# HYAKKEI open data

Every published scene of **HYAKKEI: 100 Scenes of Japan** (https://spectrip.com/hyakkei/) as structured data, plus
two separate records of the autumn color: one from a mountain hut, one from a national-park visitor center.

This is **version 2026.09.1**, covering 73 of a planned 100 scenes, built on 2026-10-13.
An archived copy of this version is kept under the DOI below; cite that rather than this page if you need something that will not move.

## How the version number works

The files here are rebuilt every time the site is, so their contents change between versions: scenes
are added at roughly one a week, and figures are re-checked and corrected as we go. The version number
moves when any of that happens, in the form `YYYY.MM.N` where `N` counts the releases in that month.
**Two downloads carrying the same version number are the same data.** If you are citing this and there
is no DOI yet, record the version and the date you downloaded it.

## Files

| File | Use |
|---|---|
| `hyakkei.json` | Everything below, one object per scene |
| `hyakkei.csv` | Same data, one row per scene, for spreadsheets |
| `hyakkei.geojson` | Points for maps (Leaflet, Mapbox, QGIS) |
| `hyakkei.gpx` | Waypoints for GPS apps and cameras |
| `hyakkei.kml` | The same points for Google My Maps and Google Earth, with a photo and a link in each pin |
| `heatmap.svg` / `heatmap.png` | Month-by-month "at full strength" chart of every scene, for embedding |
| `foliage-karasawa.csv` | A different record, not a column of the above: one row per daily autumn-color report from a mountain hut in Japan's Northern Alps, 2019 to 2025. See its own section below |
| `foliage-okunikko.csv` | Another separate record: one row per autumn-color post from the government visitor center at Oku-Nikko, 2010 onward. See its own section below |
| `CITATION.cff` | Machine-readable citation metadata |

## Fields

| Field | Meaning |
|---|---|
| `no` | Scene number. Publication order, never reassigned, capped at 100. **Numbers are not contiguous**: a scene can be published out of order, so the highest number here is 77 while the count is 73 |
| `title` | Page title of the scene |
| `place`, `prefecture`, `region` | Landmark name, prefecture, and the region used on the site. `region` is our own grouping, not an administrative one |
| `lat`, `lon` | Approximate coordinates (WGS84) for **what that scene page is about**, which is not always the settlement of the same name: No.7 is the Kuniga cliffs on Nishinoshima rather than the center of the Oki Islands, and No.39 is the bathhouse street rather than the station. Not a photo spot; see "Where to Stand" on the page |
| `coord_checked` | Date the point was cross-checked against Wikidata and OpenStreetMap **and read against what the scene page is about**. **Blank means it has not been re-checked**, which is not the same as being wrong. 36 of 73 carry a date |
| `url` | Scene page |
| `best_window` | The "Best:" badge on the page, e.g. `Dec – Feb`. Blank where the page carries no such badge, which is not a missing value: some scenes are dated events or have a fixed open season and say so in a different badge |
| `peak_months` | Months marked ◎ on the page's calendar |
| `calendar` / `cal_jan` … `cal_dec` | ◎ peak · ○ good · △ possible · × closed or not running |
| `access_level` | 1 = easiest … 5 = hardest to reach. **Our judgment, not a measurement.** It weighs transfers, walking, timetables and weather together, so it is not a stand-in for distance or journey time |
| `budget_level` | 1–5 pips on the page |
| `budget_jpy_low`, `budget_jpy_high`, `budget_basis`, `budget_text` | The "Estimated Cost" total on the page, per person, on the basis stated in `budget_basis` (usually Tokyo round trip plus one night). The individual prices are recorded from published fares; **which ones belong in the total is our judgment** |
| `tags` | Theme badges (Onsen, Heritage, After dusk …) |
| `checked` | Month we last verified that page's figures against their sources. Our own record of our own work |
| `date_published`, `date_modified` | Page dates |
| `reddit_threads_cited` | Number of Reddit threads quoted in "From the Threads". A count of what we chose to quote, not of what exists |
| `review_rating` | Visitor review score out of 5, where one is published. 1 of 73 scenes have one |
| `hero_image` | Hero photo URL (photo licences vary; see the credit on the page before reuse) |

