March in Japan: Every Region Loses Snow, and the Deepest Places Lose the Most. They Are Still Where the March Snow Is.

Corrected on 10 September 2026. An earlier version of this article, published the same day, was built on three winters of monthly data and reported that ski areas above 1,200 m keep 91% of their January snowfall in March. Measured over eleven winters the figure is 60%, and the examples we chose no longer hold: January, not March, is the biggest month at Kusatsu and Nikko Yumoto. We have rewritten the article rather than patch it, because the conclusion changed. What went wrong is at the end.

March is the month Japanese ski areas stop resembling each other, and the month the marketing is quietest about. Here is what eleven winters say, measured the same way at all 395 of them.

	Snowfall	Per day
December	68 cm	2.19 cm
January	82 cm	2.65 cm
February	58 cm	2.07 cm
March	45 cm	1.45 cm

March delivers 55% of January nationally. Nowhere in Japan is March a peak month. The question is only how far each place falls.

The fall is steepest exactly where the snow is deepest

Prefecture	Ski areas	Dec	Jan	Feb	Mar	Mar as % of Jan
Niigata	47	161	201	137	76	38%
Toyama	4	89	136	96	45	33%
Yamagata	17	84	114	68	45	39%
Gifu	16	61	72	45	27	38%
Fukushima	25	91	111	73	52	47%
Nagano	65	75	93	78	50	54%
Gunma	24	62	82	60	51	62%
Hokkaido	86	62	58	48	38	66%
Iwate	11	98	99	76	66	67%

Read the last two columns against each other. Niigata holds the worst percentage on the list and the best absolute figure on the list. Hokkaido holds the best percentage and one of the weakest absolute figures.

Niigata’s 38% is 76 cm. Hokkaido’s 66% is 38 cm. The ratio is a fact about how a region’s winter is shaped. It is not the number you book on.

Why the snowbelt falls hardest

The Sea of Japan snowbelt runs on the winter monsoon: cold Siberian air crosses comparatively warm water, loads up with moisture and unloads on the first ground it meets. That arrangement is a December-to-February machine and it shuts down. Niigata, Toyama, Yamagata and Gifu live almost entirely on it, which is why all four sit between 33% and 39%.

Hokkaido, Iwate and Gunma depend on it less. Their winters are flatter to begin with — Hokkaido’s biggest month is December, not January — so there is less height to fall from.

Where the March snow actually is

Ranked by how much snow falls in March, among ski areas getting 200 cm or more a season:

March	January	Elevation
114 cm	228 cm	780 m	Tateyama Sanroku	Toyama
106 cm	241 cm	1,063 m	Seki Onsen	Niigata
105 cm	263 cm	964 m	Okutadami Maruyama	Niigata
103 cm	252 cm	641 m	Iwappara	Niigata
103 cm	252 cm	519 m	Yuzawa Nakazato	Niigata
101 cm	256 cm	683 m	Hakkaisan	Niigata
97 cm	219 cm	1,332 m	Hakuba 47	Nagano
97 cm	219 cm	1,165 m	Hakuba Iwatake	Nagano

Eight of the best March ski areas in Japan are in Niigata, Toyama and Nagano — the same snowbelt names that lose the largest share of their season. They lose half of an enormous number and still finish ahead.

Elevation helps, but less than we thought

Sorting the 277 ski areas over 200 cm by height:

Elevation	Ski areas	Dec	Jan	Feb	Mar	Mar as % of Jan
Below 600 m	113	81	86	66	41	42%
600–900 m	63	98	109	82	57	44%
900–1,200 m	50	93	111	78	58	46%
Above 1,200 m	51	85	102	78	64	60%

The gradient is real and it runs the right way. But Hokkaido, whose median ski area sits at 225 m, holds 66% — better than the highest band. Height is one influence on how a winter is shaped and it is not the dominant one. Distance from the Sea of Japan monsoon matters more.

What to do with this

Do not plan a March trip around a percentage. A ski area that keeps two-thirds of a small January is worse in March than one that keeps a third of a huge one.

Niigata and the Hakuba valley remain the answer in March, for the same reason they are the answer in January: they start from so much more. Seki Onsen, Okutadami, Tateyama and the Yuzawa cluster all deliver around 100 cm in March, which is more than most Japanese ski areas manage in their best month.

Gifu and Toyama’s smaller areas are the ones to avoid in March. Gifu falls to 27 cm and 38% together, which is the worst combination on the list.

February remains the best base. Everything above measures snow falling, not snow lying. February has the deepest accumulated snowpack of any month because it has had December and January to build it, and that is a separate and often better reason to go — as it was before we counted.

Why this article changed {#why-this-article-changed}

The first version of this article used three winters of monthly data, because three winters is what our seasonal snowfall figures use. For a season total that is defensible. For a single month it is not, and we should have seen it sooner.

We caught it in South America. Nineteen of 24 Chilean and Argentine ski areas showed July — the middle of their season — delivering less snow than both June and August. A dip that deep, appearing at every ski area across 22 degrees of latitude, is not a climate. It is three shared years. Re-measured over twelve winters it vanished.

The same check on Japan moved figures we had already published. March fell from 58 cm to 45 cm nationally. The share retained above 1,200 m fell from 91% to 60%. Kusatsu and Nikko Yumoto stopped being March-peaking resorts.

A season total averages the same bias into every ski area, so the ranking survives. A monthly figure asks which of two months is larger, and three shared years can answer that wrongly for a whole hemisphere at once.

Two companion articles compared Japan’s season shape with the Alps and North America on the same three-winter basis. We re-measured those regions over nine to twelve winters, and neither conclusion survived. The claim that Japan and the Alps differ only by altitude was wrong — the Alps keep more of their winter at every altitude, and that article has been rewritten to say so. The claim that the Alps peak in March and the American West in February was also wrong: the Alps peak in January, the Sierra in December, and the Rockies are close to flat. That piece has been withdrawn.

How this was measured

Monthly snowfall comes from the ERA5 reanalysis, summed by calendar month across the eleven most recent winters (December to March) and divided by the number of winters returned. Every figure is a median across the ski areas in its group.

The per-day column divides by the length of each month, which is the only fair way to compare February with its neighbours — it is three days shorter.

The 200 cm floor on the two ranked tables excludes ski areas whose season total is small enough that monthly ratios become noise. The prefecture table has no floor.

A grid cell is roughly 30 km across, so nearby ski areas share a measurement. Iwappara and Yuzawa Nakazato return identical figures because they are one measurement, not two that agree.

Elevation is one representative point per ski area, not the top of the lifts.

The model understates mountain snowfall — checked against 63 Japanese ground stations it returns about 56% of what they record, so every centimetre figure here is a floor. It understates every month and every region the same way, which is why the comparisons hold even though the totals are low.