The accuracy of numerical calculations is highly dependent on the grid spacing, and the smaller the grid spacing the better the accuracy. However, the portion requiring high accuracy often corresponds to only a part of the entire model domain. Setting a small grid interval for the entire model domain to study the calculation region is undesirable from the viewpoint of computation load and computation time. In this study, we illustrate the efficiency of the adaptive mesh refinement method for wave run-up modelling in terms of computational time and accuracy using numerical experiments. In this study, we conducted numerical experiments of wave run-up to verify the efficacy of the Adaptive Mesh Refinement (AMR) method in terms of accuracy and computation load for typhoon Chaba (1618). We determined that wave propagation modelling with storm surge height using the AMR method might be a good alternative to calculated wave run-up height. The difference between the observed and AMR model values for wave run-up height was calculated to be less than 1 m. Compared to the results of non-adaptive experiments with fixed spatial resolutions, those with the AMR method produced highly accurate results, while requiring only 75% of the computation time. The estimation of wave run-up height using AMR is expected to allow the accurate estimation of coastal inundation if more accurate topography and input conditions are given.