Date choices from May to October for zero catches can reverse the within-species temperature slope at six UK beetle sites, 1994–2013

We tested the link between April–June temperature and mean May–October beetle catch dates at six UK Environmental Change Network sites in 1994–2013. The fixed sample had 12 transects and 191 complete transect-years. We corrected catches for effort, removed a local linear time trend, and split each annual date difference into dates within species and species catch shares. The mean total slope was −1.08 days per °C. Its nominal joint 95% site and year-block interval was −10.62 to +7.54. The mean direction was uncertain. The split needs a date rule for zero-catch species-years. With observed catches and reference communities fixed, we allowed each such date to vary between the first and last monthly midpoints, days 136.5 and 289.5. Exact coefficient bounds gave a within-species slope from −11.35 to +6.49 days per °C. Feasible date assignments attained both limits. The total slope stayed fixed. Common species alone also permitted both signs, from −6.35 to +1.04. These are bounds on the attribution rule, not confidence intervals or estimates of dates for absent species. The catches alone cannot set the sign of this split under the stated bounds. The package has unchanged source data, independent weather inputs, code, all results, figures and offline repeat instructions. The result concerns a selected site set and a truncated season. Pitfall catches combine activity and density. It does not test onset or identify a climate cause.

ground beetles · phenology · temperature · zero catches · rate decomposition · Environmental Change Network · reproducibility · partial identification