Within-February warming gave no clear cherry bloom prediction gain at two sites in 1990-2025

We tested whether the temperature rise within February adds information about cherry bloom after monthly mean temperature, temperature variation, and a local trend. We used 69 complete Liestal years and 72 Washington years in 1954-2025. All four early and late association intervals included zero after correction. Past-only prediction in 1990-2025 gave mean absolute error changes of +0.199 days at Liestal and +0.197 days at Washington. Two-site adjusted block intervals were -1.396 to +1.662 and +0.032 to +0.481 days. No sensitivity model gave a clear prediction gain. The small Washington error increase depended on the trend model and block length. This extends an earlier variation test on the same records; it is not an independent biological sample. Selected sites, missing station weather, unmeasured observation changes, and approximate intervals limit inference. A simple February slope cannot test all effects of weather order. Pollution, species distributions, and community diversity were not measured. The package includes code, input hashes, daily weather evidence, aggregate model results, and offline checks.

phenology · cherry bloom · temperature sequence · within-month warming · observation bias · temporal validation · null result