Dark Events Across Eight Chokepoints: Every Port Has Its Own Gap-Duration Risk Curve

30,000+ AIS dark events in 30 days across Singapore, Norfolk, Houston, Rotterdam, Antwerp, Hamburg, and the Channel entrance. The global pattern of "short gaps are riskiest" inverts at two ports and holds at the rest. Here is the full atlas.

The Setup

The dark-event detection pipeline processes global AIS transponder data continuously. When a vessel's transmission stops, a gap clock starts; when it resumes, the pipeline checks position consistency, speed plausibility, and course continuity. Over the last 30 days the pipeline recorded roughly 30,000 detectable gaps concentrated in eight geographic zones. The per-port numbers:

Singapore Strait (1.0–1.5°N, 103.0–104.5°E): 3,779 events, 1,209 carrying risk scores above 20, 447 classified as impossible kinematics. The 12–24 hour gap window peaks at 19.5 average risk — higher than the sub-12 hour bucket at 19.2.
Norfolk / Hampton Roads: 219 events, 51 high-risk, 33 impossible kinematics, 15.2 average risk. Under-72 hour gaps average 16.0.
Houston Ship Channel: 585 events, 7.2 per day, 16.0 average risk, 215 concentrated at grain terminals, 17.8 hour average gap.
Rotterdam / North Sea corridor (50.5–52.5°N, 1–5.5°E): 1,234 events across 826 distinct vessels, 13.11 average risk, 23.2 hour average gap.
Antwerp / Scheldt estuary: 342 events, 15.8 average risk, 15.1 hour average gap, 15 gaps over 72 hours.
Hamburg / Elbe: 130 events, 19.2% classified high-risk — matching Antwerp's 17.7% and roughly 2.4× Rotterdam's 8.0%.
Channel entrance / Belgian coast (51.2–51.6°N, 1.5–3.0°E): 282 events, 11 impossible kinematics averaging 124.2 hour gaps (more than five days).
Global baseline across all ports: 6,577 events distributed as 3,084 under 12 hours (15.25 risk), 1,685 in 12–24 hours (15.82 risk), 854 in 1–3 days (16.26 risk), and 954 over 3 days (10.87 risk).

The Chain

The detection pipeline is identical at every port. What differs is the kinematic context — and that context reshapes the risk curve.

At the global level, the gap-duration risk curve peaks at 1–3 days (16.26 average risk) and falls off for gaps longer than 3 days (10.87). The 3+ day bucket contains many vessels that simply exited detection zones rather than truly disappearing. The pipeline treats the 1–3 day window as the signal-rich zone because it is long enough for a vessel to deviate meaningfully, but short enough to avoid looking like a normal inter-port voyage.

Two ports invert this pattern. At Singapore, the peak shifts forward to the 12–24 hour window (19.5 risk), driven by the world's densest transit traffic — 300+ daily vessels in a narrow lane where even short gaps become operationally meaningful. The sub-12 hour bucket is also elevated at 19.2, making Singapore's entire sub-24 hour band hot. At Hamburg, the inversion is sharper: the under-12 hour bucket carries 19.9 average risk with a 57% high-risk rate, while the 12–24 hour bucket drops to 13.4 with only an 11% high-risk rate. Hamburg's riskiest events are its shortest ones. The Elbe pilot corridor is the explanation — short gaps align with pilot boarding zones, which is where the anomalies cluster.

The Channel entrance is a third pattern. Only 11 impossible-kinematics events, but those 11 average 124 hour gaps. These are not signal losses — they are long voyages with physically impossible position jumps on either end. The pattern is consistent with AIS spoofing during multi-day transits, not with momentary transponder drops.

Rotterdam's 8.0% high-risk rate is the anomaly of the European cluster. With 1,234 events across 826 distinct vessels, the Rotterdam corridor sees the highest event volume per vessel in Europe but the lowest risk rate — a function of the port's aggressive pilotage and traffic separation regime rather than any absence of opacity. Antwerp's 17.7% and Hamburg's 19.2% imply the Scheldt and Elbe channel geometries are meaningfully tighter, not that their vessels are meaningfully dirtier.

Houston's grain-terminal concentration (215 of 585 events at the grain elevators on the Ship Channel) is the US equivalent of Singapore's Jurong concentration — a commodity corridor where dwell times are long, berth assignment is opportunistic, and AIS silence does not disrupt normal operations. Norfolk's 33 impossible-kinematics events against only 219 total makes it the highest impossible-kinematics rate in the US dataset, driven by naval-adjacent traffic that uses AIS selectively.

The Implication

Sanctions screening and maritime compliance teams running a single global gap-duration threshold are mis-calibrated everywhere they screen. The under-12 hour bucket matters at Hamburg and Singapore, the 12–24 hour bucket matters at Singapore, the 1–3 day bucket matters globally, and the 124-hour impossible-kinematics signal at the Channel entrance is a pattern no gap threshold will catch.

Specifically:

Singapore-regional screening should weight the 12–24 hour window above the sub-12 hour window. The current practice of treating all sub-24 hour gaps as equivalent under-weights the riskiest band.
Hamburg-regional screening should invert that further: the sub-12 hour window is where 57% of high-risk events sit. Treat any short gap in the pilot corridor as worth investigating.
Channel-transit screening should not rely on gap duration at all. The 124 hour impossible-kinematics signal requires position-consistency checks, not timeout thresholds.
US port screening can rely on the global default (peak at 1–3 days), but Norfolk and Houston's grain-terminal corridors need concentration-based alerting — the raw event count is moderate, but it is heavily clustered in specific berth geometries.

What to Watch

Whether Singapore's 12–24 hour peak holds when the sample extends beyond 30 days. If it stabilizes across 90+ days, it becomes a permanent per-geography detection parameter rather than a monthly artifact.
The Rotterdam 8.0% high-risk rate. If it rises toward Antwerp's 17.7%, pilotage coverage may be weakening. If it falls further, the aggressive TSS regime is working harder than current stats give it credit for.
The impossible-kinematics false-positive rate — currently 98% across 1,950 flagged events yield only 39 true violations. The rate is dominated by GPS multipath in the Singapore and Houston channel geometries. A narrower kinematic threshold (e.g. excluding 1–3 NM jumps in known multipath zones) could lift the signal-to-noise ratio dramatically.
Whether the Channel entrance impossible-kinematics cluster continues to concentrate on a small number of repeat vessels, or diffuses across the wider fleet. Small-N repeat offenders are tractable; a diffusion suggests methodology replication.

Limitations

AIS coverage is densest in the Singapore Strait and Rotterdam TSS, thinner on the Strait of Malacca approaches and the Channel outer waters. Some of the longer-gap events (1–3 day and 3+ day buckets) may be vessels that legitimately left the detection zone rather than truly dark. The impossible-kinematics classification does not account for GPS multipath in narrow-channel geometries, which is why Singapore and Houston dominate the false-positive counts — multipath causes apparent sub-second jumps that no kinematic model can distinguish from genuine spoofing without ground-truth AIS receiver logs. The 30-day window is sufficient to detect patterns but not to separate seasonal variation from secular trends. Port state control inspection data is the ground truth for which flagged vessels were actually non-compliant; cross-validating against PSC records is a follow-on step this atlas does not include.

Data current as of 2026-04-23 across the eight named geographic zones. Source: Axiom Overwatch dark event detection pipeline, AIS positions from AISStream.io. Aggregates prior posts covering Singapore, Norfolk, Houston, Rotterdam, Antwerp, Hamburg, and the Channel entrance into a single canonical reference.