Excessive seepage significantly affects the stability of levees during flood seasons. Recently, infrared thermal imaging technology has been proposed to detect the seepage leakage of levees for its simplicity, high efficiency and non-contact. However, previous studies mainly focused on theoretical and laboratory tests to verify its effect, which involve many assumptions and simplification. This paper aims to study the practical application effects of infrared thermal imaging technology on levee leakage detection. A typical embankment in Xiangyin County, Hunan Province is selected as the studied area, where a specific point containing water is considered as the potential leakage point. Then, DJI M300RTK UAV equipped with Zenith H20T gimbal is adopted to collect the visible and infrared images of the embankment’s leaking point. The visible and infrared image data collected at different times and at different measurement heights are analyzed to investigate the detection accuracy of infrared thermal imaging technology. Results show that the infrared thermal imaging technology can accurately identify the potential seepage leakage point of levees. Compared with the visible light recognition method, the infrared thermal imaging technology can identify the location and scope of leakage hazards more clearly and accurately at night. During the daytime, infrared detection is less effective than visible light method. Moreover, the measurement altitude between the drone and the target leakage points significantly affects the detection accuracy of infrared thermal imaging technology.

Accurate delineation of debris-flow-endangered areas (e.g., the maximum runout distance) is a necessary prerequisite for the debris-flow risk assessment and countermeasures design. Recently, machine-learning models have been proved to be an effective tool in predicting debris-flow parameters. However, existing machine-learning models are generally developed based on a very limited number of observation data, which may result in the predictive model overfitting or underfitting. How to develop a robust model for accurate forecasting of debris-flow-endangered areas still remains a difficult task. This paper proposes a hybrid method for predicting debris-flow hazard zone by integrating machine-learning algorithms and an empirical regression model. The proposed method takes the calculated maximum runout distance obtained from the empirical model as supplementary inputs to increase the amount of training data to construct hybrid machine-learning models. Three commonly used machine-learning models (i.e., multivariate adaptive regression splines (MARS), random forest (RF), and support vector machine (SVM)) are developed based on the training datasets of a specific debris basin. Then, these three machine-learning models are combined with an empirical relationship developed using the same training datasets to generate corresponding hybrid models. Finally, the performance metrics (i.e., coefficient of determination (R²), root-mean-square error (RMSE), and mean absolute error (MAE)) of the proposed hybrid models are comprehensively investigated and compared with the single predictive model (i.e., MARS, RF, SVM, and the empirical model) under fivefold cross-validation. The proposed method is illustrated using 134 channelized debris-flow events in Sichuan province, China. Results show that compared with the three individual machine-learning models, hybridization of machine-learning algorithms and the empirical model results in R², RMSE, and MAE improved by 70.5, 32.9, and 41.1, respectively. In contrast to the empirical model, the R², RMSE, and MAE value of the proposed hybrid models are improved by 29.6, 22.3, and 32.5, respectively. The proposed hybrid models generally perform better than the single machine-learning and the empirical model, providing a promising tool for accurate forecasting of a debris-flow-endangered area.

Introduction: Extreme inflow does not necessarily raise a reservoir above its check flood level. Dam-response assessment must therefore reconcile the hydraulic boundary condition with the temporal coverage of the monitoring record. Methods: We reassessed deformation, piezometric, uplift, strain, stress, and temperature observations from a 71.50 m high double-curvature concrete arch dam in southern China. Results: Peak inflow was reported as 3,440 m ³ /s, above the 1,000 year reference inflow of 2,910 m ³ /s. The reservoir reached 160.29 m at 14:00 on 22 May 2025, 0.29 m above the design flood level and 0.13 m below the check flood level. Discrete surface surveys showed central radial differences of 5.65–5.99 mm between the reported survey epochs, whose exact dates were not retained. In the vertical-displacement table, +0.87 mm is the BD5 interval difference, whereas −0.15 mm belongs to BD1. Project-derived uplift coefficients at P1–P4 ranged from 0.07 to 0.21 at a reservoir level of 160.29 m. These values were 16%–72% below the project design reference of 0.25. The raw abutment-piezometric workbook contained no observations, and a plotted value of 165.28 m could not be reconciled with the dam datum. Automatic stress, strain, and temperature records generally stopped on 21 May and resumed on 8 June, leaving the flood peak unobserved. Discussion: The archive consequently cannot resolve the peak response, separate thermal and hydraulic effects, test recovery, or establish a formal hydraulic threshold. The available evidence documents the event but does not establish piping, elastic recovery, or structural safety. This case highlights the need for synchronized event acquisition, verified datums, calibration metadata, and targeted inspection before diagnostic or remedial decisions are made.