When Japan’s Autumn Leaves Actually Peak (According to 73 Years of Japan’s Own Records)

Affiliate disclosure: this article contains affiliate links to tour and activity partners. If you book through one, I may earn a small commission at no extra cost to you. Nothing below is timed to suit a booking window — the dates come from the Japan Meteorological Agency’s own observation record, which I have averaged myself and linked so you can check it. Averages are averages: any single year can land a week or more either side.

1. The pain point: the dates in English guides are roughly two weeks early

Search for when to see autumn leaves in Japan and you will be told, with great confidence, that Kyoto peaks in mid-November.

Japan’s own meteorological record says otherwise. Over the last ten observed years, the maples at Kyoto’s weather station reached full colour on 10 December on average. Not mid-November. Not late November. December.

That is not a rounding error, and it is not one freak year. It is the ten-year mean, and in the same decade the earliest date was 1 December and the latest was 20 December. Mid-November in Kyoto is not the peak — it is the fortnight when the trees are still mostly green and every hotel is already full.

The reason the English-language advice drifts is not laziness. It is that most of it was written once, from other English articles, and never re-checked against a measurement. Meanwhile the measurement kept moving. This article is the measurement.

2. Who this guide is not for

If your dates are already fixed by work or by school holidays, this article will only make you anxious. Skip to section 6, take the city row nearest your destination, and use it to decide where to go in the days you have rather than when to come. The gap between Sapporo and Kyoto is about a month, so almost any week in autumn has good colour somewhere in Japan — it just may not be where you first pictured.

If you want a single confident date to book around, I can’t give you one and neither can anyone else honestly. The measurement below moves about two weeks between years. Anyone quoting you a fixed peak week for next autumn is quoting an average without telling you it’s an average.

And if what you actually want is the famous illuminated temple photograph, understand that those events run to a schedule set months in advance, not to the trees. In a year that runs late, the illumination ends while the maples are still green. That is a scheduling risk, not a forecasting one, and no dataset will protect you from it.

3. What these dates actually measure — read this before you book

Japan has done something unusual since the 1950s: the meteorological agency’s regional offices observe a named sample tree at each station, and record the day it changes. The observation is standardised nationwide, which is what makes the numbers comparable across seventy years and across the length of the country.

The species for autumn colour is iroha maple (Acer palmatum), and the definition of the colour date is precise:

「対象とする植物の葉の色が大部分紅（黄）色系統の色に変わり、緑色系統の色がほとんど認められなくなった日」

The day on which most of the plant’s leaves have turned to red (or yellow) tones and almost no green tones remain.

Three consequences follow, and they matter more than anything else in this article:

This is full colour, not the beginning of colour. The “just starting to turn” stage that photographs well against green is roughly one to two weeks earlier than the dates below. If you want the mixed green-and-red look, aim before the date. If you want the saturated all-red look, aim at it.

It is one tree, at one weather station, in a city, near sea level. It is not the maple in the temple garden you saw on Pinterest. Famous gardens plant different cultivars, sit in their own microclimates, and are often several hundred metres higher. The station tree is a consistent yardstick, not a promise about a specific garden.

The record includes a second date: leaf-fall. The same programme records the day the sample tree has essentially dropped its leaves. The gap between the two is the honest answer to “how long do I have”, and I have put it in the table.

4. The table: average colour date by city

Averages of the observed dates for 2016–2025 — the most recent ten years — computed from the agency’s accumulated tables. “Earliest–latest” is the actual spread within those ten years, not a forecast range. The last two columns show how far the ten-year average has moved from the 1991–2020 average, the standard thirty-year reference period.

Nara is in the record too and behaves oddly against its neighbours — a ten-year average of 18 November, more than three weeks before Kyoto an hour away. That is a good reminder that a single sample tree can sit in its own pocket of climate, and that you should not plan a Nara day trip on a Kyoto date or the reverse.

Across all the cities above, the median spread between the earliest and latest year of the decade is 13 days, and the median viewing window between full colour and leaf-fall is 11 days.

