Talus slopes are prevalent throughout alpine mountain regions. In southeastern Tibet, along the Bomi-Ranwu section of the G318 National Highway, sand-flow talus (SFT) deposits pose a significant threat to the road. This research explores their development, environmental influences, and formation processes through field surveys, remote sensing, laboratory tests, and spatiotemporal data analysis. The findings show that factors such as precipitation, elevation, the aridity index, and stream order play crucial roles in shaping these slopes. Approximately 72 % of the slopes are found close to first-order streams. Detrital particles primarily originate from the physical weathering of granites, which are primarily made up of mineral particles such as quartz and feldspar, with minimal clay content. The weathered material comes from medium- to coarse-grained granite bedrock and is shaped by dense structural planes in cold-arid valley conditions. Daily temperature fluctuations induce thermal fatigue and freeze-thaw cycles, which primarily facilitate the physical weathering of rock. The development of SFT slopes involves rock-mass collapse, gravity-driven downward movement, and particle transport over long distances via flow zones. Hydraulic erosion processes, such as interrill and rill erosion, primarily drive the movement and accumulation of grain materials. This study offers new insights into the formation mechanisms of SFT slopes in alpine mountain environments and establishes a scientific basis for hazard prevention.

On February 8, 2025, a catastrophic landslide occurred in Jinping Village, Junlian County, Sichuan Province, resulting in severe casualties and significant property losses. Field investigations, laboratory tests, and historical image analysis indicated that the landslide was caused by progressive weakening of the strength of the rock mass under long-term dry–wet cycles. The sliding body consisted of interbedded sandstone and mudstone, forming a typical hard-over-soft lithological structure, with well-developed joints and high permeability. After 5 cycles, particles larger than 40 mm completely disintegrated, whereas after 20 cycles, the fraction of particles smaller than 5 mm increased from 0.035

On February 6, 2024, a catastrophic avalanche at Bukongla Mountain in the Southern Himalayas resulted in five fatalities and widespread destruction, extending well beyond the dense-flow deposition zone. This study investigates the geomorphic controls and dynamic mechanisms of this extreme event through field investigations, remote sensing, and analysis of historical incidents. The analysis indicates that the dynamics of the avalanche were fundamentally governed by a two-level planation surface geomorphology. The upper plateau, with an average slope of approximately 33°, acted as the principal accumulation zone, which then entered a steep, deeply incised acceleration track (approximately 50.1°). This pronounced topographical discontinuity served as the principal mechanism driving the formation of a powerful and destructive airblast. Field evidence confirms that the airblast was the main cause of damage at distal locations. After the dense snow flow ceased on the gentle lower valley floor (approximately 11.2°) at a runout distance of about 400 m, the airblast propagated independently for an additional 110 m, effectively extending the total destructive path by more than 25

Hydrodynamic pressure-driven landslides induced by precipitation and reservoir water-level fluctuations are among the most common types worldwide. Accurate multi-physical field simulations are essential for understanding the mechanisms of evolution and for reliably assessing landslide stability. In this study, a novel, practical, and simplified sequential simulation approach is proposed to model the multi-physical processes of hydrodynamic pressure-driven landslides by integrating finite element and finite difference methods. The proposed approach first employs a finite element model to simulate the time-dependent seepage field within the landslide. Subsequently, a customized interface program is developed to transfer the computed seepage field from the finite element model to the finite difference model. Finally, the finite difference model is used to simulate the resulting stress and deformation fields induced by seepage variations, while incorporating the softening effects of rainfall and reservoir water-level fluctuations on the geotechnical parameters of the landslide mass. The framework is validated using a large-scale physical model of a hydrodynamic pressure-driven landslide, and the sensitivity of key geotechnical parameters to cumulative deformation and landslide initiation time is further analyzed. The results highlight the critical role of water-induced shear strength degradation in controlling deformation behavior and landslide stability.

