2:30 PM - 2:45 PM
[SSS07-12] Application of MLTWA on small spatial scale using swarm in Hakone region
Separation of the scattering and intrinsic attenuation from the observed total attenuation is important to assess these physical properties of the attenuation. Multiple Lapse Time Window Analysis (MLTWA; Fehler et al., 1992; Hoshiba, 1993) is a quantitative estimation method of Qs-1 (scattering loss) and Qi-1 (intrinsic absorption). This method utilizes the characteristics of the two attenuation mechanisms to separate them; different lapse time of seismograms are affected by different attenuation mechanisms. Most of the previous researches use the seismic data with hypocentral distances from 20 km to 100 km in MLTWA, while short hypocentral distances of seismic data observed by dense local network is also important to know the attenuation characteristics in smaller scale.
Matsuda et al. (2023, SSJ) evaluated availability of MLTWA in small scale region by applying the shorter time windows that are considered as appropriate to extract the attenuation information. In this research, we further examine MLTWA with shorter distance events and estimate the attenuation parameters in Hakone region.
We analyze 4373 microearthquakes with -1.7 < M < 3.4 occurred in Hakone region from 20 April 2015 to 31 October 2015. Hot Springs Research Institute have been deployed dense seismic network to observe the seismic and volcanic activities around Hakone region. We use three component velocity seismograms obtained at eight short period seismographs with sampling rate of 200 Hz. Yukutake et al. (2015) relocated the hypocenters with high accuracy by double difference method (Waldhauser and Ellsworth, 2000).
We set three time windows whose starting point corresponds S-wave arrival time. Then, we compare the observed and synthetic integrated energy simultaneously in all time windows. At this time, the window length should be appropriately chosen to distinguish between direct and scattered waves. We use the same window length as Matsuda et al. (2023, SSJ) because they recommended the window length from 1.5 sec to 5.0 sec. These window lengths are suitable for microseismic data with short hypocentral distances since their amplitudes decay rapidly. We apply five bandpass filters (2-4 Hz, 4-8 Hz, 8-16 Hz, 16-32 Hz, 32-64 Hz) to the seismograms and synthesize mean square envelopes. We assume a uniform S-wave velocity structure with 3.2 km/sec. We applied grid search to estimate the attenuation parameters. The search range of each parameter is 10-5 < Q-1 <10-1, and the grid interval is equal in logarithmic as logQ-1=0.1. We evaluate the estimation error of Qi-1 and Qs-1 100 times based on the bootstrap method.
We obtain estimates of Qs-1 and Qi-1 for each frequency band and window length. Estimation accuracy is especially high in low frequency band. When frequency band is 4-8 Hz with window length of 3.0 sec, we estimate Qs-1to be 1.26x10-2 and Qi-1 to be 1.00x10-2. These values indicate stronger attenuation compared to previously studied other regions (Sato et al., 2012). Similar trend is observed for other frequency bands. This result is consistent with the observation that the attenuation parameters in volcanic area are larger than those of other area. Qs-1 is larger than Qi-1 in high frequency (8 Hz <).
In the case of very short time window, we obtain larger attenuation in high frequency. For example, with window length of 1.5 sec, Qs-1 is 4.52x10-4, while with window length of 2.0 sec, Qs-1 is 2.67x10-4. Since window length has direct relationship with the scale of analysis region, the difference of the attenuation value among window lengths implies that attenuation parameters in Hakone are non-uniform. This relationship between time window and attenuation parameters is more remarkable for Qs-1 than Qi-1.
To summarize, we obtain stronger scattering and intrinsic attenuation compared to average value of other regions. The difference of estimated attenuation parameters caused by variation of the window length imply the potential to estimate the spatial structures of attenuation.
Matsuda et al. (2023, SSJ) evaluated availability of MLTWA in small scale region by applying the shorter time windows that are considered as appropriate to extract the attenuation information. In this research, we further examine MLTWA with shorter distance events and estimate the attenuation parameters in Hakone region.
We analyze 4373 microearthquakes with -1.7 < M < 3.4 occurred in Hakone region from 20 April 2015 to 31 October 2015. Hot Springs Research Institute have been deployed dense seismic network to observe the seismic and volcanic activities around Hakone region. We use three component velocity seismograms obtained at eight short period seismographs with sampling rate of 200 Hz. Yukutake et al. (2015) relocated the hypocenters with high accuracy by double difference method (Waldhauser and Ellsworth, 2000).
We set three time windows whose starting point corresponds S-wave arrival time. Then, we compare the observed and synthetic integrated energy simultaneously in all time windows. At this time, the window length should be appropriately chosen to distinguish between direct and scattered waves. We use the same window length as Matsuda et al. (2023, SSJ) because they recommended the window length from 1.5 sec to 5.0 sec. These window lengths are suitable for microseismic data with short hypocentral distances since their amplitudes decay rapidly. We apply five bandpass filters (2-4 Hz, 4-8 Hz, 8-16 Hz, 16-32 Hz, 32-64 Hz) to the seismograms and synthesize mean square envelopes. We assume a uniform S-wave velocity structure with 3.2 km/sec. We applied grid search to estimate the attenuation parameters. The search range of each parameter is 10-5 < Q-1 <10-1, and the grid interval is equal in logarithmic as logQ-1=0.1. We evaluate the estimation error of Qi-1 and Qs-1 100 times based on the bootstrap method.
We obtain estimates of Qs-1 and Qi-1 for each frequency band and window length. Estimation accuracy is especially high in low frequency band. When frequency band is 4-8 Hz with window length of 3.0 sec, we estimate Qs-1to be 1.26x10-2 and Qi-1 to be 1.00x10-2. These values indicate stronger attenuation compared to previously studied other regions (Sato et al., 2012). Similar trend is observed for other frequency bands. This result is consistent with the observation that the attenuation parameters in volcanic area are larger than those of other area. Qs-1 is larger than Qi-1 in high frequency (8 Hz <).
In the case of very short time window, we obtain larger attenuation in high frequency. For example, with window length of 1.5 sec, Qs-1 is 4.52x10-4, while with window length of 2.0 sec, Qs-1 is 2.67x10-4. Since window length has direct relationship with the scale of analysis region, the difference of the attenuation value among window lengths implies that attenuation parameters in Hakone are non-uniform. This relationship between time window and attenuation parameters is more remarkable for Qs-1 than Qi-1.
To summarize, we obtain stronger scattering and intrinsic attenuation compared to average value of other regions. The difference of estimated attenuation parameters caused by variation of the window length imply the potential to estimate the spatial structures of attenuation.