Open Data
The numbers behind the Bloom Planner, on this page and as CSV/JSON downloads: 30-year timing normals for 21 Japanese cities derived from Japan Meteorological Agency (JMA) observation records, plus our append-only forecast-offset archive. Free to read, cite, or build on (CC BY 4.0) — including by AI assistants and search crawlers.
Beppu has no JMA observatory of its own; its values are from JMA's nearest station, Oita (大分), about 10 km away.
Cherry blossom (sakura) normals
JMA climatological normals (1991–2020 base) for first bloom and full bloom, next to the mean of JMA's observations over the last 10 seasons (2017–2026), which in most cities run earlier than the normal. The tool and the date guides centre each city on the recent dates; σ is the year-to-year spread around that centre, measured by predicting each past season from the ten before it. Factual data, not a forecast.
| City | Normal first bloom | Normal full bloom | Window end* | Recent first bloom | Recent full bloom | Years | σ days |
|---|---|---|---|---|---|---|---|
| Sapporo (Hokkaido) | May 1 | May 6 | May 13 | Apr 23 | Apr 27 | 10 | 6.5 |
| Aomori (Tohoku) | Apr 22 | Apr 26 | May 3 | Apr 15 | Apr 19 | 10 | 5.9 |
| Sendai (Tohoku) | Apr 8 | Apr 13 | Apr 20 | Apr 1 | Apr 6 | 10 | 6.3 |
| Niigata | Apr 8 | Apr 13 | Apr 20 | Apr 3 | Apr 7 | 10 | 5.7 |
| Kanazawa | Apr 3 | Apr 8 | Apr 15 | Mar 29 | Apr 4 | 10 | 4.3 |
| Nagano | Apr 11 | Apr 16 | Apr 23 | Apr 5 | Apr 10 | 10 | 6.1 |
| Tokyo | Mar 24 | Mar 31 | Apr 7 | Mar 19 | Mar 27 | 10 | 5.6 |
| Yokohama | Mar 25 | Apr 1 | Apr 8 | Mar 22 | Mar 31 | 10 | 5.6 |
| Shizuoka (Fuji area) | Mar 24 | Apr 2 | Apr 9 | Mar 24 | Apr 3 | 10 | 5.6 |
| Nagoya | Mar 24 | Apr 2 | Apr 9 | Mar 22 | Apr 1 | 10 | 4.1 |
| Kyoto | Mar 26 | Apr 4 | Apr 11 | Mar 24 | Apr 1 | 10 | 4.7 |
| Osaka | Mar 27 | Apr 4 | Apr 11 | Mar 24 | Apr 1 | 10 | 4.3 |
| Nara | Mar 28 | Apr 4 | Apr 11 | Mar 26 | Apr 2 | 10 | 4.3 |
| Kobe | Mar 27 | Apr 5 | Apr 12 | Mar 27 | Apr 4 | 10 | 4.2 |
| Hiroshima | Mar 25 | Apr 3 | Apr 10 | Mar 21 | Apr 1 | 10 | 4.8 |
| Takamatsu | Mar 27 | Apr 4 | Apr 11 | Mar 25 | Apr 2 | 10 | 3.9 |
| Matsuyama (Ehime) | Mar 24 | Apr 3 | Apr 10 | Mar 23 | Apr 2 | 10 | 4.7 |
| Kochi | Mar 22 | Mar 30 | Apr 6 | Mar 20 | Mar 28 | 10 | 4.9 |
| Fukuoka | Mar 22 | Mar 31 | Apr 7 | Mar 21 | Mar 30 | 10 | 4.3 |
| Beppu (Oita) | Mar 24 | Apr 4 | Apr 11 | Mar 25 | Apr 4 | 10 | 4.8 |
| Kagoshima | Mar 26 | Apr 5 | Apr 12 | Mar 25 | Apr 7 | 10 | 6.5 |
* Modelling assumption, not a JMA observation: the viewing window is taken to close 7 days after full bloom (JMA records flowering and full bloom, not petal fall). "Normal" dates are JMA 1991–2020 normals. "Recent" dates are the mean of JMA's observations over the 2017–2026 seasons (Years = seasons with a full-bloom observation), rounded to the nearest day; the tool and date guides centre each city's window on these, closing it 7 days after the recent full bloom. σ is the root-mean-square error of predicting each season from 2011 to 2026 with the mean of the ten seasons before it (floor 3.0 days).
