Abstract: Submarine landslides, as a natural disaster characterized by their significant destructive force and wide distribution worldwide, often pose a serious threat to human life safety. The massive waves generated by landslides can cause severe damage to marine structures, making rapid prediction and assessment of the size of landslide-generated waves a crucial part of disaster prevention and mitigation efforts. This is particularly vital for China's "Maritime Power" strategy and safeguarding the lives and property of its people. In this study, numerical simulations of landslide-generated waves were conducted using the non-hydrostatic model NHWAVE, yielding data on the waves produced by different landslides. Building upon this, a prediction model for landslide-generated waves was trained based on a one-dimensional convolutional neural network (CONV1D). The model was trained using datasets from various observation points and different types of landslides, and its predictive performance was evaluated using metrics such as Mean Squared Error (MSE). The results indicate that, even with a small amount of data, the convolutional neural network can effectively learn the patterns of landslide-generated waves. It can also predict reasonably well for features not present in the dataset, demonstrating good generalization ability. Once trained, the neural network can rapidly predict the temporal evolution of downstream waveforms based on real-time input of water level data at the location of the landslide occurrence. This neural network prediction can provide early assessment of disasters, guiding relevant departments in disaster prevention and control efforts.

Key words: Submarine Landslide, Convolutional Neural Network, Temporal Prediction, Disaster Prevention and Control