Abstract :
A considerable decline in fish catch in Switzerland triggered a 5-year project (Fischnetz) to investigate potential causes. The results of many field and laboratory studies, together with literature and historical data, indicated a need for a concise synthesis of the results. Therefore, Bayesian probability network (BPN) and weight-of-evidence (WOE) approaches were applied. Here, these both approaches are compared and evaluated. In addition, the potential reasons for the (mis)match in the two approaches were evaluated. In both studies, proliferative kidney disease (PKD), caused by a parasite, and the clinical outbreak supported by various factors, was a very probable single parameter. WOE assessed habitat and streambed quality as likely for contributing to impaired health, recruitment and abundance at single sites only, but this parameter was assessed to be the most important and ubiquitous stressor in the BPN. Mismatches suggested that these factors were either not considered or not equally assessed by the different models, which is due to different endpoints, incomplete data sets and different handling of these various data sets by the applied synthesis approaches.
