A shared annual SD error of 2.22 days permits nonpositive Meadow Brown warmth slopes in two English regions in 2000-2024

We tested how annual errors in recorded flight-date spread can change spring temperature slopes for Meadow Brown in England. We used 18,868 site-years in two regions and two periods from 1976 to 2024. The model has site effects and linear time. A linear program found the least maximum error in standard deviation (SD) that permits a nonpositive slope. Each year has one common error across sites, and corrected SD must stay nonnegative. Separate error limits were 0.059 and 0.709 days in the early northern and southern samples, and 1.219 and 2.204 days in the later samples. One shared later error sequence with maximum size 2.213 days makes both regional slopes nonpositive. If its change between years is at most one day, no such sequence exists, even with no size limit. These are fixed-data sensitivity bounds, not confidence intervals. The errors are in annual SD, not in sighting dates. We do not show that these errors occurred or that visits can produce them. Shared survey practice and shared weather limit the independence of regional results. The package gives code, source hashes, derived data, error sequences and numerical checks.

phenology · Meadow Brown · observation bias · temperature · sensitivity bounds · shared annual error · linear programming