Half the Alps Already Snows Within 2°C of Rain. In Hokkaido It Is 3%.

We have written about what warming does to Japan’s ski areas — the answer, from Hokkaido University, is that 7 of 349 keep stable natural snow at 4°C. That piece looks only inside Japan, and it reads like a story about Japanese fragility.

It is not, and the way to see that is to measure other countries the same way.

What we measured

For every resort: the mean air temperature on days it actually snowed, over the last three winters. Then how far that sits below freezing.

headroom = 0°C − mean temperature on snowing days

A place that snows at −11°C has a long way to fall before that snow arrives as rain. A place that snows at −1°C does not.

This is a measurement of the present, not a forecast. It says where the snow currently falls relative to freezing. It does not predict any particular year, and it is not a projection of which resorts will close.

Four regions

Japan is measured in full — all 395 ski areas. The Alps and North America are random samples of 120 each, drawn from the 973 and 713 in our dataset, because measuring every one takes hours and a hand-picked list of famous names would just reproduce whoever picked them.

	Areas	Grid elevation	Temp when snowing	Headroom	Rain days	Already above 0°C	Within 2°C
North America	120	496 m	−4.8°C	4.8°C	13%	0%	8%
Hokkaido	86	222 m	−4.4°C	4.4°C	8%	0%	3%
Honshu	309	874 m	−2.9°C	2.9°C	21%	4%	31%
The Alps	120	1,494 m	−2.1°C	2.1°C	16%	10%	48%

The Alps are the exposed ones

48% of the Alps sample already averages within 2°C of freezing on the days it snows. Ten per cent of it is already above freezing on average — the snow is falling at temperatures where a warm year turns it to rain.

Hokkaido’s figure for the same test is 3%. North America’s is 8%.

The Alps are doing this from 1,494 m, the highest grid elevation of the four groups. They sit 1,272 m higher than Hokkaido and have 2.3°C less room.

Among the names people know:

	Grid elevation	Temp when snowing	Headroom
Mammoth	3,356 m	−8.8°C	8.8°C
Aspen Highlands	3,122 m	−7.5°C	7.5°C
Verbier	2,531 m	−5.9°C	5.9°C
Val Thorens	2,308 m	−5.8°C	5.8°C
Zermatt	1,613 m	−1.9°C	1.9°C
St Anton	1,292 m	+0.1°C	−0.1°C
Kitzbühel	748 m	+1.0°C	−1.0°C

St Anton and Kitzbühel already average above freezing when it snows. They still get snow, because a mean is not every day — but they are working without margin.

Japan is not one place

The single most misleading thing you can say about Japanese snow is “Japan”.

	Headroom	Rain days	Within 2°C of rain
Hokkaido	4.4°C	8%	3%
Honshu	2.9°C	21%	31%

Honshu is genuinely exposed — 2.9°C of headroom and rain on a fifth of winter days already. Nearly a third of it snows within 2°C of freezing.

Hokkaido is close to the most robust group we measured, and it does it at a median grid elevation of 222 m. North America gets its 4.8°C from a sample that includes the Rockies. Hokkaido gets 4.4°C from ground barely above the sea.

That is the same result as our latitude piece, stated in a different currency. Elevation buys cold and latitude buys dry, and Hokkaido is buying with latitude what the Alps are trying to buy with altitude.

The deepest reserves in Japan are where you would now expect:

	Grid elevation	Temp when snowing	Headroom
Shiga Kogen Shibutoge	2,266 m	−12.0°C	12.0°C
Daisetsuzan Kurodake	1,407 m	−11.9°C	11.9°C
Asahidake	1,295 m	−11.2°C	11.2°C

High and northern, or northern and volcanic. Nothing in the Alps sample comes near.

What this does and does not overturn

It does not overturn the Hokkaido University finding. At 4°C, almost nothing in any of these regions is comfortable — the Alps sample would lose more than Japan’s, not less. Both things are true: Japan loses most of its ski areas under heavy warming, and Japan is not first in the queue.

What it does overturn is the ordering people assume. The story that Japanese snow is uniquely precarious, while the Alps are the established, reliable place, is the wrong way round on this measurement. The Alps are closer to the line than Honshu, and much closer than Hokkaido.

That has a practical reading for anyone deciding where to ski over the next couple of decades, and we will put it plainly: on how far the snow currently sits from rain, Hokkaido is the strongest position in the developed ski world outside the high Rockies, and it is the only one of them that does not require altitude.

Where this is weak, and it is weak in places

The elevation is the model’s, not the mountain’s. ERA5 runs on a 10–30 km grid, and each figure is the temperature at whatever elevation that cell resolves. For resorts with large vertical the cell often sits near the base: Whistler reads 672 m here against a summit above 2,100 m, and Revelstoke 515 m. Both come out above freezing as a result, which describes their valleys and not their upper mountains. Every figure in this article carries that caveat, and it is the reason the grid elevation is printed in every table.

The samples are not equivalent. Japan is complete at 395. The Alps and North America are 120 each, randomly drawn with a fixed seed. A random sample of 120 pins a median well; it does not describe any individual resort, and the two regions contain many small low areas alongside the famous high ones.

Three winters. Enough to separate −11°C from −2°C. Not enough for tenths, and not enough to speak about trend at all — nothing here measures change over time.

Mean temperature on snowing days is one number for a whole winter. It hides the distribution. A place at −2°C with a wide spread may hold more cold days than one at −2°C with a narrow one.

Headroom is not survival. Snowmaking, aspect, glaciers, wind-loading and elevation range all matter and none are here. A resort with 2°C of headroom and 1,500 m of vertical is in a different position from one with 2°C and 200 m.

We measured where snow falls, not how much. Japan’s volume advantage — 9 of the world’s 10 snowiest ski areas — is a separate axis and does not appear in this table at all.

How this was measured

Daily snowfall, rainfall and mean temperature from Open-Meteo historical data on the ERA5 reanalysis, 1 December to 31 March across the winters of 2024, 2025 and 2026. A snowing day records 1 cm or more of snowfall; a rain day records 1 mm or more of rain. Headroom is 0°C minus the mean temperature across snowing days.

No elevation correction is applied to any figure, in either direction. Our Japanese dataset holds temperatures corrected to each ski area’s own mid-elevation, but the foreign resorts in our data do not carry reliable elevations to correct against — so both sides use the raw grid value, and the grid’s elevation is shown alongside.

The script is scripts/measure-snow-headroom.mjs in this site’s repository, and the foreign measurements are saved to data/snow-headroom-world.json. The random samples use a fixed seed, so the same command reproduces the same 240 resorts.