Bloom-date shifts gave positive freezing-exposure estimates in four Liestal and Washington comparisons in 1954-2025

We separated freezing-exposure changes into weather, bloom timing and annual pairing terms. Public cherry dates and station minima cover three periods from 1954 to 2025 at Liestal and Washington. The exact method averages both orders of change and keeps a separate pairing term. Timing estimates were positive in all four successive-period comparisons, from 6.51 to 13.33 percentage points. Three adjusted intervals included zero. The fourth result became uncertain with longer bootstrap blocks and other exposure windows. Thus the tests did not establish a common timing contribution across both regions and periods. Fixed sites and complete weather seasons reduce changes in observation coverage. Unknown visit errors, station differences, few events and period choice limit inference. This is a descriptive exposure test, not a causal climate or damage estimate. The package includes original code, aggregate results, a proof, source hashes, checks and a figure. Restricted raw and record-level data remain external.

phenology · cherry bloom · freezing exposure · decomposition · observation bias · regional comparison