Summer weather adjustment reduced fitted butterfly synchrony but gave mixed held-out results in Britain in 1981-2024

We tested whether summer weather adjustment reduces shared annual abundance-index changes in nine butterfly species in England, Scotland and Wales. Six country and period samples covered 1981-2002 and 2003-2024. Weather adjustment reduced standardized pair covariance in all six same-year fits but in only four of six held-out block tests. Held-out prediction error increased in all six split samples, by 2-37 percent. April-September weather increased held-out shared variation in all six samples. Independent numerical checks reproduced the primary scores. These results do not support a common weather correction for the tested short samples. They do not show that weather has no biological effect. Shared weather, short training sets, site turnover, unknown visits, country exposure averages and missing index uncertainty limit inference. This is a retrospective validation test, not climate attribution or a pollution analysis. The package contains frozen inputs, hashes, code, all scores and held-out errors.

butterflies · weather · synchrony · cross-validation · observation bias · UKBMS