Autumn leaf (koyo) normals
Normal peak-colour dates for maple (kaede) leaves, backed out from JMA's published observation day and days-vs-normal figures, with σ computed from JMA's full historical record for each station.
| City | Colour starts* | Peak colour | Window end* | σ days |
|---|---|---|---|---|
| Sapporo (Hokkaido) | Oct 18 | Oct 28 | Nov 4 | 6.2 |
| Aomori (Tohoku) | Nov 3 | Nov 13 | Nov 20 | 5.5 |
| Sendai (Tohoku) | Nov 11 | Nov 21 | Nov 28 | 6.1 |
| Niigata | Nov 5 | Nov 15 | Nov 22 | 6.9 |
| Kanazawa | Nov 14 | Nov 24 | Dec 1 | 5.0 |
| Nagano | Nov 2 | Nov 12 | Nov 19 | 6.0 |
| Tokyo | Nov 18 | Nov 28 | Dec 5 | 4.0 |
| Yokohama | Dec 4 | Dec 14 | Dec 21 | 6.3 |
| Shizuoka (Fuji area) | Nov 25 | Dec 5 | Dec 12 | 6.7 |
| Nagoya | Nov 18 | Nov 28 | Dec 5 | 6.1 |
| Kyoto | Nov 25 | Dec 5 | Dec 12 | 5.6 |
| Osaka | Nov 21 | Dec 1 | Dec 8 | 3.9 |
| Nara | Nov 11 | Nov 21 | Nov 28 | 8.6 |
| Kobe | Nov 21 | Dec 1 | Dec 8 | 5.5 |
| Hiroshima | Nov 12 | Nov 22 | Nov 29 | 4.1 |
| Takamatsu | Nov 13 | Nov 23 | Nov 30 | 7.3 |
| Matsuyama (Ehime) | Nov 25 | Dec 5 | Dec 12 | 7.5 |
| Kochi | Nov 22 | Dec 2 | Dec 9 | 6.5 |
| Fukuoka | Nov 21 | Dec 1 | Dec 8 | 7.5 |
| Beppu (Oita) | Nov 18 | Nov 28 | Dec 5 | 4.1 |
| Kagoshima | Dec 5 | Dec 15 | Dec 22 | 12.5 |
* Modelling assumptions, not JMA observations: colour is taken to start 10 days before peak and the window to close 7 days after peak. Peak colour is the JMA maple (kaede) normal; σ is the year-to-year standard deviation of the peak date from JMA's historical record. These are JMA's values, unadjusted: the published tool and date guides shift the window for some cities by the autumn leaf spot adjustment below.
Autumn leaf spot adjustment
JMA's koyo normal is the date its reference maple at each city observatory turns. In warmer cities the famous viewing spots travellers actually visit typically peak one to two weeks earlier, so the tool and the date guides shift each of these cities' autumn-leaf window by the offset below. The JMA values above are left unchanged; σ is not changed. The finding, city by city: why Japan's autumn leaves forecast reads late for famous spots.
