Data for: Bayesian Belief Network Peat Health Expert Opinion

The elicitation of expert opinion has associated challenges due to the quantity of information required and the ability of experts to quantify subjective beliefs (Kuhnert, Martin and Griffiths, 2010). A method and programme has been developed entitled the Application for Conditional Probability Elicitation (ACE), to extract probability distributions from experts using simple questions to capture the overall shape of probability distributions (Hassall et al., 2019). The ACE approach was implemented in this project to extract probability distributions from experts. The elicitation of expert knowledge was conducted following Kuhnert, Martin and Griffiths, (2010). Seven experts were carefully chosen for CPT probability elicitation, with three industrial experts and four academic experts. Experts were interviewed online using a modification to the ACE programme. Following the Delphi method (Ling and Bruckmayer, 2021), experts were asked to define their beliefs on the state of soil functions given the observed parent values (the conditional probability distributions). These conditional probability distributions were then aggregated to create the CPTs and shared with the experts to elicit further feedback. Aggregation and incorporation of the Delphi approach allowed us to incorporate the beliefs of experts that spanned multiple fields while reducing the undue influence of a single expert. Bias was minimised by developing a document that clearly explained the states, values, and descriptions of nodes and arcs defined within the network. This document was circulated to all experts before knowledge elicitation. During the elicitation sessions, each expert was provided with a verbal summary of the document, explained the purpose of the exercise, and asked the same questions as each other.