LIMITATIONS TO USE OF TOPICAL TOXICITY DATA FOR PREDICTIONS OF PESTICIDE SIDE-EFFECTS IN THE FIELD

We consider ways in which laboratory-derived toxicity data might be us ed to predict the safety of insecticides to beneficial invertebrates. ii model test system consisting of a predator, the convergent lady bee tle, Hippodamia convergens Guerin Meneville (larvae and adults); a par asitoid, Aphidius ervi Haliday; and the bee species Apis mellifera L., Megachile rotundata (F.), and Nomia melanderi (Cockerell) was tested with diazinon, imidacloprid, and RH-7988 [ethyl (3-tert-butyl-1-dimeth yl carbamoyl-1H-1,24-triazol-5-ylthio) acetate]. We also tested the pe a aphid, Acyrthasiphon pisum (Harris), to calculate selectivity ratios for these beneficial species, which coexist with the aphid pest in Wa shington State pea and alfalfa ecosystems. Topical toxicity was estima ted for all species and ranged 0.0002-0.45 micrograms per insect for d iazinon, 0.000031-0.04 micrograms per insect for imidacloprid, and 0.0 015-6.11 micrograms per insect for RH-7988. Selectivity ratios based o n these values spanned 0.02-47.4, 12.9-1,290.3, and 13.3-4,073 for dia zinon, imidacloprid, and RH-7988, respectively. Risk assessment indice s based on probit substitution (estimate of mortality of beneficial sp ecies at LD(90) for the pest) and 2 standard methods for bees, a seque ntial testing scheme and a hazard index gave variable predictions of t he compatibility of these compounds with integrated pest management. W e conclude that predictive methods must advance to consider relative e xposure rates to pesticides, aspects of chemical fate, and behavior of the organisms concerned if they are to be useful. Above all, predicti ons must be validated with field data.