Hazard and Risk-Based Tsunami Early Warning Algorithms for Ocean Bottom Sensor S-Net System in Tohoku, Japan, Using Sequential Multiple Linear Regression

This study presents robust algorithms for tsunami early warning using synthetic tsunami wave data at ocean bottom sensor (OBS) arrays with sequential multiple linear regression. The study focuses on the Tohoku region of Japan, where an S-net OBS system (150 pressure sensors) has been deployed. To calibrate the tsunami early warning system using realistic tsunami wave profiles at the S-net stations, 4000 stochastic tsunami simulations are employed. Forecasting models are built using multiple linear regression together with sequential feature selection based on Akaike Information Criterion and knee-point method to identify sensors that improve the accuracy most significantly. The study considers tsunami wave amplitude at a nearshore location and regional tsunami loss for buildings to develop hazard-based and risk-based tsunami warning algorithms. The models identify an optimal configuration of OBS stations and waiting time for issuing tsunami warnings. The model performance is compared against a base model, which only uses the earthquake magnitude and epicenter location. The result indicates that estimating the tsunami amplitude and loss via S-net improves accuracy. For the hazard-based forecasting, adding six sensors from the S-net improves the accuracy of the estimation most significantly with an optimal waiting time of 3 min. For the risk-based forecasting, a longer waiting time between 5 and 10 min is suitable.

Automated summary

This study presents robust algorithms for tsunami early warning using synthetic tsunami wave data and multiple linear regression. The calibration of the models using S-net data improves the accuracy of the tsunami warning system. The study provides optimal configuration of OBS stations and waiting time for issuing tsunami warnings.