The products / Navigatek Meteo
02 — Navigatek Meteo
Eleven models, one answer you can act on
Meteo queries eleven forecast models, weights them by what they are worth on this particular water, and draws a consensus — with direction averaged circularly, because 350° and 10° make zero, not one hundred and eighty.
- 11weighted models, compared hour by hour
- 14,123wind observations measured on the water
- 16 dof forecast held in memory, refreshed continuously
Try it free The fleet What's new
Free and open to everyone, in the browser: meteo.navigatek.com
Why
Every weather app is right. Until they disagree.
Each app shows ONE model, with the same confidence whether it is right or not. A sailor who wants to know goes through them one by one and compares in their head. None says what the terrain around their lake does to that wind.
Yet on water ringed by hills, the forecast wind is never quite the wind you find out there. The real question is not “what wind?” but “how far can I trust it, and where will it be different?”.
Without
Five apps opened one after another, five answers, and no way of knowing which to believe on that particular lake.
With Navigatek
A weighted answer, its confidence in numbers, and a map of what the water actually does to the forecast wind.
What
The comparator, the map of the water, the race forecast.
What matters is not the average: it is the spread. When the models disagree, confidence drops and says so. You see the ten arrows side by side, and you know straight away whether this afternoon's forecast stands up.
Consensus
Eleven weighted models, direction averaged circularly, gust taken at maximum. A model that stops producing is dropped and announced — it never weighs in silence.
Candour
Every hour carries its confidence, computed from the real spread between models. A forecast the system does not believe says so, instead of asserting it with the same poise as the rest.
Three scales
Hours for the outing, half-days for the weekend, days for the week. Three questions you actually ask, one per tab.
Plain language
You ask “what wind on Sunday at one” and you get a number, a direction and a confidence. And you can have it read aloud through the boat's speaker.
The course
Plan the course before laying a single mark.
The committee places the marks on the map — start, marks, finish — and picks the race time. Meteo computes the fastest route for a J/70, in the wind forecast for that hour, brought down to the water square by square. It gives the estimated time, the distance and the number of manoeuvres, overall and leg by leg, with the point of sail of each leg.
Enough to compare two courses before going out, and to set the marks where the race will be tactical rather than one long leg with no decision. For a club, it is the difference between laying a course and designing one.
- The racing rules
Marks left to port, start on starboard tack, and a tack is only drawn if it really pays: the route does not chase every small shift.
- The real water
The wind in each square comes from the learned map of the lake, and the route accounts for shelter along the shores. Where the water has no measurements, it says so: the route there is mostly geometric.
- Assumptions in writing
A J/70 speed polar, a fixed cost per manoeuvre, wind frozen for the race, no other boats. A decision aid, not a promise — and the screen says so.
In service, free, at meteo.navigatek.com.
How
From eleven raw forecasts to race advice.
- Query
Canadian, American and European models are queried in parallel — including AIFS, the European centre's AI model — along with official site forecasts. A model that stops answering is dropped and announced.
- Weight
Each model weighs according to what it is worth here: local resolution, measured spread against the others. A model whose direction turned out to be inverted by 180° was dropped on measurement, not on a hunch.
- Put a number on confidence
The spread between models, in degrees and in knots, becomes a confidence — high, medium, low — shown hour by hour. Disagreement is not smoothed away: it is shown.
- Bring it to the water
The forecast wind is projected onto a map of hundred-metre squares, learned from real outings. Where boats have sailed, the measurement speaks; elsewhere, the terrain — and the map says so.
Three products, one loop
Every race that is judged teaches the lake's wind.
This is Navigatek's newest idea, and it only exists because the three products talk to each other. Navigatek Juge tracks the whole fleet by GPS, one phone per boat, for the start and the results. But those tracks say something else too: where each boat tacked, what heading it held, where it was headed. In other words, the wind — measured wherever the fleet has sailed, not only where one instrumented boat happened to pass.
It is 6Ai's wind AI that knows how to read that wind in a track with no wind vane. Every night it learns from the day's races, and Navigatek Meteo receives the map. The club races on Wednesday evening; by Thursday morning, the map of the lake knows more — for everyone.
- Juge: the fleet races
Each crew sends its position from its own phone. The tracks serve first for the start and the results; they also serve, aggregated and anonymous, to learn the wind. It says so on the start screen, and no boat name leaves it.
- 6Ai: reading the wind without a vane
The wind AI rebuilds the wind direction from the track alone: course over ground, the tacking angle measured in the track, and Meteo's consensus to tell upwind from downwind. Each boat is compared with the others at the same minute.
