What a reanalysis is
The chart of snowy days by month on every place page does not come from a weather station. It comes from ERA5, a reanalysis produced by the European Centre for Medium-Range Weather Forecasts for the Copernicus Climate Change Service, which we access through Open-Meteo. The methodology page has the definitions.
A reanalysis is a reconstruction of past weather. It takes every observation available for a given hour, from satellites, weather balloons, aircraft, ships and ground stations, and runs a weather model backwards and forwards through them to produce a single, consistent picture of the whole atmosphere on a grid. It does this for every hour going back decades, using the same model and the same method throughout.
That consistency is its great strength. A weather station at Alta and one at Zermatt were installed by different people, at different heights, with different instruments, and their records have gaps. ERA5 treats every place on Earth the same way, which is what makes it possible to compare them.
Why the totals are wrong in the mountains
ERA5’s grid is about 30 km wide. A single cell can contain a valley floor, a ridge, and the slopes between, and the reanalysis gives them one height and one weather. Snow in the mountains is made by air being pushed up over terrain, and a 30 km cell cannot see the terrain that does the pushing.
The result is that ERA5 badly underestimates snowfall wherever the local terrain matters, which is everywhere anyone has built a ski resort. Two examples from resorts that publish their own long-term averages:
- Alta, Utah. The reanalysis gives about 490 cm of snow a year at mid-mountain, averaged over 2016 to 2026. Alta reports an average of 543 inches a winter, about 1,380 cm. The reanalysis has just over a third of it.
- Palisades Tahoe, California. The reanalysis gives about 660 cm a year. The resort reports an average of 400 inches, about 1,020 cm. Under two thirds.
The gap is not the same everywhere, and it is not even in the same direction everywhere, which is the real problem. If the reanalysis were always half right we could double it. Instead it misses each resort by a different amount that depends on how the local terrain works, and a number we would have to correct by guesswork is not a number worth publishing.
What it gets right
Timing and proportion. Whether a day brought snow at all is a much easier question than how much, and a 30 km cell answers it well. So the site publishes the things that depend on timing and proportion and not on amount:
- how many days a month bring snow, on average
- which months carry what share of the year’s snowfall
- how each of the last ten winters compared with their average
- the coldest and warmest time of a typical day
- in how many of the last ten years there was snow on the ground at Christmas, New Year and other dates
The comparison between winters is the one that surprises people. The reanalysis may have Alta’s total wrong by a factor of nearly three, but it still sees that the winter of 2022/23 was far above the others, because the same bias applies to every year. Ratios survive when totals do not.
Counts, not percentages
Snow on the ground at Christmas is given as “8 of the last 10 years”, never “80%”. Ten years is too few to call a probability. A run of three snowless Christmases in a warming decade is a real pattern, but it is not a 30 per cent chance of anything, and printing it as one would claim a precision we do not have.
Where the record is measured
For ski resorts the record is calculated at mid-mountain, halfway between the village and the top lift, because that is where most of the skiing is and because it splits the difference between a base that may see rain and a summit that may not be lift-served. For cities it is at the elevation of the centre. The same point is used for every month and every year, so the record is consistent with itself, even though it is not the same point as the forecast, which is made for the top lift and the village separately.
How to get amounts
From the resort. Every ski resort page here shows the resort’s own official site and, where we have it, its snow report, and those figures are measured on the mountain by people who have been doing it for decades. The guide to our snow depth and the resort’s explains why even those measurements do not agree with each other.