Abstract
ABSTRACT The weights-of-evidence modelling (based on multi-class maps) was applied within a geographical information system (GIS) to prepare a landslide susceptibility map. The area selected is along the E-W highway in Malaysia where frequent landslides occur. The spatial database for factors (evidences) that influence landslide occurrence were prepared from different sources including topographical maps, geological maps, satellite data, hydrological data, soil data and field data. Eleven prepared thematic maps of evidence were: slope gradient, slope aspect, elevation, distance from road, drainage density, lithology, strata dip, foliation dip, lineament density, soil, and rainfall. All maps were subdivided into different classes by its value or feature and then were converted to raster format in the ArcGIS 9.2, each representing an independent layer of causative factor in the constructed spatial database. The conditional independence test was carried out to determine factors that are conditionally independent of each other with landslides. The results of the Chi square analysis figured out ten possible models, including combinations of different independent factors, which were used in preparing ten landslide susceptibility indexes. Among these models, model three combining data on roads, soil, lineament, rainfall, slope, lithology and foliation showed the highest prediction accuracy (AUC= 80.08).