Preparing for the Future Nankai Trough Tsunami: A Data Assimilation and Inversion Analysis From Various Observational Systems

Abstract

The future Nankai Trough tsunami is one of the imminent threats to the Japanese coastal communities that could potentially cause a catastrophic event. As a part of the countermeasure efforts for such an occurrence, this study analyzes the efficacy of combining tsunami data assimilation (DA) and waveform inversion (WI). The DA is used to continuously refine a wavefield model whereas the WI is used to estimate the tsunami source. We consider a future scenario of the Nankai Trough tsunami recorded at various observational systems, including ocean bottom pressure (OBP) gauges, global positioning system (GPS) buoys, and ship height positioning data. Since most of the OBP gauges are located inside the source region, the recorded tsunami signals exhibit significant offsets from surface measurements due to coseismic seafloor deformation effects. Such biased data are not applicable to the standard DA, but can be taken into account in the WI. On the other hand, the use of WI for the ship data may not be practical because a considerably large precomputed tsunami database is needed to cope with the spontaneous ship locations. The DA is more suitable for such an observational system as it can be executed sequentially in time and does not require precomputed scenarios. Therefore, the combined approach of DA and WI allows us to concurrently make use of all observational resources. Additionally, we introduce a bias correction scheme for the OBP data to improve the accuracy, and an adaptive thinning of observations to determine the efficient number of observations.

Tsunami Wavefield Reconstruction and Forecasting Using the Ensemble Kalman Filter

“…A low-pass filter could remove the effects of seismic and ocean acoustic waves (Saito & Tsushima, 2016). The impact of seafloor deformation on the pressure record has been explored in the context of linear potential theory (Saito, 2013), and some studies propose using bias correction based on waveform inversion (Mulia et al, 2017) or assimilating using the time derivative of pressure (Tanioka, 2017). In addition, rather than rejecting seismic and ocean acoustic waves as noise, one could potentially use them to constrain tsunami or earthquake parameters (Kozdon & Dunham, 2014).…”

Tsunami Data Assimilation Without a Dense Observation Network

“…The Optimal Interpolation method Kalnay, 2003;Maeda et al, 2015;Mulia et al, 2017) is performed as the following equations. The Optimal Interpolation method Kalnay, 2003;Maeda et al, 2015;Mulia et al, 2017) is performed as the following equations.…”

“…The tsunami wavefield at the nth time step is represented as x n (η(nΔt, x, y), M(nΔt, x, y), N(nΔt, x, y)), where η is tsunami height, M and N are velocities in two directions, Δt is the time step, and x and y are the spatial coordinates. The Optimal Interpolation method Kalnay, 2003;Maeda et al, 2015;Mulia et al, 2017) is performed as the following equations.…”

A Multiple-Parameter Methodology for Placement of Tsunami Sensor Networks

“…Mulia et al (2017) also used optimization methods, specifically a dimensionality reduction approach, to initiate the process. Unlike the previous cases where the goal was to address the performance for a wide range of sources, the goal of Mulia et al (2017aMulia et al ( , 2017b was to identify the best placing of sensors to characterize a specific, largemagnitude, target scenario. Their focus was to resolve in great detail the characteristics of nonuniform slip by maximizing the accuracy of inverting a set of stochastic scenarios on a predefined spatial domain.…”

“…Note that the performance of sensor networks has typically been quantified by analyzing their accuracy in predicting the arrival time (Bernard et al 2001;Schindelé et al 2008;Omira et al 2009), the source slip (Mulia et al 2017;Saunders 2018), or amplitude at the coast (Mulia et al 2017b;Saunders 2018), among others. Typically, these parameters are analyzed independently.…”