The Test Used to Choose Between Pesticide Survival Models Tracks Lesion Persistence, Not Hazard Structure

Abstract

Aquatic pesticide risk assessment predicts survival with toxicokinetic-toxicodynamic models that offer two hazard structures, cumulative or peak-based, which disagree about which exposure patterns are dangerous; for imidacloprid, published choices differ across eight species. The choice is made from one observation: deaths after transfer to clean water are taken to refute the peak-based structure. That inference holds only if damage is falling, which is not measured in these tests but taken from a fitted model. The observation responds to the compound and design, not the hazard structure. Across thirty-two Gammarus pulex treatments the post-exposure death rate places anticholinesterases above uncouplers, in the order of lesion persistence. For diazinon, one series returns opposite verdicts over its sixteen-day window and its first four days. For imidacloprid in Daphnia magna, one damage stage favours the cumulative structure by 78 likelihood units and two stages the peak-based one by 7. Four conditions make a design informative; thirteen of thirty-six treatments in a thirteen-compound archive meet them. Spreading a fixed exposure over longer intervals raises mortality in one of five species, which no repair law reproduces with the kinetics of the models. For neonicotinoids, predictions for such patterns are unvalidated, not conservative.