The Opening-Week Snow Reliability Index

Methodology, in full

The threshold, verbatim. A season counts as snow-ready when modelled snow depth reached at least 30 cm (12 in) on at least 3 of the 7 opening-week days at the destination's registry grid point. A destination's score is the number of snow-ready seasons out of 10. Where two destinations tie, the higher mean opening-week depth ranks first.

The opening week. Each destination's opening window is the one already published on its own destination page, written as prose such as "Late November" or "Mid December". Those windows come from each resort's own stated traditional season, checked against dated openings from recent years. Where the record disagreed with what this site had on file, the entry was corrected and the row recomputed before publication: the input gets fixed, never the threshold. The words are then resolved to dates by a fixed table applied to every destination: "Early" means the 1st to the 7th, "Mid" the 11th to the 17th, "Late" the 21st to the 27th, and a bare month name is read as the 1st to the 7th. Southern-hemisphere destinations are scored in their own winter, so their windows fall between June and September. The resolved window is printed in the table above, so you can check the reading against the resort's own dates.

Traditional, not contractual. These are the windows a mountain has historically opened in, not promises. Resorts move their opening dates for snow, for staffing, for events and for holiday calendars. Nothing here is a claim about what any resort has committed to.

Natural snow only. The model knows about weather. It does not know about snow cannons, and no snowmaking is modelled anywhere in this index. Several destinations near the bottom of the table open reliably every year on manufactured snow, and their low score says nothing against them. It says only that the sky did not do the work.

The grid-point caveat, plainly. The data is the ERA5-Land reanalysis, read through the Open-Meteo historical archive, one call per destination for the whole period. ERA5-Land models the world on a grid of roughly nine kilometressix miles, and each cell carries a single elevation that is the average of the terrain inside it. In steep mountains that average sits well below the summit: the cell used for Zermatt models terrain at about 1,600 m5,249 ft, not the 3,883 m12,740 ft top station. So these are not readings from the top of the lift. They are a consistent, reproducible measure of what the atmosphere put on each mountain's cell, applied identically to every destination, and they should be read as a comparison between places rather than as a depth you would find under your skis.

New resorts, old locations. A destination is scored over 10 seasons of weather whether or not it sold lift tickets in every one of them. The index measures the snow at a location, not the trading history of a business.

What is not in the index, and why. One destination was researched for this release and deliberately left out. Gulmarg, in Jammu and Kashmir, has the altitude and the snow record to belong here, but its gondola suffered a major failure in May 2026 that left several hundred people stranded, the closure was then extended indefinitely while a government inquiry sat, and although the lift resumed service in June no winter season had been confirmed when this index was built. There is also a standing UK Foreign Office advisory against all travel to the territory. The snow data existed; the confidence to send a reader there did not, so the row was held rather than published. It goes in when a season is confirmed.

What could make a row wrong. A reanalysis is a model, not a measurement: it can be too smooth over complex terrain, and it can miss a local storm. A grid cell can also straddle two very different aspects. Where a number here contradicts a resort's own record, the resort's own record is the better evidence for that resort, and this index is the better evidence for the comparison between them.

Take the data

Both tables are free to use and free to republish under CC BY 4.0, the same licence as the underlying weather data. A link back to this page is all the attribution we ask for.

• The ranking table, one row per destination: score, mean and maximum opening-week depth, the resolved window, coordinates, and the elevation of the model grid cell.
• The season-by-season table, one row per destination per season: the exact window dates, how many of the 7 days met the threshold, and all seven daily depths.

A second pair of tables, published 2026-09-12, takes the same ten seasons past the opening week and asks how long the snow then stayed: one row per destination and one row per destination per season, with the median start of each mountain's longest run at 30 cm12 in, how long it held and how deep it got. Every destination page carries its own block of those figures and the curve behind them, from Whistler to Perisher.

All four CSVs are in centimetres and metres, whichever units this page shows: the published dataset does not change when you flip the switch. Questions, corrections and requests for a regional cut of the data are welcome on the contact page. If you can show that a row is wrong, we would rather fix it than defend it.