At JMA stations in 1953–2020, the excess increase of phenology temperature slopes over thermal-time models is in the network-wide yearly temperature signal, not in station-specific variation

Question: a fixed thermal-time requirement makes the linear temperature slope of spring phenology (days per °C) weaker in a warmer climate. SCI 5005baf5 found that Japan Meteorological Agency (JMA) spring slopes instead became stronger from 1953–1986 to 1987–2020, but did not find the cause. We tested whether this reproduces, whether autumn leaf events show it, and where in the temperature variation it occurs. Data: JMA phenology tables (regular species) and JMA daily mean temperature and sunshine at the same stations; 556 spring series (73 stations, 20 events) and 159 autumn leaf series (48 stations, 5 events). Method: per series, a constant degree-day model fitted to all years gives model dates from the same temperatures. We compared the change of the OLS slope of date on 60-day window temperature for observed and model dates. 95% intervals come from a year bootstrap with years shared by all series. Results: spring observed minus model change was −1.00 days/°C (−1.71 to −0.39; permutation p = 0.011), which reproduces 5005baf5. Autumn leaf events had a stronger observed delay per °C than the model predicts: +0.92 (+0.28 to +1.61; permutation p = 0.073). We divided the window temperature into the mean anomaly of all stations in a year (shared part) and the station-specific rest. The spring departure was in the shared part (−1.32; −2.29 to −0.36) and not detected in the station-specific part (−0.44; −1.16 to +0.19). This held for 30- and 90-day windows and a period to 2025. Autumn was similar (shared +0.91; station-specific +0.62, CI includes zero). Thus station thermometer errors, relocations and local urban change are not a probable main cause. Sunshine as a covariate changed the spring result only to −0.86 (−1.63 to −0.18), although the spring temperature–sunshine correlation changed from −0.13 to +0.15. A chilling–forcing model also did not predict the stronger slope. Limits: one reference tree per station and event, fixed calendar windows, simple models, and wider intervals for the station-specific part; the process that acts on the network-wide signal is not identified.

phenology · temperature sensitivity · thermal time · degree-day model · Japan Meteorological Agency · cherry blossom · autumn leaf colouring · declining sensitivity · climate change · year bootstrap · sunshine duration