Intense rainfall is widely recognized as the primary trigger for landslides and is commonly used in early warning systems. Yet, many unexpected failures occur during periods of little or even no rainfall, leading to ineffective warnings and failed evacuations. The mechanisms by which landslides occur without heavy rainfall remain unexplored. Integrating field investigations, machine learning, remote sensing, and numerical simulations, we found that landslide disasters without heavy rainfall are widespread: 75.7% of 1118 catastrophic cases exhibited delayed onset, controlled by antecedent rainfall, the topographic wetness index, and landslide scale. For the first time, we identified three runoff-supply patterns (slope, gully, and creek supply) and two migration stages (surface and subsurface) that together govern strong runoff supply in confluence zones. Our work contributes an innovative perspective on the coupling of early strong runoff and weak geomaterials that triggers delayed failures, with three subsurface runoff stages-interflow, sliding-face flow, and return flow. These results clarify the hydro-geomaterial coupling behind delayed landslides and support improved early warning and prediction to reduce risk.

The Jure landslide that happened in Nepal on August 2, 2014, exemplifies a critical yet understudied type of geohazard: delayed landslides that occur days after peak rainfall. This study investigates the hydro-mechanical triggering mechanism of the Jure landslide by combining field investigations, drone-based aerial surveys, remote sensing interpretation, and coupled seepage-stability numerical modeling. The results reveal that the critical factor in the delayed failure was the confluence area behind the main scarp (1.62 times the landslide’s actual area). This area maintained high groundwater input even after rainfall ceased, leading to progressive saturation and failure 2 days after the peak rainfall. Numerical simulations using five strategic observation points demonstrated that pore water pressure (PWP) increased from −50 kPa to 43.6 kPa within 216 h, reducing the factor of safety (Fs) from 3.5 to 0.80. The landslide ultimately occurred at 194 h with a computed Fs of 0.97, when PWP values at the observation points reached critical levels. The delayed failure mechanism was facilitated by four distinct joint systems in weathered schist bedrock that created preferential flow paths, prolonging the seepage process. This research contributes to the scientific understanding of delayed landslide phenomena by quantifying the relationships among confluence-area hydrology, progressive pore-pressure development, and slope stability deterioration. These findings have significant implications for early warning systems in mountainous regions worldwide, suggesting that monitoring catchment-scale hydrogeological conditions and PWP evolution may be more effective than conventional rainfall-intensity-based systems for predicting delayed landslide failures.

Climate change is fundamentally altering debris flow triggering mechanisms in high-altitude regions, challenging the reliability of traditional early warning systems. We analyzed 214 debris flow sites in Bomi County, southeastern Tibet (1937-2024), integrating field surveys, remote sensing data, and meteorological observations. We evaluated the performance of an Intensity–Duration (I-D) threshold model calibrated on historical data (1950–2000) when applied to recent events (2001–2024), and developed a Temperature-Rainfall dual-parameter model. Our analysis reveals that debris flow frequency increased 7.1-fold from 2001 to 2010 (21 events) to 2011-2020 (150 events). The I-D model showed degraded performance: f false alarm rates increased from 12% to 28% and missed event rates from 8% to 19%. The dual-parameter model demonstrated superior performance (AUC = 0.87 vs. 0.74). We conclude that static, historically calibrated models become increasingly unreliable under warming, necessitating adaptive early warning systems with dynamic threshold adjustment, multi-parameter integration, and explicit uncertainty quantification. Our findings have critical implications for disaster risk management in cryosphere-influenced mountain regions worldwide.

A dam breach is an uncommon but profoundly destructive event that transpires when a dam collapses, releasing accumulated water downstream and leading to extensive damage. This study focuses on the Jure landslide dam, located in the Sindhupalchowk district, Nepal. The region is characterized by complex river channels and steep terrains, which are significantly influenced by flood dynamics. This study aims to establish a compressive numerical simulation of a two-dimensional dam breach unsteady flow hydraulic model to simulate the dam breach process and downstream flood propagation. The study analyzes the dynamics of the Jure landslide dam outburst flood, emphasizing the flood characteristics, inundation, and velocity hazards in the mitigation of flood impacts. The results reveal that the peak discharge of the Jure landside dam was 5336.7 m3/s, while it decreased to 1181.4 m3/s when traveling 35 km. The flood depth obtained by 2D (HEC-RAS) downstream of the dam rages between 0.0334 and 55.9 m, while the corresponding estimated peak flow velocity of simulated breaches was 21.46 m/s, demonstrating extreme hydraulic force conditions, capable of catastrophe. The proposed hydraulic simulations reveal significant variations in overflow dynamics across different terrain types, with narrower sections exhibiting faster flood progression and greater water depths. The findings underscore the necessity of accounting for terrain heterogeneity in future flood risk assessments. This work offers valuable insights into the emergency management of landslide dams in similar regions.

