The sustainability assessment of projects : an artificial intelligence approach, with application to Tasmania

This thesis develops and presents a practical method of assessing the sustainability impact of projects. The research has been motivated by the fact that sustainability is an accepted goal of resource management and planning legislation, and yet we have few tools with which to quantitatively assess sustainability impacts. The history of the sustainability imperative is reviewed, and it is proposed that sustainability be considered in terms of three themes which match the way in which most planners think about sustainability issues, and commission specialist studies: 1. Biodiversity. Biological diversity and integrity. 2. Socio-economic. Well being and equity within and between generations. 3. Physical environmental. Local, regional and global environmental quality. A sustainability assessment method is developed, whereby project impacts are measured using indicators of the sustainability issues associated with the project. This takes advantage of the indicator sets being developed by the Local Agenda 21 initiative, and by the State of the Environment reporting process. Ways are examined to aggregate indicators into indices, using both traditional approaches and expert systems. Fuzzy rule systems and neural networks are shown to offer powerful, natural alternatives to traditional aggregation methods, and case studies are presented which use these tools to aggregate information relating to roadside vegetation quality, algal blooms, urban air quality, and sewage treatment plant performance. A traditional modelling approach to predicting indicator changes would solve (numerically integrate) the differential equations governing the interactions between the indicators, computing interaction terms at each time step. However, in this instance the equations are unknown, and the inputs are often known only semi-quantitatively. A modelling approach based on a fuzzy rule system is developed that overcomes these barriers, in which indicator changes due to interactions between indicators are computed iteratively. The model is applied to a number of basic situations, and approaches to driving the model are discussed. Model performance and sensitivity tests are carried out that demonstrate the behaviour of the model to be reasonable, and an illustrative application is presented. The sustainability assessment method is further validated by case studies in mining, forestry and road transport planning. The model predictions compare well to expectations, although rigorous test data are not yet available, and the method is shown to be an effective tool for screening projects, particularly when used to compare project options. It can be used to improve the design of baseline studies, to design appropriate monitoring programs, and to examine the need for application of the precautionary principle. The use of multi-criteria decision-making methods and genetic algorithms to select and optimise a preferred project option is explored, and illustrated by application to a proposed road upgrade project. The thesis concludes by discussing follow-on areas of research, and approaches to improving the assessment method.