Micropile–anchor composite structures (MACSs) are effective in slope stabilisation by combining the active resistance of anchor cables with the passive reinforcement of micropiles. However, the reinforcement mechanism of MACS under prototype conditions remains unclear, as existing studies have focused on small-scale 1-g model tests. This study conducted three centrifuge model tests on soil slopes reinforced with double-row micropiles and double/single-row MACS, respectively, based on a highway slope in Shandong, China. The development of earth pressure, pile bending moment, anchor axial force, and slope deformation was systematically analysed, and crack propagation and failure characteristics were examined. The results demonstrate that MACS significantly improve slope performance by reducing earth pressure and restricting deformation. Lateral earth pressure is redistributed between front and rear micropile rows, and load sharing is affected by reinforcement configuration. Anchor cables enhance the plastic deformation capacity of slopes and inhibit surface crack growth, while double-row micropiles obstruct sliding surface coalescence. Furthermore, micropiles row influence slope deformation at the failure stage, with double-row MACS-reinforced slopes remaining stable with minor surface cracks, whereas single-row MACS-reinforced slopes fail as crushed blocks along a circular sliding surface initiated by main cracks. These findings contribute to understanding the reinforcement mechanism of MACS.

Geophysical granular flows, involving rapidly flowing granular materials, can exhibit volume-enhanced mobility. Lacking a mechanistic understanding of such size effects limits the applications of lab-scale findings to natural events. Using discrete element method simulations, we find that increasing granular system size suppresses energy-dissipating velocity fluctuations while promoting sustained creeping motion. This nonlocal phenomenon of granular materials enhances the mobility in various granular column collapse scenarios. This mechanism is reflected in the rheological data, which deviate from traditional rheology but are collapsed under a recent power-law rheology that incorporates velocity fluctuations. Moreover, this size-dependent power-law rheology exhibits universality in transient simulations with varied flow geometries, slope angles, and base roughness. This rheologically consistent framework, spanning inertial to quasi-static states, bridges small-scale investigations and continuum models for large-scale simulations, enabling improved predictive capability of the entire flow processes, from initiation to deposition, in natural geophysical flows.

Bedding rock slopes are prone to large-scale landslides subject to earthquakes, and their stability evaluation and prevention are key research focuses in the field of geological hazards. Taking a soft-hard interbedded bedding rock slope as the prototype, this study proposes a novel anchor with energy-dissipating and self-centering (ED&SC) functions. Large-scale shaking table tests are conducted to study dynamic responses of the bedding rock slope reinforced by two types of pile-anchor composite structures (PACS). It is found that the dynamic responses of the bedding rock slope exhibit obvious spatial differences affected by lithologic distribution, seismic intensity, waveform characteristics and excitation mode. The acceleration amplification factor (AAF) increases stepwise along multiple slope directions and evolves following a “decrease-increase- decrease” pattern within 0.050 g - 0.600 g. The AAF under the excitation of Wenchuan wave and bidirectional wave is obviously larger than those under Ludian wave and unidirectional wave. The slope exhibits progressive failure characteristics, sequentially undergoing a compaction stage of the lower soft rock layer, a tensile crack development stage of the middle hard rock layer, and a formation stage of through-going shear sliding zones along upper lithologic interfaces. Compared with conventional anchors, the ED&SC anchor improves the seismic performance of PACS by reducing the anchor axial force and rock lateral pressure. Graded energy dissipation of ED&SC anchor facilitates load redistribution, and enhances the cooperative bearing capacity and overall seismic resistance of the system. The findings provide technical support for the seismic design of bedding rock slopes in high-seismic-intensity regions.

