Bayesian Decision Networks for Management of High Conservation Assets (National Water Initiative – Australian Government Water Fund. Report 6/6 Report to the Conservation of Freshwater Ecosystem Values Project, Water Resources Division, Department of Primary Industries and Water)

Abstract

The Conservation of Freshwater Ecosystem Values (CFEV) Validation project, funded under the Australian Government Water Fund (AGWF), assessed the validity of a number of attributes of biophysical character and condition in the CFEV database for selected river and wetland assets of high conservation value in five catchments. High Conservation Value assets which are affected by existing Water Management Planning had been selected for field inspection, and observations were compared, where relevant, to values attributed to them in the CFEV database. As part of this exercise, detailed descriptions of each asset were provided, and accompanied by a description of the major ‘drivers’ of condition evident at each location, and the main management issues and priorities for the conservation of the asset’s values. These data were summarised in tabular form and were also accompanied by a summary of the priorities for Water Management Planning. Thus the project conducted an ‘on-ground’ validation of asset characteristics, and reported on the biophysical classes and condition of assets within each catchment and described their management needs. A final aspect of this project was to initiate the development of decision support for management 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, represent uncertainties in knowledge and inherently variable environments and explicitly link ecological outcomes with management activities and system context and changes. The primary aim of this project component was to develop Bayesian network models for predicting the ecological condition of stream and wetland assets, using biological condition as the primary indicator of asset condition or ‘health’. The networks developed here identify the dominant linkages between management actions and the physical and biological components of the ecosystem assets. Two generic networks have been produced that can be adapted to form the basis of a guide to making decisions about asset management, as well as to examine changes in biophysical attributes in response to controllable system changes, such as management actions.