Qualitative and quantitative landslide susceptibility assessments in Hulu Kelang area, Malaysia

Abstract

Hulu Kelang is an area in Malaysia that is very susceptible to landslides. From 1990 to 2011, a total of 28 major landslide events had been reported in this area. This paper compares and evaluates the analytical hierarchy process (AHP), probability-frequency ratio (FR), statistical index (Wi), and weighting factor (Wf), used for assessing landslide susceptibility in the Hulu Kelang area. Eleven landslide influencing factors were considered in the analyses. These factors included two indices (the stream power index (SPI) and the topographic wetness index (TWI)) and several other factor including lithology, land cover, curvature, slope inclination, slope aspect, drainage density, elevation, distance to lake and stream, distance to road and trenches found in the area. The accuracy of the maps produced from the four models were verified using a receiver operating characteristics (ROC), and relative distributions of susceptibility level and active landslide zone. All the verification results indicated that the probability-frequency ratio (FR) model which was developed quantitatively based on probabilistic analysis of spatial distribution of historical landslide events was capable of producing a more reliable landslide susceptibility map in this study area compared to its other counterparts. About 89% of the landslide locations have been predicted accurately by using the FR map. On the contrary, the analytical hierarchy process (AHP) model that relies mainly on qualitative judgment yielded the least accuracy in landslide prediction.