The use of Bayesian networks to facilitate implementation of water demand management strategies

Bayesian networks have received increasing recognition in recent years as apotentially effective tool in supporting water management decisions. Despite anumber of reports of their use, no formal evaluation of the effectiveness of Bayesiannetworks in facilitating water resources management exists. This study improvesunderstanding of the strengths and weaknesses of Bayesian networks through theirapplication in a water-stressed region in Europe where domestic sector waterdemand management is considered as a mitigation measure. The fieldwork resultsprovide a comprehensive technical and end-user evaluation of the use of Bayesiannetworks in water demand management implementation which, to our knowledge, isthe first of its kind to be reported in the academic literature. For the technicalevaluation, expert knowledge was first used to generate the structure of Bayesiannetwork models which were then populated with data collected in the case studyregion. The model development supported the examination of several researchquestions regarding the technical suitability of Bayesian network modelling tofacilitate implementation of water demand management strategies. For the end-userevaluation a survey was used to record the experiences of practitioners who appliedBayesian network models to a number of water demand management problemsduring a one-day workshop. Evaluation indicators included the effectiveness ofBayesian networks in facilitating strategic planning, technical support, transparencyof data, learning among and between stakeholders, organisational receptivity,reliance on decision, and a comparison of experiences of decision conflict, effort anddecision confidence. Results from the end-user evaluation provide evidence thatBayesian networks are particularly effective in terms of technical suitability andtransparency, and policy-makers perceived effectiveness scores were significantlyhigher than individuals from other professions. Conclusions from the technicalevaluation found that Bayesian networks can provide support in achieving cost-effectiveness in terms of sampling and data collection by focusing resources oncollecting relevant data to reduce uncertainty. Conclusions from the end-userevaluation found that, for cross-sectoral planning in the context of managing waterscarcity, their transparent representation of strengths of causes and effects betweenvariables makes Bayesian networks an effective tool for facilitating dialogue andcollaboration across science-policy interfaces.