Representing and processing uncertain soil information for mapping soil hydrological properties

Mapping soil hydrological properties for a large area (over 104 km2) is a difficult problem because the knowledge and data usually available for such a task are pervaded by uncertainty and imprecision. Because of this lack of reliable statistical data a possibility theory approach has been developed. The resulting system combines a raster type GIS (geographical information system) with a constraint satisfaction solver: the first playing the role of a data base and the latter being used to propagate uncertainty along relations among soil properties. For every point in the area covered, identification of the more or less possible soil typological units is performed on the basis of flexible matching between, on the one hand, environmental properties characterizing these units and, on the other, what is known about these properties for the location under consideration. For each candidate soil unit, hydrological soil properties, such as the water storage capacity, can be computed by exploiting numerical relationships linking them directly or indirectly to other soil properties. Due to uncertainty the possible values of these variables are obtained by a fuzzy interval calculus that is implemented as a constraint satisfaction problem (CSP) for computational efficiency. The system can produce maps of the localization of soil units and maps of the hydrological soil properties considered with respect to a threshold value. These maps are presented in a way that accounts for the inherent uncertainties in the material processed.