Keywords
ecology, bayesian decision network, elicitation, expert bias
Start Date
1-7-2006 12:00 AM
Abstract
Ecological and integrative analyses routinely involve the synthesis of a range of information sources into a single model. Bayesian decision networks (BDN) are increasingly being used for this purpose because they are flexible, transparent and relatively easy to use. Indeed, BDNs offer a scientific and pragmatic approach to improve decision-making in environmental management, directly addressing management needs, while promoting stakeholder participatory processes. However, despite their advantages, many BDNs developed to meet such needs are not being developed or applied to their full potential. The majority of BDNs published to date rely only on expert opinion to parameterise and evaluate ecologically relevant endpoints. In contrast, environmental processes in BDNs, such as water quality, are optimised using quantitative data. In this paper we discuss the need to better exploit the Bayesian aspect of BDNs. We use examples to discussed the contrast between probability networks and BDNs, the need to use existing data where possible for parameterisation and evaluation (in conjunction with knowledge and weighting information sources), and the need to incorporate BDNs into an iterative cycle of updating. We argue that the alleged advantages of BDNs in improving the robustness and scientific credibility of ecological decision-making are questionable if ecological data is not better used in sustainability and risk assessments.
Bayesian decision networks - going beyond expert elicitation for parameterisation and evaluation of ecological endpoints.
Ecological and integrative analyses routinely involve the synthesis of a range of information sources into a single model. Bayesian decision networks (BDN) are increasingly being used for this purpose because they are flexible, transparent and relatively easy to use. Indeed, BDNs offer a scientific and pragmatic approach to improve decision-making in environmental management, directly addressing management needs, while promoting stakeholder participatory processes. However, despite their advantages, many BDNs developed to meet such needs are not being developed or applied to their full potential. The majority of BDNs published to date rely only on expert opinion to parameterise and evaluate ecologically relevant endpoints. In contrast, environmental processes in BDNs, such as water quality, are optimised using quantitative data. In this paper we discuss the need to better exploit the Bayesian aspect of BDNs. We use examples to discussed the contrast between probability networks and BDNs, the need to use existing data where possible for parameterisation and evaluation (in conjunction with knowledge and weighting information sources), and the need to incorporate BDNs into an iterative cycle of updating. We argue that the alleged advantages of BDNs in improving the robustness and scientific credibility of ecological decision-making are questionable if ecological data is not better used in sustainability and risk assessments.
