Mapping of earthquake hotspot and coldspot zones for identifying potential landslide hotspot areas in the Himalayan region

Landslide and earthquake are two of the dangerous natural hazards in the Himalayan mountainous ranges. This study focused on small-scale landslide susceptibility (LS) maps as well as hotspot zones of earthquake and landslide on the Himalayas. Remote sensing, geological, and hydrological data were collected and processed using the geographic information system (GIS) for this purpose. Seventeen landslide influential causative factors (LICFs) were used, which included slope, aspect, elevation, profile curvature, plan curvature, topographic wetness index (TWI), stream power index (SPI), terrain ruggedness index (TRI), drainage density, distance to drainage, seismic zone, soil texture, geology, and land use and land cover (LULC). The machine learning (ML) neural network and deep learning models, namely, linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), artificial neural network (ANN), deep learning neural network (DLNN), and Getis-Ord GI*, have been applied for landslide susceptibility (LS) maps, and earthquake and landslide hotspot maps respectively of the Himalayas. The LS maps were compared with the earthquake and landslide hotspot zones, and there is valid agreement among these. The results of the susceptibility mapping were validated by the confusion matrix table and sensitivity, specificity, accuracy, precession, F score, and kappa statistical indices. The validation result showed that all the landslide susceptibility models (LSMs) have a near 90% prediction rate. Among the four models, the DLNN models have the highest prediction accuracy followed by LDA, QDA, and ANN. The small-scale LS maps of the Himalayas may help planners and developers for macro-scale mitigation of landslide hazard and land use planning.

Automated summary

In this study, landslide susceptibility and earthquake and landslide hotspot maps were created for the Himalayan mountainous ranges using remote sensing, geological, and hydrological data processed with GIS. Machine learning and deep learning models were applied to analyze the data, with high prediction accuracy observed. These small-scale maps can assist planners and developers in mitigating landslide hazards and land use planning.