Assessing landslide susceptibility in the Upper Ravi river catchment, Himachal Pradesh, India: a comprehensive analysis using the logistic regression model

Abstract Background The Himalayan region has witnessed a notable increase in landslide occurrences due to changing human-environment relations and rising anthropogenic pressures. These geomorphic hazards are frequently triggered by extreme weather events, such as intense monsoon rainfall and cloudbursts, resulting in loss of life, infrastructure damage, and widespread socio-economic disruptions. Himachal Pradesh, in particular, remains highly vulnerable to such events. Objectives This study aims to assess the spatial distribution and susceptibility of landslides in the Upper Ravi River Catchment, Himachal Pradesh, using a logistic regression model. The primary objective is to identify high-susceptibility landslide zones and understand the underlying geospatial factors affecting landslides in the Himalayan regions. Methods A landslide inventory of 513 events was prepared using visual interpretation of high-resolution satellite imagery from Landsat, Sentinel, PlanetScope, and Google Earth and field work from 2001-2020. A total of 22 thematic layers were generated using ArcGIS Pro 2.5 and Erdas Imagine 2014, covering topographic, hydrological, geological, and anthropogenic variables. Logistic regression modeling was implemented in the R environment. Model performance was evaluated using pseudo-R² indices (McFadden’s, Cox & Snell, and Nagelkerke) along with the Area Under the Receiver Operating Characteristic Curve (AUC) to assess predictive accuracy. Results The logistic regression model showed strong predictive capability with an AUC value of 0.855, indicating excellent model performance. Approximately 8.65% of the catchment area—equivalent to around 280 sq. km—is classified as having high to very high landslide susceptibility. The spatial analysis revealed that susceptibility is greatest in the western and central parts of the catchment, particularly along valley floors, river-adjacent slopes, and human-inhabited areas. A total of 192 villages is identified as being exposed to potential landslide risks, along with vulnerable infrastructure such as roads, agricultural lands, and residential settlements. Conclusions The study successfully maps landslide-susceptible zones using logistic regression and multi-source geospatial data. It provides actionable insights for local authorities, planners, and disaster risk managers. The findings emphasise the need for targeted interventions in highly susceptible areas to reduce hazard exposure and enhance community resilience in the Upper Ravi River Catchment. The methodology presented can be replicated in other mountainous regions facing similar challenges.