Representation and propagation of vague knowledge: what the theories of uncertain can bring to environmental sciences
Congress title :17èmes Journées Scientifiques de la Société d'Ecologie Humaine
Congress town :Arles
Congress date :23/11/2005
The taking into account of uncertainties became a stage impossible to circumvent in the évaluation and risk management process. However, the realization of an analysis of uncertainties raises many technical difficulties because it often runs up against the limits of scientific knowledge and the formalization of some sources of uncertainties rests on an incomplète or subjective knowledge. Moreover, the propagation of vague knowledge, i.e. the characterization of its effect on the computation resuit, requires the use of théories still badly known in the field of environmental sciences, the théories of the uncertain. The présent communication proposes to présent the work completed for several years by the Institut de Radioprotection et de Sûreté Nucléaire (IRSN) in collaboration with the Bureau de Recherches Géologiques et Minières (BRGM), the Institut National de l'Environnement Industriel et des Risques (INERIS) and the Institut de Recherche en Informatique de Toulouse (IRIT - Université Paul Sabatier) on the représentation and the propagation of incomplète or vague information and its application in the case of the contamination of the food chain by a pollutant. The main objective of this work is to promote cohérence between the way in which information is mathematically represented and information really available. To achieve this objective, alternative methods are proposed to represent and propagate uncertainties. In particular, this work highiighted that the systematic use of probability distributions to represent environmental knowledge is often subjective and arbitrary and can resuit in underestimating the risk. To exceed this problem, the possibilities of use of various théories of uncertain were studied (probabilities, possibilities, functions of belief, p-boxes and fuzzy numbers). This utilisation required mathematical and numerical developments to elaborate tools usable for the représentation and the propagation of knowledge available in environmental sciences. Thèse tools will be described through their application to the case of the contamination of milk following a chronic strontium 90 deposit. This concrète example makes it possible to compare the impact of the différent mathematical methods and to discuss the difficulties of the use of the various kinds of results produced within the framework of the decision-making.