Wind at your working height: the instrument

This page is a working instrument, not a weather app. Every figure on it carries its height, its source and its uncertainty. This guide explains each element in the order you meet it. Nothing here is fetched: it is the same text for everyone, written to stay true whatever the wind is doing.

What the hero number is

The large figure is the modelled wind speed at your working height for the hour selected on the time strip, in the unit you chose. It is not the 10 m value that most forecasts print. A forecast quoted at 10 m understates what a crane jib, a drone at 100 m or a rotor hub actually meets, often by a third or more on an open site.

The forecast models publish wind on fixed levels: 10, 80, 120 and 180 m above ground. If your height sits on a level, you get that level directly and the instrument calls it native. If it falls between two levels, the speed is interpolated on a logarithmic height scale between the levels either side. That is the log-height method: wind speed in the lowest hundred metres rises roughly with the logarithm of height, so the weight given to each level follows ln(h/h1) / ln(h2/h1), not a straight line. If an intermediate level is missing from the data, the note says so.

Below 10 m or above 180 m there is no level to interpolate between. The instrument then extrapolates from the nearest level with a neutral log law and an assumed roughness length of 0.03 m, which describes open, grassy terrain. That is a stated assumption, it ignores stability, terrain and obstacles, direction is held constant, and the Agreement Spine marks the height tick as failing. Treat an extrapolated number as a starting point to confirm on site, never as a measurement.

Gust is different. The models define gust at 10 m only. The instrument does not invent a gust aloft; where a gust limit is applied at another height it scales the 10 m gust by the mean-speed ratio and says that it did.

What P(>limit) means

Set a limit and the instrument shows P(>limit): the share of ensemble members that reach or exceed it in the selected hour. An ensemble is a set of forecasts started from slightly different initial conditions. Where they bunch together, the atmosphere is predictable; where they fan out, it is not. The three ensembles read are NOAA GEFS, ECMWF IFS ENS and DWD ICON-EPS.

Member counts matter, so they are always shown. At 10 m all three ensembles publish, which gives a pool of about 122 members. At 80 m the upstream data carries fewer members, around 40 on a typical run, so the pool is smaller and each member weighs more. The ensemble service names speed only at 10, 80, 100 and 120 m and gust only at 10 m, and it does not interpolate between ensemble levels: the instrument reports the nearest level it actually used. If a model returns no members for the hour, the readout says ENSEMBLE INCOMPLETE with received over expected, and with fewer than 10 members it gives no probability at all.

The figure is uncalibrated. It is a raw member fraction, not a probability tuned against past outcomes. Ensembles are often slightly overconfident, so a displayed 5% can understate real risk. Use it to compare hours and to see when the odds start to climb, not to quote a percentage as a certainty. Hours at or above 40% are marked on the time strip as worth a look.

The Agreement Spine, tick by tick

Under the hero sits a row of five small ticks. A filled tick means the check holds, a hollow tick means it is weak, and an ember tick means it fails. Hover or tap a tick to read its reason. Together they answer one question: how much weight can this number bear?

• Obs age. How old the nearest METAR or buoy report is. Under 30 minutes holds, up to 90 minutes is weak, older or stale fails. No usable station within 150 km also fails.
• Model spread. Whether the deterministic models agree. A score of 70 or more holds. It also fails when the primary forecast and the ensemble median fall on opposite sides of your limit.
• Ensemble IQR. The distance between the 10th and 90th percentile members. Under 6.7 mph reads as tight; wider reads as wide. No members at all fails.
• Height fidelity. Holds within 10 m of a model level, weak when interpolated further away, fails when extrapolated.
• Lead time. Holds to 24 hours ahead, weak to 72 hours as skill falls, fails beyond that.

More ticks is better, and one ember tick is enough reason to open the detail. The deeper explanation is in The Agreement Spine.

The exceedance fan

The fan plots the ensemble through time at your height. The thin central line is the median, the band is the 10th to 90th percentile, and your limit is a horizontal rule across it. Hours where the band crosses the rule are the hours to read closely. A narrow band close to the limit is a different message from a wide band that straddles it: the first says the models agree and it will be near the line, the second says nobody knows yet. See the exceedance fan and ensembles and probability.

The map field and the AGREE / SPLIT chip

The map draws a regional wind field at 10 m for the selected hour: moving streaks show flow direction and colour shows speed. Wind direction is always given as the direction the wind blows from, so a 270 degree wind is a westerly. The chip at the top compares the three deterministic models, ECMWF IFS, NOAA GFS and DWD ICON, for the same hour. It reads AGREE when they sit close and SPLIT when they do not, with a count such as 3/3 of models that returned data. A split is common around fronts and in the shelter of headlands, and it is the cue to widen your margin or wait for the next run. For more detail read reading the live map.

Observations versus model

Models forecast. Stations measure. The observation line shows the nearest METAR airport or marine buoy, its age in minutes, its measured wind, and what the model said for the same hour at 10 m. When the two disagree by a lot, trust the measurement for the present and be sceptical of the model for the next few hours. The line turns ember when the report is stale or older than 90 minutes. Stations measure at their own height, usually 10 m, which is another reason to compare against the 10 m model value rather than your working-height number. See reading observations against the model.

Route mode and golf mode

Route lets you click waypoints on the map to draw a path of up to 60 points and about 5,000 km. It reports headwind and crosswind per segment and per hour at the time you would actually be on each segment, given your speed, not one timestamp for the whole journey. It also suggests a best departure. Golf walks a round hole by hole from the first tee: for each hole it shows the time you will be there, the crosswind component and the odds that the crosswind passes your limit, approximated from the 10 m ensemble with direction held at the primary model, and labelled as an approximation. Route uses the 10 m level and says so. Neither mode recommends a club, a line or a departure as a decision: they show the numbers the decision is made from.

Freezing a record

Freeze record stores a fixed, shareable page holding the place, hour, height, limit, hero number, probability, ensemble member counts, the Agreement Spine and the data sources. It does not change when the forecast does. Use it before a lift or a spray window, so that a later question has a fixed answer. Read more in evidence records.

What this instrument will never say

• It will never say it is safe, cleared or approved. There is no clearance field.
• It holds no statutory or manufacturer limits. Your limit comes from your document, and the instrument only measures forecast against it.
• It will never show zero for missing data. A missing value is UNKNOWN, with no probability implied. Read UNKNOWN versus zero.
• It will not blend away disagreement. Models are never silently substituted and a lower member count is shown, not hidden.
• It will not claim a forecast is an observation. Modelled values say modelled.

Where the data comes from

Wind and gust come from Open-Meteo's modelled best match, built on DWD ICON, NOAA GFS and ECMWF IFS. Ensemble members come from NOAA GEFS, ECMWF IFS ENS and DWD ICON-EPS through the Open-Meteo Ensemble API. Observations come from aviationweather.gov METAR reports and NOAA NDBC buoys, and on the island of Ireland from Met Éireann and the Marine Institute buoy network. Climatology and extremes come from ERA5 reanalysis (Copernicus). The base map is OpenFreeMap with OpenStreetMap data. Open-Meteo data is licensed CC BY 4.0. Every response keeps its provenance, including which model run it came from, and the attribution sits in the footer of every page.

Want to go further? Browse the Learn library, the live charts, or the industry pages for limits explained by trade. Prefer a quieter view? Open cab mode for a single large readout.