Flying Opportunities

Where these numbers come from

The five ridge sites read a cascade: the sharpest model that still covers the day, by lead time.

• Wind speed and direction — D+0 Open-Meteo best match, D+1–2 HARMONIE 2.5km. Direction comes from the same model as the speed so the pair stays mutually consistent.
• Gusts — D+0 Open-Meteo best match, D+1–2 HARMONIE 2.5km.
• Rain — six-model agreement, not one model. It is NO-GO when at least 3 of 6 models expect rain in the window; 1–2 models is a heads-up warning only. best match alone (the most rain-conservative model) missed about half of all rain hours.
• Sunset-collapse check (the rust strip under the green direction ribbon) — Open-Meteo best match, which is the only source here carrying boundary-layer height and 80 m wind. It runs at Paldiski, Türisalu and Tabasalu only.

The three ground-handling sites do not read one blended forecast. Each number comes from the model that verifies best against the Harku station for the question that number answers:

• Wind and direction — MET-Nordic 1 km. Picked for rarely missing the 3 m/s minimum you need to keep a wing up.
• Gusts — AROME 2.5 km (KNMI). Picked for rarely crying wolf: ground handling is a comfort call, not a safety call, so a false “too gusty” costs you more mornings than a missed one.
• Rain — six-model agreement, not one model. NO-GO when at least 3 of 6 models expect rain in the window; 1–2 models is a warning only. This is the opposite asymmetry from gusts: a wet wing beats a wasted trip, so rain is ranked to catch misses, not to avoid false alarms.

Why the model changes with the day: the sharpest models are short-range and only reach about two days out, so the later columns fall back to the best global model that covers them.

Forecast timing

Built 2026-10-04 02:30 EEST (23:30 UTC) · next update ≈ 03:30. The page is rebuilt when any model below publishes a new run — at most once an hour, and at least every 6 hours.

“Run” is the model’s initialisation time: the forecast knows nothing observed after it. Next-run times assume the next run takes as long to publish as this one did.

Symbols

How it works

Score points: +2 wind direction in optimal window · +1 acceptable direction · +2 wind speed in sweet spot · +1 flyable speed · +1 no low stratus (thermals only) · +1 CAPE/BLH bonus (thermals only) · +1 dry soil on cool days (thermals — Bowen ratio)

Ridge and ground handling sites use civil twilight to determine the flyable period. Each hour is checked individually; consecutive flyable hours form windows. Tap or hover cells for details and all windows.

Ridge and ground-handling cells show the hourly forecast, not a verdict. Solid blue is the mean wind at 10 m, dashed orange the gust, and the blue bars in the bottom lane show how many of six forecast models expect rain that hour — a tall solid bar means most models agree, a short faint one means only one or two do. Rain that only one model sees is real but uncertain; agreement is the honest signal for patchy showers. Hover an hour for the count and the expected amount in mm. The scores above still decide which days reach the “best opportunities” list and the alert email — they are just not printed on the cell, because a single word cannot say whether a day suits your flying.

The lane under each chart is wind direction. full height = the site's optimal window; half height = flyable but off centre (ridge sites only — ground handling has one window, so it only ever shows the full lane). Harku has no lane at all because every direction works there. Ground-handling windows: Stroomi NW 280°–320°, Kakumäe S–N via W 180°–360°, wind ≥3 m/s, gusts ≤9 m/s, no precipitation.

Diagonal hatching means “too close to call”: the forecast puts that hour outside the window, but by less than the models are typically wrong about direction. Every model we archive sits 11–18° off on direction, and there is no better one — direction is a large-scale field, so a 13 km model matches a 1 km one. An hour 4° outside a window is not a different answer from an hour 4° inside it, and the lane no longer pretends otherwise. Hover the hour for the margin in degrees. Two limits. That error figure is a property of the models, not of the site, and it comes from a summer 2026 window only. And the windows it is measured against are not all equally solid: Paldiski, Türisalu and Tabasalu were fitted to recorded flights against Paldiski Põhjasadam's wind observations, while Toila and Alajõe are geometric — drawn perpendicular to the slope, never checked against flights — so a margin quoted there is a distance from a drawn line, not from a verified one. How often the window itself calls right or wrong has only been measured at Paldiski. Hours under 3 m/s get no hatching at all: at that speed the direction reading is noise.

The thin rust strip below it is a sunset-collapse warning: those evening hours are forecast to decouple from the flow aloft, so the ridge can stop working while the 10 m wind still looks flyable. It is checked at Paldiski, Türisalu and Tabasalu only — an empty strip at Toila or Alajõe means the check does not run there, not that the evening is safe.

Data: Open-Meteo + FMI HARMONIE-AROME 2.5km. Source code

Disclaimer: Assessments are automated and may be inaccurate. Always check actual weather conditions before flying. The author assumes no responsibility for decisions made based on this data. About