Debris avalanches generally have massive mobility and extensive deposited area, which have strong relation to the granular effect. The granular effect is an outcome from the interaction (based on frictional mechanism) among particles. Discrepancy of grain composition, for instance the difference of particle size and their proportion of particles, is a key factor to influence granular effect. A series of laboratory tests have been conducted to research the granular effects on depositional processes of debris avalanches containing fine- and coarse-grained particles. The velocity of debris front, the morphology characteristics and runout distance of the debris avalanche are considered in detail, and the mass parameter and the size parameter are introduced in analysis for the research of granular effect. The experimental results indicate that the granular effects can decrease the dissipation of momentum and increase the accumulation region, leading to further runout distance and more extensive deposited areas. Further, the velocity, the range of depositions of debris avalanches are larger than that in other conditions when the proportion of fine-grained particles is 30%, especially the materials consist of heterogeneous granular. It means there have a critical mass parameter in experiments which lead to the high efficiency of granular effects, and the granular effect is weakening in homogeneous granular composition.

The task of landslide recognition focuses on extracting the location and extent of landslides over large areas, providing ample data support for subsequent landslide research. This study explores the use of UAV and deep learning technologies to achieve robust landslide recognition in a more rational, simpler, and faster manner. Specifically, the widely successful DeepLabV3+ model was used as a blueprint and a dual-encoder design was introduced to reconstruct a novel semantic segmentation model consisting of Encoder1, Encoder2, Mixer and Decoder modules. This model, named DeepLab for Landslide (DeepLab4LS), considers topographic information as a supplement to DeepLabV3+, and is expected to improve the efficiency of landslide recognition by extracting shape information from relative elevation, slope, and hillshade. Additionally, a novel loss function term—Positive Enhanced loss (PE loss)—was incorporated into the training of DeepLab4LS, significantly enhancing its ability to understand positive samples. DeepLab4LS was then applied to a UAV dataset of Baihetan reservoir, where comparative tests demonstrated its high performance in landslide recognition tasks. We found that DeepLab4LS has a stronger inference capability for landslides with less distinct boundary information, and delineates landslide boundaries more precisely. More specifically, in terms of evaluation metrics, DeepLab4LS achieved a mean intersection over union (mIoU) of 76.0% on the validation set, which is a substantial 5.5 percentage point improvement over DeepLabV3+. Moreover, the study also validated the rationale behind the dual-encoder design and the introduction of PE loss through ablation experiments. Overall, this research presents a robust semantic segmentation model for landslide recognition that considers both optical and topographic semantics of landslides, emulating the recognition pathways of human experts, and is highly suitable for landslide recognition based on UAV datasets.

Toppling failure is a fundamental mode of instability in rock slopes and occurs predominantly in reservoir bank anti-dip bedded rock masses. Reservoir impoundment changes seepage conditions and weakens slopes, whereas discontinuity non-persistence introduces uncertainty and complicates the identification of coupled toppling–sliding mechanisms. To address this, a probabilistic framework using the Goodman–Bray limit equilibrium method is developed. Equivalent strength parameters are introduced to unify the strength contrast between unsaturated and saturated segments along a common basal surface. Basal discontinuity connectivity is modeled as a random variable, and a Monte Carlo simulation is used to derive failure mode probabilities and a probability-weighted factor of safety. The framework is applied to the Huangcaoping anti-dip slope in the Dagangshan reservoir area at a normal water level of 1130 m. The most probable scenario has a probability of 0.116, involving sliding at 1120–1420 m and toppling at 1420–1550 m, with a probability-weighted mean factor of safety of 0.978. Predicted failure characteristics and deformation intervals are consistent with engineering observations, confirming the method’s effectiveness. This integration enables the simultaneous characterization of stability levels and the evolution mechanism. The approach provides mechanism-explicit mode likelihoods and a robust stability metric to support hazard assessment, monitoring placement, and reinforcement design.