Using fuzzy logic to generate conditional probabilities in Bayesian belief networks: a case study of ecological assessment

The survival of rare animals is an important concern in an environmental impact assessment. However, it is very difficult to quantitatively predict the possible effect that a development project has on rare animals, and there is a heavy reliance on expert knowledge and judgment. In order to improve the credibility of expert judgment, this study uses Bayesian belief networks (BBN) to visually represent expert knowledge and to clearly explain the inference process. For the case study, the primary difficulty is in determining a large amount of conditional probabilities in the BBN, because there is a lack of sufficient data concerning rare animals. Therefore, a new method that uses fuzzy logic to systematically generate these probabilities is proposed. The combination of the BBN and the fuzzy logic system is used to assess the possible future population status of the Pheasant-tailed jacana and the associated probabilities, which have been affected by the construction of the Taiwan High-Speed Rail. The analysis shows that a restoration program would successfully preserve the species, because in the restoration area, the BBN model predicts that there is a 75.49 % probability that the species will flourish in the future.

Keywords
Pheasant-tailed jacana; Future population status; Expert judgment; Artificial intelligence