SPATIAL INTEGRATION OF GEO-ENVIRONMENTAL VARIABLES FOR MODELLING OF LANDSLIDE HAZARD AND RISK, CENTRAL NEPAL HIMALAYA

GIS-based comprehensive computations have been carried out to quantify landslide hazard and risk in the central Nepal Himalaya where an extreme weather event of 19–21 July 1993 triggered a large number of landslides, debris flows and floods that caused extensive infrastructure damages and human fatalities. Spatial integration of geo-environmental variables has focused on digital elevation model, geo-lithology, slope gradient, land use and intense rainstorm. Database was acquired semi-automatically from different sources: existing digital data, image processing and field surveying. Characteristic features of landslides were evaluated based on frequency statistics and analytical combinations. Analysis reveals that most of the landslides were strongly triggered by topography and complex geological characteristics of the terrain; slope gradient and bed rock layering factors are statistically significant slope instability factors. Natural slope angle is the most distinct pre-disposing factor for likelihood, distribution and characteristics of landslides. The slope angles for landslide initiation varied from 15° to 35° with critical angle of 27°. Bedrock layering is another important explanatory variable which caused the spatial localization of slope failures.

A multivariate model was developed to assess landslide hazard and risk by a logistic regression method that uses the input factors as independent variables and the presence/absence of landslides as a dependent variable. Results of the model have been optimised by adjusting the input variables in an iterative process. Cross verification of landslide hazard map pointed out those areas of high susceptible zones show significant indications of slope instability marked by landslides, erosion and subsidence. Good correlation exists between the areas predicted as ‘high’ probability of hazard and the known landslides. Areas in high to very high hazardous zones that are devoid of landslides indicate potential landslide zones. Risk probability of the study area exhibited that 15.4 per cent of population lives on highly hazardous landslide-prone areas. Blending infrastructure and settlement areas with the hazard/risk delineated map has enabled the assessment of buildings, highway and foot trails distribution according to different levels of vulnerability scenarios.