Landslide risk assessment by integrating hazards and vulnerability indices in Southeast Bangladesh

Abstract

Landslide risk assessment (LRA) is crucial to develop sustainable risk reduction and response measures. Although Southeast Bangladesh is prone to landslides, there is insufficient comprehensive investigation that incorporates susceptibility and vulnerability assessments with LRA. This study first calculated the Landslide Susceptibility Index (LSI) using the frequency ratio (FR), verified it using the ROC-AUC curve, and then measured the Social Vulnerability Index (SoVI) using Principal Component Analysis (PCA). Finally, the research integrated the LSI and SoVI map by normalizing and weighting to compute a Landslide Risk Index (LRI) map. The region's LSI reveals 10.19 % very high, 28.10 % moderate, and 13.41 % very low landslide susceptibility. According to the FR model, slope, elevation NDVI, and rainfall predispose to landslides. However, the SoVI revealed 39.3 % of Upazilas experienced medium landslide social vulnerability, which is hazard independent. The LRI statistics classify landslide risk as very high (23 %), high (12 %), moderate (28 %), low (20 %), or very low (17 %) for a total area of 19163.53 km². A district-wise risk analysis ranks landslide risk as Chattogram > Cox's Bazar > Rangamati > Khagrachhari > Bandarban due to short-term or prolonged rainfall, natural drainage changes, unplanned development, deforestation, hydroelectric plant effects, and population growth. The landslide risk is highest in Teknaf, Ukhia, and Ramu Upazilas and lowest in Thanchi, Rowangchhari, Juraichhari, and Langadu. LSI performed satisfactorily with an AUC-based prediction rate of 81.8 % and success rate curve of 87.0 %. Finally, decision-makers can implement this macro-scale regional landslide risk analysis for southeast hilly Bangladesh for developing sustainable risk reduction schemes.