Evidence for a change in the spring interval slope depended on model and record filters in matched US lilac records, 1967-2008

We tested whether a fixed set of reporting units established a change in the link between spring temperature balance and the interval from first leaf to first bloom. We used 2,394 common-lilac records from 135 US units with repeated reports in both 1967-1979 and 1980-2008. The main later-minus-earlier slope change for April-minus-March temperature at fixed spring mean was -0.697 days per C. Its 95% state-and-year interval was -1.544 to +0.149. Allowing unit-by-period baselines gave -1.114 (-1.995 to -0.234). Removing flagged pairs gave -1.003 (-1.750 to -0.255). All point estimates were negative, but evidence for a change depended on model and record filters. This is an exploratory follow-up using an archive that overlaps prior campaign work. Unknown visits, changing completeness, selected managed plants, weather after events, and 11 state clusters limit inference. The result does not identify a causal climate or pollution effect. The package contains code, raw-source hashes, a reusable paired table, estimates, figures and numerical checks.

phenology · lilac · observation bias · temperature · matched reporting units · model sensitivity