How raceday.ski calculates the snow temperature

Wax charts refer to the snow temperature — yet in everyday practice almost everyone measures the air temperature. The two can be several degrees apart. At night the snow surface radiates heat towards the sky and cools below the air temperature. During the day, sun position, aspect and slope angle warm the surface again.

That is why raceday.ski does not estimate the snow temperature from rules of thumb but calculates it physically — hour by hour, from the previous evening up to the selected start time. What you see is not guessed, but derived.

Data sources

• Open-Meteo — hourly weather forecasts: temperature, humidity, cloud cover, wind, precipitation, snowfall, snow depth plus direct and diffuse radiation. On top of that, temperature and geopotential on eight pressure levels.
• WSL Institute for Snow and Avalanche Research SLF — measurements from the IMIS stations, including the measured snow surface temperature (data: WSL/SLF, CC BY 4.0).
• Tyrol Avalanche Warning Service (Land Tirol) — measurements from the Tyrolean stations, including the measured surface temperature: open instantaneous values and the seasonal archives of the measurement series. Data source: Land Tirol - data.tirol.gv.at, CC BY 4.0.
• Digital elevation models (SRTM, EU-DEM, Copernicus GLO-90) — for horizon profiles, shading and the validation of the calculation points.

The 3-layer energy-balance model

The core of the calculation is a snowpack model with three layers (surface ~3 cm, intermediate layer ~17 cm, base ~80 cm) that solves the energy balance of the surface hour by hour:

• Shortwave radiation: direct and diffuse solar radiation, corrected for slope angle, aspect and horizon shading; the albedo ages with the snow (CLASS scheme) and is reset by fresh snow and the morning slope grooming.
• Longwave radiation: atmospheric counter-radiation after Brutsaert (1975) with the Crawford-Duchon cloud correction; plus the emission of the snow surface itself — the mechanism that makes clear nights so cold.
• Latent heat: sublimation and condensation as a function of humidity, wind and altitude (saturation vapour pressure over ice).
• Heat conduction: between the layers, with density-dependent conductivity after Sturm et al. (1997) — groomed piste snow conducts very differently from loose fresh snow.

The snow type itself — fresh snow, old snow, slush, machine-made snow … — is classified by raceday from the last 48 hours of weather history. It determines the density, albedo and conductivity of the model. The moisture class follows the international snow classification (ICSSG). So the model does not run with a standard snow, but with the snow that is actually lying there.

Terrain: altitude, horizon, shading

raceday converts the air temperature to the target altitude via a dynamic lapse rate — that is, with the actual temperature profile over altitude from pressure-level data instead of a fixed gradient. This also detects inversions, where it is warmer up high than down in the valley.

Measuring-station calibration: point validation (SLF-IMIS & LWD Tyrol)

Every physical model has blind spots. To find ours, we compare exactly where measurements are taken: every two hours the model computes a twin at each station site — with the station's altitude, flat ground and horizon. This raw, still uncorrected model value competes against the measurement of the same hour. Only this difference shows what the model can really do.

The yardstick: 132 snow stations of the SLF IMIS network in the Swiss Alps and 145 stations of the Tyrol Avalanche Warning Service. In addition, every run archives the forecasts for +24 and +48 hours — this yields the reported error per lead time.

Outside the two measurement networks the model runs without measurement anchoring — this is honestly reflected in the displayed uncertainty.

How this model is tested against 18 winters of measured data and refined in a falsification-driven way is explained in the article How the raceday model is validated.

Note on the Tyrolean stations: the open dataset only provides instantaneous values on a 10-minute grid (no measurement history). They are paired to the full hour and therefore yield fewer pairs per day than the IMIS stations with their 30-minute means; the methodology is unaffected. Data source: Land Tirol - data.tirol.gv.at, CC BY 4.0.

Uncertainty: the ± band

What goes in? Every calculation yields a confidence score from seven factors: cloud cover, wind, temperature stability, measuring-station proximity, terrain profile, lapse-rate quality and snow-classification confidence. Special situations such as foehn, rain on snow or inversion apply fixed deductions.

