Missing indices permit both signs of a rain timing contrast for butterfly crashes in four England and Wales samples, 1984-2023

We tested whether dry summers precede more severe Ringlet declines than Meadow Brown declines at the same sites. Four England and Wales samples in 1984-2023 used 14,307 baseline-eligible site pairs. Observed pairs had known indices for both species. A crash meant a decline of at least 75 percent. We compared prior and future summer rain-deficit slopes with current weather, temperature and time controls. Observed timing contrasts ranged from -2.33 to +6.41 percentage points per 100 mm deficit. All four corrected intervals included zero. When unknown current crash states could be either zero or one, sharp fitted-contrast bounds allowed both signs in every sample. These fixed-record possibility bounds are not confidence intervals. They expose selection uncertainty for eligible sites but do not cover missing prior indices or index error. Small Welsh samples, country weather, changing detection and short time series limit inference. The result does not establish a causal drought or pollution effect. The package gives source hashes, code, evidence, bounds, checks and figures.

butterflies · drought · severe declines · missing outcomes · observation bias · partial identification · regional replication