Forecasting of Rainfall-Induced Landslides Along a Major Road Corridor of Darjeeling Himalayas, India

Forecasting the occurrence of rainfall-induced landslides involves examining the statistical relationship between previous landslide events and rainfall characteristics. Therefore, a catalog of rainfall and landslide events have been prepared for forecasting landslides along one of the severe hazard-prone road corridors of the eastern Himalayas, i.e., National Highway-10 (NH-10). Daily rainfall data for the period of 2011 to 2018 have been collected from four rain gauge stations, adjacent to the NH-10. In contrast, landslide records were obtained for the same period from Public Works Department (PWD), media reports and literature. The NH-10 identified a total of 288 rainfall events that triggered 675 landslides. The rainfall event-duration model has been applied for different exceedance probability (1% to 50%) to estimate the rainfall threshold for initiation of landslides. The model was calibrated for the entire landslide prone section of NH-10 first and then for each lithological unit separately on which the highway is passing. The estimated rainfall threshold for NH-10 is E = 0.15D1.03 with 24 ≤ D ≤ 408, where E is representing the cumulated event rainfall (mm) and D is the rainfall event duration (hours). The result shows, the chlorite sericite schist is comparatively resistant to rainfall-induced slope failure than other rocks. Finally, the proposed threshold for NH-10 has been validated with 48 rainfall events using receiver operating characteristics (ROC) curve. The acceptable prediction accuracy (65%) recommends the model’s efficiency for forecasting of rainfall-induced landslides along NH-10.