Cherry bloom intervals shortened at 16 northern Japan stations in 1953-1988, with no clear trend in three other samples

We tested trends in the time from first cherry flowers to full bloom at a fixed set of Japanese stations. The panel had 2,774 paired station-years at 38 stations in 1953-2025. Four tests split the panel at 36 N and at 1989. The early northern interval shortened by 0.292 days per decade, with an adjusted block bootstrap interval of -0.541 to -0.044. This result persisted in setting and coverage checks. The later northern test and both southern tests gave intervals that included zero. First flowers and full bloom both moved later in the early northern sample, so the interval result does not show that warming accelerated flower opening. The complete-series rule can exclude failed full bloom. Missing visit records, tree changes, and the exploratory design limit the claim. No climate cause or pollution effect was identified.

phenology · cherry · observation bias · Japan · bloom interval · fixed station panel