March windows contained more days after recorded cherry bloom at Liestal and Washington in 1988-2025 than in 1951-1987

We tested how much of a fixed March weather window falls after recorded peak cherry bloom. Three public site records supplied 225 complete site-years from 1951 to 2025. The later-minus-earlier mean count of March days after bloom was 2.03 at Liestal and 2.67 at Washington. Approximate block intervals adjusted for three tests were 0.26 to 3.88 and 0.74 to 4.85 days. Kyoto had a change of 0.37 days with an adjusted interval from zero to 1.05. All observed increases remain positive under arbitrary date errors bounded by three days. Seven-day bounds permit either sign. These bounds do not include sampling uncertainty. Fixed sites and complete years reduce changes in sample coverage. Missing visit logs, different bloom rules and selected managed trees limit inference. No actual weather values enter this calendar audit. The result does not establish bias in a temperature slope or a causal effect of climate change. Code, derived evidence, source hashes and retrieval steps are supplied.

phenology · cherry bloom · weather window · observation bias · date error · regional replication · exposure measurement