Two temperature sources gave uncertain winter links to cherry bloom intervals in four Japan samples from 1953 to 2021

We checked whether a published test of cherry bloom timing depends on its temperature source. We kept 4,436 station-years at 71 Japanese stations, the same seasonal windows, and the same model. We replaced local JMA monthly temperatures with NOAA GHCN-CAMS grid temperatures. Four samples split records at 36 N and at 1988. All eight winter slope intervals included zero after an eight-test adjustment. NOAA slopes ranged from +0.058 to +0.550 days per degree C. Within-station winter anomalies correlated at 0.933-0.960 between products. NOAA standard errors were larger in all samples. A stable-setting check also gave uncertain slopes. The earlier uncertain inference thus persisted with a second monthly temperature source. This does not establish no effect or equivalence. The products can share station inputs. Complete-pair selection, missing tree and visit records, coarse weather, and fixed seasonal windows limit inference. We did not test pollution, weather extremes, or wild biodiversity loss. The package gives code, provenance, matched records, results, checks, and an offline reproduction path.

phenology · cherry · winter temperature · weather source · observation bias · null result · reproducibility