9:45 AM - 10:00 AM
[SSS06-04] Near-field Waveform Modeling and Source Characterization of Small Earthquakes in Northern Ibaraki Prefecture
Keywords:source process, moment rate function, near-field term, moment tensor inversion
In the northern part of Ibaraki Prefecture, seismicity has been highly active in the crust since the 2011 M9 Tohoku Earthquake. In this area, two Hi-net stations are located within a few kilometers of the epicenters (, although one channel was not measured correctly for a while after the Tohoku earthquake). The observed waveforms at these near-source stations are very simple, and we can readily see the near-field term and surface reflection above the downhole sensor. It appears possible to obtain information on the source time information of earthquakes in detail directly from the observed waveforms. This talk will report our attempts to obtain information about the source processes of small earthquakes and crustal structure based on this near-field observation data.
First, we roughly compared the observed waveforms of Mw3.5-4.5 earthquakes in the NIED F-net catalog with synthetic waveforms. The synthetic waveforms were computed using moment tensor solution and a simple 1-D velocity structure by Kubo et al. (2002). We used the code of Zhu & Rivera (2002) to calculate the Green functions. The instrumental responses were removed from the observed waveforms. We confirmed that observed waveforms (displacement) are generally consistent with the synthetic waveforms even without low-pass filtering.
We then refined the 1D velocity model to allow more detailed comparisons between the synthetic and observed waveforms. We used the joint determination method of Kissling (1994) to determine the 1-D velocity model and hypocenters of 2000 M>2.5 earthquakes in this region. We also relocated hypocenters of 111418 earthquakes in the JMA catalog based on the obtained 1-D velocity model. Due to the use of nearby observations, we obtained a high-resolution hypocenter distribution without using a relative hypocenter relocation method. We recognize a large, west-dipping planar structure (approximately 10 km) and multiple smaller planar structures (approximately a few km) oriented in various directions. We adjusted the Q-values so that the details of observed waveforms of the MJMA<2 events match those of the synthetic waveforms.
Using the refined model, we performed the moment tensor inversion for the MJMA1-4 earthquakes with a single station (N.JUOH) in the frequency band of 0.5-1.4 Hz. This frequency band is much higher than the frequency band usually used for moment tensor inversion in Japan (< 0.1 Hz). The higher frequency band allows for the analysis of small earthquakes. The moment tensor solutions were estimated via grid search, assuming a double-coupled source. Variance reductions exceed 70% for 3384 of 4123 MJMA>2 events analyzed. The obtained focal mechanisms are in good agreement with the stress field and independently estimated focal mechanisms from first-motion polarity data. As is often reported, a systematic tendency for MJMA to be smaller than Mw was observed for earthquakes with MJMA<3.
We computed the synthetic waveform with a short-duration source for each MJMA2-4 earthquake using the estimated moment tensor. The source time function of each earthquake is estimated by the deconvolution of the observed waveforms with the synthetic waveforms. We used the entire frequency range from 0.2 Hz to 45 Hz. The results clearly show that Mw2-4 earthquakes have various shapes of source time functions. Many of them are characterized by a triangular one-pulse shape, roughly in agreement with the source model commonly used for small earthquakes. However, a few tens of percent of the earthquakes show significantly more complex source time functions. These features differ significantly from the simplistic source models (e.g., Brune, 1970; Sato & Hirasawa, 1973) often used for small earthquakes. This result implies that an analysis that does not assume a specific source time function shape is desirable to characterize small earthquake sources.
First, we roughly compared the observed waveforms of Mw3.5-4.5 earthquakes in the NIED F-net catalog with synthetic waveforms. The synthetic waveforms were computed using moment tensor solution and a simple 1-D velocity structure by Kubo et al. (2002). We used the code of Zhu & Rivera (2002) to calculate the Green functions. The instrumental responses were removed from the observed waveforms. We confirmed that observed waveforms (displacement) are generally consistent with the synthetic waveforms even without low-pass filtering.
We then refined the 1D velocity model to allow more detailed comparisons between the synthetic and observed waveforms. We used the joint determination method of Kissling (1994) to determine the 1-D velocity model and hypocenters of 2000 M>2.5 earthquakes in this region. We also relocated hypocenters of 111418 earthquakes in the JMA catalog based on the obtained 1-D velocity model. Due to the use of nearby observations, we obtained a high-resolution hypocenter distribution without using a relative hypocenter relocation method. We recognize a large, west-dipping planar structure (approximately 10 km) and multiple smaller planar structures (approximately a few km) oriented in various directions. We adjusted the Q-values so that the details of observed waveforms of the MJMA<2 events match those of the synthetic waveforms.
Using the refined model, we performed the moment tensor inversion for the MJMA1-4 earthquakes with a single station (N.JUOH) in the frequency band of 0.5-1.4 Hz. This frequency band is much higher than the frequency band usually used for moment tensor inversion in Japan (< 0.1 Hz). The higher frequency band allows for the analysis of small earthquakes. The moment tensor solutions were estimated via grid search, assuming a double-coupled source. Variance reductions exceed 70% for 3384 of 4123 MJMA>2 events analyzed. The obtained focal mechanisms are in good agreement with the stress field and independently estimated focal mechanisms from first-motion polarity data. As is often reported, a systematic tendency for MJMA to be smaller than Mw was observed for earthquakes with MJMA<3.
We computed the synthetic waveform with a short-duration source for each MJMA2-4 earthquake using the estimated moment tensor. The source time function of each earthquake is estimated by the deconvolution of the observed waveforms with the synthetic waveforms. We used the entire frequency range from 0.2 Hz to 45 Hz. The results clearly show that Mw2-4 earthquakes have various shapes of source time functions. Many of them are characterized by a triangular one-pulse shape, roughly in agreement with the source model commonly used for small earthquakes. However, a few tens of percent of the earthquakes show significantly more complex source time functions. These features differ significantly from the simplistic source models (e.g., Brune, 1970; Sato & Hirasawa, 1973) often used for small earthquakes. This result implies that an analysis that does not assume a specific source time function shape is desirable to characterize small earthquake sources.