Modélisation et prévision de l'exposition de l'écosystème aux pesticides de bain rejetés par les exploitations piscicoles marines : une première perspective

Abstract

This document focuses on models in relation to the discharge of chemical active ingredients associated with bath pesticide treatments used in net-pen finfish aquaculture operations. The document includes a brief overview of the context and associated conceptual processes to be modelled, specific modelling challenges, a review of modelling efforts to date, and a description of some simple models. In general, modelling for bath pesticides for discharge and dispersal is in an early stage of development. There have been few modelling efforts and few extensively calibrated or validated models. Furthermore, few models have been incorporated into pesticide management considerations. Models range in complexity from models that make many simplifying assumptions, such as a constant current, to models that use more realistic representations, for example temporally and spatially varying currents. In general, outputs from models which use non-spatially varying currents may not be representative when displacement distances are greater than 500 m. This is problematic since displacement distances of bath pesticide discharges are often greater than 500 m. Hydrodynamic models are one potential solution; however, their implementation and operation is resource intensive and predictions are imperfect representations of real-world situations. Nevertheless, their use should be considered when predicted displacement distances exceed 500 m. All models have uncertainties and sensitivities associated with them. In general, models for predicting exposure to bath pesticides have not been extensively explored and quantified, including sensitivities to hydrographic detail and resolution, initial condition specification, and active ingredient behaviour. Predictive models, whether they are forecast or hindcast models, are subject to the validity of the assumptions made about future and past conditions; since these are often not well known, any model output needs to be interpreted cautiously and with appropriate respect for the uncertainties. Calibration and validation of models are challenging due to the difficulties in obtaining spatially and temporally extensive observations associated with discharges from multiple farm sites and multiple treatment scenarios. Despite the uncertainties, models can be useful for regulatory decision support. Model selection depends on the decision maker’s needs; and the interpretation of model results requires clearly defined decision rules. Specific recommendations on model selection will require further clarification of management needs, more detailed evaluations of model uncertainties and sensitivities, as well as verification and validation of chosen models.