Guide · Passage · risk
How do you measure the forecast risk along a route?
Count scenarios, not vibes. One forecast run gives you a single future; an ensemble gives you dozens. Passage checks every ensemble member against your declared limits on every leg and reports the count — in the Cherbourg–Plymouth example, all 51 scenarios exceed the skipper's 28-knot gust limit on three consecutive mid-Channel legs. And when models disagree with each other, that disagreement is itself the finding: wait for the next run.
View as table
| Leg | Gusts to | Vs 28 kt limit |
|---|---|---|
| L1 · Passe de l'Ouest | 25 kt | within limit |
| L2 · Omonville | 25 kt | within limit |
| L3 · Casquets TSS | 32 kt | 51/51 scenarios above, ~16:00 UTC |
| L4 · Mid-Channel | 35 kt | 51/51 scenarios above, ~19:00 UTC |
| L5 · Eddystone approach | 34 kt | 51/51 scenarios above, ~17:00 UTC |
| L6 · Plymouth entrance | 27 kt | within limit |
Three signals, one verdict
Route risk in a Passage briefing reduces to three measurable signals. First, limit exceedance: for each leg and hour, how many ensemble scenarios break a limit you declared in advance — here, 51 of 51 on three legs. Second, model divergence: on the Mid-Channel leg the deterministic models disagree for an hour with a spread of 7.7 knots, and divergence says wait, whatever the average looks like. Third, run age: the briefing names when the next model cycle is expected (~22:30 UTC in the example) so "check again" has a time on it. The demo's overall verdict follows mechanically: the forecasts disagree too much to assess the passage — reassess after the next run.
Where the forecast data comes from
Live Passage briefings draw on openly published model data — NOAA GFS, the GEFS ensemble and GFS-Wave, ECMWF open data, and Copernicus GLO12 currents — ingested on a fixed schedule into pre-processed forecast tiles. The worked example on this page uses a fixed demo scenario so that the arithmetic can be inspected end to end.
Methodology
Every ensemble member is evaluated hourly along the route, course-relative, across each leg's ETA occupancy window, and compared against the skipper's declared limits. Exceedance is reported as a raw member count. Deterministic models are cross-checked against each other, and their divergence is reported with its duration and spread rather than averaged away.
Limitations
A 51-member count is a scenario fraction, not a calibrated probability — 51/51 means "no modeled future stays under your limit," not "100% chance." Agreement between runs is not proof of accuracy. Wave forecasts are deterministic only. The example is a fixed demo scenario; for a live passage, run a current briefing and read the official warnings.