ObjectiveIn situations involving mountain floods, intense rainfall-induced erosion, and the breaching of clusters of landslide dams that lead to upstream soil and water loss, the inflow to landslide dams often becomes a sand-laden flow. Compared to clear water inflow, sand-laden flows produce more complex infiltration, erosion, and deposition processes, substantially influencing both the dam breaching process and the peak flow rates. This study investigates the mechanisms through which the sediment concentration and particle size of sand-laden flows influence dam breaching processes and examines the erosion-deposition dynamics and seepage-clogging phenomena induced by sand-laden flows during landslide dam breaching.MethodsFive sets of model experiments were conducted using a flume measuring 5 m in length and 0.4 m in width. Clear water inflow served as the control group, while the effects of different volumetric concentrations of sand-laden flow (0.01 and 0.03) and various maximum sediment sizes (0.006 5, 0.125 0, and 0.500 0 mm) on the dam breaching process were investigated. The primary method involved pre-experimental testing to adjust the composition of the sand-laden water flow, supplemented by thorough mixing in the water supply tank and the flow stabilization tank to achieve a stable and target sand concentration in the dam body area. The design of the landslide dam model was based on the range of dimensionless parameters related to dam height, dam volume, and reservoir volume derived from a database of landslide dam cases, and the material selection referenced empirical data from the 2008 Tangjiashan landslide dam. The experiments primarily used water level gauges to record the breach outflow process, while high-resolution cameras and grid systems monitored changes in the erosion rates of the landslide dams. Timed quantitative sampling methods tracked variations in sediment concentration. In addition, sampling, drying, and sieving techniques collected data on the moisture content and grain size distribution of the residual dam. The main steps of the experiment included material preparation, placement of the movable bed, layered construction of the dam body, execution of model tests, recording of experimental data, and subsequent sample analysis.Results and DiscussionsThe failure mode of the landslide dams in this series of experiments was consistently classified as overtopping failure, and the breaching process was primarily characterized by layered erosion accompanied by slope instability. The breach evolution process was divided into two stages: the initiation stage and the development stage. During the breach initiation stage, headward erosion predominated, and the dam height showed little noticeable reduction. The breach development stage was characterized by a rapid decrease in dam height, and peak flow rates were reached early in this stage. The water level in front of the dam reached its peak at the end of the breach initiation stage, whereas the peak flow rate occurred early in the breach development stage, and a secondary peak developed due to upstream slope instability that partially blocked the breach. The concentration of sand-laden flow in the dam area was inversely correlated with the breach flow rate, whereas the concentration measured at the end of the flume area showed a positive correlation. The mean particle size of the residual dam decreased longitudinally within the dam area and remained relatively stable in the downstream movable bed area. Comparative tests of sand-laden flows with different volumetric concentrations (Tests 1~3) showed that as the initial volumetric sand concentration increased from 0 to 0.03, the peak breach flow rate increased from 3.2 to 3.6 L/s, the time to peak increased from 76 to 90 s, the maximum water level in front of the dam increased from 24.6 to 26.9 cm, the residual dam height increased from 6.5 to 8.2 cm, and the final average breach width increased from 18.8 to 20.6 cm. The analysis indicated that higher sediment concentrations in the flow enhanced the erosion capacity during the breach development stage; however, the depositional layer formed during the process influenced the downward erosion rate during both the breach initiation stage and the late breach development stage. Comparative tests of sand-laden flows with different maximum particle sizes showed that as the maximum particle size increased from 0.006 5 to 0.500 0 mm, the peak breach flow rate slightly increased from 3.3 to 3.4 L/s, the time to peak increased from 79 to 90 s, and the maximum water level in front of the dam increased from 26.0 to 26.5 cm. The residual dam height remained relatively constant at approximately 7.5 cm, and the final average breach width also showed minimal variation at approximately 19.5 cm. The analysis indicated that the particle size of the sand-laden flow had minimal influence on the erosion rate of the landslide dam, mainly because the effect of particle size on the viscosity of the flow became negligible when particle sizes exceeded 0.1 mm. The comparative analysis of clear water inflow (Test 1) and sand-laden flows (Tests 2~5) showed that during the breach initiation stage and the late stage of breach development, low-velocity and high-concentration sand-laden flows generated strong depositional effects. These effects caused a slight reduction in erosion rates, an increase in the maximum water level in front of the dam, and a greater residual dam height. In contrast, during the early to middle stages of breach development, high-velocity sand-laden flows demonstrated greater erosive capacity and reduced deposition, which was reflected in increased erosion rates and higher peak outflow. In addition, sand-laden flows during the dam breaching process produced a pronounced clogging effect, which resulted in a multilayered structure within the dam body composed of depositional layers, retention layers, and original layers. The dense structure of the retention layer reduced the permeability of the sand-laden flow, lowered the position of the saturation zone on the breach slopes, and increased the scale of slope failures.ConclusionsThe experimental results indicate that, compared to clear water, the viscosity coefficient of sand-laden flow is higher, which enhances its erosive capability and also intensifies deposition. As the concentration increases, the viscosity of the sand-laden flow increases markedly, resulting in higher erosion rates, larger breach sizes, and greater peak outflow rates. As the particle size increases, the viscosity coefficient of the sand-laden flow exhibits negligible variation, leading to only minor changes in erosion rates and peak outflow rates. The seepage-clogging effect refers to the accumulation of sand particles within the surface layer of the dam body during the infiltration of sand-laden water flow, accompanied by the retention of these particles inside the dam. This retention layer decreases the infiltration rate, which lowers the saturated zone along the side slope of the breach. This change results in an increased scale of collapses on the side slope.

