Comparing seismic susceptibility models of the Himalayan terrain

Abstract

The ongoing intra-continental collision between the Indian and Eurasian plates along the Himalayas has resulted in many damaging earthquakes with severe damages to man-made structures and natural landscapes due to ground shaking and ground failure, which in turn depends on geomorphological, geological and geophysical variables. Seismic susceptibility models are developed for Gangtok City by combining all the three variables using both knowledge-driven and data-driven methods on facet and grid cell terrain units. Finally, the results are critically evaluated by validation with the earthquake intensity data recorded during earthquake events. First-stage modelling attempt using different knowledge-driven methods on different terrain units shows bi-modal data distribution with low predictability due to extremely rugged topography with wide altitudinal variations within short distances. Second-stage modelling of separated population by using the same methodologies increases model predictability in which one model method describes the higher topographic levels better and the other model method is found to be better for lower topographic levels. Seismic susceptibility of the area is best described by composite models, combining different best methods of fine classification for lower and higher topographic levels having the same mapping/terrain units. Comparison of the composite models shows that the terrain unit does not play a significant role but the type of models selected determines the best possible seismic susceptibility map of the area.