Informing the management of multiple stressors on estuarine ecosystems using an expert-based Bayesian Network model.

Abstract The approach of applying stressor load limits or thresholds to aid estuarine management is being explored in many global case studies. However, there is growing concern regarding the influence of multiple stressors and their cumulative effects on the functioning of estuarine ecosystems due to the considerable uncertainty around stressor interactions. Recognising that empirical data limitations hinder parameterisation of detailed models of estuarine ecosystem responses to multiple stressors (suspended sediment, sediment mud and metal content, and nitrogen inputs), an expert driven Bayesian network (BN) was developed and validated. Overall, trends in estuarine condition predicted by the BN model were well supported by field observations, including results that were markedly higher than random (71–84% concordance), providing confidence in the overall model dynamics. The general BN framework was then applied to a case study estuary to demonstrate the model's utility for informing management decisions. Results indicated that reductions in suspended sediment loading were likely to result in improvements in estuarine condition, which was further improved by reductions in sediment mud and metal content, with an increased likelihood of high abundance of ecological communities relative to baseline conditions. Notably, reductions in suspended sediment were also associated with an increased probability of high nuisance macroalgae and phytoplankton if nutrient loading was not also reduced (associated with increased water column light penetration). Our results highlight that if stressor limit setting is to be implemented, limits must incorporate ecosystem responses to cumulative stressors, consider the present and desired future condition of the estuary of interest, and account for the likelihood of unexpected ecological outcomes regardless of whether the experts (or empirical data) suggest a threshold has or has not been triggered.