SnowScope

How accurate is our snow depth map? First check against 2,012 stations

Snow regionsGuidesData of 9 November3 min read

A snow map is only useful if it is right. So from now on we check ours every day against the measuring stations of the Alps, the Pyrenees and North America, and we publish the result. Here is the first score card, from 9 November 2026.

See the live snow forecast for Alps in the SnowScope appSnow depth, powder alerts and the 3D map, free.
How accurate is our snow depth map? First check against 2,012 stations

The headline numbers

We compared the map with 2,012 stations that reported a snow depth (2,516 stations from 16 networks were collected that day). On average the map was 0.5 cm too shallow, the RMSE (the typical size of an error, with large misses counting more) was 7.7 cm, and the map agreed with the stations on snow or no snow (5 cm or more) 88 % of the time.

Early November is easy in one way: many stations have no snow yet, and the map agrees. At the 780 stations where either the station or the map had snow, the RMSE was 12.4 cm, the bias -1.3 cm and the typical error 3.3 cm.

Region by region

US West, 1,063 stations (mostly SNOTEL): bias 0.1 cm, RMSE 4.6 cm, snow or no snow right 89 % of the time. The Alps, 917 stations: bias -1.1 cm, RMSE 10.3 cm, 86 % agreement. The Pyrenees only have 14 reporting stations so far, RMSE 3.0 cm.

One station explains much of the Alps number: a glacier site at 3,499 m measured 320 cm while our map shows 28 cm. Glaciers keep old snow and ice that a seasonal snow map should not count, so we flag such sites rather than chase them.

What we are not good at yet

Thin early-season snow. Where a station measured snow, the map showed snow only 61 % of the time, and where the map showed snow, 23 % of those stations had none. Patchy snow of a few centimetres is hard for any model at 500 m, and it is the first thing we work on.

Warm storms. This week we found that our model turned rain into snow when the freezing level was high. We fixed it: for the rain of 12 November at 2,300 m in the southern Alps, the forecast new snow fell from 2.7 cm to 0.1 cm, in line with the other weather models. New snow over 24 hours already scores well: RMSE 1.0 cm over 1,890 stations.

Read these numbers honestly

The stations also help build the map, so this check shows how well the final map fits the measurements, not a fully independent test. A station measures one point, the map gives an area, and the snow can change a lot within a few hundred metres (wind, forest, sun). We will add a test with stations kept out of the model.

What comes next

We now run this check twice a day and keep the history, so you will see the numbers move through the season. Coming soon: a sharper 100 m snow map, more stations (Spain and Norway), a better rule for patchy early snow, and a 'likely range' around each value so you know when the map is less sure.

Tips for skiers

  • Use the map to compare places and heights; check the resort's own report before you drive.
  • Turn on the 'Measured snow' layer to see the stations next to the map.
  • Early in the season, trust the high, north-facing slopes and expect thin snow lower down.

About these numbers

Method: every station reading of the day (quality-checked) is paired with the final 500 m snow depth at the station (interpolated from the 4 nearest pixels), for the model day closest to the reading time. Bias = map minus station. RMSE = root mean square error. Snow or no snow uses 5 cm. Stations also feed the model.

Open the live snow map →

Cite this report

For research, teaching or the press: the numbers and charts are free to reuse under CC BY 4.0 with this citation.

SnowScope (2026), How accurate is our snow depth map? First check against 2,012 stations, https://snowscope.app/blog/en/2026/11/how-accurate-is-our-snow-depth-map-november-2026.html, accessed 2026-11-09

SnowScope (Atlas Labs). (2026, November 9). How accurate is our snow depth map? First check against 2,012 stations. SnowScope snow reports. https://snowscope.app/blog/en/2026/11/how-accurate-is-our-snow-depth-map-november-2026.html

BibTeX
@misc{snowscope2026howaccurateisoursnowdept,
  author       = {{SnowScope (Atlas Labs)}},
  title        = {How accurate is our snow depth map? First check against 2,012 stations},
  year         = {2026},
  month        = nov,
  howpublished = {\url{https://snowscope.app/blog/en/2026/11/how-accurate-is-our-snow-depth-map-november-2026.html}},
  note         = {Published 2026-11-09 19:55 UTC; accessed 2026-11-09}
}

· CC BY 4.0 · Methods, data and tools →

Follow Alps

Free: a message when fresh snow is coming, and the forecast every morning.

SnowScope on your phoneSnow depth for every resort, powder alerts and the 3D map, free.
Download the app