What the two ends are
Every snowfall total on the site comes from the same three models, ECMWF, GFS and ICON. For any period we add up what each model forecasts, then show the lowest total and the highest. The methodology page has the rounding rule and the thresholds. This guide is about why.
We do not average the three into one number. An average hides the thing you most need to know. Two models at 5 cm and one at 30 cm average to about 13 cm, and so do three models at 13 cm. The first forecast says a big fall is possible but two of the three do not believe it. The second says everyone expects a moderate fall. You would plan those two days differently, and one number cannot tell them apart.
Two Likely days that were not alike
On the morning of 21 September 2026 the site carried forecasts for Thursday 24 September at two Andean resorts, both marked Likely because all three models showed snow.
At Las Leñas in Argentina the range was 20–35 cm. The lowest model gave about 22 cm and the highest about 32 cm. The models disagreed on the size of a big fall, not on whether there would be one.
At Valle Nevado in Chile, the same Thursday read 1–35 cm. The lowest model gave a little over 1 cm and the highest about 30 cm. All three had the storm arriving, which is why the day was Likely, but they were nowhere near agreement on what it would drop at that spot. The week’s total for the resort read 10–45 cm.
A single number would have called both days something like 25 cm and 15 cm. The ranges say that one is a snow day to plan on and the other is a storm to watch.
How to read the width
- Narrow and Likely: plan on it. The models agree on the storm and roughly on the amount.
- Wide and Likely: the storm is coming, the amount is open. This is common in steep terrain, where the models’ grids handle the mountain differently, and near the freezing level, where one model’s snow is another’s rain. Check again the day before.
- Wide with a zero at the bottom: at least one model has no snow at all. The day will be marked Possible rather than Likely. Treat the top of the range as a chance, not a plan.
- Ranges close up as the day approaches. A 1–35 cm day a week out is normal. The same range the night before would be unusual. The guide on how far ahead a forecast can be trusted has more on that.
Why we round
Ranges of 10 cm or more are rounded outwards to the nearest 5 cm, so that the 1.3 to 30.3 cm above appears as 1–35 cm. The models do not know the amount to a tenth of a centimetre, and printing one would suggest they did. Totals below 1 cm are simply “Under 1 cm”.
What a range cannot tell you
The range is the spread between three models, not a probability. It does not mean the true amount is equally likely anywhere inside it, and the truth can fall outside it. Nor does it say which model is right this time. Over a season each of the three has its good and bad weeks, which is the reason for comparing them at all, and the guide to the three models explains their differences.
We keep every forecast the site has published. When there is enough history, this guide will say how often the snow that arrived fell inside the range.