Development And Application Of A Bayesian Decision Support Tool To Assist In The Management Of An Endangered Species

Bayesian decision support tools are becoming increasingly popular as a modelling framework that can analyse complex problems, resolve controversies, and support future decision-making in an adaptive management framework. This paper introduces a model designed to assist the management of the endangered Camphora swamp eucalypt (Eucalyptus camphora). This tree species is found in the Yellingbo Nature Conservation Reserve (YNCR), an isolated patch of forest in the Yarra Valley (Victoria, Australia). The eucalypt community provides both habitat and food for a variety of threatened and endangered flora and fauna. Over the last 20 years the E. camphora has become increasingly threatened by dieback. In order to maintain and rehabilitate existing trees and encourage regeneration, management strategies and action plans have concentrated on restoring the hydrological regime, which has been altered due to agricultural activities within the catchment. However, research suggests that nutrient enrichment from surrounding horticulture and livestock is having a greater impact on the health of the trees. The Bayesian decision support tool has been used to examine the differences between these two hypotheses. The tool will also promote future integrative and iterative monitoring and research in the YNCR. This project was undertaken as part of a larger piece of work developing risk-based assessment guidelines for natural resource management, a process known as Ecological Risk Assessment. The Woori Yallock Creek Catchment, of which YNCR is a part, was chosen for the case study due to the high ecological assets identified in the catchment, the diverse land use and activities in the region, and the large body of knowledge and data available. The case study aims to identify environmental assets at greatest risk from ecological degradation in the Woori Yallock Creek Catchment and subsequently identify options for managing these risks. Phase One of the risk assessment process is problem formulation. Input was sought from a wide range of interest groups via one-on-one interviews and a workshop. The priority environmental assets in the catchment were subsequently identified, along with the hazards that threatened these assets. The Sedge-rich E. camphora community within YNCR was identified as one of the priority threatened environmental assets within the Woori Yallock Creek Catchment. For the purposes of this case study the condition of E. camphora was identified as the management end point for which YNCR will be managed. Phase Two (risk analysis) utilised Bayesian networks (BN) to quantify the risks to E. camphora. BNs are probabilistic networks that support reasoning under uncertainty. BNs are used to establish causal relationships between key factors and final outcomes, and maintain clarity by making causal assumptions explicit (Stow and Borsuk 2003). They are particularly useful for uncertainty analysis as they have the ability to consider inadequate knowledge or understanding of system processes, inherent randomness, subjective judgment and vagueness in parameter estimation, disagreement, measurement error and sampling error (Morgan and Henrion 1990). Risks to the condition of E. camphora have been prioritised and key knowledge gaps identified, while accounting for predictive uncertainties. Using the networks, the outcomes of a range of management scenarios have also been tested. The parameters that are most influential in determining E. camphora condition according to the model are generally soil nutrients, soil cations, pests or disease and inundation patterns (duration and frequency). The findings show E. camphora responses are different for each region in the model, and findings are specific for each survey. Much of the data used to parameterize the model was patchy and qualitative. This has contributed to significant knowledge gaps. The results of this study should be viewed as a guide to further work.