Two-year removal reversed lilac interval slopes in 2 of four US samples from 1965 to 2008

We tested whether any two observed years could change the sign of a spring temperature slope. We used paired lilac leaf and bloom dates from a verified published table. Four region and period samples contained 5,407 site-species-years. Models used unit and year effects and February temperature. Every pair of observed years was removed and the model was fit again. Interval slopes changed sign in 2 of four samples. Leaf and bloom slopes had 0 sign reversals across the pair-removal fits. These ranges are deterministic checks, not confidence intervals. Unknown visits, selected pairs, shared sites, different plant types and coarse monthly weather limit inference. We do not test air pollution, daily extremes or causal climate change.

phenology · lilac · temperature · observation bias · year deletion