History

30 days of forecast vs. what actually happened at the lake. Click a single day to see its details.

Last 30 days

Rule-based forecast (actually used)
ML-based forecast โ€” logistic (experimental)
ML-based forecast โ€” HistGradientBoosting (black-box model, experimental)
Actual measurements (session โ‰ฅ 1 h)
Wed Jun 24 Thu Jul 23
session (โ‰ฅ 1 h โ‰ฅ 11 kt) marginal (โ‰ฅ 1 h โ‰ฅ 8 kt) no wind no data thunderstorm (forecast rows: predicted ยท actual row: observed)

Jul 23, 2026

For today, Jul 23, 2026

NO GO

700 hPa crossflow 25 kt โ€” valley decoupled

๐Ÿค– ML โ€” logistic experimental NO GO GO 26% ยท MAYBE 14% ยท NO GO 60%

Strongest inputs โ€” (+) morning solar; (โˆ’) pressure ฮ” (thermik), dew-point spread.

๐Ÿค– ML โ€” HGB ยท black box experimental NO GO GO 0% ยท MAYBE 0% ยท NO GO 100%

Most distinctive inputs: morning solar, Innsbruck pressure, Bolzano pressure.

Experimental shadow models โ€” they never drive the verdict. How good are they? โ†’

Previous forecast method

How the day was forecast at the time, before the rules were recalibrated โ€” shown for comparison only.