Ice-rich permafrost slopes on the Qinghai-Tibet Plateau (QTP) are becoming increasingly unstable under climatic warming. However, the instability mechanisms of permafrost slope associated with ground-ice melting remain insufficiently understood. This study integrates remote sensing, field investigations, and large-scale physical modeling to systematically elucidate the failure mechanisms of these slopes. A regional inventory of 1298 landslides shows a marked surge in slope instability in 2016, coinciding with record-high air temperatures. Field investigations further indicate that ice-rich layers are widely developed at the base of the active layer within failed slopes, exerting a critical control on slope stability. The physical experiment reproduced a three-stage retrogressive failure sequence-toe deformation, translational sliding, and headwall retreat-demonstrating that active-layer detachment (ALD) and retrogressive thaw slump (RTS) represent successive stages of a single interface-controlled failure process. The ice-rich layer acts both as a thermal buffer that delays heat penetration and as a hydrological barrier that promotes interfacial saturation and perched-water accumulation at the slope toe. Under thermal disturbance, phase change, meltwater redistribution, and interfacial weakening jointly initiate sliding and drive subsequent retrogressive collapse. Field evidence from ALD and RTS cases supports the experimental observations. The study therefore provides a multi-scale, process-based framework for understanding, monitoring, and mitigating thaw-driven slope instability in rapidly warming permafrost terrain.

Surface-wave (SW) and electrical resistivity tomography (ERT) surveys form a widely used and complementary geophysical pair for geological condition characterization, owing to their compatible acquisition geometries and contrasting sensitivity to stiffness and fluid–electrical pathways. However, the two methods respond to different physical properties and often produce inversion results that are difficult to reconcile, especially in heterogeneous soils and fractured rock masses. To obtain a more reliable characterization of shallow subsurface materials, we integrate SW and ERT results within a probabilistic modeling framework that quantifies and reconciles method-specific uncertainties. ERT data are inverted using ensemble Kalman inversion (EKI) with level-set parameterization to derive zoned resistivity structures and posterior uncertainty. Surface-wave dispersion curves are inverted using competitive particle swarm optimization (CPSO) to generate inversion ensembles of shear-wave velocity. Both inversions are transformed into voxel-wise probabilistic fields and integrated through an entropy-based weighting strategy that accounts for their relative reliability. Applied to the Enziping #2 landslide, the integrated probabilistic model delineates three principal material domains, including gravelly clay, weathered limestone, and a nodular limestone block. The block exhibits a low-resistivity but high-velocity signature that appears contradictory when ERT and SW are interpreted separately; the integrated result resolves this mismatch and identifies it as a relatively intact limestone fragment within the slide mass. This material configuration indicates that past slope movement occurred through a staged process involving block detachment and reassembly rather than a single translational slip surface. The embedded limestone block also modifies the internal mechanical and seepage structure of the slope, potentially creating stress concentrations or local deformation barriers that influence failure mode and effective strength parameters. Overall, this probabilistic modeling framework provides a quantitative linkage between geophysical responses and geological media, transforming independent single-method inversions into a unified probabilistic representation and advancing uncertainty-aware characterization in engineering geophysics and engineering geology.

