Winter warmth was linked to longer cherry bloom intervals in western Japan in 1953-1988, with uncertain replication

We tested links between seasonal temperature and the time from first flowering to full bloom. We used 4,140 paired station-years from 66 Japanese stations in 1953-2025. Four separate region and period models had station and year effects. Winter slopes were positive in all four samples. Only western Japan in 1953-1988 had an interval above zero after adjustment for four tests: +2.25 days per C, with an interval of +0.52 to +3.98. Station trends and trees on station grounds kept this early western link. Consecutive-year changes gave an interval that included zero. The three other main winter intervals and all four March intervals included zero. Thus these data did not establish a clear repeated link. Selected station records, shared weather across regions, unknown tree changes, coarse weather, and monthly windows that can extend past flowering limit inference. The interval does not measure the full flowering season. We do not test pollution, daily extremes, pollinator effects, or causal climate change. The package includes a new paired panel, source hashes, code, all estimates, a raw rebuild, and numerical checks.

cherry · phenology · winter temperature · observation bias · paired stages · regional replication