The stability and dynamics of Thwaites Glacier depend on the structural properties of its marine terminus; however, the relationship between these variables on the floating ice tongue is poorly understood. Here we present a six-year record of ice speed, derived from satellite observations starting in 2015, showing two large-magnitude (approximately 30–45%) and prolonged (approximately one to two years) cycles of speed variation across the ice tongue. Using an automated, deep learning-based method of extracting high-resolution fracture maps from satellite imagery, we detail periods of increasing fracture development and subsequent reconsolidation in the ice tongue shear margin that coincide with the observed speed changes. Inverse modelling using the BISICLES ice-sheet model indicates that the variation in ice speed can be accounted for by these observed changes to the spatial pattern of fracturing. This study provides further evidence of direct coupling between fracturing and dynamic variability in West Antarctica but indicates that increased fracturing and associated speed changes are reversible on one- to two-year timescales. We suggest that fracturing does not necessarily lead to positive feedback with glacier acceleration on these timescales and that damage process modelling is important for accurately predicting the evolution of the Antarctic Ice Sheet.

Thwaites Glacier 的稳定性和动态取决于其海洋终点站的结构特性；然而，人们对浮冰舌上这些变量之间的关系知之甚少。在这里，我们展示了从 2015 年开始的卫星观测得出的六年冰速记录，显示了整个冰舌的两个大幅度（大约 30-45%）和长时间（大约一到两年）速度变化周期。使用基于深度学习的自动化方法从卫星图像中提取高分辨率裂缝图，我们详细描述了冰舌剪切边缘裂缝发展增加和随后再固结的时期，这些时期与观察到的速度变化相吻合。使用 BISICLES 冰盖模型的逆向建模表明，冰速度的变化可以通过这些观察到的压裂空间模式的变化来解释。这项研究提供了南极洲西部压裂和动态变化之间直接耦合的进一步证据，但表明增加的压裂和相关的速度变化在一到两年的时间尺度上是可逆的。我们认为，在这些时间尺度上，压裂不一定会导致冰川加速的正反馈，并且损伤过程建模对于准确预测南极冰盖的演变很重要。这项研究提供了南极洲西部压裂和动态变化之间直接耦合的进一步证据，但表明增加的压裂和相关的速度变化在一到两年的时间尺度上是可逆的。我们认为，在这些时间尺度上，压裂不一定会导致冰川加速的正反馈，并且损伤过程建模对于准确预测南极冰盖的演变很重要。这项研究提供了南极洲西部压裂和动态变化之间直接耦合的进一步证据，但表明增加的压裂和相关的速度变化在一到两年的时间尺度上是可逆的。我们认为，在这些时间尺度上，压裂不一定会导致冰川加速的正反馈，并且损伤过程建模对于准确预测南极冰盖的演变很重要。