China's southwestern region is characterised by active geological structures and frequent seismic activities. These conditions frequently trigger landslides that obstruct rivers and form landslide dams. The failure of these dams represents a significant threat to downstream populations, as exemplified by the breach of the Tangjiashan landslide dam during the Wenchuan earthquake. This study focuses on the overtopping breach type of landslide dams. It conducts large-scale experiments to investigate the impacts of various grading materials on the breach process. Furthermore, the DABA (Dam Breach Analysis) numerical simulation model is employed to conduct an in-depth analysis of the breach process in landslide dams. Based on the experimental and simulated results, the following characteristics of landslide dam breaches were analyzed. Under identical inflowing conditions, the peak discharge of the fine-grained dam is 1.6 times that of the widely-graded dam, which has a significantly higher susceptibility to breaching. Although both dams failed due to overtopping, their erosion mechanisms differed substantially. For the widely graded dam, coarse particles led to scouring and retrogressive erosion, significantly prolonging the breaching process. In contrast, the fine-grained dam primarily failed through layered scouring. The results show a high degree of consistency between the DABA numerical simulation outcomes and the large-scale experiment data, thus validating the model's reliability. The parameter sensitivity analysis revealed that breach development duration and peak discharge were significantly influenced by dam height, dam crest width, and initial water level. Scientific simulation models can more precisely predict the breach time and impact range of landslide dams, aiding in the development of more effective prevention methods.

Landslides are complex geological hazards that require monitoring and predictive models capable of both high accuracy and physical interpretability. Traditional data-driven methods often treat monitoring sensors as isolated points, ignoring the spatial interactions that drive slope failure. This study proposes a domain knowledge-constrained dynamic spatio-temporal graph convolutional network (KD-STGCN) to monitor and predict slope displacement by explicitly modeling these spatial dependencies. The framework integrates geotechnical prior knowledge, specifically regarding traction, types of failure mechanisms and rainfall-induced lags, into a dynamic graph structure, ensuring the model adheres to physical laws. Application to a real-world bedding, controlled landslide reveals that the model successfully captures the progressive failure process. Analysis of the learned graph connections identifies a consistent 12-21-h time lag in the transmission of deformation from the slope toe to the mid-slope. Additionally, the model detects a rainfall-intensity-dependent response, where heavy rainfall triggers a rapid reaction within 12 h, while minor rainfall leads to a delayed response exceeding 48 h. A signif icant increase in the betweenness centrality of mid-slope nodes is also observed prior to acceleration, providing a quantifiable indicator of stress redistribution. Evaluation using data from 12 GNSS stations demonstrates that KD-STGCN achieves high predictive precision, with a Root Mean Square Error (RMSE) of 0.355 mm and an R2 of 0.996. The model maintains robust performance for prediction horizons ranging from 15 minutes to 12 h, with errors at the critical slope toe remaining below 1 mm. The one-year validation further shows that the framework remains effective under seasonally varying conditions. These results confirm that integrating domain knowledge with deep learning not only improves prediction accuracy but also provides interpretable insights into deformation patterns, offering a practical and reliable tool for landslide monitoring.

