[HDS08-12] Operational use of tsunami source inversion in near-field tsunami warning by JMA
Keywords:Tsunami, Real-time forecast, Tsunami warning
In our tsunami-forecasting method (tFISH: Tsushima et al., 2009, JGR), offshore tsunami-waveform data are used to estimate spatial distribution of initial sea-surface displacement in tsunami source area. Then linear combination of the estimated source and the precomputed tsunami-waveforms from sea-surface elements (i.e. Green’s functions) are calculated to forecast tsunami waveforms at coastal sites of interest. Actual OBPG data contain pressure variations not only due to tsunami, but also due to other natural phenomena such as permanent seafloor deformation (coseismic offset) and seismic waves (short-period pressure fluctuation). To reduce effect of these non-tsunami components we applied revision of tsunami Green’s functions (Tsushima et al., 2012, JGR) and real-time noise reduction such as recursive digital filter (Saito 1978). The retrospective application to the 2003 Tokachi-oki earthquake (Mw 8.3) and the 2011 Tohoku earthquake (Mw 9.0) (e.g. Tsushima et al., 2011, EPS) showed its good tsunami-forecasting ability.
The forecasting accuracy by tFISH depends on station coverage (Tsushima et al., 2009, JGR). Since the present offshore tsunami network, especially S-net, is distributed widely, densely and homogeneously, tFISH expected works well. In the actual pressure records, however, coseismic offset and long-wave-period variations that cannot be interpreted as seafloor deformations and tsunamis are sometimes measured (Wallace et al., 2016, JGR; Kubota et al. 2020, GRL). For example, when the 2016 off Fukushima earthquake (Mw 6.9) occurred, coseismic step appeared in the records of the OBPGs around source area, but the step amount is too large to be explained as coseismic seafloor deformation (Kubota et al., 2020, this meeting). The application of tFISH to these data showed significant tsunami-source artifact, resulting in degrading accuracy of tsunami forecast (overestimation of coastal tsunami height). The source artifact can be decreased substantially by applying the revised tFISH inversion that is proposed by Tsushima and Yamamoto (2017, SSJ Meeting). In the observation equation of the revised inversion, we put assumption that observed waveform at OBPGs can be represented by not only tsunami and seafloor deformation, but also coseismic step function and liner trend function. The reason why we chose these functions is that actual waveforms of non-tsunami signal resembles these functions: i.e. we do not have physical image. We applied the revised method to the S-net OBPG data of the 2016 Fukushima earthquake and succeeded in improving source model and coastal tsunami forecasts.