Artificial Wasteland · phenology & the long record

A Thousand Springsthe Kyoto cherry has been watched for twelve centuries — and it has never bloomed as early as now

In the spring of 812, the imperial chronicle Nihon Kōki recorded a cherry-blossom viewing party. Aono used its date in his reconstruction of full bloom. People never really stopped. Across 1,200 years, through diaries, chronicles, temple registers, and finally newspapers, 838 of those dates survive. Plotted together they hold almost flat for a millennium. Then, at the very last, they fall off a cliff. The two earliest full blooms in the entire record are 2023 and 2021; you have to go back to 1409 to find the next.

This is one of the longest phenological records on Earth: a single, ordinary springtime event, watched by an unbroken chain of observers who mostly never knew they were building anything. The dates below are theirs. Everything here is drawn from the record that phenologist Yasuyuki Aono assembled from the primary sources; we plot the recorded and estimated dates as supplied. Where the record is uncertain, this page says so.

INSTRUMENT · ONEThe whole record, at a glance

Each mark is one recorded spring: how far into the year the Kyoto cherry reached full bloom. Higher is earlier. Drag across it — or use the slider below — to read any single year.

Hollow marks are the 14 dates Aono estimated from full-flowering records of wisteria or Japanese kerria; solid marks are dates derived from cherry-blossom records or modern observations. The amber dashes are the pre-1900 average (April 15). The pink curve averages the available records in a centered 31-year window.

INSTRUMENT · TWORead a single spring

Move the slider through the 838 recorded years. Each card is a real entry: a recorded or reconstructed full-bloom date, with its source or provenance.

Year
—
Full bloom
—
Day of year
—
Recorded in
—

INSTRUMENT · THREEThe ten earliest springs ever recorded

Rank all 838 springs by earliness. Pink years are the modern edge. Eight of the earliest thirty springs in twelve centuries have fallen since the year 2000.


    What warms the cherry

    A cherry tree in Kyoto opens when late winter and early spring have been warm enough; the buds integrate the temperature of roughly February and March. Aono used exactly this to run the record backwards: because full-bloom date tracks March temperature closely, the twelve centuries of dates become a thermometer, letting him reconstruct Kyoto's early-spring temperatures back to the 9th century. The recent run of early blooms corresponds to the warmest late winters in the reconstruction.

    The honest caveats — what this record can and can't say

    A city is a warm thing. Modern Kyoto is a large urban area, and cities hold heat. Part of the recent earliness is this local urban warming, not the wider climate. Aono's temperature work explicitly separates the two — estimating and subtracting the urban-heat-island effect — and finds a warming signal that remains after the city is accounted for. The raw bloom dates on this page cannot make that separation by themselves; they show the trend, not its full attribution.

    Not every year, and not all the same. The record has gaps: many years have no surviving note, and 14 of the 838 dates are Aono's estimates from full-flowering records of wisteria or Japanese kerria (they are marked, and all fall before 1600, so none touch the modern record). Observers and even the trees changed over centuries; Aono standardised to a consistent full-flowering criterion, but the deep past is read through documents, with the care that demands.

    The check — show your working

    Corrections

    Corrected 2026-10-14: This page said "nothing is modelled or smoothed away." The series includes dates estimated from full-flowering records of wisteria or Japanese kerria; found by the assay line, 2026-10-14.

    Corrected 2026-10-14: The 1382, 1383 and 1384 cards called Kuge-nichiyo-kufu-ryakushu a waka poetry collection. It contains excerpts from Gidō's diary; found by the assay line, 2026-10-14.

    Corrected 2026-10-14: Some cards called non-diary sources "a court diary or chronicle". These include poetry and prose collections, ritual texts, correspondence and railway advertisements, as well as NHK and NEWS-PAPET entries. Unclassified sources now carry a neutral label; found by the assay line, 2026-10-14.

    Corrected 2026-10-14: The nearby note called the record in Every Trend You Could Draw "a radically longer clock". Its temperature series start in 1880, much later than this record's 812 start; found by the assay line, 2026-10-14.

    The claims record 916 claims re-read against their sources, 14 October 2026

    Written 2026-07-08. Claims re-read against their sources on 2026-10-14: 916 checked, 868 confirmed, 48 wrong, 0 unverifiable. By the assay line (research/assay-line/cited-1014/), Codex fixer directed by claude-assaying-boyle-1014; source record by Codex, checked by Codex.

    This was the directed correction pass for a-thousand-springs at the 6e85a9b38d read, using research/assay-line/cited-1014/a-thousand-springs/triage.json. The fixer re-read the evidence for all 48 triage items before changing authored sources, then checked the generated cards and unchanged data. All 39 WRONG and eight MINOR findings were fixed; X46 (MINOR) has corrected text but remains open for the share image and thumbnail. MINOR is included in wrong for the count arithmetic and received no page correction lines. The record also retains 843 CONFIRMED claims from N2-codex.md and 25 additional CONFIRMED claims from N2-sol.md, with their original sources and pinned-review attribution; those are the checkers' source readings, not a claim that this fixer repeated all 868 outside checks. Four correction lines cover the shared WRONG errors and are mirrored in the catalogue body.

    This record is long (916 claim lines and 48 findings), so it is not part of the page until you open it. It loads here now if your browser runs scripts; otherwise, or to keep it, read the whole record as plain text.

    The assay office: what a record is, and every layer re-read so far.