With the increasing frequency of extreme weather events, port infrastructure faces higher risks from vessel collisions. This study develops scaled physical models of segmented approach bridges and wharf structures, using a comprehensive monitoring system to capture key parameters like acceleration, displacement, pile strain, and pore-water pressure. The dynamic responses of the two structures under varying collision intensities were analyzed. The collision process of the approach bridge was divided into four stages: initial impact, decoupling separation, secondary impact, and energy dissipation. Due to insufficient stiffness, the approach bridge may experience secondary impacts during rebound, increasing its vulnerability. In contrast, the wharf structure, constrained by embankment soil, shows a shorter collision duration with no significant secondary impact, demonstrating higher resilience. The study also observes transient stiffness enhancement in the soil behind the piles due to pore-water pressure. Additionally, stress concentration during impact energy transfer is highlighted by the combined responses of acceleration and pile strain. Finally, the study suggests using momentum as a performance evaluation indicator for collision-resistant design and risk management of port infrastructure.

Under global warming, increasingly severe freeze-thaw cycles in cold regions trigger frequent rockslides, posing significant threats to infrastructure safety. This paper presents an engineering-oriented thermal-mechanicaldamage coupled smoothed particle hydrodynamics (SPH) method to simulate the progressive failure of freezethaw-induced rockslides, prioritizing practical applications over theoretical complexity. The freeze-thaw evolution within fractured rock masses is analyzed by integrating phase transition with porous media heat transfer. Frost heave pressure and thermal stress are incorporated into the Cauchy stress equation, effectively coupling the frost heave and thaw settlement between rock fractures and the intact matrix. The proposed method is rigorously validated and applied to investigate the failure mechanisms of jointed rock slopes on the Qinghai-Tibet Plateau under various engineering conditions: number of freeze-thaw cycles, temperature range, joint dip angles, frost heave coefficient (beta f), and elastic modulus (E). Results demonstrate that the improved SPH framework successfully captures the complete failure process, including thermal thaw weakening, frost-heave cracking, icewedge propagation, sliding-toppling failure, and large deformation. The joint dip angle, temperature amplitude, freeze-thaw cycling mode, beta f, and E collectively control the failure mode transition, instability timing, damage extent, and sliding magnitude. Notably, rectangular freeze-thaw cycles induce the most severe damage, while the synergistic effect of beta f and E shows a linear correlation with slope displacement and damage intensity. This work overcomes key limitations of traditional numerical methods in simulating freeze-thaw rockslide mechanisms, providing a reliable and practical tool for geohazard risk assessment and disaster mitigation in critical cold-region engineering projects.

Rock-ice avalanches and debris flows transport substantial amounts of fresh debris-ice mixture from high-elevation cold regions to warmer lowlands. These fresh deposits undergo pronounced settlement upon ice melt, driving landscape evolution and potential secondary hazards. Yet the evolution and underlying mechanisms of thaw-induced deformation are known. This study aims to (1) elucidate the thaw-settlement process in fresh debris-ice mixtures, (2) reveal the governing mechanisms, and (3) develop a model to predict thaw-induced deformation. The Sedongpu hazard in 2018 in the Yarlung Tsangpo Grand Canyon, Southeastern Tibet, was analyzed as a benchmark via remote sensing, field investigations, and laboratory tests. Results show that (150 ± 6.3) × 106 m3 of sediments detached from the 9° valley floor. The mass flow fan blocked the river, forming a barrier dam 55 × 106 m3 in volume. The average deposit thickness was 70 m. The primary settlement of the deposit was completed within two years, with a vertical strain of 32.5%. The settlement mechanism differs fundamentally from that of ice-cemented soils. The ice in fresh deposits exists as discrete interstitial fragments rather than bonding soil particles, forming a loadbearing ice-soil skeleton. Melting of ice slags triggers skeleton collapse and particle rearrangements, resulting in large deformation. The process follows a three-stage pattern: rapid initial, uniform, and decelerated settlement. We developed a novel ice-content-dependent settlement model that captures both the magnitude and temporal dynamics of thaw-induced settlement, validated against field and experimental data. The findings provide insights into the evolution of debris-ice fans and scientific basis for hazard mitigation.

