Station temperature explains 1953–2025 trends in spring plant phenology at Japanese weather stations, but animal first-detection dates and autumn leaf dates became later than temperature predicts
Question: does local temperature explain the long-term change in seasonal event dates? We used Japan Meteorological Agency phenology records (47 events, 20-92 stations each, 136,444 station-years, 1953-2025; animal records end in 2020) and GHCN-M v4 monthly temperature from the same stations. For each event, we calculated the date-temperature slope from year-to-year variation within stations and the slope from the differences between stations. We used one calendar temperature window per event. Then we divided the 70-year date trend into a temperature part (interannual slope multiplied by station warming) and a residual part. The 95% CIs come from a joint bootstrap that resamples stations and 5-year blocks of years. Spring plant events (16) advanced by 0.78 days/decade. The temperature part was -0.95 days/decade, and the mean residual was +0.17 days/decade (CI -0.18 to +0.55). The residual for clonal Somei-yoshino cherry flowering was -0.15 (CI -0.36 to +0.09). Animal first detections (20 events) had a mean residual of +1.85 days/decade (CI 1.50 to 2.21). Their observed trend was +1.26 days/decade, although warming predicted -0.58. The residual CI was above zero for 15 of the 20 events. Autumn leaf colouring and leaf fall (5 events) had a residual of +1.55 days/decade (CI 0.97 to 2.08), which was positive in all 5 events. A step of about 5 days near 1987 fits this residual better than a linear trend. Across all events, the spatial and interannual slopes had a correlation of 0.79. For spring plants they agreed on average (difference -0.01 days per degree C). The clonal red spider lily had slopes of opposite sign, so genetic uniformity does not make the two slopes equal. For animals, a larger increase in non-detection years at a station came with a larger residual delay, but this effect was small. The results stayed the same with homogenised temperature, a fixed window, multi-window models and the two halves of the period. One exception: the autumn residual was not significant within either half. Limits: monthly temperature windows, one reference tree or observer per station, and linear trends. Thus temperature-based trend attribution is reliable for spring plant events in these records. Animal first-detection trends contain a non-temperature delay of nearly 2 days/decade, so they need a correction for detection before use as climate responses.
phenology · climate change · space-for-time substitution · temperature sensitivity · Japan Meteorological Agency · first detection · autumn leaf colouring · cherry flowering · GHCN-M · trend attribution · detection bias · Somei-yoshino
References
10- Earlier budburst did not cause earlier autumn yellowing or leaf fall in the same ginkgo and mulberry trees at Japanese weather stations in 1953–2020 when autumn temperature is controlled sci:sha256:11f49f6ad83249631ee87728c5812f778dc56e0d8946e8e047891f31291f2783
Uses the same JMA phenology records to test spring-to-autumn carry-over in ginkgo and mulberry; complementary question.
- supportsdoi:10.1038/s41559-021-01459-3
Reports that JMA ended animal observations because animals became difficult to observe near urban stations; agrees with our non-temperature delay in animal first-detection dates.
- supportsdoi:10.1016/j.biocon.2009.06.003
Earlier report of delayed animal and advanced plant phenology in JMA records; we quantify the delay after removing the temperature part.
- doi:10.1016/j.biocon.2012.06.024
Review of delayed animal phenology in Japan.
- doi:10.1111/j.1365-2486.2008.01619.x
Shows that first-detection dates depend on population size; candidate mechanism for the animal residual delay.
- doi:10.1007/s00484-026-03308-z
Space-time equivalence test of flowering phenology in European PEP725 data.
- doi:10.1016/j.agrformet.2023.109571
Somei-yoshino bloom responses along the latitude gradient of Japan.
- doi:10.1111/j.1466-8238.2008.00398.x
Earlier analysis of JMA leaf colouring and leaf fall responses.
- usesdoi:10.1175/JCLI-D-18-0094.1
GHCN-M v4 station temperature.
- doi:10.1270/jsbbs.57.1
Evidence that Somei-yoshino is a single clone.
Cited by
13 publicationsPublic catalog entries on Sciscape.
- 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
It found that JMA autumn leaf dates became later than temperature predicts; our residual time series shows the same level shift, and we add the change in slope.
- In 1953–2020, JMA stations permanently lost 16% of animal and 6% of plant phenology series (frog 64%), unrelated to nearby population; animal dates also became later than temperature predicts
[Legacy relation: reproduces] We reproduce its non-temperature delay in JMA animal first dates with daily station temperature windows and a cherry-flowering thermometer, and we add the permanent-loss analysis and the split by detection trend.
- The latitude correlation of cherry first flowering dates decreased at 48 Japanese stations in 1980–2025
This related study uses the JMA phenology archive for spatial and annual temperature-response analysis. Our analysis tests geographic date order and spread.
- 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
Earlier analysis of the same JMA records (temperature attribution of trends).
- February warmth was linked to longer first-to-full cherry bloom intervals in 69 Japanese station records, 1953–2025
This prior SCI study tested temperature links for separate JMA event dates. Our study tests the paired first-to-full bloom interval.
- The fitted swallow–cherry gap trend is +0.83 days per decade in 54 Japanese records from 1953–2020, but a 10-year block interval includes zero
This prior analysis reports individual JMA event trends from the same underlying data. It establishes that the direction of our full-period contrast is already known.
- Warmer winters predict longer cherry bloom intervals at Japanese stations with mild winters in 1953–2025
Prior analysis of separate JMA event dates and their temperature-related trends. The present study tests the paired first-to-full interval.
- 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
Temperature attribution of JMA phenology trends in the same records.
- 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
Same JMA records; found that one linear temperature slope explains spring plant trends. We show that the slope itself changed between 1953-1986 and 1987-2020.
- A fixed thermal-time model predicts lower spring temperature sensitivity at warm US lilac and honeysuckle sites in 1956–2009, but not the observed increase in sensitivity over time
Related attribution of spring phenology trends to station temperature.
- supportsA heat-sum model fitted to year-to-year variation predicts 84% of the spatial flowering gradient of clonal Somei-yoshino cherry at 81 Japanese stations in 1953–2025; adding chilling gives 94–98%
We reproduce its finding that linear within- and between-station slopes agree for spring plants, including the clone, and that the red spider lily has slopes of opposite sign; our thermal-time test shows a mismatch that the linear test does not detect.
- supportsJMA animal first-detection dates in 1953–2020 were later next to station-wide gaps in animal records, unlike plant dates, but these gap terms did not explain the long-term animal delay
We reproduce its non-temperature delay of JMA animal first-detection dates (+1.51 days/decade here) and its item-level link to non-detection with a different model; we add year-scale detection tests.
- Long-distance migrant birds had weaker first-arrival responses to local spring temperature than short-distance migrants at 114 sites west of 60° E, 1961–2018, but not at 12 sites east of 60° E
SCI study that found animal first-detection dates in Japan became later than temperature predicts; related to our residual-trend result.