Assessment of landslide zonation by multi variant statistical method: In the case of Iran

Abstract

This research attempts to zonate the Marbor River domain in the south of Semirom city with an area of 1438 km2 using multivariate statistical methods because of increasing instability of the slopes and landslide phenomena, to prevent it from human and financial damages. Therefore, by collecting required information on landslides dispersion in the region, landslide dispersion map is provided by preparing six informative layers which includes slope, aspect, lithology, land use, distance from the road, and precipitation rate. These layers were then intersected, whereby 1365 homogeneous land units were obtained. After that, the homogeneous units map was intersected with the landslide dispersion map. Finally, a 1365 × 22 matrix was prepared with values 0 and 1.0 denotes orders with landslide and 1 without landslide. Then, this matrix was included into SPSS application and linear regression and discriminate analysis methods were used to calculate and analyze the data. Afterwards, we analyzed the accuracy and precision of each method (P, QS). In conclusion, it is clear that linear regression method is more accurate and precise than the discriminate analyses. It is recommended, thus, to use the former in landslide hazard zonation. © 2010 IEEE.