Assumed three-day bloom date errors permit both signs of freezing exposure change at Liestal and Washington in 1951-2025

We tested how assumed bloom date errors change a period contrast in freezing exposure. We used annual cherry bloom records and daily station minima at Liestal and Washington in 1951-2025. The primary measure was any station minimum below 0 C in 14 days from bloom. If each annual date can differ by up to three days, exact change bounds are -16.5 to +35.0 percentage points at Liestal and -3.1 to +28.7 at Washington. Both signs are possible. At Washington, a common offset of up to seven days gives only positive fixed-record changes. Sampling intervals still cross zero. Date errors are assumed, not measured. Different window and temperature definitions do not all give both signs. These are exposure bounds, not frost damage or causal climate effects. The package provides code, aggregate results, source hashes, proof, checks, and a figure. Restricted raw data and annual derived tables remain external.

phenology · cherry bloom · frost exposure · date uncertainty · observation bias · partial identification · regional comparison