Seven-day delay bounds kept negative spring slopes in two western US lilac samples in 1965-2008

We tested calendar alignment and reporting-delay sensitivity in 9,015 US lilac first-bloom records from 1965-2008. Four models used separate East and West samples in two periods, with plant and year effects and winter temperature controls. Exact linear-regression bounds allowed each reported date to be zero to seven days after the event, with delay allowed to depend on temperature. Western spring-slope ranges stayed negative: -6.428 to -0.290 and -6.804 to -0.679 days per degree C. Eastern ranges included zero. All four ranges included zero at a fourteen-day cap. These are finite-sample sensitivity ranges, not confidence intervals or evidence that actual delays are at most seven days. The calendar audit found 25.94% weekend dates in the early East, against a two-in-seven reference. Its adjusted rotation p-value was 0.011, but rose to 0.093 with equal site weights. The other three samples had no clear weekend departure. Thus, the audit did not establish a repeated weekend pattern. Unknown visits, selected long records, different regional plant types and bloom rules, coarse weather, and few clusters limit inference. We do not measure pollution, biodiversity loss, or a causal climate effect. The package includes exact input hashes, a byte-identical raw rebuild, independent regression checks, bound proofs and endpoint checks, a repeated analysis, figures, and offline reproduction instructions.

phenology · observation bias · lilac · reporting delay · sensitivity bounds · weekend reporting · spring temperature · temporal replication