Landslides present remarkable hazards in the Himalayan region, particularly in areas with young and fragile topography. Mitigating vulnerability requires assessing susceptibility, which relies heavily on the accuracy of susceptibility maps generated through various approaches that consider different conditioning factors at various resolutions. This study, conducted in Jajarkot District within the Karnali Province of Nepal and covering 2230 km2, aims to identify suitable conditioning factors at appropriate resolutions. Sixteen factors, encompassing topography, hydrology, geology, and anthropogenic activities, were analyzed alongside a landslide inventory of 159 occurrences compiled from satellite imagery, the literature, and field surveys. A genetic algorithm (GA) was employed to determine the optimal set of conditioning factors, while Maximum Entropy (Maxent) modeling produced landslide susceptibility maps (LSM) at spatial resolutions ranging between 12.5 and 200 m. Resolution selection was guided by Receiver Operating Characteristic (ROC) curve and Area Under the Curve (AUC) analyses. Multicollinearity testing identified 15 influential factors, with land use ranking highest at 22.7%, followed by stream power index (SPI), drainage density, and aspect. The GA consistently highlighted land use and slope as effective factors across subset sizes. The results indicated resolutions finer than one hundred meters enhanced discrimination between landslide and non-landslide areas, emphasizing the need to balance resolution with computational resources and data availability. This study emphasizes the intricate interplay of conditioning factors, the GA’s efficacy in subset selection, and the crucial role of resolution in the improvement of susceptibility models. The findings provide practical insights for policymakers and disaster management authorities, aiding evidence-based decision making in the mitigation of landslide risk in Jajarkot and similar regions.

Water Poverty Index (WPI) is an effective tool to assess water stress by linking physical approximations of water availability with socioeconomic drivers of poverty at household and community levels. It also helps in determining the priority needs for interventions and makes the better use of water resources without compromising the environmental integrity at the policy level. This study presents a comparative analysis of WPI to assess the water stress in two local units i.e., Mahakali Municipality-4 (Upstream) and Bhimdatt Municipality-13 (Downstream) of the Darchula and Kanchanpur districts, respectively along the Mahakali River Basin (MRB), Sudurpaschim Province, Nepal. Five components, i.e., resource availability, use of resources, capacity to utilize water sources, accessibility to water sources, and environment were applied to appraisal the water poverty. Under these components, 28 indicators were selected to visualize the water poverty condition. The information for acquiring these indicators was collected by making the field visit, semi-structured questionnaire survey, and secondary information from relevant sources. The results showed that the downstream segment has a higher WPI value (65.43) in comparison to the upstream segment (52.60) of the MRB, indicating relatively water-advantaged and stressed situations in down- and up-streams, respectively. Interestingly, despite having higher resource availability, the upstream region exhibited more stressed situations reflecting spatial heterogeneity in terms of capacity, environmental conditions, and infrastructure development to use the available water resources. Thus, in order to reduce the water stress conditions, priorities for intervention must be given to the water use capacity, especially in the upstream segment of the MRB. This study could be useful for academia and policymakers for the sustainable use of water resources in the MRB, Susurpaschim Province, Nepal.

An attempt has been made in this research to assess soil erosion and its spatial distribution by the revised universal soil loss equation (RUSLE) model at Rangun Khola Watershed, western Nepal. The erosion factors were acquired from multiple sources (Satellite images, ALOSPALSAR DEM, SOTER soil database, Esri 2020 land cover map, rainfall database of DHM) and an integrated analysis was carried out in raster format of GIS. A landslide inventory was generated on the basis of satellite images and past literature to validate soil erosion intensity in the area. The result map of the RUSLE model was categorized into six levels based on the erosion severity, and 9.06% of the area was found to be under extremely severe soil erosion risk (> 80 ton ha^− 1year^− 1) indicating urgent consequences. The frequency ratios for each level of potential erosion susceptible to landside exhibited a linear relationship depicting reasonable and satisfactory level of agreement between the landslide event/location data and the erosion map that validates the model result. The result of this study will be helpful to detect the sensitive zones presenting a priority of protection and offer valuable information that aids decision-makers and user agencies in creating adequate conservation planning programs to stop soil erosion and maintain the natural balance.

Landslides pose significant threats to biodiversity, life, and infrastructure in mountainous regions, making susceptibility mapping essential for disaster risk reduction. This study presents a comprehensive multi-model comparison of statistical, machine learning, deep learning, and ensemble approaches for landslide susceptibility mapping in the eastern Chure region of Nepal, characterized by fragile Sub-Himalayan lithology, intense monsoon precipitation, and expanding infrastructure. Ten models were evaluated: Frequency Ratio (FR), Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), XGBoost, LightGBM, Artificial Neural Network (ANN), Deep Neural Network (DNN), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM). Twelve conditioning factors were analyzed, including soil type, topographic wetness index, elevation, slope, stream power index, aspect, lithology, distance to roads, land use, distance to rivers, and curvature. The dataset included 123 training and 53 testing landslide locations with balanced non-landslide samples. Random Forest showed the best performance (Test AUC = 0.928; Accuracy = 0.838), followed by XGBoost (Test AUC = 0.881) and LightGBM (Test AUC = 0.844). Deep learning models showed lower performance (ANN ≈ 0.773; CNN ≈ 0.766; DNN ≈ 0.765), while LSTM produced the weakest result (Test AUC = 0.622). The Top 3 ensemble (RF, XGBoost, LightGBM) produced coherent spatial patterns and strong class separability. Classification of ensemble outputs into five susceptibility zones indicated that 22.7% of the watershed falls within High and Very High susceptibility classes. These results demonstrate that ensemble machine learning approaches provide robust landslide susceptibility assessments and can support land-use planning and disaster risk management in landslide-prone regions.