The 1995 drought collapse of six British butterfly species did not repeat at similar size after the 2003, 2006, 2018 or 2022 droughts on UK transects, 1989-2024

Oliver et al. (2015) used the 1995 drought to select six drought-sensitive British butterflies. Their extinction projections assume that each future drought has the 1995 effect. We tested this assumption with four later droughts of equal or higher April-September aridity in England: 2003, 2006, 2018 and 2022. We used UK Butterfly Monitoring Scheme site indices (standard transects, 236-734 sites per event). For each site and species, the response was the log index in the year after the drought minus a six-year pre-drought trajectory. We compared the six species with 33-34 other species and calibrated each statistic against 28 non-drought placebo years. In 1996 the six species were 0.83 log units below the other species (95% CI -0.94 to -0.65), lower than in all placebo years. After 2003 and 2018 the contrast was -0.32 and -0.25 (14th placebo percentile). After 2006 and 2022 it was +0.16 and +0.02. The 2022 drought was the most arid year before 2025 and had a rainfall deficit similar to 1995. The mean contrast of the two post-publication droughts was at the 15th placebo percentile. The species order of the 1995 response agreed with 2003 (Spearman rho 0.54) and 2018 (0.38), but not with 2006 or 2022 (about 0). Within each drought, local E-OBS aridity did not predict larger local declines. Ringlet was below trajectory after all five droughts, but by less after the later events. Limitations: only five events, a fixed-reference aridity index that recent warming inflates, no habitat-fragmentation analysis, and a growing non-random transect network.

drought · butterflies · UKBMS · extreme events · out-of-sample test · population collapse · climate change · placebo years · Lepidoptera · Great Britain · E-OBS · replication