This study describes the development, validation, and testing of a tsunami forecast model for Hilo, Hawaii. Based on the Method of Splitting Tsunamis (MOST) model, the forecast model is capable of simulating 4 hr of tsunami wave dynamics at a resolution of 2 arc sec (~60 m) in 10 min of computational time. A reference inundation model at a higher resolution of 1/3 arc sec (~10 m) was also developed in parallel, to provide modeling references for the forecast model. Both models were tested for 16 past tsunamis and a set of 18 simulated magnitude 9.3 tsunamis.
The error of the maximum wave height computed by the forecast model is within 35% when the observation is greater than 0.5 m; when the observation is below 0.5 m the error is less than 0.3 m. The error of the modeled arrival time of the first peak is within ±3% of the travel time. The good agreement between the model computations and observations, along with the numerical consistency between the model results for the maximum amplitude and velocity, provide a quantitative validation and reliable robustness and stability testing of the forecast model.
The validated Hilo forecast model was further applied to hazard assessment from 1435 scenarios of simulated tsunami events based on subduction zone earthquakes of magnitude 7.5, 8.2, 8.7, and 9.3 in the Pacific Ocean basin. The results show an impressive local variability of tsunami amplitudes even for far-field tsunamis, and indicate the complexity of forecasting tsunami amplitudes at a coastal location. It is essential to use high-resolution models in order to provide accuracy that is useful for the practical guidance of coastal tsunami forecasts.