An example to illustrate the potential use of probabilistic modelling to estimate operator exposure to pesticides.

To illustrate the potential for probabilistic simulation modelling of operator exposure a probabilistic model was constructed using the draft EUROPOEM II database. The model also incorporated actual pesticide use data. Although the EUROPOEM II database is not complete, and the resulting simulation has to be regarded as only provisional, the data were sufficient to illustrate some of the principles involved. The model was constructed using non-parametric empirical input distributions. Associations were observed in the input exposure data and were reflected in the model. The exposure and usage data were all variable and showed highly negatively skewed distributions and the technique avoided concerns about having to define single representative values for point estimates. Increasing the number of iterations improved the stability of the output, but as expected, resulted in higher exposures being predicted due to a combination of rare events. Analysis of the inputs for individual iterations giving high results indicated that, in this example, these were due to unrepresentatively high input values for inhalation exposure. Simulations that excluded these values were more stable and showed lower exposures, illustrating the need to ensure the validity of the input distributions.