Hijiori Onsen-kyo

Find a place to stay · Open in maps

"Find a place to stay" searches Rakuten Travel and contains advertising; we may earn a commission on bookings made through it. The snowfall figures and ranking are calculated independently of that.

We do not link to an official site. We have not verified URLs for all 79 towns, and guessing would send you somewhere wrong. They will be added once checked.

Hijiori Onsen-kyo is the collective name for three springs — Hijiori, Kogane and Ishidaki. It sits in Okura, a village of about 3,000 people in northern Yamagata, the smallest municipality in the prefecture.

The spring is said to date from 807. It grew as a starting point for pilgrims climbing the Three Mountains of Dewa, and it still feels like a toji town — a place people come to soak for days rather than overnight.

And it is buried in snow. Of the 184 Japan Meteorological Agency stations we checked, Hijiori was second only to Sukayu — 985 cm a winter averaged over three seasons. The village has turned that into events: a festival of giant snow figures and fireworks, a corridor carved through the snowpack, and — genuinely — a World Cup for digging. Competitors dig down through more than four metres of snow, and whoever reaches bare ground first wins.

Beside the town stands the Hijiori dam, a 1952 cobble-concrete flood-control structure listed as a registered tangible cultural property since 2009. Water falling over it has become part of the view.

The main water is a sodium chloride–bicarbonate spring. Kogane, 1.5 km west, is named for the gold mine whose workers settled there; its bath was in use from the late Meiji period into the 1920s.

Sources: 環境省 肘折温泉郷 国民保養温泉地計画書（令和5年3月） · 湯の里 肘折温泉

Up to five within 40 km, straight-line, nearest first. Quality rank is out of the 395 Japanese ski areas.

The snowfall shown here is an estimate produced by a weather model, not a record of what actually fell.

We compared it against 63 Japan Meteorological Agency weather stations to see how far off it is. On average it comes to about six tenths of the real figure — it runs low. And it runs low by different amounts in different places: some read barely a third of what was measured, others match almost exactly. Mountains tend to read lowest.

We tried scaling everything up by a fixed amount to correct for it. It barely helped, because the error differs from place to place.

So treat these numbers as a guide to roughly where somewhere sits, not a precise figure. There is no meaning in the gap between first and second, or between 30th and 35th. Large differences between regions are real and can be read as they are.

We are working on replacing these estimates with observed values.

← Back to the full list