Excavated spillways are widely used for emergency drainage of landslide dams because they can rapidly lower reservoir water levels and reduce breach-induced flood hazards. However, inadequate spillway design or construction may trigger undercutting erosion and lateral slope instability, leading to excessive peak discharge and increased downstream flood risk. This study proposes flexible geobag chains as a novel spillway protection structure. Flume experiments were conducted to evaluate their protective performance and examine how layout configuration influences breaching evolution. The geobag chains enhance local erosion resistance and reduce flow velocity, thereby mitigating bottom erosion and regulating spillway expansion. As a result, peak breaching discharge is reduced and its occurrence is delayed relative to unprotected conditions. Their flexible configuration also allows adaptation to progressive morphological changes, providing sustained protection during breaching. Four functional dam areas are identified, and their responses to different layout configurations are evaluated. Reducing chain spacing at the dam crest improves control over spillway expansion and suppresses abrupt collapse. In contrast, shortening chains downstream intensifies slope erosion, whereas crest shortening prolongs the effective protection duration and enhances regulation of lateral widening. Based on the response of each dam area, an optimized protection scheme is proposed. This scheme shows clear advantages in reducing peak discharge, delaying peak occurrence, and controlling spillway morphological evolution. These findings improve understanding of landslide dam breaching processes and provide a practical reference for emergency spillway protection.

Landslides composed of particles of varying sizes undergo particle-size segregation during motion, forming heterogeneous dams that influence breaching processes and resultant flooding. To elucidate the mechanisms of such heterogeneity, this study develops a three-dimensional Riemann-based smoothed particle hydrodynamics framework to simulate bidisperse landslide dam formation by coupling a nonlocal granular fluidity model that accounts for local and nonlocal effects in flow behavior transitions with a segregation model governed by competing gravity-driven percolation and diffusion. During high-velocity motion, the landslide behaves as a fully mobilized granular flow with pronounced particle-size segregation. After impacting the valley floor, it transitions to a quasi-static regime with reduced segregation. Higher flow mobility in the fully mobilized flow state exerts a stronger influence on the enhancement of segregation and dam non-uniformity than that in the quasi-static state. The evolution of segregation mechanisms results in two stages: a percolation-dominated stage, in which small particles migrate downward while large particles rise to form an inverse grading structure, followed by a percolation-diffusion coupled stage in which diffusion promotes particle mixing and forms a large-small particle transition zone. Increasing the particle size enhances percolation in the initial stage but strengthens diffusion more strongly in the latter stage, leading to reduced heterogeneity. The model is further applied to a field-scale Hsiaolin heterogeneous landslide dam and reproduces the inverse grading patterns, consistent with field investigations. These findings enhance the understanding of the formation mechanisms of heterogeneous landslide dams and contribute to the prediction of dam breaching and disaster mitigation.

Landslide-damming-breaching-flood (LDBF) events represent complex cascading geohazards in mountainous regions, involving strong fluid-soil coupling and multi-stage evolution, which pose significant challenges for fullprocess simulation. This study presents a unified three-dimensional (3D) Riemann-based fluid-soil coupled Smoothed Particle Hydrodynamics (SPH) framework for simulating the complete LDBF process. The framework employs a consistent elastoplastic constitutive model throughout all stages and adopts an energy-based criterion to identify stage transitions, enabling its application from laboratory to field scales. The effects of key topographic factors on cascading hazard evolution are systematically investigated. Valley geometry is found to exert primary control on breach intensity, causing variations in peak discharge exceeding 45% and governing both barrier lake capacity and breaching mechanisms. In contrast, slide geometry mainly influences dam morphology and overtopping location, resulting in comparatively minor changes in peak discharge of approximately 5%. Quantitative relationships between topographic conditions and peak discharge are established, providing a basis for rapid assessment of cascading hazard severity. This model is further applied to the 2018 Baige LDBF event and successfully reproduces key features, including the near bank breach and a peak discharge with an error of only 3.8%. The proposed framework advances a chain-based understanding of how topographic conditions govern flood intensity and provides improved support for hazard assessment and risk mitigation in mountainous regions.