Debris flow events are frequent in Tajikistan, yet comprehensive investigations at the regional scale are limited. This study integrates remote sensing, Geographic Information System, and machine learning techniques to evaluate debris flow susceptibility and associated hazards across Tajikistan. A dataset comprising 405 documented debris flow points and 14 influencing factors, encompassing geological, climatic-hydrological, and anthropogenic variables, was established. Three machine learning algorithms—Random Forest, Support Vector Machine (SVM), and Multi-layer Perceptron—were applied to generate susceptibility maps and delineate debris flow risk zones. The results indicate that the areas of higher and high susceptibility accounted for 20.43% and 4.41% of the national area, respectively, and were predominantly concentrated along the Zeravshan and Vakhsh river basins. Among the evaluated models, SVM model demonstrated the highest predictive performance. Beyond conventional topographic and environmental controls, drought conditions were identified as a critical factor influencing debris flow occurrence within the arid and semi-arid mountainous regions of Tajikistan. These findings provide a scientific basis for regional debris flow risk management and disaster mitigation planning, and offer practical guidance for selecting conditioning factors in machine-learning-based susceptibility assessments in other dry mountainous environments.

Landslides pose a significant threat in the mountainous regions of Nepal. Landslide susceptibility maps are commonly used to identify potential landslide zones by statistically analyzing geological, topographical, and hydrological factors, assuming that similar conditions may trigger future failures. While such maps provide valuable insights into landslide-triggering conditions, they are limited in assessing risk to settlements and infrastructure located downslope or in valley bottoms. This study integrates machine learning based landslide susceptibility with numerical runout modeling to provide a comprehensive landslide hazard assessment in the Bhotekoshi watershed, overcoming the limitations of traditional models that focus solely on statistical susceptibility. To conduct the susceptibility analysis, a total of 439 landslides were mapped from 2012 to 2021 using satellite images. Of these, 70% were used for training two machine learning (ML) models: random forest and Xtreme Gradient Boosting (XGBoost), and the remaining 30% were used for validation. Among the two ML models, Random Forest model demonstrated slightly superior performance, achieving higher predictive accuracy. After the machine learning susceptibility analysis, the study transitions into a regional-scale landslide runout analysis. First, a back analysis of the past landslide event was conducted to fine-tune the model parameters (internal angle of friction and basal friction angle) and validate performance of the runout model. Following the back analysis, the regional-scale numerical modeling of landslide runout was conducted by designating areas classified as the highest susceptibility class in the Random Forest susceptibility map as potential release zones. This approach allows for a detailed examination of landslide propagation and potential impacts along the downslope settlements and infrastructures. The analysis clearly demonstrates that integrating both machine learning and numerical runout methods significantly increases the estimated exposure of population, buildings, and roads within the very high hazard class compared to relying solely on susceptibility methods. Specifically, population exposure rises from 360 to 7743, buildings increase from 97 to 2771, and road exposure expands from 41 to 251 km. This result highlights the significant risk of underestimating exposure in the analyses that solely rely on landslide susceptibility models. Integration of susceptibility and runout analysis improves landslide risk assessment, aiding in land-use planning and disaster mitigation strategies.

Frequent human engineering activities on the Qinghai-Tibet Plateau have resulted in widespread shallow landslides that pose serious threats to infrastructure and safety. This study investigates the distribution and failure mechanism of shallow landslides in Jiacha County, taking the Zhaxuecun landslide, which occurred on October 24, 2022, as a representative case. The landslide, with a volume of approximately 6.27 × 10⁴ m³, is a typical shallow landslide in the middle reaches of the Yarlung Zangbo River. Through field investigations, in-situ tests, laboratory experiments, and numerical simulations, the material composition, structural characteristics, degradation of mechanical properties, and failure mechanism were systematically analyzed. Historical deformation identified from Google Earth Pro imagery indicated that slope instability began after highway excavation. In-situ permeability tests showed a high permeability coefficient (297.06 m/day) within the landslide. Direct shear tests demonstrated a marked decrease in the shear strength of gravelly soil under saturated conditions, and rebound tests indicated weak anisotropy between bedding and vertical planes. The long-term effects of water and freeze-thaw cycles promoted the formation of fissures and infiltration channels, accelerating slope failure. Numerical simulations revealed the evolution of velocity and energy during sliding. Combined with meteorological data, it is inferred that the deformation of the Zhaxuecun landslide was jointly influenced by excavation, freeze-thaw cycles, and snowmelt. The landslide evolution can be divided into four stages: excavation-induced deformation, long-term freeze-thaw degradation, critical stability formation, and sliding failure. These findings provide insights into the deformation mechanisms of shallow landslides along the Yarlung Zangbo River.

