10:45 〜 12:15
[SSS10-P17] Double-difference法による大量地震の効率的な震源再決定と高精度震源分布のリアルタイム・モニタリング
キーワード：DD法
Precise earthquake hypocenter information provides important clues regarding external forces that drive seismicity, such as fluid diffusion and aseismic slip (e.g., Yoshida & Hasegawa, 2018; Matsumoto et al., 2021). To obtain highly accurate hypocenter information, it is important to use relative hypocenter relocation methods with large amounts of arrival time difference data of seismic waves between different earthquakes. The widely used double-difference method (Waldhause & Ellsworth, 2002) simultaneously determines the hypocenters of all earthquakes in a cluster, thus effectively using the arrival time difference information precisely determined by waveform correlation. On the other hand, when this type of method is used for a large number of earthquakes (>30000), the computational cost becomes very high, and the computation takes a great deal of time (maybe weeks). Furthermore, when new earthquakes are added to the data set, it is necessary to relocate all the earthquakes again, not just the new ones, which is time consuming. In recent years, it has become possible to detect very small earthquakes, but it is often challenging to fully utilize their location information due to this computational cost problem.
In order to efficiently relocate a large number of events one after another, we propose the use of a double-difference algorithm with a simple modification that fixes the hypocenters of specified earthquake. We first divide earthquakes used in the DD algorithm into two groups: (1) those for which the hypocenters are to be newly relocated and (2) those for which the hypocenters are not to be relocated but used for reference. From the unknown parameters (hypocenter and origin time for each earthquake), we remove the parameters related to the earthquakes in the second group. Still, we use time difference data obtained both within the first group and between the first group and the second group. By combining this function with the following algorithm, we reduce computational cost and attempt to quasi-real time monitoring of relocated hypocenters. Note that Waldhauser et al. (2009) approximate the hypocenter fixation by introducing a high damping parameter. In the present study, the parameters related to fixed earthquakes are directly removed from the equations so that a larger number of earthquakes can be relocated at once.
We use the following procedure to relocate all the earthquakes in a target cluster. First, all events are divided into multiple groups consisting of a small number (e.g., 2000) of events. Next, within any one earthquake group, the DD method is applied in the usual way. The hypocenters of the next earthquake group are then relocated with time-difference information among this group and those with the earthquakes that have already relocated (they are fixed at this time). The determination of earthquake groups and the order of relocation is arbitrary, but assuming quasi-real-time monitoring, it is conceivable to group the earthquakes in the order they occurred and relocate them in that order. By sequentially adding newly occurring events, quasi-real-time hypocenter relocation is possible.
We applied this algorithm to the earthquake swarm that has been occurring in the northeastern Noto Peninsula since around the end of 2020. Yoshida et al. (2022, SSJ) relocated the hypocenters in the region but limited their coverage to M>1 events. As the result of our new relocation, we estimated the hypocenters of 35,000 events without discarding very small events (M<1). The resulting hypocenters revealed a detailed fault structure and movement, which helps understand the mechanism of their occurrences.
The fixation of earthquake hypocenters in the DD method is very simple but could be used for different purposes. For example, it would be possible to use events whose absolute locations are independently constrained (such as repeating earthquakes or earthquakes whose depth is well constrained by depth phase) to relocate nearby events.
In order to efficiently relocate a large number of events one after another, we propose the use of a double-difference algorithm with a simple modification that fixes the hypocenters of specified earthquake. We first divide earthquakes used in the DD algorithm into two groups: (1) those for which the hypocenters are to be newly relocated and (2) those for which the hypocenters are not to be relocated but used for reference. From the unknown parameters (hypocenter and origin time for each earthquake), we remove the parameters related to the earthquakes in the second group. Still, we use time difference data obtained both within the first group and between the first group and the second group. By combining this function with the following algorithm, we reduce computational cost and attempt to quasi-real time monitoring of relocated hypocenters. Note that Waldhauser et al. (2009) approximate the hypocenter fixation by introducing a high damping parameter. In the present study, the parameters related to fixed earthquakes are directly removed from the equations so that a larger number of earthquakes can be relocated at once.
We use the following procedure to relocate all the earthquakes in a target cluster. First, all events are divided into multiple groups consisting of a small number (e.g., 2000) of events. Next, within any one earthquake group, the DD method is applied in the usual way. The hypocenters of the next earthquake group are then relocated with time-difference information among this group and those with the earthquakes that have already relocated (they are fixed at this time). The determination of earthquake groups and the order of relocation is arbitrary, but assuming quasi-real-time monitoring, it is conceivable to group the earthquakes in the order they occurred and relocate them in that order. By sequentially adding newly occurring events, quasi-real-time hypocenter relocation is possible.
We applied this algorithm to the earthquake swarm that has been occurring in the northeastern Noto Peninsula since around the end of 2020. Yoshida et al. (2022, SSJ) relocated the hypocenters in the region but limited their coverage to M>1 events. As the result of our new relocation, we estimated the hypocenters of 35,000 events without discarding very small events (M<1). The resulting hypocenters revealed a detailed fault structure and movement, which helps understand the mechanism of their occurrences.
The fixation of earthquake hypocenters in the DD method is very simple but could be used for different purposes. For example, it would be possible to use events whose absolute locations are independently constrained (such as repeating earthquakes or earthquakes whose depth is well constrained by depth phase) to relocate nearby events.