Search

Abstract: Wetlands are essential components of floodplain–river ecosystems that often suffer degradation due to river regulation. To this end, the application of environmental water is increasingly being seen as an important amelioration strategy. However, decisions regarding the delivery of water to maximise environmental benefits, including native fish population health, are complex and difficult. This paper describes the development of a Bayesian Belief Network (BBN) model as part of a Decision Support Tool for assessing inundation strategies to benefit native wetland fish. Separate, albeit closely related, BBNs were developed for three native (golden perch Macquaria ambigua, carp gudgeon Hypseleotris spp., Australian smelt Retropinna semoni) and one alien fish species (common carp Cyprinus carpio carpio). The model structure was based on a conceptualisation of the relationships between wetland habitats, hydrology and fish responses, with emphasis on the types of inundation activities undertaken by managers. Conditional probability tables for fish responses were constructed from expert opinion and the model was validated against field data. The predictive ability and sensitivity of the model reflected the inherent high variability in relationships between wetland characteristics, hydrology and fish responses, but was nonetheless able to address satisfactorily such complexities within a holistic framework. As the model was designed in conjunction with managers and evaluated by them, its application will be enhanced by on-going engagement between managers and scientists. [Copyright &y& Elsevier]

Bayesian belief networks (BBNs) are a widespread tool for modelling the effects of management decisions and activities on a variety of environmental and ecological responses. Parameterisation of BBNs is often achieved by elicitation involving multiple experts, and this may result in different conditional probability distribution tables for the nodes in a BBN. Another common use of BBNs is in the comparison of alternative management scenarios. This paper describes and implements the ‘belief index’ (BI), an empirical measure for evaluating outcomes in BBN modelling that summarises the probabilities (or beliefs) of any one node in a BBN. A set of four species-specific BBNs for managing watering events for wetland fish is outlined and used to statistically assess between-expert and between-species variability in parameter estimates by means of the BI. Different scenarios for management decisions are also compared using the % improvement measure, a derivative of the BI. [Copyright &y& Elsevier]