Abstract: The impulse wave is a universally existent hazard induced by landslides nearby water basins. Predicting the characteristics of the impulse waves is of great importance for risk assessment of landslide generated waves. Previous studies often focused on the instant wave parameters and selected several landslides parameters such as thickness and velocity to build predictive models for wave characteristics such as height and amplitude. The time series relations between landslide parameters and wave parameters need to be determined. This paper gives insights into the time series variation of the wave characteristics. Experiments are first conducted by using granular particles, and a time series panel database is built by using experimental data. Then a time series prediction model is established for wave characteristics based on a random coefficient panel data model, and the model is tested with the support of experimental data. The proposed model provides a temporal predictive method, which compensates for the deficiency of conventional models that can only predict the instantaneous wave parameters.