[S-CG45] Science of slow-to-fast earthquakes

[SCG45-P25] Broadband source time function estimations of seismic slow earthquakes at the shallow plate boundary in Nankai

*Shunsuke Takemura¹, Kentaro Emoto², Suguru Yabe³ (1.Earthquake Research Institute, the University of Tokyo, 2.Kyusyu University, 3.National Institute of Advanced Industrial Science and Technology)

Keywords:Slow earthquake, Nankai Trough, Source time function

We tried to estimate broadband source time functions of shallow slow earthquakes in Nankai, which can reproduce observed seismograms in both frequency bands of very low frequency earthquakes (VLFEs) and tectonic tremors. To achieve precise estimations using DONET data, we prepared Green’s function dataset in the local 3D model of Takemura et al. (2020) using OpenSWPC (Maeda et al. 2017). In our method for STF estimation, we assume the broadband characteristics of seismic slow earthquakes (e.g., Ide et al., 2008; Masuda et al., 2020). First, we estimated STFs of shallow VLFEs (0.03-0.06 Hz) via the simulated annealing method (e.g., Takemura et al., 2022ab). Due to near-source DONET observation, we well constrained STFs and locations of shallow VLFEs compared to previous studies (Takemura et al., 2022ab; Yamamoto et al., 2022).
However, estimated VLFE STFs cannot reproduce tremor envelopes due to a lack of high-frequency components. Then, we added random temporal fluctuations into an obtained smooth VLFE STF. This approach is similar to previous studies for regular earthquakes (e.g., Koyama, 1985; Hisada, 2000). We assumed that fluctuations of source time function can be characterized by a von Kármán autocorrelation function, which is controlled by characteristic time of t₀, RMS value of e, and decay rate of k. A decay rate parameter (k) of a von-Kármán autocorrelation function (ACF) can model various decay rates at high-frequency ranges. We tested the case of a von-Kármán ACF with e²/t₀ = 0.04 and k = 0.1. High-frequency seismic energies from STFs in this case were several tense times larger than those without temporal fluctuations.
After correcting site amplifications at DONET stations, based on the grid search inversion, we will estimate the optimal parameter set of e²/t₀ and k, which can reproduce both VLFE seismograms and tremor envelopes.