FUZZY CLASSIFICATION OF SITES SUSPECTED OF BEING CONTAMINATED

One part of today's environmental pollution problem is referred to by the term contaminated soils. In Germany specialists expect to find app roximately 260 000 sites suspected of being contaminated after complet ion of the registration phase. This large number, the urgency of carry ing out measures to protect the environment and limited personnel and financial resources result in the necessity for a computer support. De tailed considerations of deficiencies of existing computer systems and formal estimation methods lead to a bipartite knowledge-based approac h supporting the relative estimation of hazard. Multidimensional class ification, the core of this approach, is mathematically well-founded a nd serves as an extension of methods already used successfully in prac tical operation. The unsuitability of formal risk analysis methods and various sources of incompleteness, uncertainty and vagueness of the w hole research field motivate the use of fuzzy methods, and in particul ar the use of fuzzy classification providing a rough ranking method. F eature generation, the other main part of the approach, allows selecti ng, valuating and tuning the properties of the sites in such a way to ensure an optimal classification. For maximizing the expressive power of the system's results, the user is enabled to compromise between a d etailed survey of a site and an easy to survey representation of a sit e with resulting loss of information caused by a certain a priori aggr egation of properties.