Spring plant phenology at 81 Japanese weather stations became more sensitive to temperature from 1953–1986 to 1987–2020, while fixed thermal-time models predict less sensitivity

Question: is the change in the linear temperature sensitivity of spring phenology in Japan the change that a fixed thermal-time requirement predicts? A fixed degree-day requirement makes the linear sensitivity (days per °C) decrease when the climate becomes warmer (Wolkovich et al. 2021). We used 587 station-event series of Japan Meteorological Agency phenology records (81 stations, 23 flowering and budburst events, 1953–2020) and the daily mean temperature of each station. For each series, we fit a fixed model to the 1953–1986 dates only. Then we calculated the slope of the event date on the 60-day pre-event temperature in 1953–1986 and in 1987–2020, for observed and model dates. The preseason temperature increased by 1.05 °C. The fixed degree-day model predicted a weaker response: +0.42 days/°C (95% CI +0.27 to +0.57, station bootstrap). The observed response became stronger: −0.52 days/°C (−0.83 to −0.18), from −4.78 to −5.30. The difference was −0.94 (−1.24 to −0.63) and was outside the range of fixed-model simulations with resampled residuals (p < 0.001). Sigmoid forcing and chilling–forcing (alternating) models predicted +0.32 and +0.66, so a chilling limit does not explain the result. The result was the same with model fits in 1987–2020 or 1953–2020, a late period to 2025, 45- and 90-day windows, detrending, homogeneous temperature records only, on-site trees only and all filters together. With a 30-day window, the observed change was −0.03 and the model change was +0.55. Cherry flowering and full bloom, wisteria and ginkgo budburst became 0.7–1.0 days/°C more sensitive than the model predicts. Japanese apricot sensitivity decreased, but less than predicted. Japanese lawn grass was the only event with a less sensitive observed response than predicted. Limits: one reference tree per station, often in a city; station air temperature in fixed calendar windows; simple grid-search models; the cause of the increase is not identified. Thus, in these records the predicted artefact of declining sensitivity is not present, and the observed response became stronger.

phenology · temperature sensitivity · thermal time · degree-day model · Japan Meteorological Agency · cherry flowering · climate change · chilling · budburst · out-of-sample test