Probabilistic analysis of impulse waves generated by large deposits in mountain reservoirs

Landslide-generated impulse waves (LGIWs) usually have a severe impact on the normal operation of the reservoir area. This study conducts probabilistic analysis of LGIWs generated by Rongsong (RS) deposit by combining the energy equivalent method, large-scale physical similarity model experiment and machine learning surrogate model. The energy equivalent method utilizes a wave generation model to get the initial waveform and employs a wave propagation model to evaluate the LGIWs in the study area. The two models are connected by the initial wave, ensuring that the total energy and the energy release rate transferred from the landslide to the water body remain consistent with the simplified landslide. This method is used to study the LGIWs hazard chains induced by RS deposit, and a large-scale physical similarity model experiment is carried out to verify. Subsequently, taking into account the uncertainties of landslide velocity and the volume of unstable mass, a surrogate model is established to get the probability distribution of the impulse waves runup height on the dam. The results show that the energy equivalent method is able to accurately simulate LGIWs hazards. The runup height on the Rumei (RM) dam is concentrated between 11.02 and 12.62 m, accounting for 0.76. It is recommended to install a 1.2 m wave wall on the dam crest. This study provides a novel method for conducting probabilistic analysis of LGIWs in mountain reservoirs, and offers valuable references for disaster prevention and control in the reservoir area.