Warmer winters predict longer cherry bloom intervals at Japanese stations with mild winters in 1953–2025

We tested the interval from first flowers to full bloom in Somei-yoshino cherry at 72 Japanese weather stations in 1953–2025 (4,760 paired station-years). A station model controlled year, first flowering day, and temperature during the first seven days of bloom. At a station with a 1961–1990 winter baseline of 8 °C, a 1 °C winter increase predicted a 1.15-day longer interval (95% block bootstrap interval 0.86 to 1.42). At a baseline of 3 °C the increase was 0.27 days (0.08 to 0.46). A 1 °C bloom-week increase predicted a 0.59-day shorter interval (0.53 to 0.65). The warm-station winter link persisted in checks with station trends and after removal of the hottest station. Models trained on 1953–2000 data had a test RMSE of 2.04 days without winter terms and 1.86 days with them in 2001–2025. The error change was −0.18 days (95% interval −0.35 to −0.07 with model refitting). The pooled interval trend was uncertain: +0.10 days per decade (−0.05 to +0.29). This wider-network test supports earlier southern field evidence of longer intervals after mild winters. It is a conditional statistical result. Observed bloom-week weather is used, so the test is not a forecast before flowering. Missing full bloom records and changes to reference trees limit the result. The package preserves source bytes, provenance, code, data, and figures.

phenology · cherry flowering · winter chilling · Japan · JMA · climate response · reproducibility · temporal validation