## What is recorded and what is judged

Some columns can be checked against somebody else. Others are our editorial reading, written in a
numeric shape because that is the shape that is useful. They are not measurements and should not be
used as if they were.

| Kind | Columns |
|---|---|
| **Recorded from a source** | `lat`, `lon`, `prefecture`, `url`, `date_published`, `date_modified`, the individual prices behind `budget_jpy_low` / `budget_jpy_high` |
| **Our own record of our own work** | `coord_checked`, `checked`, `reddit_threads_cited`, `review_rating` |
| **Our judgment** | `calendar` / `cal_jan`…`cal_dec`, `peak_months`, `best_window`, `access_level`, `budget_level`, `tags`, `region`, what goes into a `budget_jpy_*` total, and **which places are on the list at all** |

Columns not named above (`no`, `title`, `place`, `hero_image`, `budget_basis`, `budget_text`) are
descriptive labels rather than values.

## The Karasawa autumn record (`foliage-karasawa.csv`)

The Karasawa Hyutte, a mountain hut at 2,300 m in Japan's Northern Alps, posts a dated note on the
autumn color most days of the season and keeps a public archive year by year. We read 111 of those
notes, from 2019-09-20 to 2026-10-11, and wrote down what each one said. One row per report.

As far as we can find, no equivalent record exists in English, and no public body keeps one: Japan's
national phenological observations are made at weather stations in towns, so nothing covers the color
at this altitude at daily resolution.

| Column | Meaning |
|---|---|
| `date`, `year` | The date of the report |
| `colour_pct` | The percentage the hut gave that day, from its figure in tenths (割). A range is averaged, so "5 to 6 tenths" becomes 55. **Sometimes the hut gives this figure for the upper part of the bowl only, and this column does not distinguish that**; 4 of the 22 figures are scoped that way. Blank where it gave no figure |
| `at_best` | 1 when the hut used its own word for at its best (見頃) or peak, **including** when it limited that to part of the bowl |
| `whole_valley` | 1 only when the hut used that word with nothing limiting it. The scene page calls this place a bowl, or cirque, which is what the word describes |
| `past_best` | 1 when it said the color was past, with nothing limiting it and not as a sign-off |
| `season_closed_marker` | 1 when the post carried the hut's end-of-series mark. These posts are a sign-off rather than an observation, so they do not set `past_best` |

### How we read it

The dates and the words are the hut's. **The rule for reading them is ours**, and it is this:

- A day counts as `at_best` when the text contains 見頃, ピーク, 最盛期, 真っ盛り or 極み.
- It counts as `whole_valley` only when none of 上部, 下部, 一部, 場所により, 所もあ, 付近 or 方面 also
  appears, since each of those limits the claim to part of the bowl.
- It counts as `past_best` when the text contains ピークは過ぎ, 終盤, 晩秋, 散り or 移ろい, **unless**
  the same day also says the color is continuing (見頃 followed within a clause by 続 or 残), or the
  post is a sign-off.
- A sign-off is （終）, （完）, or a bare 終 or 完 at the very end of the text.
- Words that appear in only one season are not used, because a rule that fires once is not a rule.

We publish this so you can disagree with it. Run the rule over the archive and you should get this file.

### Limits worth knowing before you use it

- **The record ends before the color does.** The hut stops posting when its season closes. In
  4 of the 7 complete years it never said the color was past; in the other
  3 it did, and 2 of those were its last post of the year. Read the end of a season as
  the end of the reporting.
- **14 of the 111 rows have no text**, because the post was a photograph. They appear here
  with every flag at 0, which is indistinguishable from a day the hut wrote something that matched
  nothing. Do not use the row count as a denominator without allowing for them.
- **The 2026 season is incomplete.** Its reports are in the file up to 2026-10-11, the last one
  read when the file was built. In the 7 complete years the hut's first report came between 20 and
  27 September, so a file built before then that ends in the previous year means the season has not
  started, not that the record was abandoned.
- **One hut.** Karasawa is here because that hut keeps an archive going back to
  2019. That also means this is one location and one observer, not a network. The
  Oku-Nikko file below is a second place and a second observer, read with its own rule.

