ATLASCOMMAND
METHODOLOGY · HOW THE RISK SCORES ARE PRODUCED
← LIVE CONSOLE
The Cable Report correlates live, open event feeds against the geometry of the world’s submarine cable network every 10 minutes. Each qualifying event near cable infrastructure becomes an alert scored 0-100; alerts combine with a documented structural baseline for each corridor (the standing risk that is present even on a quiet day) and roll up into corridor indices, route stress and a single Cable Risk Index. Scores use a soft-saturation scale, so the top of the range is reserved for genuinely extreme conditions rather than clipping. This page states the exact weights in production so the numbers can be cited with confidence. The scoring code is version-controlled and weights change only with a corresponding update to this page.
OPEN SOURCE · INDEPENDENTLY VERIFIABLE

The scoring logic described on this page is published in full as runnable code, with every worked example below asserted in an automated test. We open the method so the numbers can be read, run and challenged; the data and intelligence stay proprietary.

github.com/thecablereport/methodology

THE PIPELINE
1Ingest: seismic, conflict, cyclone and vessel feeds are pulled on 3-15 minute cycles; cable and landing-station geometry refreshes twice daily.
2Filter: each feed has hard eligibility gates (below) before anything is scored; most raw events never become alerts.
3Correlate: surviving events are measured against every cable route and landing station; events with no infrastructure inside their threat radius are dropped.
4Score & review: alerts are scored and tiered; an AI analyst pass reviews every conflict alert and suppresses geocoding noise before publication.
5Aggregate: the top 40 alerts drive corridor indices, route stress, the Cable Risk Index, the live console and the daily brief. Every pass is archived.
DATA SOURCES
Cable & landing geometryThe Cable Report clean-room dataset (independently sourced schematic routes)
SeismicUSGS: magnitude 4.5+, rolling 7 days
ConflictGDELT 2.0: material-conflict events, ~2.5 h window, city-level geocodes only
Tropical cyclonesNOAA National Hurricane Center: active storms
Vessels (AIS)digitraffic (Finnish Transport Infrastructure Agency): Baltic Sea, terrestrial
Marine weatherOpen-Meteo
Strategic facilitiesCurated: interconnection hubs and semiconductor supply-chain sites
SEVERITY TIERS

Every alert’s score maps to a tier used across the console, indices and brief:

HIGH score ≥ 70   ELEVATED 40-69   MONITORING below 40
EARTHQUAKES

Seismic events and the turbidity currents they trigger are among the best-documented causes of cable breaks (Taiwan 2006, Tonga 2022). Eligibility: magnitude 4.5+, and at least one cable, landing station or strategic facility inside the threat radius.

threat radius = min(400 km, 60 × (M − 3.5))
score = (22 × (M − 4) + 8 if depth < 70 km + 18 if a strategic facility is in range − 0.15 × km to nearest asset) × decay
decay = 1.0 for the first 24 h, declining linearly to 0.3 by day 7

Worked example: an M6.2 at 15 km depth, nearest cable 80 km away, no facility in range → 22 × 2.2 + 8 − 12 = score 44, ELEVATED.

CONFLICT EVENTS

Machine-geocoded news events, filtered hard before scoring. Eligibility: material-conflict class only; city-level geocodes only (country and province centroids are rejected because they put “national unrest” dots in the sea); clustered to ~0.25°; kinetic categories only (assault, armed clash, mass violence); minimum 3 events per cluster; and the cluster is either within 40 km of a landing station or inside a monitored maritime corridor. Proximity to a schematic cable route line alone does not qualify an event — this is what keeps misgeocoded inland political news off the map.

score = min(60, 6 × events) + 30 near a strategic facility / 25 near a landing station / 10 otherwise − 0.2 × km to nearest asset

An AI analyst pass then reviews every conflict alert before publication: geocoding artefacts and metaphorical “clashes” are suppressed entirely; alerts judged credible gain +15 and are re-tiered. The daily brief states how many alerts were excluded this way.

TROPICAL CYCLONES

Cyclones stress cables in shallow water and keep repair ships in port, which extends outage duration even when they cause no break. Eligibility: any active NHC storm with infrastructure in range.

threat radius = 350 km (cables) · 250 km (landing stations)
score = 0.9 × intensity (kt) − 0.05 × km to nearest asset
VESSELS · AIS (BALTIC)

Cable-repair and cable-lay ships are identified by name against a curated fleet list (AIS has no “cable ship” type code); tankers by AIS ship type, surfaced only within 25 km of a cable route. A watchlist covers vessels tied to documented Baltic cable incidents, and EU-sanctioned shadow-fleet vessels are matched against the sanctions list. Two behaviour signatures are detected, each an explicit rule, never an attribution of intent:

Loitering requires persistence, not a snapshot: speed under 2 kn beside a cable sustained for at least 30 minutes across successive observations (a 20 minute gap or movement resets the episode).

