|
Ecological Risk Assessment (ERA) |
||
|
Navigation:
Background |
The Bayesian Network, which is the subject of this thesis, is in the ecology domain and is part of a broader Ecological Risk Assessment (ERA) specific to the Goulburn Broken Catchment. The objective of this ERA is to develop and test a generic framework to be used in the assessment of ecological risks associated with Australian irrigation activities. This project is funded by the National Program for Sustainable Irrigation and is currently being conducted at the Water Studies Centre (Monash University), under Dr Carmel Pollino. The model component of this project incorporates physical and fishery knowledge to assess the native fish abundances and diversities resulting from management interventions in the Goulburn river system. The model is to be used in an adaptive management scheme and will continue to undergo development, using the methods identified in this thesis. The study is focused on the Goulburn Broken Catchment. The geographical areas within the catchment are divided into regional and local scales. The local scale chosen was the Goulburn Weir/Lake Nagambie, and the regional scales are the Goulburn River reaches from Eildon to Seymour and Murchison to the Murray River. Temporal scales to be considered are 1, 5, 10 years, and 30 to 50 years; these were selected as they accurately reflect the life history of the different fish species in the area. During the problem formulation phase of the ERA, a conceptual model was created to show possible factors influencing the native fish population and diversity. A simplified version of this model is shown here, the double arrow links are included to demonstrate possible interaction between factors:
Reverse Engineering the Prototype Model:The four portions identified in this conceptual model were built into the prototype causing the query variables named Native Population Abundance and Native Population Diversity. Due to the multiple data nodes in each of the model portions, the results were integrated into a descriptor node designed for the model. These descriptor nodes for each portion were named Water Quality, Overall Change in Flow Regime, Structural Habitat Quality and Competition. These descriptor nodes were based purely on expert elicitation. The only nodes to be trained via data were those that were identified as the variables for which data was available. Below is the prototype Bayesian Network:
Prototype Model
|