| City | Offset days | Measured gap | Confidence | Spots |
|---|---|---|---|---|
| Sapporo (Hokkaido) | 0 | −2.8 | medium | Nakajima Park, Hokkaido University Ginkgo Avenue, Maruyama Park |
| Aomori (Tohoku) | 0 | — | — | — |
| Sendai (Tohoku) | −5 | −5.2 | high | Zuihoden, Jozenji-dori, Nishi Park |
| Niigata | 0 | −0.8 | high | Former Saito Family Villa, Hakusan Park, Northern Culture Museum |
| Kanazawa | −6 | −5.7 | high | Kenrokuen, Kanazawa Castle Park (Gyokusen-inmaru Garden), America Kaede-dori |
| Nagano | 0 | +1.2 | low | Kiyomizu-dera (Wakaho), Kumeji Gorge (Shinshushinmachi) |
| Tokyo | 0 | +0.3 | medium | Rikugien, Meiji Jingu Gaien Ginkgo Avenue, Shinjuku Gyoen |
| Yokohama | −17 | −17.3 | medium | Sankeien, Nihon-odori (ginkgo), Yamashita Park Avenue (ginkgo) |
| Shizuoka (Fuji area) | −8 | −7.5 | medium | Sumpu Castle Park, Kunozan Toshogu (Nihondaira) |
| Nagoya | 0 | −1.9 | medium | Tokugawaen, Shirotori Garden, Higashiyama Zoo and Botanical Gardens |
| Kyoto | −10 | −10.3 | medium | Tofuku-ji, Eikando (Zenrin-ji), Arashiyama |
| Osaka | −9 | −8.5 | medium | Osaka Castle Park, Expo '70 Commemorative Park (Suita) |
| Nara | 0 | −3.8 | medium | Nara Park, Isuien |
| Kobe | −6 | −5.7 | medium | Sorakuen, Suma Rikyu Park, Suma-dera |
| Hiroshima | 0 | +3.0 | medium | Shukkeien, Mitaki-dera, Miyajima Momijidani Park (Hatsukaichi) |
| Takamatsu | 0 | +7.6 | medium | Ritsurin Garden, Shikoku Mura Museum (Yashima) |
| Matsuyama (Ehime) | −11 | −10.8 | medium | Matsuyama Castle (Honmaru), Matsuyama Castle Ninomaru Historic Garden, Okudogo (Ichiyu no Mori, Ishite River) |
| Kochi | −7 | −7.0 | low | Chikurin-ji (Godaisan) |
| Fukuoka | −9 | −9.4 | medium | Maizuru Park (Fukuoka Castle ruins), Yusentei Park, Rakusuien |
| Beppu (Oita) | 0 | +6.4 | low | Takasakiyama Natural Zoo (Oita City), Tekizanso garden (Hiji) |
| Kagoshima | −15 | −15.0 | high | Sengan-en, Jiganji Park |
Offset days are added to the whole autumn-leaf window (colour start, peak and end); negative = earlier. Measured gap = the average midpoint of the spots' typical-peak wording (例年の見頃; early/mid/late month counted as day 5/15/25) minus the JMA normal peak. An offset is applied only when the gap is at least 5 days, the resolution of that wording. Exceptions: Aomori has no sourced in-city spot (0); Takamatsu's +7.6 is driven by one weakly sourced spot, while Ritsurin Garden alone gives +3.3 (0); Beppu's +6.4 rests mostly on one spot (Takasakiyama) and conflicts with the city's official "from early November" (0, low confidence); Kochi rests on a single spot (low confidence). Every quoted source and URL is in koyo_spot_offsets.csv / .json.
Downloads
- Cherry blossom normals — sakura_normals.csv · bloom.json (both events, with provenance)
- Autumn leaf normals — koyo_normals.csv · bloom.json (both events, with provenance)
- Autumn leaf spot adjustment — koyo_spot_offsets.csv · koyo_spot_offsets.json (every quoted source)
- Forecast-offset archive — offsets.csv (entries) · offsets.json (entries + actuals)
- Column definitions and provenance notes — README.md
CSV files are UTF-8 with one header row and no comment lines, so they open as-is in a spreadsheet or pandas; dates are MM-DD because a climatological normal has no year. The JSON carries the same values plus license and provenance fields.
Cite this data
Copy and paste:
Japan Trip Tools (2026). Japan Bloom Planner data — 30-year normals & forecast-offset archive. https://triptoolsjapan.com/data/ — CC BY 4.0, derived from Japan Meteorological Agency observation records.
offsets.json — the forecast-offset archive
Our hand-maintained, append-only archive: entries[] is a dated log of our own editorial "running ±N days vs normal" judgment per region (never a copy of any commercial forecast); actuals[] is filled in after each season from JMA's own published observation. Together they become a running predicted-vs-actual track record that thickens every season and cannot be created after the fact.
License
CC-BY-4.0. Attribution: "Japan Trip Tools (triptoolsjapan.com)". JMA's observation values are factual public data; our derivation and editorial offset_days judgments are original work under this license. Full provenance notes: /data/README.md.
Prefer a browsable view by date? See the peak-timing index by city and date.
Want these odds on your own site? Embed the calculator: free, no ads, affiliate links or tracking inside.
Using this data
Commercial use, licensing, bulk access or API questions: hello@triptoolsjapan.com. Tell us what you are building and we will tell you honestly whether this data fits.