The expansion of potentially hazardous glacial lakes is a symptom of global warming during this interglacial period. A pertinent example is the Jiongpu glacial lake in southeastern Tibet, the area of which has expanded approximately fivefold in the last half-century. However, recently, the glacier tongue has retreated to a high steep slope, and the rate of retreat of the glacier and expansion of the lake have temporarily slowed. The risk of a glacier tongue landslide after glacier detachment and subsequent glacial lake outburst flood (GLOF) needs to be assessed. In this study, we employed a combination of unmanned aerial vehicle (UAV), sonar, geological radar, remote sensing, field investigation, sampling, drilling, and dating techniques to determine the critical parameters of potential GLOFs, including glacier tongue geometry, lake bathymetry, and moraine dam geometry and composition. Utilizing empirical models and multiphase flow models, we identified the most hazardous triggers and simulated the processes of a glacier tongue landslide into the lake, moraine dam overtopping by a displacement wave, and subsequent flood evolution. The results showed that the most hazardous trigger in volume is a glacier tongue landslide, accounting for 58.29 % of all triggers associated with potential GLOFs. Lapped by the largest glacier tongue landslide impulse wave, the moraine dam would not fail because the minimum safety factor is approximately 1.66 +/- 0.7 according to empirical methods and geological slope simulation. However, overtopping would occur, resulting in a peak discharge of approximately 9740 +/- 4137 m3/ s at the moraine dam based on r.avaflow calculations. The flood would reach the densely populated Jinling township and inundate approximately 46 +/- 4.55 % of the houses according to HEC-RAS. Reducing the water level of the glacial lake represents an effective strategy for mitigating potential losses. This concise, physics-based method effectively assesses GLOF triggers and processes and can be applied to risk assessments of other expanding glacial lakes.

Landslides often occur unexpectedly in unexpected locations and can cause significant human casualties and property losses. Although landslides are less common in winter, a catastrophic landslide occurred in Liangshui Village, Zhenxiong County, Yunnan Province, China, on January 22, 2024, at 5:51 am, resulting in the burial of 18 houses and the death of 44 people. In this study, the basic characteristics, deformation and failure mechanisms, and motion processes of the landslide were analyzed based on field investigations, drone aerial surveys, laboratory tests, and numerical simulations. The results indicate that the Liangshuicun landslide is a smaller-scale landslide within the category of medium-sized landslides, with a total volume of approximately 119,000 m3. The rear part of the landslide formed a catchment area, where groundwater infiltrated directly into the lower part of the landslide provenance zone. The lower part of the provenance zone, consisting of mudstone acting as an aquitard, facilitated the accumulation of groundwater, creating a water-rich zone that provided favorable conditions for the development of the landslide. The sliding mass is composed of siltstone and mudstone interbeds of Feixianguan Formation of Triassic. The strength along the bedding planes is low, and the strength deteriorates significantly after being subjected to water action. The lithological conditions in the area provided favorable conditions for the development of the landslide. The occurrence of the landslide coincided with the phase of nocturnal temperature decline when the provenance zone might have experienced the greatest frost heave forces. The low-temperature in freeze–thaw process provided conditions for the shear failure of the provenance zone. During the initiation of the landslide, the superficial rock mass in the provenance zone and the scraping sliding zone exhibited much higher sliding velocities compared to the lower regions, indicating a slower material initiation, higher sliding velocity, and shorter duration in the rear part, while the front edge experienced rapid initiation and a longer duration of movement. The findings of this study provide insights into the evolutionary mechanisms of landslides occurring in winter.