Anchor drag is the documented cable-damage mechanism and the single highest-signal case: not a ship sitting still, but one dragging its anchor over distance (the Eagle S dragged roughly 90 km across the Gulf of Finland in December 2024). We flag it when a vessel’s own 48-hour track shows a slow run over distance whose path passes close to a cable, and we weight it up when the vessel is also sanctioned or known, and in rough seas, since drag faults correlate with bad weather. Position history for surfaced vessels is retained for 48 hours and drawn as tracks on the console.

anchor-drag rule = path ≥ 8 km at 0.5-6 kn over ≥ 1 h, passing within 6 km of a cable
rough seas = wind ≥ 25 kn (or gust ≥ 33 kn) or wave height ≥ 2.5 m (Open-Meteo)

base = 68 anchor drag near a cable · 62 EU-sanctioned vessel loitering · 58 known incident vessel · 50 EU-sanctioned vessel active · 42 tanker loitering · 22 repair ship on station
anchor drag adds +8 if also sanctioned/known, +8 in rough seas
score = base − 0.25 × km to nearest cable

depth gate (anchor mechanisms only) = × 1.0 at ≤ 200 m, tapering to × 0.15 by 1000 m (GEBCO bathymetry)

Water depth gates the anchor and loiter signals, because an anchor can only threaten a cable in water shallow enough for it to reach the seabed: the great majority of anchor and fishing faults occur on the continental shelf (under 200 m), and by roughly 1,000 m an anchor cannot reach the cable at all. Depth comes from the public-domain GEBCO grid; where it is unknown the score is left unchanged, and a sanctioned or known-incident vessel keeps an ELEVATED watch floor regardless of depth.

STRUCTURAL BASELINE · STANDING RISK

The live feeds above capture acute events. But a corridor carries risk even when the news is quiet: a chokepoint stays a chokepoint, and the most common cause of cable faults worldwide (anchor drag and trawling, and natural turbidity currents) is not visible in a conflict or seismic feed at all. Each of the seven corridors therefore has a structural baseline that sets its standing risk. Every input is documented and the weights are fixed in advance, so an elevated corridor follows from its stated inputs rather than a single hand-set score. The inputs are expert-assigned and reviewed on a quarterly cadence against the observable record; they are not yet fitted model parameters, and we say so in the limitations below.

structural = 0.28 × chokepoint + 0.26 × geopolitical + 0.16 × physical + 0.30 × incident-base
  then reduced by redundancy, and capped at 55 (structural alone never exceeds ELEVATED)
chokepointTraffic concentration & network value-at-risk (EIA; Nature Communications 2025)
geopoliticalThird-party hostility index (GPR-C / Verisk / EIU class), not our own opinion
physicalDepth band, seismic and turbidity-current exposure
incident-baseDocumented cable-fault history for the corridor, weighted for recency and severity (expert-assigned from our incident record)
redundancyProtective: alternative systems & landing diversity (lowers the baseline)

The per-corridor input values are published in the open scoring code (STRUCTURAL_BASELINE). Worked example: the Strait of Hormuz has no documented subsea-cable cuts, yet reads elevated standing risk because its chokepoint concentration and third-party geopolitical inputs are high, while its score stays below the Baltic’s and Taiwan Strait’s, whose documented incident base rates are higher. The corridor’s score follows from these published inputs and weights, so it can be reproduced and challenged.

CONFIDENCE

Magnitude and confidence are tracked as separate quantities: a high-magnitude, low-confidence event is a different object from a corroborated one. Each alert carries a source-confidence weight that scales its contribution to the corridor and global indices (it is not silently folded into the alert’s own score). Physical instruments rate highest (USGS seismic 0.95, NHC cyclones 0.90); vessel behaviour is 0.85 when tied to a specific cable and 0.70 otherwise; machine-geocoded conflict inherits the AI analyst verdict (credible 1.0, monitor 0.6, untriaged 0.7).

AGGREGATE INDICES

Corridor index (0-100): seven strategic chokepoints (Red Sea/Bab el-Mandeb, Hormuz, Eastern Mediterranean, Black Sea, Taiwan Strait, South China Sea, Baltic). The index is the structural baseline plus the confidence-weighted acute pressure, passed through a soft-saturation curve that approaches 100 but never clips:

acute = Σ score × w × confidence   w: HIGH 1.6 · ELEVATED 1.0 · MONITORING 0.45
corridor = 100 × (1 − e−(structural + acute) / 120)

Because the scale saturates softly, 100 is reserved for genuinely extreme concurrency; a handful of alerts reaches the low 90s, not a pegged 100. The console shows the structural / acute split so every score explains itself. (A future release replaces the soft curve with an Extreme Value Theory return-period map, so “100” will correspond to a stated 1-in-N-year state.)

Route stress: for each of 12 inter-market routes (e.g. New York⇄London, Europe⇄Singapore), the share of serving cable systems currently named in an ELEVATED or HIGH alert. The headline figure is the average across routes.

Cable Risk Index (0-100): the single global number on the console and brief. A standing structural floor (chokepoint risk is never zero) plus the acute load, soft-saturated:

CRI = 100 × (1 − e−(structural floor + 8×HIGH + 2.5×ELEVATED + 0.5×MONITORING + 1.2×route stress) / 68)
KNOWN LIMITATIONS
CITING & UPDATE POLICY

Cite figures as “The Cable Report (thecablereport.com)” with the reading time, since scores move as events develop. The Cable Risk Index and corridor indices are designed to be quoted directly (e.g. “Baltic corridor risk stood at 62/100”).

Scoring weights are version-controlled alongside the engine code. Material changes to weights, filters or sources are reflected on this page in the same release. Method questions: intel@thecablereport.com.

Live console: thecablereport.com · Daily brief: /brief · Risk history: /history · Open scoring code: github.com/thecablereport/methodology · Cable geometry © The Cable Report
The Cable Report · Methodology