An approach for integrating fuzzy rule system in hydrological modeling

摘要

When uncertainty is important, the fuzzy set theory offers a framework for processing qualitative information. In such a model, variables are treated as fuzzy numbers and calculations are replaced by inference rules. Therefore, we built some tools in order to introduce qualitative knowledge in the field of hydrology.rnFirstly, a C library has been developed, including basic fuzzy tools.rnSecondly, we built a fuzzy rule model software in which several algorithms are taken into account : defuzzification, aggregation rule, logical operator definition. The case-study concerns the determination of the ten year runoff coefficient in the Sahel area for which an evaluation method is based on three variables : the area and the slope of the basin, and an infiltrability index. In the classical ORSTOM method, the runoff coefficient is determined by interpolation between these variables on predefined curves. In the fuzzy approach, the runoff coefficient is computed by combination of fuzzy rules defined by the expert knowledge. The comparison shows a good agreement between the two methods.rnLastly, a specific fuzzy library has been implemented in a GIS in order to apply fuzzification and inference rules operations to raster maps.