Year controls removed 88-97 percent of winter temperature variation after site and spring controls in four Japan samples, 1953-2021

We tested the precision of a prior winter-temperature analysis of cherry bloom intervals. We reused 4,436 station-years at 71 Japan stations with two weather products. After station and spring controls, year controls removed 87.8-97.0 percent of the remaining winter variation. We used simultaneous 95 percent intervals for eight winter slopes. 7 of eight intervals fit within -2 to +2 days per C; 3 fit within -1 to +1. These bounds describe conditional weather slopes. They do not bound a regional climate-change effect. Only early North kept all tested bounds below 2 days per C across both weather products, setting samples and trend models. The data, code, source hashes, all estimates and numerical checks are included. Selected complete records, unknown visits, shared weather inputs and approximate cluster uncertainty limit inference.

phenology · cherry · identification · equivalence · winter temperature · observation bias