Landslide mapping along the Karnali highway, Nepal using high resolution imagery

The Karnali highway is a vital transport link connecting the remote Karnali region to the capital in Province 6 of Nepal. This area becomes inaccessible by roads during every rainy season due to landslides. However, reliable information is not available on the location of areas along the highway susceptible to landslides. In this study, we use object-oriented analysis to map landslides along the Karnali highway with a 6 km buffer using imagery obtained from high resolution sensors of the DigitalGlobe constellation in 2012 and Sentinel 2 in 2017. In total 1103 landslides were detected, ranging from 3.39 to 30,496 m² in area. Landslides were located at lower elevations, on steep south facing slopes. Landslides were common near highway and drainages but not near faults. The Kuncha geologic formation contained the majority of landslides, as did the forest and agricultural land cover types. Based on the 2012 landslide inventory, susceptibility mapping along the highway was conducted using logistic regression. This map was validated with the 2017 landslide inventory. The validation highlighted good agreement between the susceptibility map and the location of the 2017 landslides. The very high and high susceptibility zones make up 28.84 % of the total study area and includes lower altitude slopes of the Kuncha formation, banks of the Karnali River and areas closer to the highway. This work provides first estimates of high-risk areas to landslides along the highway. It can be incorporated into development projects seeking to retrofit or mitigate hazards along this dangerous highway.