A New Algorithm for Real-Time Tsunami Forecast Using a Dense Network of Cabled Ocean-Bottom Pressure Gauges

Monday, 15 December 2014: 5:15 PM
Naotaka Yamamoto, Shin Aoi, Kenji Hirata, Takashi Kunugi, Hiromitsu Nakamura and Wataru Suzuki, NIED National Research Institute for Earth Science and Disaster Prevention, Tsukuba, Japan
We started to develop a new algorithm for real-time tsunami forecast based on offshore tsunami observations with 150 cabled ocean-bottom pressure gauges of the Seafloor Observation Network for Earthquakes and Tsunamis (S-net), under construction along the Japan Trench (Kanazawa et al., 2012, JpGU; Uehira et al., 2012, AGU). The most important concept on the new algorithm is involving any type and/or form uncertainties in the tsunami forecast, which cannot be dealt with any of standard linear/nonlinear least square approaches. We first construct a tsunami scenario bank (TSB). It contains offshore tsunami waveforms at the 150 stations and maximum coastal tsunami heights, calculated using nonlinear long-wave theory with runup boundary condition from any possible tsunami sources (fault models) that affect target coastal regions. From TSB, then we quickly explore a range of several suitable tsunami scenarios, that can explain offshore observations. At the same time, maximum possible tsunami heights along the target coastlines, coupled with selected scenarios, are predicted. In the near future, it is possible to forecast real-time tsunami inundation by adding its component in TSB under the same strategy.

In this study, we focus on near-field tsunami occurring off the Pacific coast of Tohoku and Hokkaido. Provisionally, we generate 1848 tsunami scenarios, prepared for a research project of nationwide Probabilistic Tsunami Hazard Assessment for Japan (Hirata et al., 2014, AOGS), to construct TSB. For a given pseudo "observed waveforms", the developing algorithm rapidly picks up an allowable range of tsunami scenarios from TSB. In this procedure, we use multiple indexes such as correlation coefficient, sum of squared residual, as well as geometric mean and geometric standard deviation in ratios of scenarios to observations. Use of multiple indexes rather than any single index as linear inversion does reinforce to obtain robust tsunami forecast.