- At night: everything is relearned
A finished race or a new outing triggers, the following night, a relearning of the map and then its delivery to Meteo. A failed step is retried the next night; nothing is sent until the map has been rebuilt successfully.
- Meteo: the forecast on the lake
The map covers more water, and the forecast gets translated: the diagram shows side by side the direction the models announce and the one the lake will make of it. With a south-westerly regional wind, for instance, the lake blows 48° further south.
Without
A wind map simulated from the terrain, the same whatever sailors have actually seen on the water.
With Navigatek
A measured map that grows richer with every club race: the more the fleet races, the more water it covers, for everyone.
- 99.6 %correct tack read from a track with no vane, checked against an instrumented boat's vane
- 92 → 133hundred-metre squares covered in the north of the lake, at 8 knots, once the fleet is added
- 19.2° → 14.6°between forecast wind and lake wind, judged out of sample over 24 h of real forecasts
What the fleet does not bring yet is written down too. Where the instrumented boat can check, the map's accuracy goes from 83.8 to 84.8 %: the fleet widens the map, it does not yet make it more accurate. Each new race is added to the history, and the measurement is redone.
The forecast, turned into a route
Draw the course on the map; Meteo finds the fastest way round it.
A forecast tells you the wind. The map of the water tells you what that wind becomes, square by square, from thousands of measurements taken while sailing. Put the two together and a question becomes answerable that no weather app asks: given this course and this hour, which way round is fastest? You drop the marks on the map — start, marks, finish — pick the race hour, and Meteo routes a J/70 through the flows, square by square.
The answer is a line on the map and a number: here 41:08 for 5.74 km and 5 manoeuvres, with each leg detailed — downwind to the mark in 18:47 with one gybe, upwind back in 22:21 with four tacks. It is computed from the J/70's ORC polar, 8 seconds per tack or gybe, and the wind of the chosen hour as the lake will make it, frozen for the race, without other boats. A decision aid, not a promise — and it says so under every result.
- The wind of the hour, on the lake
The eleven models give one direction for the region. The map of the water, built from measurements and from the fleet's races, says what the lake makes of it in each hundred-metre square — the flows you see drawn.
- The course, as you will sail it
You trace it on the map in order: start, marks, finish. The same marks the race committee drops, the same the boat's coaching knows.
- The fastest route, square by square
Meteo moves a J/70 through the flows with its polar, counts each tack and gybe, and keeps the fastest way round. The line on the map is the recommendation; the numbers are its proof.
Without
One wind for the whole lake, and the route is worked out in your head on the dock.
With Navigatek
The wind as the lake will make it, and a route through it, timed leg by leg — before leaving the dock.
- 41:08estimated for 5.74 km and 5 manoeuvres on the course of 1 October 2026 at 16:00
- 210° → 168°the south-westerly the models announce, and the south-south-easterly the lake makes of it
- 49,288wind observations on the water behind the map that the route is computed on
What it does not do is written down too: the wind is frozen at the chosen hour, the other boats are not there, and where the lake was never sailed the flow is only deduced from the relief — the map says so, and the route inherits that uncertainty.
What's new
What Meteo does that no weather app does.
Five positions taken, all checkable against the data.
- Disagreement as information
Others show a value; Meteo shows the spread too. When eleven models agree, you build a race plan on it. When they diverge, you know before leaving the dock — often the most useful fact of the day.
±1.4 kn and ±9° between models → HIGH confidence
- A map measured, not simulated
The map of the water comes from thousands of wind measurements taken while sailing, not from a terrain model. And it says what it does not know: the share of the lake never sailed is displayed in plain words.
16.8 % measured · 76.7 % never sailed — and it says so
- Pockets, not halves
“The left side pays”: measured over eight wind directions, the difference between the two halves of the lake is only 0 to 2.5°, and it changes sign. There is nothing to bet on. Pockets, though, are real — so the advice names pockets: “in the centre, +20° over twelve neighbouring zones”.
2,353 observations behind that one piece of advice
- An answer that cites its sources
No language model on the forecast path: the same question gives the same answer. Each piece of advice says whether it comes from the MODELS or from the WATER, and when nothing separates two options, it says so instead of inventing a bet.
Same question, same answer
- A route on the forecast
Weather apps stop at the wind. Meteo takes the course you draw and the wind as the lake will make it, and recommends the fastest way round — timed leg by leg, tacks and gybes counted, from the J/70's polar.
41:08 · 5.74 km · 5 manoeuvres, before leaving the dock
Sixteen days of forecast held in memory, refreshed continuously. The question can land on any hour — and if the data is not there, the answer says so.