration risk curves invert at Singapore: short gaps there are less risky than long ones, opposite the global pattern. Re-running the same data through the STS filter explains why. The 4,496 Singapore gaps include ~2,920 with the rendezvous fingerprint, and those carry a mean risk score of 21.6 — well above the 17.6 average for all Singapore gaps and well above the global STS-event mean. The same composition does not appear at Rotterdam, where STS-flagged gaps are shorter (15.6 hours) and lower-risk (18.1) — consistent with bunkering and cargo lightering rather than sanctions-evasion patterns. Houston sits between the two: its 51% STS share with 27.9-hour gaps and a risk of 19.4 looks more like Singapore than Rotterdam, which is the new finding.

The Implication

If you are reading STS exposure as a single global number, you are missing the regime split. Singapore's STS share is structural, embedded in how that geography is used commercially — the right baseline for compliance-driven risk modeling is per-port, not global. Houston's STS profile has begun to converge on Singapore's by gap duration and risk score even though its event volume is one-tenth the size; the directional read is that some share of Gulf STS activity is no longer routine bunkering. Persian Gulf coverage at 101 events is too thin for confident inference and reflects coastal AIS receiver density, not absence of behavior.

What to Watch

• Monthly STS share at Houston Ship Channel — a move from 51% toward 60%+ would mark a regime shift, not noise.
• Mean gap duration at Rotterdam — sustained drift above 20 hours would indicate the bunkering profile is changing.
• Persian Gulf event volume — if pipeline coverage there improves and the 74% share holds, the implied STS density rivals Singapore.
• Co-occurrence of identity-change flags with STS flags in Singapore — that is the sanctions-evasion signature; the atlas did not separate it.

Limitations

The STS family is a heuristic match, not a confirmed transfer — false positives include parallel anchorage clusters, fishing vessel pair-trawling, and dense traffic separation lanes where reappearance proximity is structural rather than meaningful. Persian Gulf and Norfolk samples are coverage-limited; absent results there are not absent behavior. The risk score weights STS at one component among twelve, so the headline 65% share is "STS fingerprint fired," not "STS-driven risk." Geographic boxes are coarse — the Singapore Strait box captures both the Strait proper and the OPL anchorages east of it without distinguishing them.

Data current as of 2026-04-27. Source: dark_events table, n = 26,370 events / 30d. Read the broader chokepoint atlas at /blog/dark-events-chokepoint-atlas-overwatch-2026-04-24.