An approach for GIS-based statistical landslide susceptibility zonation: with a case study in the northern part of El Salvador

The present study deals with evaluation of landslide prone zones in the northern part of El Salvador. The study area falls onto a tectonically and seismically active zone of Central America with on-going neo-tectonic activities. Focus has been put on applying the technique that allows a fast assessment of large regions. The analysis was based on digital data sets including various derivatives of digital elevation models (DEMs) as well as Landsat-based information such as micro-lineament density and landcover; seismic database, geological and morphological maps. Spatial multi-layered information has been used for landslide susceptibility analysis. Here, an inventory map of 363 landslides induced in 1998 by hurricane Mitch were used to produce a dependent variable, the statistical hazard analysis has been carried out while the zonal statistics was used to assign the weights for individual classes of the studied factors. Thus, all the relevant thematic layers representing various independent factors (slope, aspect, relative relief, lithology, drainage density, micro-lineament density and land cover) were relatively weighted and classified due to its disposition to cause landslides. Principle Component Analyses (PCA) was used as a multivariate statistical method that allowed decorrelation of the individual hazard triggers. It has been observed that the high potential zones were found to have very high lineament density, high relative relief and drainage density areas. On the young volcanic pyroclastic deposits, heavy rainfall and sparse vegetation cover cause persistent recurrence of landslides along this region. As result, a landslide susceptibility map integrating morphological, lithological and hydrological information was computed. Delineated hazard zones were again validated with the landslide inventory map and both, the model and terrain mapping, showed a good agreement as the highest class occupied the 64% of the landslide areas and the two highest classes together occupied 90% of the landslide areas, on the other hand none of the landslides fell into the lowest class.