Changes in rainfall, groundwater levels, and reservoir water levels exacerbate the deformation of water-involved landslides, accelerating the transition from landslide evolution to extinction. Extracting the destruction patterns of landslides from extensive monitoring data, and understanding their overall deformation mechanisms are crucial for geological hazard prevention and control. Herein, we took the Jiuxianping landslide in the Three Gorges Reservoir area as an example and proposed a deformation mechanism analysis model for water-related landslides based on monitoring data mining techniques. Using Granger causality testing, the study analyzes the spatiotemporal characteristics of GPS displacement data from three different profiles, which confirms that Jiuxianping exhibits a traction destruction mode. By comparing GPS displacement data and their Granger causality relationships across different profiles, we reveal that segmented sliding features of the landslide's front, middle, and trailing during its evolution. Furthermore, the impact intensity of triggering factors (rainfall and reservoir water level changes) on landslide displacement was identified. Based on GPS displacement data from profiles II–II′, an empirical mode decomposition–long short-term memory-regression model (EMD-LSTM-regression) is developed for multisource prediction of landslide displacements. The Shapley additive explanations algorithm is used to analyze the influence of rainfall and reservoir water level changes on periodic displacements at different positions of the landslide. Owing to the large area of the Jiuxianping landslide, the response to triggering factors varies across different locations. In the context of global warming and frequent extreme weather events, these findings offer important insights for preventing and mitigating water-related landslides in the Three Gorges Reservoir area, while also providing new perspectives for the analysis of global water-involved landslide deformation.

中文翻译：

---

降雨与水库水位效应下的滑坡变形机理及耦合效应分析


降雨量、地下水位和水库水位的变化加剧了涉水滑坡的变形，加速了滑坡从演变到灭绝的转变。从广泛的监测数据中提取滑坡的破坏模式，并了解其整体变形机制，对于地质灾害防治至关重要。本文以三峡库区九仙坪滑坡为例，提出了一种基于监测数据挖掘技术的涉水滑坡变形机理分析模型。利用格兰杰因果关系检验，该研究分析了 3 个不同剖面的 GPS 位移数据的时空特征，证实了九仙坪表现出牵引破坏模式。通过比较 GPS 位移数据及其在不同剖面中的格兰杰因果关系，我们揭示了滑坡在演变过程中前部、中部和尾部的分段滑动特征。此外，还确定了触发因素 （降雨和库水位变化） 对滑坡位移的影响强度。基于剖面 II-II′ 的 GPS 位移数据，开发了一种经验模态分解-长短期记忆回归模型 （EMD-LSTM-regression），用于滑坡位移的多源预测。采用 Shapley 加法解释算法分析降雨和水库水位变化对滑坡不同位置周期性位移的影响。由于酒仙坪滑坡面积大，不同地点对诱发因素的响应存在差异。 在全球变暖和极端天气事件频发的背景下，这些发现为预防和缓解三峡库区与水有关的滑坡提供了重要的见解，同时也为分析全球涉水滑坡变形提供了新的视角。