What do you see? Instead of an abstract percentage, raceday.ski displays an uncertainty band on the snow temperature (e.g. −6.5 ±1.5 °C) in steps from ±0.5 to ±4 °C. The mapping is deliberately conservative; from the start of the season we review it against the collected measurement-model pairs. Special situations additionally appear as a warning flag. The less certain the situation, the wider the band.

From snow temperature to wax recommendation

The recommendation selects from 127 products by Swix, Toko, Holmenkol, HWK and Rex. Every wax receives a score from four criteria: How centred is the calculated snow temperature within the manufacturer's range? Does the snow type match? Does the moisture match? And how have users rated this wax under similar conditions? The best result wins. Structure and brushing advice follow curated templates per brand — the wax choice itself is always made by the score, never by the template. We verified the temperature ranges in the database product by product against published manufacturer specifications, apart from a few marked exceptions — every verified product carries its source and verification date. Where manufacturers publish moisture bands (currently only Swix's Pure race line), they feed directly into the scoring.

The result is deliberately verifiable: the transparency block of every recommendation shows the derivation — from the air temperature via the night minimum to the snow temperature, including physics parameters and score breakdown.

Limits of the model

• The calculation is based on weather forecasts — their errors propagate, especially in unstable situations.
• Race-specific preparation (water injection, salting) can change the piste locally to a large degree and is not modelled.
• Foehn, rain on snow and inversions violate model assumptions — they are detected and flagged with a wider uncertainty band, but remain difficult.
• The systematic comparison of model vs. measuring stations runs as an ongoing measurement campaign over the winter season — the evaluation is disclosed on the accuracy page and fills in automatically as soon as season data are available.

References

• Brutsaert, W. (1975): On a derivable formula for long-wave radiation from clear skies. Water Resour. Res. 11(5), 742–744.Atmospheric emissivity (clear sky)
• Crawford, T. M. & Duchon, C. E. (1999): An improved parameterization for estimating effective atmospheric emissivity for use in calculating daytime downwelling longwave radiation. J. Appl. Meteor. 38(4), 474–480.Cloud correction of the incoming longwave radiation
• Juszak, I. & Pellicciotti, F. (2013): A comparison of parameterizations of incoming longwave radiation over melting glaciers. J. Geophys. Res. Atmospheres 118, 3066–3084.Validation of the longwave parameterisation in high mountains
• Verseghy, D. L. (1991): CLASS — A Canadian land surface scheme for GCMs. I. Soil model. Int. J. Climatol. 11, 111–133.Ageing scheme of the snow albedo
• Sturm, M., Holmgren, J., König, M. & Morris, K. (1997): The thermal conductivity of seasonal snow. J. Glaciol. 43(143), 26–41.Thermal conductivity as a function of snow density
• Alduchov, O. A. & Eskridge, R. E. (1996): Improved Magnus form approximation of saturation vapor pressure. J. Appl. Meteor. 35, 601–609.Saturation vapour pressure over water and ice (sublimation, dew point)
• Fierz, C. et al. (2009): The International Classification for Seasonal Snow on the Ground (ICSSG). UNESCO-IHP, Paris.Moisture classification of the snow
• NOAA Solar Position Algorithm (nach Meeus, Astronomical Algorithms).Solar position (elevation/azimuth) per hour
• Bartelt, P. & Lehning, M. (2002): A physical SNOWPACK model for the Swiss avalanche warning. Part I: numerical model. Cold Reg. Sci. Technol. 35(3), 123–145 · Part II: Lehning, M., Bartelt, P., Brown, B., Fierz, C. & Satyawali, P., 147–167 · Part III: Lehning, M., Bartelt, P., Brown, B. & Fierz, C., 169–184 · snowpack.slf.chSnow-cover process modelling — background literature

Snow data: WSL Institute for Snow and Avalanche Research SLF (www.slf.ch), CC BY 4.0 · Weather data: Open-Meteo · Terrain data: SRTM, EU-DEM, Copernicus GLO-90 (© European Union).