Snow Fallback

When a model predicts precipitation but provides no separate snowfall (e.g. AROME), the fallback algorithm is used:

$s = R \times K(T) \times M(v) \times D(z_f,\, z_p)$

Kuchera Ratio $K(T)$

The snow-water equivalent depends strongly on temperature. Colder air produces lighter, more voluminous snow:

Values between reference points are linearly interpolated.

Wind Modifier $M(v)$

Strong wind compresses snowflakes and increases density. The modifier reduces the ratio:

• < 20 km/h: factor 1.0 (no influence)
• 20–40 km/h: linearly decreasing
• ≥ 40 km/h: factor 0.7 (30% reduction)

Freezing Level Damping $D(z_f, z_p)$

When the freezing level is just above the location, snow is unlikely. The damping smoothly reduces the fallback:

• Freezing level ≤ location: factor 1.0 (full fallback)
• Freezing level ≥ location + 300 m: factor 0.0 (no fallback)

The freezing level is computed as the average of ICON EU, ICON D2, and GFS.

Precipitation Phase

Each model is individually classified into snow, rain, or mixed phase:

1. ECMWF precipitation type takes priority (freezing rain, wet snow)
2. Freezing level relative to location elevation (±300 m transition zone)
3. Ratio of snowfall to rain in the model output

Only models in the “snow” or “mixed” phase contribute to the snow average, and vice versa for rain.