Spring phenology temperature sensitivity at Japanese weather stations did not decrease in 1953–2025 beyond a forcing-only model, which also explains lower cherry sensitivity at warmer stations

Question: does the apparent temperature sensitivity (days per °C) of spring phenology decrease with warming, as reported for European leaf-out, or is the decrease a result of heat-sum (forcing) arithmetic? Data: Japan Meteorological Agency records 1953–2025 (regular species only) for eight spring events: cherry 'Somei-yoshino' (one clone) first flower and full bloom, plum, dandelion, wisteria and azalea flowering, ginkgo and mulberry budburst; 36–80 stations and 1,718–4,776 station-years per event; GHCN-Daily temperature from the same stations. Method: per-station OLS slopes of event day on mean temperature in a fixed 60-day or 30-day pre-event window. A growing-degree-day null model (common start day and base temperature, station-specific requirement) simulated event days from the same daily temperatures, and we put the simulated days through the same regressions. 95% intervals come from a 1,000-replicate bootstrap of whole years shared by all stations. Robustness: two windows, detrending, an instrumental-variable slope with neighbour-station temperature, on-premises cherry records only, and a null fitted on the colder half of stations. Results: the null predicts lower sensitivity at warmer stations for all events (+0.14 to +0.27 days/°C per °C of winter temperature). For cherry first flower, the observed gradient was +0.25 (0.15 to 0.33) and the null gradient was +0.24 (0.19 to 0.28), with a difference of 0.00 (−0.06 to 0.07). Only the four warmest stations (winter mean 9–12.5 °C) were clearly less sensitive than the null. In 32 tests (8 events × 4 specifications), no event had a significant decrease in sensitivity between 1953–1988 and 1989–2025. Wisteria had a significant increase. The observed-minus-null change was never significantly positive. The plum decrease (+1.5 days/°C, not significant) agreed with the null. Between 1980–1994 and 1999–2013, the cherry first-flower sensitivity changed by +13% (−27% to +62%; positive = stronger). In 20-year moving windows, cherry sensitivity became stronger until about 1995 and then weaker, so two-period comparisons can show either sign. Limits: one urban reference tree per station, a simple forcing null with little skill for plum and dandelion, fewer stations after 2010, and no chilling model, so the cause of the excess at the warmest stations is not tested.

phenology · temperature sensitivity · growing degree days · cherry blossom · Somei-yoshino · Japan Meteorological Agency · GHCN-Daily · climate change · null model · spring flowering · budburst