Date-defined temperature windows gave negative shuffle medians in four US shrub samples from 1961 to 2008

We tested a bias that can occur when temperature is averaged between observed leaf and bloom dates. We used 9,831 site-species-years in four US region and period samples from 1961 to 2008. We exchanged paired dates across years within site-species and kept weather years fixed. Temperature between the exchanged dates gave median interval slopes of -1.52, -1.44, -0.90 and -0.23 days per degree C. Three central 95 percent draw ranges were below zero. The later western range included zero. Fixed March-April controls gave medians near zero. Fixed seasonal climate curves also gave negative slopes in all four samples. A linear seasonal-ramp identity explains how unequal leaf and bloom date variance can create this result. The test shows a risk from date-defined exposure windows. It does not measure a causal temperature effect or correct observed slopes. Selected garden records, monthly weather, unknown visits and exchange assumptions limit wider use. Code, paired records, weather extracts, all draws, source hashes and numerical checks are included.

phenology · exposure window · observation bias · seasonal temperature · date exchange · lilac · honeysuckle