The excess spring temperature sensitivity at Japanese stations over a thermal-time model continued in 2002–2025, did not increase with local warming, and came with stronger winter-warmth delays

Question: sci:sha256:a372304ce3a1aadd55322628f187e2740d8c2ae143952729f53d47badda6abbb found that cherry, wisteria and ginkgo at Japanese weather stations became more sensitive to pre-event temperature after 1986 than a fixed thermal-time model predicts. sci:sha256:a30b18e0b9dbed22d7ae3539df648fe0cae472a7014dbd9770521051485120f2 found that raw cherry sensitivity became weaker after about 1995. We tested if the excess continues, which part of the pre-event temperature record causes it, and if it is larger where stations became warmer. Data: Japan Meteorological Agency records for 9 spring events (79 stations north of 30 N, 30,109 station-years, 1953-2025; most events end in 2020) and GHCN-Daily station temperature with neighbour gap filling. Method: per-station OLS slopes of event day on the mean temperature in a fixed pre-event window. A forcing-only degree-day model and a chill-forcing model were fitted to 1953-1986, and their predicted dates went through the same regressions. 95% CIs come from a bootstrap of 3-year blocks of years. Results: from 1953-1986 to 1987-2025, the observed-minus-model change was -0.67 days/C (CI -1.16 to -0.23) for cherry first flowering, -0.57 (-1.08 to -0.14) for full bloom and -1.01 (-1.67 to -0.49) for wisteria. This reproduces sci:sha256:a372304ce3a1aadd55322628f187e2740d8c2ae143952729f53d47badda6abbb. Plum sensitivity decreased (+1.15), as the model predicts (+1.74). In three periods with the same stations, raw cherry sensitivity became weaker after 2001 (+0.49), by about the model prediction (+0.64). The excess in 2002-2025 relative to 1953-1978 continued: -0.72 (-1.59 to -0.13) for cherry and -1.71 (-2.54 to -1.13) for wisteria. In 15-day block regressions, warmth 76-120 days before the event delayed 6 of 9 events more after 1987 (+0.54 to +0.77 days/C). Warmth 46-60 days before advanced cherry, wisteria and ginkgo more (-0.41 to -0.48). The fitted models did not reproduce these changes. The station-level excess was not correlated with station warming in the pre-event window or in winter (Spearman rho -0.09 to +0.16, all p >= 0.29). Limits: one urban reference plant per station, simple grid-search models with base temperatures at the grid edge, one chill-forcing form, and exploratory single-block tests. The cause is not identified. Because the excess does not scale with local warming, a common change in time (observation practice, circulation, tree age) remains possible.

phenology · temperature sensitivity · thermal time · growing degree days · chilling · cherry blossom · Japan Meteorological Agency · climate change · reproduction · GHCN-Daily