Research on early warning methods for step-type landslides in reservoir areas

Step-type landslides in reservoir areas exhibit distinct characteristics of sudden occurrence and staged deformation. However, the current lack of systematic identification methods and targeted early-warning mechanisms hinders its effective prevention and control. To address this challenge, this study investigates the Linbang landslide at the Houziyan Hydropower Station on the Dadu River. We employ K-means clustering analysis combined with a local extremum method to identify step-type deformation intervals within the monitoring data. Specifically, we systematically analyze the response mechanisms between seven recorded step-type deformation events (occurred between 2018 and 2022) and three key indicators: effective reservoir water level fluctuation (H_effect), modified tangential angle (α), and deformation rate (v). Based on this analysis, we propose an early-warning method for step-type landslides in reservoir areas. The results indicate that the warning threshold values for the three indicators (H_effect, α, v) of the Linbang landslide are 4.69 m, 45°, and 3.87 mm/d, respectively. The quantitative identification criteria for step-like deformation are the appearance of a stable inflection point combined with H_effect < 1.42 m and v < 3.87 mm/d. A warning is issued when all three indicators simultaneously reach their thresholds; if the step-type deformation identification criteria are satisfied, the event is confirmed as step-type deformation and the alert is lifted. The research findings can provide a reference for early warning work concerning step-type landslide hazard.