Landslide Susceptibility Mapping, Using Fuzzy Inference System andGIS (Case Study: Sections of Mazandaran Province)

Landslides are natural disasters that can damage human lives and various properties annually.Unplanned development expansions of cities results in locating inhabited area with high risk of landslides. Consequently, it is essential to generate landslide susceptibility maps for city expansions.The objective of this paper is to propose a method for discovering knowledge (fuzzy membership functions and fuzzy rules) that can be used for predicting landslide locations in Mazandaran province.Landslide phenomenon is a spatial one, and there are numerous sources of uncertainty in spatial data.Therefore, there is a need to integrate a Fuzzy Rule Based Inference System (FRBIS) into a Geographic Information System (GIS) for mapping such events. The first step towards forming such a system is clustering which is exercised by Fuzzy C means algorithm. To determine the optimum number of clusters, two Cluster Validity Indexes (CVI) and three criteria- namely; accuracy, completeness, and consistency-are proposed and then used. By projecting clusters onto perpendicular axes, fuzzy membership functions and fuzzy rules are obtained. At the end, the system is improved by adding experts’ knowledge. The accuracy of the landslide susceptibility map is estimated over 80%. Also, by examining the histograms of the landslide susceptibility maps, it is inferred that 13.5% of the region are presumably faced with high risk of landslides. The results indicated that the landslide susceptibility intensity has the most dependency with three factor maps; lithology, distance to roads and landuse. This is done by calculating the relations between the attained landslide susceptibility zonation and the factor maps.