On May 1, 2024, a small embankment collapse occurred in the early hours of the morning on the Meida Highway in Meizhou City, Guangdong Province, resulting in 48 fatalities. The small-scale collapse caused massive casualties and garnered widespread attention. In detail, there is a significant lack of precipitation at the time of the “5·1” Meida collapse disaster, lagging 10 h behind the peak precipitation. The collapse occurs on a mountainous slope, with a hollow catchment area located above the embankment. Multiple potential streams converge in the area, contributing to the water flow towards the slope. Within the western zone of the Lianhua Mountain fault, the collapse area is crossed by fault lines at approximately 800 m on the upper side and 650 m on the lower side. Bedrock fractures formed by faults act as water conduits. The combination of catchment topography and potential faults enriches the water around the embankment slope, contributing to its instability. The disaster site is situated within granite formations. The refilling soil, composed of weathered granite, exhibits poor hydro-mechanical properties, making the slope particularly susceptible to failure due to the effects of multi-source water infiltration. A key insight from this research is that potentially unstable embankment slopes should be identified by considering the interaction between multi-source water and soil/rock. Greater emphasis should be placed on factors such as fault development and hollow topography above the slope, which influence the effects of multi-source water. These factors should be quantified in future studies to improve the assessment of unstable highway slopes in mountainous regions. The findings and strategies outlined in this study can serve as a valuable reference for assessing both embankment and natural slopes in mountainous areas.

Heavy and intense rainfall commonly triggers debris flow events, but the relationship between gentle rainfall, snowmelt water, and freeze–thaw erosion in high-altitude cold regions of northern Pakistan remains unexplored. A devastating debris flow from Bicharh Nallah hit the village of Sherqilla in northern Pakistan on July 5, 2022, and killed 8 people and destroyed 10 houses completely and 250 houses partially. The Bicharh Nallah debris flow (Sherqilla village) in Ghizer district is used as a case study to find out what makes this size debris flow happen in a cold region and how it is affected by rainfall, snowmelt water, and freeze–thaw erosion. This work was done through fieldwork investigation, laboratory work, and statistical analysis. The debris flow dynamic characteristics such as peak discharge, velocity, and density were calculated as 430.82 m3/s, 5.11 m/s, and 1.74 g/cm3, respectively. We determined that rainfall acts as a direct triggering factor. Snowmelt water was quantitatively calculated, which contributed to triggering the debris flow. This study revealed that gentle rainfall and snowmelt water are direct triggering factors in the debris flow outbreak in Bicharh Nallah of Sherqilla village. Long-duration, extreme, and severe freeze–thaw erosion provided enough loose soil conditions for the triggering and formation of debris flow. The regional tectonics, geology, and local topography indirectly contributed to the development and amplification of Bicharh Nallah debris flow.

On May 1, 2024, a small embankment collapse occurred in the early hours of the morning on the Meida Highway in Meizhou City, Guangdong Province, resulting in 48 fatalities. The small-scale collapse caused massive casualties and garnered widespread attention. In detail, there is a significant lack of precipitation at the time of the "51" Meida collapse disaster, lagging 10 h behind the peak precipitation. The collapse occurs on a mountainous slope, with a hollow catchment area located above the embankment. Multiple potential streams converge in the area, contributing to the water flow towards the slope. Within the western zone of the Lianhua Mountain fault, the collapse area is crossed by fault lines at approximately 800 m on the upper side and 650 m on the lower side. Bedrock fractures formed by faults act as water conduits. The combination of catchment topography and potential faults enriches the water around the embankment slope, contributing to its instability. The disaster site is situated within granite formations. The refilling soil, composed of weathered granite, exhibits poor hydro-mechanical properties, making the slope particularly susceptible to failure due to the effects of multi-source water infiltration. A key insight from this research is that potentially unstable embankment slopes should be identified by considering the interaction between multi-source water and soil/rock. Greater emphasis should be placed on factors such as fault development and hollow topography above the slope, which influence the effects of multi-source water. These factors should be quantified in future studies to improve the assessment of unstable highway slopes in mountainous regions. The findings and strategies outlined in this study can serve as a valuable reference for assessing both embankment and natural slopes in mountainous areas.

