YMM滑坡是三峡库区近坝库段距离大坝最近、规模最大的崩滑堆积体，因规模巨大、位置敏感且一旦失稳后果严重而备受关注。近20 a的观测数据表明，滑坡变形持续缓慢增长且未收敛。该文采用逐步线性回归方法和水动力触发机制力学模型，深入探究滑坡地质条件、外部影响因素与变形特征之间的关系。基于外部触发因素作用效应的阶段性转换时间节点，将滑坡体自蓄水以来的变形演化过程划分为3个阶段：1) 2003年6月至2006年9月，受库水位大幅抬升影响，滑坡以前缘牵引变形的方式复活。2) 2006年10月至2018年9月，YMM滑坡的变形模式由蓄水初期的前缘牵引转变为整体蠕滑，其实质为滑坡岩土介质因库水浸泡产生的劣化效应导致的变形调整；滑坡变形速率随着劣化效应的衰减而逐渐减小；在这个阶段库水位波动的影响要明显大于季节性降雨的影响。3) 2018年10月至2024年2月，水岩作用引起的岩土介质物理力学性状劣化过程基本完成，滑坡体适应了地下水环境改变所产生的影响，滑坡整体变形速率进一步明显降低，季节性降雨对滑坡变形的影响超越库水位波动的影响。研究表明，滑坡整体呈稳定的收敛型蠕滑，极端降雨将是滑坡产生局部失稳的重要触发因素。识别滑坡的变形演化阶段并确定各阶段居主导地位的外部影响因素是滑坡研究的关键，该文对此提供了一种有效的研究范式。

Objective: The YMM landslide is the largest landslide accumulation body nearest to the dam in the near-dam reservoir section of the Three Gorges Reservoir area. The landslide, located on the north bank of the Yangtze River within Zigui County, Yichang City, Hubei Province, with a surface area of approximately 0.48 km² and has a total volume of approximately 2 000 × 10⁴ m³. It is situated 17 km upstream of the Three Gorges Dam. Owing to its massive scale, sensitive location, and severe consequences of potential instability, the YMM landslide has attracted significant attention. Nearly 20 years of observation data indicate that although the landslide's deformation has been slow, it has continued without convergence. Methods: This study comprehensively considers the relationships among geological conditions, external influencing factors, and deformation characteristics of the landslide. A stepwise linear regression method is applied to analyze the observational data. Combined with a mechanical model of the hydrodynamic triggering mechanism of reservoir bank landslide deformation, the study quantitatively decomposes the roles and effects of various external triggering factors in the landslide's deformation process. Based on the phase-transition time nodes of these effects, the deformation evolution process due to reservoir impoundment is divided into three stages. Results: The study shows that the YMM landslide was stable before the impoundment. The reservoir impoundment led to its reactivation, which was followed by a complex deformation adjustment process. In the first stage (June 2003—September 2006), the landslide was reactivated in a retrogressive mode by a significant rise in the reservoir water level. In the second stage (October 2006—September 2018), the deformation mode shifted from front retrogressive to overall creep deformation, mainly due to the deterioration of the landslide rock-soil medium caused by reservoir water infiltration. The deformation rate gradually decreased as the deterioration effect weakened, and reservoir water level fluctuations had a more significant influence than seasonal rainfall during this period. In the third stage (October 2018—February 2024), the deterioration process of the physical and mechanical properties of the rock-soil medium induced by water-rock interaction was essentially complete. The landslide adapted to changes in the groundwater environment, resulting in a further significant reduction in the overall deformation rate. During this stage, the influence of seasonal rainfall on landslide deformation exceeded that of reservoir water level fluctuations. In terms of geological conditions, landslide characteristics, and deformation patterns, time-dependent deformation mainly convergent creep indicates that the landslide is generally stable. However, extreme rainfall remains a key triggering factor for potential local instability of the YMM landslide. Conclusions: This study provides a robust framework for interpreting the long-term deformation evolution of large-scale reservoir landslides by integrating monitoring data, statistical modeling, and mechanical analysis. Identifying stage-specific deformation patterns and dominant triggers enhances the understanding of landslide behavior in response to external forcing. These insights are crucial for improving early warning systems and developing targeted mitigation strategies in similarly high-risk reservoir environments.