Multi-criteria decision analysis in Bayesian networks - Diagnosing ecosystem service trade-offs in a hydropower regulated river

The paper demonstrates the use of Bayesian networks in multicriteria decision analysis (MCDA) of environmental design alternatives for environmental flows (eflows) and physical habitat remediation measures in the Mandalselva River in Norway. We demonstrate how MCDA using multi-attribute value functions can be implemented in a Bayesian network with decision and utility nodes. An object-oriented Bayesian network is used to integrate impacts computed in quantitative sub-models of hydropower revenues and Atlantic salmon smolt production and qualitative judgement models of mesohabitat fishability and riverscape aesthetics. We show how conditional probability tables are useful for modelling uncertainty in value scaling functions, and variance in criteria weights due to different stakeholder preferences. While the paper demonstrates the technical feasibility of MCDA in a BN, we also discuss the challenges of providing decision-support to a real-world habitat remediation process. [Display omitted] •Multi-criteria decision analysis (MCDA) of environmental flows and physical habitat restoration measures in a hydropower regulated river.•Conditional probability tables used to model uncertainty in value scaling functions and criteria weighting.•Object-oriented Bayesian network used to integrate expert judgement and quantitative model simulations in MCDA.