[J04-P] Poster

[J04-P-04] Waveform matching for the ocean bottom pressure data toward real-time tsunami forecast

^○Wataru Suzuki, Shin Aoi, Naotaka Yamamoto (National Research Institute for Earth Science and Disaster Resilience, Tsukuba, Ibaraki, Japan)

Large-scale and dense ocean bottom observation networks have been maintained by NIED for regions along the Japan Trench (S-net) and along the Nankai Trough (DONET). To take an advantage of the ocean bottom observation network, we are developing a real-time tsunami forecast system including inundation for the Pacific coast of Chiba prefecture, using the real-time ocean bottom pressure data observed by S-net (Aoi et al., 2017). In this system, the several tsunami scenarios that reasonably explain the observed ocean bottom pressure data are selected according to the multi-index method (Yamamoto et al., 2016) from the pre-calculated database. The forecast information is generated from the coastal tsunami height and inundation information related to the selected tsunami scenarios. To advance the robustness of forecast and warning, it is better to implement several different approaches for real-time tsunami detection and forecast. In this study, therefore, we examine the matching for the time series of the ocean bottom pressure change at each station for selecting the tsunami scenarios that well explain the observation well using the simulated data for 150 S-net stations. When we use L1 and L2 norms to evaluate the fitness between the observed and scenario pressure data, the scenarios that give the smallest L1 and L2 norms do not provide the sufficient forecast result in the point that they underestimate the coastal tsunami height as well as the ocean bottom pressure data. This is probably because L1 and L2 norms are sensitive in the slight difference of phase information. Therefore, we pose the condition for the selected scenarios that the number of the stations where the maximum absolute amplitude of scenario is equal or greater than the observed one exceeds a criterion value before the evaluation according to L1 and L2 norms. This condition makes the waveform matching between the observed and scenario data better and the forecast result more satisfactory.