A one-day annual change limit for delays capped at 14 days kept negative spring slopes in four US lilac samples in 1965-2008

We tested whether slow changes in reporting delay can remove negative spring-temperature slopes in 9,015 selected US lilac records from 1965-2008. Four region and period models used plant and year effects and winter temperature. With delays capped at 14 days and changing by at most one day per calendar year within a plant, all four upper slope bounds stayed negative, from -2.341 to -0.557 days per degree C. Without the annual change limit, all four ranges included zero. The smallest change limits that permitted zero were 1.35, 1.77, 3.34 and 4.29 days per year. Flag removal and alternative weather terms retained negative upper bounds at the one-day limit, with smaller margins in March-only models. These are exact sample bounds from linear programs, not population confidence intervals or evidence that actual delays obey the limits. Unknown visits, selected records, different regional plant types and coarse weather limit inference. We do not measure pollution, daily extremes or biodiversity loss. The package includes code, data provenance, a raw rebuild, endpoint assignments, numerical checks and a figure.

phenology · lilac · observation bias · reporting delay · linear programming · sensitivity bounds · spring temperature