11:00 AM - 1:00 PM
[SCG55-P04] Automatic Tsunami Inversion System, Marlin, for Nankai Trough Region Using DONET Data
Keywords:Real-time Tsunami Forecast, Tsunami Inversion, Tsunami Source, Seafloor Observation Network, Ocean Bottom Pressure Data
Real-time tsunami forecast method based on automatic tsunami source inversion using ocean bottom pressure data proposed by Tsushima et al. (2009, JGR) has become plausible because dense seafloor real-time observation networks, S-net and DONET, have been deployed and in operation. The automatic tsunami source inversion also contributes to a rapid evaluation of rupture area of mega-thrust earthquakes, which is required in responses to the hypothetical Nankai Trough earthquakes. Suzuki et al. (2020, JpGU) reported that the NIED constructed the automatic tsunami inversion system named Marlin using the S-net data. We have improved Marlin to accelerate the calculation time and to extend the analysis areas to the Nankai Trough region using the DONET data.
Main analysis procedure follows Suzuki et al. (2020). We have prepared the tsunami Green’s functions between element tsunami sources distributed southern off of Shikoku and Honshu islands and DONET stations. Spacings of the element tsunami sources are approximately 24 km for great earthquakes and approximately 8 km for M7-class earthquakes or smaller ones. It was found in test run of the previous system that it took much longer time to read the Green’s function data than had expected. We then have modified the system to accelerate the reading procedure. Forward simulation of tsunami propagation and inundation based on the inverted tsunami source is automatically performed using GPUs of the NIED supercomputer. This enables to forecast coastal tsunami heights and inundation depths with a 90-m resolution for the Nankai Trough region from western Shikoku to Aichi prefecture using the DONET data and for East Japan from Chiba prefecture to eastern Hokkaido using the S-net data.
Performance of Marlin for the Nankai Trough region is examined using the simulated DONET data assuming the source models proposed for the past Nankai Trough mega-thrust earthquakes. For the 1707 Hoei earthquake, which is considered to have ruptured whole the Nankai Trough plate boundary (Furumura et al., 2011, JGR), the inverted result using the first 10-min data shows rather uniform uplift over the source area. The subsequent analyses using longer data (>=30 min) pose the larger uplift off the eastern Shikoku than the other areas, which is consistent with the assumed source model. The examination for the last events of the 1944 Tonankai (Baba et al., 2006, Tectonophysics) and 1946 Nankai earthquakes (Baba et al., 2002, PEPI), which ruptured eastern and western sides divided by off Kii peninsula, shows that the analysis using the first 10-min gives overall tsunami source images and the subsequent analyses can resolve localized uplift and subsidence.
Main analysis procedure follows Suzuki et al. (2020). We have prepared the tsunami Green’s functions between element tsunami sources distributed southern off of Shikoku and Honshu islands and DONET stations. Spacings of the element tsunami sources are approximately 24 km for great earthquakes and approximately 8 km for M7-class earthquakes or smaller ones. It was found in test run of the previous system that it took much longer time to read the Green’s function data than had expected. We then have modified the system to accelerate the reading procedure. Forward simulation of tsunami propagation and inundation based on the inverted tsunami source is automatically performed using GPUs of the NIED supercomputer. This enables to forecast coastal tsunami heights and inundation depths with a 90-m resolution for the Nankai Trough region from western Shikoku to Aichi prefecture using the DONET data and for East Japan from Chiba prefecture to eastern Hokkaido using the S-net data.
Performance of Marlin for the Nankai Trough region is examined using the simulated DONET data assuming the source models proposed for the past Nankai Trough mega-thrust earthquakes. For the 1707 Hoei earthquake, which is considered to have ruptured whole the Nankai Trough plate boundary (Furumura et al., 2011, JGR), the inverted result using the first 10-min data shows rather uniform uplift over the source area. The subsequent analyses using longer data (>=30 min) pose the larger uplift off the eastern Shikoku than the other areas, which is consistent with the assumed source model. The examination for the last events of the 1944 Tonankai (Baba et al., 2006, Tectonophysics) and 1946 Nankai earthquakes (Baba et al., 2002, PEPI), which ruptured eastern and western sides divided by off Kii peninsula, shows that the analysis using the first 10-min gives overall tsunami source images and the subsequent analyses can resolve localized uplift and subsidence.