A step-like landslide identification and prediction method based on trend speed ratio

The displacement prediction of step-like landslides is the simplest and most reasonable method for assessing their potential destructiveness. Over the years, machine learning methods have been progressively developed and optimized, and are now extensively used by researchers for predicting the displacement of step-like landslides. However, these methods, often referred to as "black box" models, fall short of explaining the physical processes that lead to landslide displacement, resulting in a lack of interpretability in the prediction of results. Here, we propose the use of the Trend Speed Ratio (TSR) as a novel method to identify step points in step-like landslides. A step in the landslide is observed when TSR > 1.0 and Delta TSR > 0. When TSR > 2.0, the landslide is deemed to have experienced failure. Additionally, TSR is employed to predict the displacement of secondary steps following landslide deformation. In the application cases of the Baishuihe and Baijiabao landslides in the Three Gorges Reservoir area, the accuracy of the step point identification method based on TSR reached 100%, and the mean absolute errors (MAEs) of the step post-displacement prediction method based on TSR were 31.60333 mm and 25.68056 mm, respectively, and the coefficient of determination values were 0.91043 and 0.99378, respectively. Compared to traditional methods, this approach provides practical physical insights and is more straightforward, sensitive, and stable, thus providing new technical support for onsite engineers to assess the potential risks of step-like landslides.