The text of the reports is not reproduced here: this is a record of dates and of our reading, not a
mirror of the source. Read the source at <https://karasawa-hyutte.com/autumn_colors_cat/>.

**The reports belong to the Karasawa Hyutte.** The classification is our work and carries no
endorsement from the hut; we have not asked for one, and nothing here should be read as the hut's
own conclusion.

## The Oku-Nikko autumn record (`foliage-okunikko.csv`)

The Nikko Yumoto Visitor Center, run for the Ministry of the Environment at the top of the Oku-Nikko
valley north of Tokyo, posts a photograph and a line or two on the autumn color through the season.
It does so on nine pages, eight for one place each and one for everywhere else, and keeps every past
season online. Its pages carry 1,655 posts from 2010-09-14 to 2026-10-01; 15 of them repeat another
post's date and text, which leaves 1,640. We read all of those and wrote down how each one used the
word 見頃 (*migoro*, "at its best"). One row per post.

This is the record behind the Oku-Nikko table on our autumn page, which is built from this file, and
the table on the Oku-Nikko scene page, which is written by hand and checked against this file on every
build: the build stops if the two disagree.

| Column | Meaning |
|---|---|
| `date`, `year` | The date the center put on the post. Usually the day the photograph was taken; at least one post says its photograph is from the day before |
| `spot` | The center's own page the post sits on: `yumoto` Yumoto Onsen, `yunoko` Lake Yunoko, `yutaki` Yutaki Falls, `senjyo` Senjogahara marsh, `odashiro` Odashirogahara meadow, `ryuzu_u` above Ryuzu Falls, `ryuzu` Ryuzu Falls from the viewing platform below them, `chuzenji` Lake Chuzenji, `other` everywhere else |
| `mentions_migoro` | 1 when the post contains 見頃 or 見ごろ at all, in any sense |
| `at_best` | 1 when at least one use of the word says the place is at its best at the time of the post. The rule is below |
| `not_yet` | 1 when the word is used to say it is not yet: "not at its best yet", "a step short of it", "should be next week" |
| `past_best` | 1 when the word is used to say it has passed: "past its best", "over" |
| `part_only` | 1 when the word is used about only part of the place (一部, 所々, 部分) |
| `subject` | On `senjyo` and `odashiro` only: `grass` when the counted use is about the grass on the marsh (草紅葉), `larch` when it is about the larches (カラマツ), blank when it names neither. A clause naming both counts as `larch`; a post with counted clauses about each would read `grass+larch` (none does so far) |

A post can carry more than one flag: "the grass is past its best, but the larches are at their best"
sets both `past_best` and `at_best`, with `subject` = `larch`.

### How we read it

The dates and the words are the center's. **The rule for reading them is ours**, and it is this:

- We split each post into sentences at 。, ！, ？, ! and ?, and each sentence into clauses at が、, ものの
  and けど, and read every use of 見頃 or 見ごろ inside its own clause.
- Each use is tested in this order, and the first test it meets decides it: past, then not yet, then
  part only; a use that meets none counts as at its best. In the lists below, a character in parentheses
  may or may not be there, and "six characters" means six Japanese characters.
- A use is **past** when the word is followed by 過ぎ or すぎ (with or without は, を or が first), 終わ,
  終り, は終, はそろそろ終 or (の)ピークは過ぎ.
- A use is **not yet** when the word is followed by には, まで, に近づ, の予想, (は)まだ, (は)もう,
  (は)そろそろ, (は)これから, (は)当分, (は)目前, (は)間近, (は)先, (の)ピークはまだ, (の)一歩手前, 手前,
  近く, に(は)早, になりつつ, を迎えつつ, を迎えていく, を迎えるはず, を迎えるので, を迎えそう,
  を迎えるでしょう, になりそう, になるまで, に入りつつ, かも, は来週 or は少し; or when, within six
  characters before it, the clause has これから, もうすぐ, 間もなく, 来週には, 今週中に, 週末には,
  あと少しで, もう少しで or そろそろ.
- A use is **part only** when its clause contains 一部, 所々 or 部分. A clause that names a shore or a
  side ("the south shore is at its best", "on the east side of the lake") is not treated as part,
  and counts.
- Any other use counts as **at its best**. "At its best and continuing" counts. ピーク on its own
  never counts: unlike the Karasawa hut, the center mostly uses it for the approach to the best and
  the passing of it.
- Unlike the Karasawa rule, we do not drop phrases that appear in only one season. Some phrases in the
  lists appear once, and a few appear in no post yet. Each of them can only mean "not yet" or "past",
  and dropping a one-off would turn that post into "at its best".

