BackgroundLandslides, among the most catastrophic natural hazards, result from natural and anthropogenic factors, causing substantial financial losses, infrastructural damage, fatalities, and environmental degradation. Uttarakhand, with its unique topographical and hydrological conditions, unplanned human settlements, and changing precipitation patterns, is highly susceptible to landslides.MethodsThis study evaluates landslide susceptibility for Uttarakhand, a Himalayan state in India, by employing bivariate analysis, multi-criteria decision-making, and advanced machine learning models, such as Random Forest and Extreme Gradient Boosting (XGBoost). A total of sixteen landslide influencing factors were used for performing landslide hazard susceptibility zonation, including the innovative use of geomorphons for detailed terrain analysis.ResultsApproximately 18.47% of the study area was classified as high to very high landslide susceptibility zones, and 21% was classified into the moderate susceptibility category. High to very high susceptibility zones were concentrated in the Uttarkashi, Chamoli, and Pithoragarh districts of the Lesser and Higher Himalayas, areas characterized by rangelands and high annual rainfall. Conversely, very low to low susceptibility zones were predominantly located in the Tarai-Bhabar and Sub-Himalayan districts, including Haridwar and Udham Singh Nagar. The Random Forest and XGBoost models demonstrated superior predictive performance.ConclusionsThe spatially explicit landslide susceptibility maps provide critical insights for urban planners, disaster management agencies, and environmentalists, aiding in developing effective strategies for landslide risk reduction and promoting sustainable development in Uttarakhand. This study exemplifies applying advanced analytical techniques to address landslide susceptibility and related soil erosion and water resource management challenges in Uttarakhand.

The probabilistic methods are receiving increasing recognition in assessing the hazards due to landslides, owing to the ability of these methods to consider the estimation uncertainties and geographical heterogeneity of geomorphological, geotechnical, geological, and seismological components. Therefore, the present study developed a probabilistic method to model the parametric uncertainties of modified Newmark's method using the Monte Carlo simulation technique. The proposed methodology was applied to evaluate the hazard potential for co-seismic landslides in the Uttarakhand state, located in the Indian Himalayan region. The modified Newmark model considered in the study incorporates the rock joint shear strength properties instead of soil shear strength parameters in the permanent displacement computation of slopes. The simulations were done pixel-by-pixel by seamlessly integrating into the current GIS computational settings. When analyzing these data, statistical distributions were used to account for uncertainties and variations in the input parameters. Monte Carlo simulations were employed to generate various probability density functions for each individual pixel within the study area. These simulated distributions were then maintained consistently across the entire computational workflow, ensuring that the generated samples were preserved throughout the analysis. With no limitations on the symmetry or complexity of the underlying distributions, the resultant numbers were then turned into probabilistic hazard maps. In the final step, a hazard map was produced, where each pixel indicates the probability that slope displacement will surpass the 5 cm threshold. Values of probability are distributed between 0.1 and 1, with elevated values primarily observed in the upper Himalayan region, emphasizing the greater likelihood of co-seismic landslides in this zone. This seismic landslide hazard map serves as a valuable tool for local planners and authorities, enabling them to assess regions vulnerable to seismic landslide hazards and implement measures to mitigate potential losses.