Reservoir landslides pose significant risks to hydropower projects, potentially leading to catastrophic disasters that threaten downstream lives and properties. Landslide susceptibility assessments are critical for effective regional disaster prevention and mitigation. However, the complexity, model uninterpretability, and data scarcity related to reservoir landslides, particularly when adapting models across diverse geographic regions, present significant challenges. This study proposes an interpretable Deep Transfer Learning model coupled with multi-source data and Physical methods (DTLP). The model is trained on multi-source data from the Three Gorges Reservoir Area (TGRA) and Lower Jinsha River Basin (LJRB), tested in Baihetan Reservoir Area (BHT), addressing the issues of limited data and cross-regional generalization. The physical method captures the effect of dynamic water level changes on slope stability. SHAP values are used to interpret the model, providing clear insights into its internal mechanisms. Results demonstrate that DTLP outperforms TrAdaBoost in data-scarce regions, achieving higher accuracy (AUC=0.953, Accuracy=0.941) with better feature generalization and susceptibility zone identification. Incorporating dynamic water level changes into the physical model enhances identification of high-susceptibility areas and reduces misclassifications. SHAP analysis indicates that elevation, lithology, and distance to river significantly influence the model decisions. Using TGRA as the source domain further validates the superiority of DTLP framework. However, due to the initial discrepancies between TGRA and the target domain, the transferability is constrained to some extent, resulting in models trained on LJRB data outperforming those trained on TGRA data.

Landslide-river blockage-dam breaching-flood (LRDF) events originate from a single landslide and subsequently evolve into a sequence of cascading hazards, often resulting in severe environmental and socio-economic consequences. This study develops an integrated Bayesian network framework that combines physics-based and data-driven approaches to assess the probability of LRDF. The framework consists of four interlinked subnetworks, each representing a distinct stage of the cascade and explicitly identifying the dominant control parameters and their causal dependencies. Parameters uncertainties are systematically incorporated to quantify prior probabilities, and network-based inference is employed to evaluate sensitivity and probabilistic propagation across stages. The proposed method is applied to assess the cascading hazards of LRDF sourced from 305 loose deposits from during the M.S. 8.0 Wenchuan earthquake in the PR303 region. The results demonstrate that the demonstrate framework effectively addresses challenges in quantifying multi-hazard interactions and multisource parameters uncertainties in probabilistic assessment. The cascading hazards are governed by a quantitative compound control mechanism, in which rainfall intensity, slope gradient, geotechnical strength, and landslide volume jointly influence the transition probabilities between successive stages. Slope angle and soil strength parameters primarily govern slope instability, while interactions among soil and hydraulic parameters indirectly modulate intermediate nodes, such as landslide volume, sliding velocity, and sliding distance through chain-like propagation, further influencing river blockage and landslide-dam morphology, ultimately affecting dam-breaching flood distribution and downstream risk. Overall, the proposed framework provides a unified and interpretable approach for quantifying systemic risk and safety associated with cascading geo-hazards, offering a probabilistic basis for early warning and risk-informed decision-making.

Anti-slide piles exhibit remarkable reinforcement efficacy in slope stabilization under seismic conditions. Drawing on the failure characteristics of conventional anti-slide pile structures, this study proposes a novel seismic-resistant anti-slide pile system (double rows of anti-slide piles connected with flexible structures, FSDRP). Subsequently, taking the Shaba rock slope along the Duyun - Shangri-La highway in Yunnan, China as the research case, the mechanical properties, reinforcement effects, and internal mechanisms of single row of anti-slide piles (SRP), double rows of anti-slide piles (DRP), and FSDRP are analyzed via centrifuge shaking table tests and numerical simulations, under peak ground accelerations (PGAs) ranging from 0.05g to 0.20g. Key findings indicate that the seismic dynamic response of rock slopes with weak interlayers is jointly influenced by rock mass properties, earthquake intensity, and reinforcement structure types. All three reinforcement systems effectively suppress slope deformation; notably, FSDRP demonstrates superior stabilization performance compared to DRP and SRP. For SRP and DRP, the maximum shear forces and bending moments are concentrated at the interface of the weak interlayer. In contrast, the flexible connecting structures in FSDRP alter the force distribution pattern, thereby transforming the pile-rock interaction from passive to active reinforcement. This enhancement in the integrated "pile-rock-pile" structure enables coordinated deformation between the piles and the surrounding geotechnical mass, ultimately achieving more efficient dissipation of seismic energy. These findings provide critical theoretical support for evaluating reinforcement strategies for stratified landslides in seismic-prone regions, and offer significant implications for geohazard mitigation and engineering practice.

