Landslide Hazard Assessment for Fayzabad District, Badakhshan Province, Afghanistan

Abstract

Fayzabad District is one of those most impacted by landslide hazards in Afghanistan, accounting for 71% of all national landslide fatalities reported between 2012 and 2017. Necessary elevation data did not cover the very south of Fayzabad District; consequently, this study focuses on the northern two thirds of the district, where data were available. A landslide inventory was developed by mapping landslides using DEMs and high-resolution satellite imagery to aid in development and assessment of both Heuristic and bivariate statistical models of landslide susceptibility. Landslide statistics, including length, area, width, and pertinent relationships to geology, elevation, aspect, slope, and proximity to faults and streams were quantitatively calculated using geoprocessing tools. Hazard maps were produced using landslide susceptibility and proximity of villages to mapped landslides. Mapped susceptibility results indicate that in this part of Afghanistan landslides occur primarily on north to northwest aspects in loess or soil media over gneiss bedrock. Landslides are concentrated between 1500 m and 2000 m elevation and on 18° to 45° slopes within 60 m of a stream channel and or within 1 km of a fault. Landslide dimensions plot linearly on log-log scales, simplifying the development of predictive associations. Model results encapsulate a high proportion of landslide pixels within areas of high susceptibility, although there were significant variations between Heuristic and bivariate methods. Bivariate methods performed better universally, but may be over trained when the entire dataset is used to produce statistical weights. Use of subset of data to develop weights results in a more even distribution of landslides between low- to high-susceptibility zones. Findings in both the landslide inventory and susceptibility models are supported by prior studies of landslide behavior in Afghanistan. Programmatic workflows allowed for rapid production of many model components after initial reclassification and will facilitate further research in Afghanistan, and application of the methodology elsewhere. Map products potentially provide a new tool for hazard planners and aid groups in northeastern Afghanistan, and supplemental code will allow for rapid incorporation of new datasets as they are developed.