Those two numbers are the whole planning problem in one line: the window is about eleven days long, and it moves about thirteen days.

5. Three things the table tells you that the guidebooks don’t

5-1. Kyoto is a December destination now

The single most repeated piece of autumn advice about Japan — go to Kyoto in the second half of November — puts you there, on average, about two weeks before full colour. Even the earliest year in the last decade only reached 1 December.

This is genuinely useful, because it cuts against the crowd. Late November in Kyoto is peak price and peak crowd, driven by an expectation that the measurement does not support. The first ten days of December are cheaper, quieter, and closer to the average colour date.

The trade-off is real and you should know it: December is colder, daylight is shorter, and if the year runs early you will arrive to trees that have already dropped. That is what the earliest–latest column is for.

5-2. Plan a window, not a day

A thirteen-day median spread means a fixed one-day plan is a coin flip. The practical response is not to guess better — it is to build the trip so that being a week off does not ruin it.

Two ways to do that:

• Stack cities that are a week apart on the list. Sendai (Nov 23) and Tokyo (Nov 28) are five days apart in the average; Tokyo and Kyoto are twelve. A trip that moves down the list is a trip that catches colour somewhere, whichever way the year breaks.
• Aim at the late edge of your window, not the middle. Arriving slightly late costs you leaves on the ground; arriving early costs you green trees and there is nothing to be done about it. The leaf-fall column tells you how much late you can afford — twelve days in Tokyo, six in Kanazawa, four in Sapporo.

5-3. It has been getting later, and unevenly

Compare the ten-year average against the 1991–2020 reference and most of the country has moved later — Sapporo by eleven days, Nagano by eight, Hakodate by seven. Two decades of guidebook advice have quietly gone out of date, which is the deeper reason the English-language dates now read early.

The shift is not uniform. Some southern stations moved the other way. Do not treat “everything is two weeks later than it used to be” as a rule; use the row for the city you are actually going to.

6. The one adjustment the table cannot make for you: altitude

Every station in that table is a city station near sea level. The places people actually photograph — the gorge walks, the alpine routes, the mountain temples — are higher, and higher means earlier. Sometimes dramatically earlier: a high alpine route can be finished with autumn before Tokyo has started.

This dataset cannot tell you by how much, and I am not going to invent a conversion factor. What works instead:

• Use the nearest city row as your floor, never your target. If you are heading into the mountains near Nagano (Nov 20 in the city), the mountains are earlier than that, not later.
• Read the operator, not the guidebook. Ropeways, national park visitor centres and mountain railways publish their own colour updates through the season, usually with photographs, and they are the only source with an incentive to be right. They are also the ones who will tell you when a road or a trail is closed.
• Assume the mountain season ends before the city season starts. If your trip is one week long and you want both a mountain and a city, do the mountain first.
• Or hand the altitude problem to someone who watches it daily. A private full-day guide is the one booking that adapts to the year rather than to a schedule set in advance — if the colour has moved up or down the mountain, the route moves with it. Nikko is the obvious case, since it spans roughly a thousand metres of elevation within one day trip: Nikko full-day private tour.

If you are building a mountain leg into the trip, the Japan packing list covers the temperature gap between a Tokyo afternoon and a gorge at 8am, which in late autumn is larger than most visitors expect.

7. How to actually build the trip around this

Decide which of the two looks you want. Mixed green-and-red is roughly a week or two before the date. Saturated red is the date. Leaves on the ground and thin canopy — which photographs beautifully and is nearly always uncrowded — is the week after the date.

Book flights against the demand calendar, not the leaf calendar. Autumn is structurally expensive into Japan, and the expensive weeks are set by Japan’s own holidays as much as by yours. The cheapest time to fly to Japan guide covers which weeks are unavoidable and which are just habit.

