The models are coarse
The three models behind every forecast here work on grids of roughly 10 to 25 km. A mountain the size of the Matterhorn is a bump in that grid, and a ski resort is a fraction of one cell. The models see the storm, the cold air and the wind that push snow onto the mountain, and they are good at that. They do not see the particular gully, bowl or ridge that turns the same storm into 10 cm on one side of a resort and 40 cm on the other.
The practical sign of this is disagreement. On 21 September 2026 the three models gave Valle Nevado between 1 and 35 cm for the same Thursday, and all three were sure it would snow. The storm was clear. The amount at that one point was not. The guide on why snowfall is a range is about reading that.
What to check instead: the range and the number of models, not the top of the range. For where snow actually lands on the mountain, the resort’s own report the morning after.
Skill drops off after a few days
Days one and two are usually right about whether it will snow and roughly how much. By day five the day of a storm may have moved. By day seven, the storm may not exist. The 14-day outlook on each page is made from two models only and is labelled as an outlook because at that range it can show a pattern arriving, such as a cold unsettled spell, and little more. The guide on how far ahead a forecast can be trusted goes through this day by day.
What to check instead: the trend across runs. The site rebuilds every three hours. A day that stays snowy run after run, and gains models, is the one to believe.
The wind is the model’s wind
Wind figures are the models’ wind 10 metres above their own terrain, which is smoother and lower than any real ridge. Gusts on an exposed top station or a chairlift over a col are usually stronger than the forecast says, sometimes much stronger. The site flags gusts from 40 km/h and marks down the ski-day grade for them, but the threshold at which a lift closes is the resort’s decision and differs from lift to lift.
What to check instead: the resort’s lift status page on the morning, which every resort page links to where we have it.
Snow depth is an estimate
The “snow on the ground” figure is the middle of three model estimates for a grid cell, not a measurement. It is usually below what a resort measures on its pistes, and it knows nothing about grooming or snowmaking. The guide to our snow depth and the resort’s has a worked example.
What to check instead: the resort’s snow report for piste conditions. Use ours for the trend and for comparing places.
The snow record is about timing, not amount
The ten-year record on each page comes from a reanalysis on a 30 km grid. It is sound on how many days a month bring snow and how one winter compared with another, and it underestimates mountain snowfall so badly that we do not publish its totals at all. The guide on the snow record and centimetres explains, with numbers.
What to check instead: the resort’s own long-term averages, for amounts.
We do not know what is under the snow
Nothing on this site can tell you whether a slope is safe. Avalanche danger depends on the layers already in the snowpack, which a weather forecast does not see, and it is assessed by people who dig into it every day. Where we have a link, each resort page points to its regional bulletin. The main services for the countries we cover are the SLF in Switzerland, Lawinen.report for Tyrol, South Tyrol and Trentino, Météo-France for the French Alps and Pyrenees, AINEVA for the rest of Italy, Avalanche Canada, the regional centres in the United States such as the Utah Avalanche Center and the Colorado Avalanche Information Center, the Japan Avalanche Network, and the New Zealand Avalanche Advisory.
What to check instead: the bulletin, every day you go off-piste, and before you go.
We are not a warning service
A national weather service issues warnings for storms, extreme cold and heavy snow, and it has forecasters, radar and local observations that a site built from global models does not. When one of them warns, that overrides anything here.
How to use the site well
- Today and tomorrow: the headline, the range and the freezing level are reliable enough to act on. Read the resort’s report as well.
- Three to five days: trust the pattern and the label. Expect the amounts and the exact day to shift.
- Six to seven days: a week’s total and the trend across runs are useful; a single day’s figure is not.
- Days 8 to 14: an outlook. Use it to see whether a cold spell is coming, not to book a day off.
- Months ahead: the snow record. It says when it usually snows and how variable the winters are, which is more use than any forecast at that range.