[S-SS09] Seismic wave propagation: Theory and Application

[SSS09-P07] Crack source and volcanic structural heterogeneity inferred from oscillatory waveforms of long-period seismic events: Application at Kusatsu-Shirane Volcano

*Nakano Masayuki¹, Hiroyuki Kumagai¹ (1.Nagoya University)

Long-period (LP) events observed at active volcanoes are interpreted as oscillations of fluid-filled resonators in magma and hydrothermal systems. The fluid-filled crack model has been used to interpret LP waveforms. However, the crack source was estimated for only specific LP events due to technical difficulty in synthetic waveform calculations and poor understanding of volcano heterogeneity. To solve these problems, we developed a method to estimate crack sources of LP events and volcano heterogeneity easily by using multiple spectral peaks in LP events. We used peak-to-peak amplitudes of observed and synthetic waveforms to obtain stable inversion results. First, we performed numerical tests to verify the waveform inversion method using the analytical solution of Green's functions in an infinite homogeneous medium. We used OpenSWPC (Maeda et al., EPS, 2017) based on a finite difference method (FDM) accommodating topography and heterogeneity, and calculated synthetic waveforms for a horizontal crack, which were regarded as observed waveforms. By assuming the source mechanism as a crack, we conducted inversion using synthetic waveforms calculated with the analytical solution. As a result, the source depth was estimated to be shallower, but the crack inclination was estimated to be within errors of 10º. Thus, the analytical solution may be used to estimate LP sources, although there are some errors. Next, we applied our inversion method to an LP event observed on 26 April, 1992 at Kusatsu-Shirane volcano, Japan. Our result indicates that the estimated crack angles are similar to those of Nakano and Kumagai (GRL, 2005) which used synthetic waveforms calculated with a homogeneous half-space, although the source depth is slightly shallower than the previous estimate. Our application of the inversion method to other three LP events at this volcano, of which source mechanisms were not estimated before, indicates that the crack angles of the four LP events are similar, suggesting that these events were excited by the same crack. Taguchi et al. (JGR, 2018) considered that steam generated by degassing from magma entered into a crack repeatedly. Therefore, our results support this source model. By assuming that the source mechanism for multiple spectral peaks is same as the crack for the lowest frequency peak, we estimated normalized residuals by fitting observed waveforms at individual spectral peaks to synthetic waveforms calculated with the analytical solution. We found that the estimated residuals increased with increasing frequency. To evaluate this trend, we conducted FDM simulations using a von Kármán-type heterogeneity model characterized by the correlation distance (a) and heterogeneity fluctuation (ε). Volcano heterogeneity is characterized by smaller a and larger ε than those in the crust, and a = 50 m and ε = 0.2 were used by Morioka et al. (JGR, 2017). Wegler (JGR, 2004) and Kumagai et al. (JGR, 2018) indicated that strongly heterogeneous layers at volcanoes are limited to a depth of 1 km. Using a heterogeneous layer with a = 50 m and ε = 0.2 to a depth of 1 km from topography surface, we calculated synthetic waveforms with FDM for individual spectral peaks in 3−8 Hz, and fitted them to the synthetic waveforms calculated by the analytical solution. Their normalized residuals were, however, much smaller than those estimated for the observed LP events. The mean free path (l₀) for a = 50 m and ε = 0.2 around 5−10 Hz is about 1700 m, but l₀ at volcanoes around 5−10 Hz has been estimated to be about 100−200 m (Wegler, JGR, 2004; Kumagai et al., JGR, 2020). The heterogeneity used in our simulations may not be appropriate and weaker than the actual heterogeneity. We cannot use large values of ε because such FDM calculations are unstable. Monte Carlo simulations (e.g., Yoshimoto, JGR, 2000) based on the radiative transfer theory may be required to estimate the heterogeneity.