Move by rail, and do the pass maths before you buy. An autumn trip that chases colour down the country is exactly the shape of trip where a nationwide pass can pay — and exactly the shape where people buy one out of habit and lose money. Run the numbers in the JR Pass guide, and pick up an IC card either way for the local legs, which the Suica, PASMO and ICOCA guide explains.

Book the timed things last. Illuminated evening openings, guided walks and mountain transport sell out in autumn, but they are also the things you most want to move if the year runs early or late. Day trips out of the big cities are the flexible layer here — Klook’s Tokyo day-trip listings are where I would look first for the ones you can rebook without losing the whole booking, and the best Klook tours in Tokyo guide goes through which ones are worth it.

If you would rather not gamble on the date at all, book a tour that is built around the foliage itself. These exist for exactly the reason this article does — the operator is watching the colour and the itinerary is chosen for it, which transfers the timing risk off you:

• Out of Tokyo: Mt Takao and Showa Kinen Park autumn foliage tour. Takao is high enough to run ahead of central Tokyo’s Nov 28 average, which makes it a sensible hedge if you arrive early in the window.
• Out of Fukuoka: Saga autumn foliage and shrine day tour with lunch. Kyushu is the latest region in the table — a Dec 4 average in Fukuoka — so this is the one that still works when everywhere north of you has finished.
• For the postcard rather than the leaves: Mt Fuji, Chureito Pagoda, Lake Kawaguchi and Oshino Hakkai. The pagoda-and-Fuji frame is the most photographed autumn composition in Japan, and it is a long, awkward day to assemble by public transport.
• At the other end of the calendar: Asahiyama Zoo and Biei day tour from Sapporo. Hokkaido’s colour is finished by early November, so this is an October booking, not a November one.

Getting to the mountains is usually a bus problem, not a train problem. The high places that colour first are frequently off the rail network entirely — Japan by highway bus and ferry covers which ones, and why the seats sell out in exactly the weeks you want them.

Weekdays, always. An eleven-day window that includes two weekends is effectively a seven-day window, because the two weekends will be unusable at anywhere famous.

8. Where these numbers come from, and how to check them yourself

I have deliberately not asked you to take any of this on trust. The underlying record is public:

• The colour dates and leaf-fall dates are from the Japan Meteorological Agency’s accumulated phenological tables, which run from 1953 to 2025.
• The ten-year averages, the earliest–latest spreads and the window lengths in the table are my own calculations from those tables, not figures the agency publishes in that form.
• Where the record notes a date in the following January — which happens in western Japan for leaf-fall — I have counted it as belonging to that autumn rather than to the start of the next year.

If you want a specific city not in the table, the tables cover roughly ninety stations and are searchable by station name.

One honest limitation. Ten years is a short base for an average, and I chose it deliberately over the thirty-year standard because the thirty-year figure is now measurably out of date for much of the country — that is the whole point of the shift column. A ten-year average is more current and more volatile. Both of those are true at once, and the earliest–latest column is there so you can see the volatility rather than have it hidden inside a mean.

9. Conclusion

The single most useful correction this data makes is the simplest one: for most of Japan, “autumn colour” is a late-November-to-mid-December event, not a mid-November one, and the further west and south you go the later it gets. Kyoto’s average is 10 December.

Plan an eleven-day window rather than a day. Expect the year to land within about a week either side of the average. Go a week early if you want green in the frame, a week late if you want quiet. And for anywhere in the mountains, treat every date in this article as the last possible date rather than the target.

One more thing worth building into an autumn plan: the cold end of the season is when a hot-spring stop earns its place in the itinerary, and several onsen towns can be done as an afternoon rather than an overnight. Which ones, and what a day visitor actually pays, is in the onsen towns you can bathe in without booking a room.

Sources: JMA — phenological observation records, maple colour (1953–2025, PDF), JMA — phenological observation records, maple leaf-fall (1953–2025, PDF), JMA — maple colour dates, 2024–2025, JMA Hiroshima Local Meteorological Office — observation definitions. Averages and window lengths are my own calculations from the accumulated tables. Confirm live prices and opening dates before booking.