Evaluation of probabilistic risk assessment of pesticides in the UK: chlorpyrifos use on top fruit.

The use of probabilistic risk assessment (PRA) for examining chemical impacts has become an important area of debate within the European Union. This paper describes a case study on probabilistic techniques to assess pesticide risks in the UK aquatic environment. The main aim of this paper is to demonstrate both the potential strengths and weaknesses of PRA for assessing pesticides when compared with the conventional deterministic approach, and to examine whether PRA is useful within the European regulatory context. The organophosphate insecticide, chlorpyrifos, was selected as a model compound and toxicity exposure ratios calculated using Monte Carlo analysis and different distributions of spray drift and toxicity values following application to top fruit. Chlorpyrifos is highly toxic to arthropods but less toxic to fishes. Species sensitivity followed a log-normal distribution when fitted to all toxicity data. Toxicity data quantity had little influence on species sensitivity distribution model parameters when n was greater than 10 species. Below this, estimates were less accurate and precise, possibly because of the inclusion of data from many different sources. Estimates of chlorpyrifos exposure derived from the standard spray drift model differed substantially from measurements of chlorpyrifos in European surface waters. When a distribution based on measured concentrations was used in a PRA, the risk of acute fish mortality was low, and the risk of acute arthropod mortality was lower than in other scenarios, although not negligible. If PRA is used to assess pesticides, risk managers need further guidance on how to conduct a PRA and what constitutes 'unacceptable risk' under EC Directive 91/414/EEC, as judgement is required when simple trigger values are no longer used.