Debris flows induced by heavy rainfall are a major threat in Northwest and Southwest China, due to its abrupt occurrence and long runout. In light of this, this work presents the runout simulation and risk assessment of the Boshuigou debris flow under different rainfall conditions in Wudu district, Gansu Province, Northwest China. Based on field reconnaissance, the geomorphological feature and main source of the Boshuigou debris flow were described. With the application of the FLO-2D simulation, the potential flow depth and flow extent of the Boshuigou debris flow under 100-year return-period rainfall and 50-year return-period rainfall were calculated. The maximum flow velocities of the Boshuigou debris flow under the 100-year return-period rainfall and 50-year return-period rainfall were 5.46 and 5.18 m/s, respectively. Accordingly, the maximum flow depths were 5.85 and 5.57 m. Then, the hazard zonation was conducted in combination of the construction and other properties within the potential impact zone, and the risk assessment of the Boshuigou debris flow under the 100-year return-period rainfall and 50-year return-period rainfall was finally completed. This work presents a method for debris flow risk assessment considering the solid source and water flow, which can provide a basic reference for mitigation and reduction of geohazards induced by torrential rainfall.

This study aims to improve the accuracy and interpretability of flood susceptibility mapping (FSM) by integrating automated machine learning (AutoML) with explainable artificial intelligence (XAI) frameworks. Using spatial data from Seoul, South Korea, ten topographic and environmental factors were selected as model inputs. Initially, the Tree-based Pipeline Optimization Tool (TPOT), an evolutionary AutoML algorithm, was employed to construct baseline models using Gradient Boosting (GB), Random Forest (RF), and XGBoost (XGB). Subsequently, Bayesian optimization with Optuna was applied for hyperparameter tuning, resulting in enhanced performance, with the GB model achieving the highest AUC of 0.966. To ensure interpretability, SHAP (SHapley Additive exPlanations) was used to analyze both global and local feature contributions. The SHAP summary and dependence plots identified elevation (DEM), slope, stream distance, stream density, and built-up areas (LULC_7.0) as primary drivers of flood risk. Force and waterfall plots provided instance-level insights into model behavior, validating prediction logic across various geospatial contexts. In addition, Optuna’s visualization tools, such as optimization history, parallel coordinate plots, and hyperparameter importance graphs, offered transparency into how parameter spaces affect performance. These findings highlight the critical role of hyperparameter tuning in FSM and demonstrate the synergy of combining TPOT, Optuna, and SHAP within an XAI framework. The proposed approach not only improves predictive performance but also ensures interpretability, supporting more transparent and data-driven flood risk management.

In the context of global climate change, shallow landslides induced by strong typhoons and the ensuing rainstorms have increased significantly in China’s eastern coastal areas. On 27 May 2022, more than 700 liquefied landslides were induced by the rain gush in Wuping County, Longyan City, Fujian Province, SE China. In light of their widespread occurrence and the severe damage caused, detailed field investigations, UAV surveys, trench observations, in situ tests, and numerical simulation are conducted in this work. The cascading landslides are classified as channelized landslides and hillslope landslides. Long-term rainfall, the influence of vegetation roots under wind load, and differences in the strength and structure of surficial soil are the dominant controlling factors. The sliding surface is localized to be the interface at a depth of 1–1.5 m between the fully weathered granite and the strongly weathered granite. Kinetic analysis of a channelized landslide shows that it is characterized by short runout, rapid velocity, and strong impact energy. The maximum velocity, impact energy, and impact force of the Laifu landslide are 29 m/s, 4221.35 J, and 2110 kPa. Effective excavation is usually impossible in this context. This work highlights the escalating issue of shallow landslides in eastern China’s coastal areas, exacerbated by climate change and extreme weather events like typhoons. By conducting comprehensive investigations and analyses, the research identifies key factors influencing landslide occurrence, such as rainfall patterns and soil characteristics. Understanding the dynamics and impact of these landslides is vital for improving risk assessment, developing effective early warning systems, and informing land management policies in this region. Further exploration concerning hydro-meteorological hazard early warning should be encouraged in this region.

Background: On 18 December 2023 (CST), an Ms 6.2 earthquake struck Jishishan County, Gansu Province, northwestern China. The catastrophic Caotan loess landslide–mudflow, located approximately 14.8 km from the epicenter, was triggered simultaneously. Because the landslide caused extensive damage, affected a large area, and resulted in substantial casualties, we conducted a detailed investigation comprising field surveys, UAV surveys, borehole drilling, trench observations, and in situ tests. We analyzed the topography, geological setting, hydrological conditions, and geometry of the landslide. Results: Strong seismic shaking initiated slope failure in the source area. The displaced mass rapidly transitioned into a mudflow and traveled down the Qijia Valley, producing an unusually long runout over gentle terrain. We discuss the probable initiation mechanism, the transition process from sliding to flow, and the movement characteristics using evidence from the field investigation. Conclusions: We identify the principal factors controlling the initiation and long-runout movement of the Caotan landslide. These findings provide insight for mitigating similar hazards in northwestern China and other loess-covered regions worldwide.