Plant trends reduced excess consecutive-year gap similarity in four US lilac samples in 1961-2008

We tested consecutive-year similarity in the gap from first leaf to first bloom. Four US lilac samples in 1961-2008 contained 5,521 plant-years and 4,394 consecutive-year pairs. After plant, year, and spring temperature effects, normalized covariance exceeded the mean shuffled reference by 0.173, 0.156, 0.110, and 0.017. Three four-test p values were below 0.05. Plant trends reduced all four excess estimates. With trends, only the two early-period tests gave p values below 0.05. The method compares observed year order with 1,999 within-plant residual shuffles and refits the controls. It accounts for negative residual dependence caused by estimated plant means. Fitted-residual exchangeability is an assumption, not a fact established by these data. Persistent reporting errors, changing observation effort, missing years, coarse weather, and plant selection limit interpretation. The result does not identify biological memory or a causal climate effect. The package includes the panel, code, all reference draws, source hashes, figures, and independent numerical checks.

phenology · lilac · temporal persistence · observation bias · residual permutation · temperature