Bellemont — 10-Day Snowfall Outlook

How to read this chart

Each card is one forecast day, running 7am to 7am America/Phoenix local time — the conventional NWS/COOP observation window for daily snow totals, not midnight to midnight. The rows show each forecast model's snowfall estimate in inches: NBM (NOAA's National Blend of Models), GFS, IFS (the European model), and ICON (the German model). The thin bracket on each bar shows the range across that model's alternate calculation methods — genuine forecasting uncertainty, not a data-quality issue (see below). Locations are matched to each model's nearest forecast grid point.

Snowfall isn't a direct model measurement the way rainfall is — it's calculated from liquid-equivalent precipitation multiplied by a snow-to-liquid ratio that depends on temperature and, for some models, the model's own internal tracking of precipitation type. NBM uses its own built-in ratio, already more advanced than the simpler alternatives. GFS, IFS, and ICON use the Kuchera method, which derives the ratio from the warmest temperature in the atmospheric column above each site, because we couldn't independently confirm that Open-Meteo passes through those models' own internal snow calculations unmodified. The thin bracket on each bar shows how much the estimate would change under the other computable methods (a fixed 10-to-1 ratio, and the model's own native snowfall field where available) — a wide bracket means the models themselves are less certain about exactly how much of the precipitation will fall as snow, not that the data is unreliable.

Flagstaff, Munds Park, and Bellemont sit in complex, high-elevation terrain that the coarser global models (GFS is roughly 28km resolution, IFS roughly 9km) don't fully resolve. If a model's nearest grid point sits at a different elevation than the actual site, snowfall can be systematically off — often by more than the snow-to-liquid-ratio uncertainty described above. This is one of the biggest sources of error in this product, not a footnote to it.

The alert banner above is driven by the multi-model mean of each model's primary estimate, summed across the full 10-day window: under 0.1″ shows “no snowfall expected,” 0.1″–1.0″ shows “snowfall possible,” and 1.0″ or more shows “snowfall expected.” These thresholds are a starting point, not a final calibration — they'll be revisited against real storms as they occur.

This product's totals won't cleanly reconcile against the companion precipitation outlook's QPF numbers by a simple ratio. Both use the same 7am-to-7am day boundary, so that's not the reason — it's because this product applies a different snow-to-liquid-ratio assumption per model than a simple conversion would, on purpose, to better reflect each model's actual behavior. A mismatch between the two products is expected, not a bug.