1 Enregistrement scientifique n° : 766 Symposium n° : 17 Présentation : oral Mapping soil hydraulic properties from imprecise information contained in a small scale pedological map Cartographie de propriétés hydriques de sol à partir de l’information imprécise contenue dans une carte pédologique à petite échelle CAZEMIER Durk (1), LAGACHERIE Philippe (1), MARTIN-CLOUAIRE Roger (2), BORNAND Michel (1) (1) INRA, Science du Sol, 2 Place Viala, 34060 Montpellier Cedex, France (2) INRA, Biométrie et I.A., BP 27, 31326 Castanet-Tolosan Cedex, France 1 Introduction Environmental research or town and country planning are fields where generally soil related information is required. Soil maps of different scale have been produced by soil survey institutes in order to meet this need. Small-scale soil maps (i.e. 1:250,000 - 1:1,000,000) can be considered as the main source of soil information when a large area of land is concerned, e.g. a European region or a country as a whole. Compared to detailed soil maps, these maps generally provide an approximate and synthetic description of geographical reality. Consequently, regional soil databases contain imprecise information about the soil properties characterizing a soil type and about the geographical location of the soil types. The term regional soil database includes the computerized system for storage and displaying the small scale soil map together with the body of related data like soil profiles. The application of information extracted from regional soil databases still requires further development in dealing with the existing imprecision: in particular to specify it explicitely in the representation of knowledge and to propagate it to the output returned to the user as answer to her/his queries. This study discusses the estimation of the water storage capacity of the soil on the basis of the 1:250,000 regional soil database of the French Languedoc-Roussillon region (Bornand et al., 1989). The water storage capacity of the soil is considered in relation to its agricultural importance: the estimation of this soil property over a region is performed as part of a European project on the prediction of the impact of climate change on regional European agriculture. The use of existing conventional soil maps for predicting soil hydraulic properties has been demonstrated by Leenhardt et al. (1994) and Voltz et al. (1997). The statistical framework used in these studies for handling uncertainty is not relevant here because of the lack of reliable frequency data and because of the fuzzy nature of the geographical objects composing the soil maps. This paper advocates the use of possibility theory (Zadeh, 1978) to represent the qualitative uncertainty that corresponding to the kind of soil information available in practice. The possibilistic representation of uncertainty essentially relies on the order over the plausibilities of the elementary events rather than