The tables then take, for each year, the first post with `at_best` = 1:

| Table column | Rows used |
|---|---|
| Marsh grass | `senjyo` or `odashiro`, with `subject` = `grass` |
| Lake Yunoko | `yunoko` |
| Ryuzu Falls | `ryuzu` (the view from the viewing platform below the falls; `ryuzu_u`, above them, is in this file but not in the table) |
| Lake Chuzenji | `chuzenji` |
| Larches | `senjyo` or `odashiro`, with `subject` = `larch` |

A year with no such post is blank. **We do not fill a blank from a photograph or from a different
word.** For example, in the 15 seasons from 2010 to 2025 that have a date for Lake Yunoko, the
first fell between 9 October and 20 October.

We publish this so you can disagree with it. Applied to the center's pages as they stood on
2026-10-03, the rule gives this file; if you get something different, tell us.

### Limits worth knowing before you use it

- **The center posts every few days, not daily.** A first date is the first post that called the place at its best,
  so in a year with a gap between posts the real start may be earlier.
- **A blank in a table means the center did not call that place at its best that year, not that
  the color failed.** In some years it described a place at its best in other words, used the word
  only to say the color had passed, or, on the marshes, did not say whether it meant the grass or the
  larches.
- **The places are not fixed points.** Lake Chuzenji in particular is photographed from different
  shores in different years, and the color does not arrive on all of them on the same day.
- **442 of the 1,640 posts use the word at all, and 284 count as at its best.** Many posts,
  especially in the early years, are a place name and a short line, and carry every flag at 0.
- **The 2026 season is incomplete.** The center's pages were read on 2026-10-03; this season's posts are included, but
  the tables only show seasons that had ended when it was read. The Karasawa file does the same.

The text of the posts is not reproduced here: this is a record of dates and of our reading, not a
mirror of the source. Read the source at <https://www.nikkoyumoto-vc.com/nature/kouyou.html> (in
Japanese).

**The posts belong to the Nikko Yumoto Visitor Center** (Ministry of the Environment; run by the
Nikko branch of the Natural Parks Foundation). The classification is our work and carries no
endorsement from the center; we have not asked for one, and nothing here should be read as the
center's own conclusion.

## How to cite

Suzuki, Y. (2026). *HYAKKEI: 100 Scenes of Japan (open data)* (Version 2026.09.1) [Data set]. SPECTRIP. https://doi.org/10.5281/zenodo.22861242

That DOI resolves to the archived copy of this version, which does not change. Every version has its own DOI, so cite the one you actually used.

If you are crediting the data in a product or an article rather than citing it in a paper, the credit
line is "SPECTRIP HYAKKEI (spectrip.com)" with a link, and you do not need the form above.

## Corrections

When we find an error we fix it and say so: corrections are listed in the release notes for the next
version, and errors in the coordinates are recorded scene by scene with the reason. We would rather
hear about a mistake than not, at hello@spectrip.com.

## License

Data: **CC BY 4.0**. Attribute as "SPECTRIP HYAKKEI (spectrip.com)" and link to the scene page or to
https://spectrip.com/hyakkei/. Photos are not covered by this license; each page carries its own credit and license.

Questions and corrections: hello@spectrip.com