Application of statistical and machine learning techniques for landslide susceptibility mapping in the Himalayan road corridors

Abstract Landslides are frequent geological hazards, mainly in the rainy season along road corridors worldwide. In present study, we have comparatively analyzed landslide susceptibility by employing integrated geospatial approaches, i.e., data-driven, knowledge-driven, and machine learning (ML), main of Muzaffarabad district. The inventory three is developed to evaluate susceptibility, eleven causative factors (LCFs) were analyzed. After statistical significance analysis, these LCFs generated models using WoE, AHP, LR, RF. Distance from roads, landcover, lithological units, slopes considered more influential LCFs. performance matrix different LSMs evaluated through area under curve (AUC-ROC), overall accuracy, Kappa index, F 1 score, Mean Absolute Error, Root Square Error. AUC-ROC for RF techniques Neelum 0.86, 0.82, 0.91, 0.97, respectively, Jhelum Valley 0.83, 0.81, 0.93, 0.95, while Kohala 0.89, 0.88, 0.92, respectively. produced ML (i.e., LR) showed better prediction accuracies than WoE AHP corridors. categorized into very high, moderate, low susceptible zones roads. LSM hybrid can facilitate concerned local agencies implement mitigation policies landslide-prone