Bayesian Decision Networks for Management of HighConservation Assets (National Water Initiative – Australian GovernmentWater Fund. Report 6/6Report to the Conservation of Freshwater EcosystemValues Project, Water Resources Division, Departmentof Primary Industries and Water)

The Conservation of Freshwater Ecosystem Values (CFEV) Validation project,funded under the Australian Government Water Fund (AGWF), assessed the validityof a number of attributes of biophysical character and condition in the CFEV databasefor selected river and wetland assets of high conservation value in five catchments.High Conservation Value assets which are affected by existing Water ManagementPlanning had been selected for field inspection, and observations were compared,where relevant, to values attributed to them in the CFEV database. As part of thisexercise, detailed descriptions of each asset were provided, and accompanied by adescription of the major ‘drivers’ of condition evident at each location, and the mainmanagement issues and priorities for the conservation of the asset’s values. These datawere summarised in tabular form and were also accompanied by a summary of thepriorities for Water Management Planning.Thus the project conducted an ‘on-ground’ validation of asset characteristics, andreported on the biophysical classes and condition of assets within each catchment anddescribed their management needs.A final aspect of this project was to initiate the development of decision support formanagement of these assets by developing preliminary Bayesian Decision Networks(BDN’s) informed by the learnings from the field surveys and from the ‘expert’project team (Dr Peter Davies, Dr Lois Koehnken and Dr Philip Barker).Bayesian networks are able to integrate different sources of evidence, representuncertainties in knowledge and inherently variable environments and explicitly linkecological outcomes with management activities and system context and changes. Theprimary aim of this project component was to develop Bayesian network models forpredicting the ecological condition of stream and wetland assets, using biologicalcondition as the primary indicator of asset condition or ‘health’.The networks developed here identify the dominant linkages between managementactions and the physical and biological components of the ecosystem assets. Twogeneric networks have been produced that can be adapted to form the basis of a guideto making decisions about asset management, as well as to examine changes inbiophysical attributes in response to controllable system changes, such asmanagement actions.