Quantifying impacts of salmon lice on wild salmon: roles of migration timing and tolerance of juvenile salmon to salmon lice

Assessments of salmon louse impacts on wild salmon populations typically involve quantitative modelling of impacts and sensitivity analysis of potential effects of uncertain model assumptions. A Bayesian network approach allows such quantitative results to be combined with more subjective expert assessments of the belief in these assumptions. Using this approach, we quantified probability distributions for effects of salmon lice on mortality of out-migrating juvenile wild salmon in Norway. These distributions captured uncertainty quantified in the stochastic part of the model and uncertainty associated with 2 model assumptions: the timing of migration and the tolerance of juvenile salmon to salmon lice. Results showed that while the calculated impact was sensitive to the assumptions made about timing and tolerance, the probability distribution was only marginally wider when the uncertainty about timing and tolerance was included. The results support assessments of the uncertainty of salmon louse-induced mortality of wild salmon, an environmental impact indicator used in region-level biomass regulation of Norwegian aquaculture. The findings also have broader implications: when multiple model assumptions are uncertain but equally likely to deviate in opposite directions, their combined effect on overall uncertainty may be smaller than thought a priori , emphasising the importance of quantifying joint uncertainty in complex models.