We measure ourselves. Publicly.

Sichtflug: wind at launch altitude, the value the pilot sees on the launch row of the altitude wind chart. ICON-D2 (or CH1 in the Alps) model level, vertically interpolated to the launch's actual elevation.

Open-Meteo: fixed 80 m AGL wind (wind_speed_80m) from the public ICON-D2 API.

Burnair: wind strength at the model level whose AMSL height is closest to launch altitude, from the public Burnair point API.

MeteoParapente: wind at ground level, the lowest level of the public windgram (z[0] = model terrain height). This is the value the pilot reads at the bottom of the MPP windgram chart.

For each provider, we use the public wind value shown in that provider's forecast surface.

Weather stations across the DACH region, picked by siting. We prefer summit masts of national weather services and stations mounted cleanly at the launch itself.

What we deliberately leave out: sensors in the wind shadow of vegetation or terrain edges. They would produce an apparently large forecast error. In reality it would be local sheltering, not the model. An honest comparison only happens when the sensor actually measures what the model predicts.

A point forecast cannot replace what is actually happening in the valley, behind the ridge, or at the venturi. The deviation number is an average difference to sensors in open terrain. It says nothing about what may come through the saddle at your specific launch today.

Values under 5 km/h sit inside the shared model and measurement uncertainty. 5 to 8 km/h is typical for days with a clear synoptic situation. Spikes above that are usually foehn transitions, isolated convection, or sunrise/sunset. Situations no weather model captures reliably.

Live every ten minutes between 05:00 and 21:00 local time. When data from a provider is missing, we show no number. Better a gap than something stale. We correct factual errors within 48 hours; email address is below.

It describes how close on average the model forecast was to reality at the stations used here. At your specific launch the deviation may be smaller or larger. Use the forecast as a starting point. Your on-site observation remains decisive.

Sichtflug compares its own forecast against three others. That is not neutral, but it is transparent. Source code, station list, and methodology are public. External validation is welcome.

Yes, available for download below. Columns: date, provider, horizon, n, mean deviation wind speed, mean deviation wind direction, bias.

Not currently. Each additional provider is hand work. The comparison starts with the ones most commonly consulted in practice.