Landslides triggered by intense rainfall pose persistent threats to infrastructure and communities, particularly in data-scarce mountainous regions. Predicting such hazards at a regional scale remains challenging due to sparse dated landslide inventories and limited rainfall monitoring networks. This study presents an integrated spatiotemporal framework for landslide hazard prediction in the Muzaffarabad region, NW Himalayas, Pakistan. Landslide records (2006–2023) and satellite-based IMERG rainfall data were used to derive empirical cumulative event–duration (ED) thresholds using a frequentist approach to distinguish triggering from non-triggering rainfall conditions. Poisson-based modeling of threshold exceedance frequencies enabled estimation of temporal landslide probabilities for 1- to 20-year return periods. In parallel, spatial landslide susceptibility was modeled using machine learning techniques (Random Forest and XGBoost), with Random Forest outperforming XGBoost (AUC-ROC = 0.94; accuracy = 89%). Temporal probabilities were then integrated with spatial susceptibility to produce regional-scale landslide hazard index maps. This study is an early application of satellite-derived rainfall threshold exceedance in a probabilistic landslide hazard framework for data-scarce, mountainous regions. Results indicate a clear intensification of relative hazard across return periods: mean hazard increased from 0.13 to 0.18. The mean relative temporal factor (Pt) per IMERG pixel ranged from 0.19 to 1 across return periods, reflecting higher potential triggering during extreme rainfall events. From 1-year to 20-year return periods, the relative area of very high, high, and moderate hazard zones increased by 83.1%, 33.7%, and 11.5%, respectively, while low and very low hazard zones decreased by 13.4% and 7.4%. These trends underscore a spatial shift toward higher hazard zones, with significant implications for risk management, land-use planning, and infrastructure resilience. The proposed framework offers a scalable and transferable solution for landslide hazard assessment in mountainous, data-scarce regions, contributing to improved prediction, proactive mitigation, and safer engineering outcomes.

Non-destructive testing (NDT) based on ultrasonics is widely used for internal defect detection. To enhance the efficiency of defect detection, Deep Learning (DL) has been incorporated into ultrasonic NDT workflows. However, conventional methods rely on computer-vision-based DL to identify defects from post-imaging data, and their performance is highly related to the quality of the image and ultrasonic data. Consequently, end-to-end strategies are more suitable, as they can reconstruct the velocity field directly from ultrasonic data and detect defects by the difference of velocities. End-to-end methods have demonstrated excellent performance in numerical tests. However, they lack direct and precise comparison between predictions and true models in field application. Thus, this study presented an end-to-end, imaging-free framework for inner void detection by a Fully Convolutional Network (FCN). This net was tested by numerical tests and model experiments. The methodology in this study includes FCN, SH-wave forward modeling, sparse traces frequency-wavenumber (F-K) filtering, and image processing. The sparse traces F-K filtering mitigates boundary reflections in model experiments, and image processing realizes the mapping from limited ultrasonic data to a high-resolution velocity model. The FCN was trained on 1,200 randomly generated void models, achieving a median Intersection over Union (IoU) of 0.70 (with a maximum of 0.92, a minimum of 0.11) in numerical tests. Furthermore, the model was validated using experimental data acquired via an array ultrasound device on a homogeneous glass block with preset voids. The FCN successfully predicted the internal velocity model directly from the experimental ultrasonic data, accurately identifying the spatial distribution of the voids. The model experiment results demonstrate a successful transition from synthetic data to physical measurement, highlighting the potential of imaging-free FCN for multi-scale void detection in engineering scenarios.