One-day date delays permit both signs of lilac flowering-duration temperature slopes in four western US samples

We tested whether uncertain event dates can change spring temperature slope signs for common lilac flowering duration. Four plant and year models used 3,565 paired reports from the western US historic network. Each true date could precede its report by up to one day. Sharp coefficient bounds permitted negative and positive slopes in all four samples. The smallest common delay allowances that permitted zero ranged from 0.250 to 0.583 days. These are fixed-record possibility bounds, not confidence intervals or estimates of actual errors. Equal delays at both bloom stages cancel. The sign changes require differential errors related to model weights. Selected records, unknown visits, coarse monthly weather and shared protocols limit inference. We did not test pollution, daily extremes, wild biodiversity loss or causal climate effects. The package includes source hashes, the input panel, endpoint evidence, code and separate solver checks.

phenology · lilac · flowering duration · observation bias · date uncertainty · partial identification