Using Bayesian Networks to complement conventional analyses to explore landholder management of native vegetation

Influencing the management of private landholders is a key element of improving the condition of Australia's natural resources. Despite substantial investment by governments, effecting behavioural change on a scale likely to stem biodiversity losses has proven difficult. Understanding landholder decision-making is now acknowledged as fundamental to achieving better policy outcomes. There is a large body of research examining landholder adoption of conservation practices. Social researchers are able to employ a suite of conventional techniques to analyse their survey data and assist in identifying significant and causal relationships between variables. However, these techniques can be limited by the type of data available, the breadth of issues that can be represented and the extent that causality can be explored. In this paper we discuss the findings of a unique study exploring the benefits of combining Bayesian Networks (BNs) with conventional statistical analysis to examine landholder adoption. Our research examined the landholder fencing of native bushland in the Wimmera region in south east Australia. Findings from this study suggest that BNs provided enhanced understanding of the presence and strength of causal relationships. There was also the additional benefit that a BN could be quickly developed and that this process helped the research team clarify and understand relationships between variables. However, a key finding was that the interpretation of the results of the BNs was complemented by the conventional data analysis and expert review. An additional benefit of the BNs is their capacity to present key findings in a format that is more easily interpreted by researchers and enables researchers to more easily communicate their findings to natural resource practitioners and policy makers.