Visit intervals kept one of four spring temperature slopes negative for selected US lilac open-flower dates in 2009-2014

We tested whether visit-date uncertainty preserves the sign of a spring temperature slope for lilac open-flower dates. We used 696 plant-years at 139 US sites in four region and period samples from 2009 to 2014. Models controlled plant, year, and January-February temperature. Actual absence-to-presence brackets permitted both coefficient signs in three samples. The later eastern bound stayed negative, from -6.196 to -1.019 days per C, and remained negative after each site omission. Its common-lilac-only bound included both signs. Gaps above seven days contributed 73-95% of bound width. A seven-day filter changed western samples enough to give positive bounds. Thus gap filters change the selected association as well as date precision. These are exact fixed-record bounds under correct reports and separate date choices, not confidence intervals or causal effects. Selected plants, six years, missing raw status histories, coarse weather, and possible missed episodes limit ecological use. The package gives source hashes, code, date and weather extracts, extreme date assignments, independent checks, and reproduction instructions.

phenology · lilac · observation bias · interval censoring · spring temperature · partial identification · reproducible dataset