1 Recognition of erosion risk areas using neural network technology: an application to the island of Corfu Gournelos Th., Vassilopoulos A., Evelpidou N. Geography & Climatology Department, Faculty of Geology, University of Athens, Panepistimiopolis, Zografou, 157 84, Athens, Greece, evelpidou@geol.uoa.gr Abstract There is a wide range of alternative approaches to study erosion processes. In this paper the construction of a model based in the interaction of Geographical Information System (GIS) and Artificial Neural Networks (ANN) is described. The neural model uses supervised competitive learning process. The whole procedure starts with the digitization of the data and the definition of the input variables: such as slope form and gradient, lithology and vegetation - landuse. The neural model transforms the input variables into the erosion risk output variable. Thus, the last stage regarded the creation of an erosion risk zones map. For case study was chosen the island of Corfu (Greece). The island consists of lithologies very vulnerable to erosion and receives considerable amounts of rainfall, especially if compared to the rest of the Greek territory. Finally, the whole model was tested and the proper function of the model was confirmed by field data observations. Key Words: Neural networks, geomorphological process, GIS Introduction Erosion is a very complicated geomorphological process. Before erosion weathering, takes place. The thickness of the weathered material is interdependent of climatic variables such as temperature and rainfalls substratum rocks’ lithology, vegetative cover and topography. The erosion process carries away the weathered material, but there are many factors controlling that procedure. Recognition of erosion risk zones involves a series of different stages: field work, air-photos interpretation, satellite image analysis, digitization of geological, topographical and drainage system maps and finally the definition of the input parameters. There is a vast bibliography concerning soil erosion or gravity movement processes making all kinds of different approaches (Brundsen et al., 1975; Carrara et al., 1977; Malgot and Mahr, 1979; Ives and Messerli, 1981; Carrara, 1983; Morgan, et al. 1984; Carrara et al., 1991; Morgan, 1996; Marinos et al., 1997; Binagli et al., 1998; Stassopoulou et al., 1998; Gournelos et al., 2004). In this study the main variables used to estimate erosion risk zones are geomorphology of slopes, morphological slope gradient, lithology, vegetation and land use. The geomorphology of slopes, depends of the slope form, the elements of which reflect the exogenous processes involved, such as weathering and erosion. A typical three division of slope’s elements has been adopted: upslope, mid-slope and footslope. The second processed variable is the morphological slope gradient. It is obvious that the slope’s steepness is critical for the intensity of erosion. The slope characteristics are also influential. Sediment removal is estimated as a function of divide distance multiplied by a power of slope gradient (Kirkby, 1978). Where slopes are steeper, gravitational movement and downslope sediment delivery are higher. In general, a linear or power relation between slope gradient and erosional products has been noted (Kirkby, 1969; Schumm, 1977). Lithology is an important factor controlling the hardness of rocks, while the combination of rock’s composition and texture, expressed by physical resistance, also affects erosion rates (Meybeck, 1987).