Full-bloom updates lowered mean end-bloom error by 1.10-1.79 days in four western US lilac samples in 1975-2003

We tested whether a full-bloom report improves an estimate of end bloom beyond a first-bloom report. We used 2,228 complete common lilac records from 335 reporting units in two western US regions and two test periods from 1975 to 2003. Each prediction used at least five earlier complete records from the same unit. Full-bloom updates reduced mean absolute error by 1.10-1.79 days. Three of four adjusted bootstrap intervals excluded zero. The later Interior interval included zero. Median observed lead fell from 20-25 days at first bloom to 11-14 days at full bloom. Mean gains remained positive in warm-year records and after date-flag removal, but some uncertainty intervals included zero. Visit histories and report receipt times are absent. These tests predict reported dates in selected long records. They do not establish live forecast skill, true flower life, or a causal climate or pollution effect.

phenology · end bloom · lilac · forecast · observation bias · lead time · seasonal weather