Deep-learning-based landslide early warning method for loose deposits slope coupled with groundwater and rainfall monitoring

Focusing on the stability of loose deposits slopes, a deep learning-based early warning method is proposed to predict the slope displacements along the monitoring depth at the same time that can predict a potential sliding depth range, taking into account groundwater and rainfall. To predict the periodic displacement induced by underground water and rainfall, a time series data decomposition technique is proposed herein by integrating the long short-term memory (LSTM) models and Prophet. To illustrate the specific process of early warning method, a case study is conducted on a loose deposits slope in Southwest China undergoing excavation. The performance evaluation of the early warning method involves comparing it with various models in terms of slope displacement prediction and the result shows that the prediction accuracy can significantly increase with due consideration of groundwater and rainfall. At last, the safety condition of the slope is analyzed based on risk evaluation indicators and the proposed prediction method described in this paper, which can be applied to other regions as effective measures to mitigate landslide losses.