Run-by-run snow forecast

1. Four global models (ECMWF, GFS, ICON, GEM) are pulled hourly from Open-Meteo and blended. Their disagreement on summit snowfall sets the confidence level.
2. Every run is mapped from OpenStreetMap, and its top, bottom, aspect and steepness come from a digital elevation model.
3. Elevation: temperature follows a lapse rate fitted to the freezing level, so the rain/snow line lands on the right runs. Precipitation grows ~4% per 100 m, and colder snow is fluffier (Kuchera snow-to-liquid ratio).
4. Wind: new snow is loaded onto lee slopes and scoured off windward ones, mostly above treeline.
5. Sun: the real sun angle on each slope drives melt, crusts and spring corn cycles.
6. Segments: every run is simulated at its top, middle and bottom, so a long run can show powder up high and crust or slush lower down.
7. High-resolution models (HRRR, HRDPS, AROME, ICON-D2, JMA MSM) join the blend at double weight where they cover the resort.
8. Verified daily: forecasts are saved every day and scored against SNOTEL snow stations near US resorts.
9. People: groomed runs reset at 4 am; skier traffic tracks out powder and builds bumps through the day.

Early-stage model with approximate resort data. Resort and patrol reports always win: check them before committing to steep terrain, and read the local avalanche forecast before going off-piste.