Northern Pakistan is a rough, mountainous region with high gradients, disintegrated lithology, glaciers on the highest peaks, and a seismically active area. District Astor is among the most susceptible locations, with yearly landslides due to various causes. This research has developed a comprehensive landslide inventory and a susceptibility model for the chosen region. Frequency Ratio is the most generally utilized probabilistic method; a moderately Analytical Hierarchy Process (AHP). The Frequency Ratio (FR) model technique has been used to ascertain the connection between both variables that cause landslides and landslides that have been mapped. A persistent Scattered Interferometry Radar (InSAR) technique was employed to investigate deformation movement in the vulnerable zones of the extracted models, finding a high Line-of-Sight (LOS) displacement velocity in both models' extremely sensitive areas. The derived Landslide Susceptibility Index (LSI) models had a prediction accuracy of 84.4% and 78.0% for the FR and AHP methods, respectively, calculated by applying the Area Under Curve (AUC) derived from the Receiver Operating Characteristic (ROC) approach. Finally, five susceptibility classes were assigned to both Landslide hazard index maps. Because the research region is prone to landslides, these susceptible models will be useful in delineating hazardous zones for future landslide catastrophes and utilized in decision-makers' planning strategies for Development initiatives in the studied region.

The challenge of obtaining landslide susceptibility zoning in Tibet is compounded by the high altitude, extensive range, and difficult exploration of the region. To address this issue, a novel evaluation approach based on Stacking ensemble machine learning is proposed. This study focuses on Jiacha County, adopts the slope unit as the evaluation unit, and picks up 14 evaluation factors that symbolize the topography and geomorphology, environmental and hydrological features, and basic geological features. These landslide conditioning factors were integrated into a total of 4660 Stacking ensemble learning models, randomly combined by 10 base-algorithms, including AdaBoost, Decision Tree (DT), Gradient Boosting Decision Tree (GBDT), k-Nearest Neighbors (kNNs), LightGBM, Multilayer Perceptron (MLP), Random Forest (RF), Ridge Regression, Support Vector Machine (SVM), and XGBoost. All models were trained, using the natural discontinuity method to classify landslide susceptibility, and the AUC value, the area under the ROC curve, was taken to evaluate the model. The results show that the maximum AUC values in the 9 models performing better reach 0.78 and 0.99 over the test set and the train set. Most of the areas identified as high susceptibility and above show consistency with the interpretation of the existing geological field data. Thus, the Stacking ensemble method is applicable to the landslide susceptibility situation in Jiacha County, Tibet, and can provide theoretical support for disaster prevention and mitigation work in the Qinghai–Tibet Plateau area.

Study region: The Zeketai River Basin of Ili Region, China Study focus: In seasonally frozen regions, landslides induced by snowmelt have increasingly become a prevalent disaster phenomenon. However, the specific effects of snowmelt on landslide dynamics remain not well understood. This paper selects the Zeketai River Basin in the Tian Shan as study area, aiming to explore the deformation processes and failure mechanisms of snowmelt-induced landslides. New hydrological insights for the region: A total of 242 loess landslides were identified in the Zeketai River Basin of the Ili Region, China. 61 % of landslides are scattered across northward-facing slopes, and 70 % of the recorded landslide events (40 occurrences) transpired during spring between 1990 and 2010. The deformation and failure of these loess slopes generally manifest in late autumn (October to December) and early spring (February to April). During the snowmelt season, surface deformation exhibits a stepwise increase at a rate of approximately 120 mm/yr. Snow accumulation and melting processes are identified as the paramount driving factors of landslide dynamics. Daily and monthly surface displacements demonstrate moderate and strong positive correlations with atmospheric temperature in early spring (r = 0.72 and 0.69, respectively) and late autumn (r = 0.47 and 0.64, respectively). Furthermore, monthly snow depth shows strong negative correlations with monthly surface displacements and atmospheric temperature in early spring (r = -0.85 and −0.82, respectively) and late autumn (r = -0.7 and −0.86, respectively). Both temperature-dependent snow accumulation and melt processes, in conjunction with accompanying freeze-thaw cycles, potentially exert a combined influence on the types and mechanisms of landslides throughout different seasons. Future investigations will emphasize near-surface and subsurface thermo-hydro-related measurements and experiments to elucidate the intricate interaction mechanisms of these landslides.