Visit brackets permit both signs of spring temperature slopes for lilac flowering intervals in four US samples in 2009-2014

We tested whether actual visit records fix the sign of the spring temperature slope for a lilac flowering interval. We used 530 repeated plant-year records at 103 US sites in 2009-2014. Four separate plant and year models controlled winter temperature. Every allowed event date lay between a prior absence report and a presence report. We also required bloom to start before it ended. All four sharp coefficient bounds permit negative and positive slopes. Bounds range from -24.21 to +22.20 days per C in the small early western sample and from -4.17 to +5.63 in the later eastern sample. The largest ten percent of record contributions account for 42-68 percent of bound width. A seven-day visit filter gives opposite slope signs in the two eastern periods and removes usable weather variation from the early western model. These are fixed-record possibility bounds, not confidence intervals or causal effects. Correct reports and one ordered flowering episode are necessary assumptions. Selected garden plants, missing records, coarse weather and short periods limit wider use. The package gives raw source hashes, the visit and weather panel, attainable date assignments, code, figures and independent numerical checks.

phenology · lilac · flowering interval · temperature · observation bias · interval censoring · partial identification · regional replication