9:45 AM - 10:15 AM
[SSS09-02] Finite fault modeling based on ABIC: joint inversion using geodetic and tsunami waveform data
★Invited Papers
Keywords:2024 Noto Peninsula earthquake, Joint inversioin, ABIC, Tsunami, GNSS, SAR
Finite fault modeling is one of the most popular analysis methods for estimating the earthquake source process. Tsunami records, observed by a tide gauge or an ocean-bottom pressure gauge, are useful to constrain the slip amount on the offshore fault (e.g., Satake, 1987). Combining tsunami records with other observations such as seismic and geodetic records, therefore, enables us to reveal the fault rupture process associated with a tsunamigenic earthquake (e.g., Yokota et al., 2011).
The purpose of this study is to construct the finite fault model associated with the 2024 Noto Peninsula earthquake. We carry out the joint inversion analysis using GNSS, SAR, and tsunami records based on ABIC (Akaike’s Bayesian Information Criterion; Akaike, 1980). When conducting the joint inversion, we have to determine the weight among data and a priori constraints such as smoothing or damping on model parameters. Yabuki and Matu’ura (1992) shows that ABIC is useful for searching the optimal weight of a priori constraints. ABIC can be used to determine the weight of data as well (Asano et al., 2005; Funning et al., 2014).
Some tsunami studies used ABIC to estimate fault slip distribution (e.g., Gusman et al., 2010; Kubota et al., 2018). These studies, however, used an arbitrary weight for data or a non-negative constraint on slip amount, which was inappropriate for the ABIC-based inversion. In this study, we propose a new ABIC-based joint inversion scheme using tsunami waveform and geodetic data.
For the data weight, we followed Funning et al (2014); employing the relative variances of GNSS data and SAR data to tsunami data. In addition, our inversion method considered the covariance component, the variance of Green’s function, and the direct constraint (Matu’ura et al., 2007; Fukahata and Write, 2008; Yagi and Fukahata, 2011). Since the number of hyperparameters was increased, we used the Optuna library to search for the minimum value of ABIC. Note that, in the inversion analysis, we used tsunami records not in the time domain but in the frequency domain by the Fast Fourier Transform.
As a result, we obtained the finite fault model with large amounts of slip at NT4 and NT6, the northeast and northern faults of the Noto Peninsula. This model is consistent with the previous studies such as Fujii and Satake (2024) and Okuwaki et al. (2024). While we did not use the non-negative constraint in this inversion, the minimum amplitude was less than -0.5 m. We therefore considered that our proposed inversion scheme succeeded in finding the optimal weights of constraints.
The purpose of this study is to construct the finite fault model associated with the 2024 Noto Peninsula earthquake. We carry out the joint inversion analysis using GNSS, SAR, and tsunami records based on ABIC (Akaike’s Bayesian Information Criterion; Akaike, 1980). When conducting the joint inversion, we have to determine the weight among data and a priori constraints such as smoothing or damping on model parameters. Yabuki and Matu’ura (1992) shows that ABIC is useful for searching the optimal weight of a priori constraints. ABIC can be used to determine the weight of data as well (Asano et al., 2005; Funning et al., 2014).
Some tsunami studies used ABIC to estimate fault slip distribution (e.g., Gusman et al., 2010; Kubota et al., 2018). These studies, however, used an arbitrary weight for data or a non-negative constraint on slip amount, which was inappropriate for the ABIC-based inversion. In this study, we propose a new ABIC-based joint inversion scheme using tsunami waveform and geodetic data.
For the data weight, we followed Funning et al (2014); employing the relative variances of GNSS data and SAR data to tsunami data. In addition, our inversion method considered the covariance component, the variance of Green’s function, and the direct constraint (Matu’ura et al., 2007; Fukahata and Write, 2008; Yagi and Fukahata, 2011). Since the number of hyperparameters was increased, we used the Optuna library to search for the minimum value of ABIC. Note that, in the inversion analysis, we used tsunami records not in the time domain but in the frequency domain by the Fast Fourier Transform.
As a result, we obtained the finite fault model with large amounts of slip at NT4 and NT6, the northeast and northern faults of the Noto Peninsula. This model is consistent with the previous studies such as Fujii and Satake (2024) and Okuwaki et al. (2024). While we did not use the non-negative constraint in this inversion, the minimum amplitude was less than -0.5 m. We therefore considered that our proposed inversion scheme succeeded in finding the optimal weights of constraints.