Every snow site answers what is the weather. Alpcast tries to answer the question you actually have on a Friday evening: will it be snow or rain at the height I’m riding, will the lifts run, and which day is the one to take.
Weather models disagree, and the disagreement is information. Alpcast reads every global model it can reach — ECMWF and its AI variant, GFS, ICON, the UK Met Office, Météo-France ARPEGE, JMA, Canada’s GEM, CMA and MET Norway among them — and combines them into one outlook, then tells you how much they agreed. Two models promising a metre is a very different thing from eleven promising thirty centimetres, and a single forecast can never show you which one you have.
Snow at the summit and rain at the village is the most common way a good forecast turns into a wasted trip, and a single resort-level number hides it completely. Alpcast works at each mountain’s real elevation bands — road gate, base, village, mid-mountain, summit — and shows where the rain turns to snow through the day, so you can see the boundary moving up and down the hill rather than reading one figure and hoping.
Headline snow totals are quoted at the top of each resort’s lifts, because a “village” is a valley floor at one resort and a ridge at another. The difference between the top and the village is whether the same precipitation falls as snow or rain. The forecasts are for one point on a coarse grid, with no extra precipitation added for height, so a bigger figure up top never means more falls there.
A metre of new snow is worth nothing behind a wind hold. Alpcast tracks forecast gusts against each mountain’s exposure and, where there is a record of what actually held, against the thresholds those lifts have really stopped at. The result is a risk through the day rather than a single wind speed you have to interpret yourself.
Forecast winds are corrected to each resort’s own weather station. Global models read the wind over a smoothed, lower mountain than the real one, and at most resorts they under-call summit gusts, worst of all in strong wind. Alpcast learns that gap from the station’s own record and removes it, sometimes from the wind higher in the atmosphere rather than the near-ground gust, whichever has tracked that station better on weeks held back from the fit. Where a resort has no station of its own, the wind is shown as forecast, uncorrected. A lift’s hold point is marked as an estimate until that lift has a season of recorded holds.
Every model is checked continuously against what the alpine weather stations actually recorded, and scored on error at each forecast lead time. The scorecard is published so you can see who has been right here lately. It does not change the outlook: each family of related models counts once, equally, because a few weeks of temperature and wind scores are not good enough evidence to tilt a snow forecast.
One honest exception: snowfall is not scored. Australian gauges do not catch snow well enough to judge a model on it, so models are ranked on temperature, wind, gusts and humidity, where the observations are trustworthy. Anyone claiming to rank models on Australian snowfall is scoring against a broken ruler.
Japan is different. The Japan Meteorological Agency measures snow depth every hour at stations across the snow country, so for Japanese resorts Alpcast shows how much the depth has changed since the lifts closed. Those stations sit in the towns below the lifts, so every reading is labelled with its station and height, and a change in depth is called that: new snow settles as it lands, so it always reads a little under what fell.
Snow forecast a day ahead for the village, against what each resort reported for the same days. Resorts measure at a height and time they do not always state.
| Resort | Reported | Forecast | Snow days called |
|---|---|---|---|
| Charlotte Pass | 10cm | 6cm | 2 of 2 |
| Falls Creek | 20cm | 10cm | 3 of 4 |
| Mt Buller | 8cm | 11cm | 3 of 3 |
| Mt Hotham | 23cm | 19cm | 4 of 4 |
| Perisher | 41cm | 16cm | 3 of 7 |
Alpcast is not an official forecast and does not replace one. Model agreement is not the same as accuracy — models share ancestry and can be confidently wrong together, which is why a confidence figure sits next to the outlook instead of being folded into it. Ten days out, treat any of this as a shape rather than a plan. For warnings, road closures and anything safety-critical, use the Bureau of Meteorology and the resort’s own channels.
It is a personal planning tool, built because no existing site answered the question in the form it gets asked.