MULTI-CRITERIA ANALYSIS OF LANDSLIDE SUSCEPTIBILITY, AFGHANISTAN

In Afghanistan, diverse geologic settings result in slopes susceptible to various forms of failure that pose considerable risk to the nation’s growing population. Poor land management, such as overgrazing, deforestation, flood irrigation, unmaintained dirt roads, and unplanned urban sprawl exacerbate slope-stability problems and hazards. Landslide-susceptibility modeling (LSM) provides a tool for determining landslide risk potential and is critical for hazard planning and implementation of early warning systems. Multi-criteria models using remote sensing and GIS are ideal for regions such as Afghanistan that are too dangerous or ill-equipped to gather sufficient ground-based data for quantitative analyses. Lithology, slope, fault density, earthquake density, vegetation (NDVI), river density, aspect, and road density in Afghanistan were used as input criteria for the model. Each criteria was assigned relative susceptibility values based on established methodologies and modified considering previous field and remote studies of landslides in Afghanistan. The model was performed as a weighted overlay to produce a landslide-susceptibility index (LSI). Model performance was assessed by examining studies that used similar multi-criteria models that assume relative contributions of environmental factors to landslide susceptibility and more rigorous statistical analyses of landslide inventories. Model results were further validated with qualitative ground assessment by comparing areas modeled to be susceptible to those with failures described in previous studies. The most susceptible modeled areas in Afghanistan are in granite and metamorphic outcrops at steep slope angles in areas of high fault and earthquake densities. Loess deposits are modeled as second-most susceptible to landslide events. Low slope angle and dense vegetation cover appear to be stabilizing factors resulting in lower modeled LSI values of otherwise highly susceptible lithologies such as loess. Findings are consistent with expectations based on modeling literature and are supported by other studies of landslide distribution in Afghanistan. Such modeling efforts will help improve hazard assessment in Afghanistan and could improve our understanding of mass-movement controls in similarly rugged and arid landscapes.