Transferring scientific knowledge to natural resource managers using artificial intelligence concepts

Abstract An expert system shell, GEM-II, has been developed specifically for classification problems in environmental management. Specialist information can be represented as production rules, mathematical models and time-independent facts held in a database. All information, i.e., the rules, models and facts, can be assigned to positions in Cartesian space. The shell has been coupled to a knowledge base containing information on fire behaviour and the effects of fire on vegetation in part of Kakadu National Park, Northern Australia. This system is now installed at the Park headquarters. The information has been derived from experiments run over ten years and from general experience of the region. The first part of the base contains a model of fire behaviour believed to be applicable throughout the Park; the second part (fire effects) is restricted to regions typified by the experimental site. The expert system will be run and assessed by Park managers over the next few years. The system is flexible and allows the managers to modify rules and change ‘facts’ as further knowledge of fire behaviour and effects becomes available and to add mathematical models of fire behaviour when these have been calibrated for northern Australia. Whilst the suitability of the system for the Park managers is yet to be fully assessed, it is already cler that expert systems can assist scientists to develop models of the processes they study.