Quasi-Real-Time Hypocenter Relocation and Monitoring in the Northeastern Noto Peninsula

doi:10.21203/rs.3.rs-4645791/v1

Download PDF

Research Article

Quasi-Real-Time Hypocenter Relocation and Monitoring in the Northeastern Noto Peninsula

https://doi.org/10.21203/rs.3.rs-4645791/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

The seismicity rate markedly increased in the northeastern Noto Peninsula of Ishikawa Prefecture around the end of 2020, with an M_w6.2 event on 5 May, 2023, followed by many aftershocks. Previous earthquake relocation studies have detected upward migration of microearthquakes via multiple faults and clusters, suggesting the involvement of crustal fluids in this sequence. Since some active faults exist near the source region, there was concern that the sequence could lead to a larger earthquake; this became a reality with the M_w7.5 earthquake on 1 January 2024. The objective of this study is to develop an algorithm to precisely relocate the microearthquake hypocenters in quasi-real time for better monitoring. A fine view of seismicity requires relative relocation methods such as the Double-Difference (DD) method with numerous and accurate arrival time difference data derived from the waveform correlation analysis. However, the standard DD method has the disadvantage of huge computational costs when data increases, making it unsuitable for real-time monitoring in such situations. We developed a quasi-real-time algorithm that divides earthquake data into multiple time windows and performs the DD relocation each time new time window data is added. The major improvement is that our method incorporates a traditional simple relative relocation method and preserves constraints between different time windows; the relative locations of new events are constrained from reference events that were already relocated in the previous time windows. We tested a daily relocation algorithm on 11,546 events from 19 June, 2022, to 31 May, 2023, in the Noto Peninsula earthquake sequence. We found that our modification substantially reduced artificial hypocenter offsets between different time windows and succeeded in resolving the fine fault structures from the cloud-like distribution of initial hypocenters. If we do not impose constraints between different windows, the relocated hypocenters are scattered and do not show fine planar structures. Moreover, our algorithm greatly reduces the computational cost, allowing for quasi-real-time earthquake relocation and monitoring. We hope this algorithm will help monitor the spatio-temporal distribution of future earthquake sequences.

Real-time monitoring

Double-Difference earthquake relocation

Noto Peninsula

Earthquake swarm

Seismicity

The northeastern part of the Noto Peninsula in Ishikawa Prefecture, Japan, has been experiencing pronounced seismicity since December 2020 (Fig. 1). It first showed the appearance of an earthquake swarm (Nakajima et al., 2022; Amezawa et al., 2023; Yoshida et al., 2023a; Nishimura et al., 2023; Okada et al., 2024), and then an M_w6.2 event occurred on 5 May, 2023, followed by many aftershocks. The Japan Meteorological Agency (JMA) unified earthquake catalog lists more than 20,000 Mj ≥ 1.0 events over approximately three years between December 2020 and December 2023 (Fig. 1b), including eight Mj ≥ 5.0 events (four of which occurred on 5 May, 2023), where Mj is the local magnitude scale used by the JMA.

Notably, the source region migrated over the past three years during the swarm period. Previous studies have suggested the involvement of crustal fluids in this swarm (Nakajima et al., 2022; Amezawa et al., 2023; Nishimura et al., 2023; Yoshida et al., 2023a and b; Kato et al., 2024). The precise hypocenter relocation results revealed that seismicity moved from deep to shallow via multiple planar structures (Yoshida et al., 2023a and b; Kato et al., 2024).

Several kilometers offshore north of the source region was the trace of an active fault called the Suzu-Oki segment Fault (Inoue & Okamura, 2010). Still, the earthquakes before 2024 occurred on faults different from the Suzu-Oki segment (Yoshida et al., 2023a and b; Kato et al., 2024). There was concern that this sequence could lead to a larger earthquake. This became a reality on 1 January 2024. On that day, the M_w 7.5 2024 Noto Peninsula Earthquake occurred and caused significant damage in a wide area of the Hokuriku region, including Wajima and Shika, Ishikawa Prefecture. The earthquake had a maximum intensity of 7 and claimed more than 200 victims.

The objective of this study is to develop an algorithm to precisely relocate the microearthquake hypocenters in quasi-real time for better monitoring in the above situation. The Japan Meteorological Agency (JMA) releases automatically processed hypocenters immediately after the earthquake and the preliminary JMA unified hypocenter catalog about half a day to two days later. These are very important as basic information on very recent activity. However, the above catalogs locate hypocenters based on picked arrival time data, from which it is difficult to accurately assess the fault structure and migration of shallow earthquakes. Indeed, unless a station is fortunately located directly above the source region, the hypocenter distribution of shallow earthquakes obtained by the standard method is usually blurred due to estimation errors (e.g., Yoshida et al., 2016). The combination of precise arrival time difference data derived from waveform correlation analysis with a relative earthquake relocation method is necessary to resolve the fine fault structure of shallow earthquakes.

One of the most commonly used methods for earthquake relocation is the Double-Difference (DD) method (Waldhauser & Ellsworth, 2000), and previous studies in the region (Yoshida et al., 2023a and b; Kato et al., 2024) and other regions around the world (Waldhauser & Ellsworth, 2002; Fukuyama et al., 2003; Hauksson & Shearer, 2005; Yukutake et al., 2011 and 2022; Shelly & Hill, 2011; Shelly et al., 2013; Naoi et al., 2015; Ruhl et al., 2016; Ross et al., 2017; Yoshida & Hasegawa, 2018; De Barros et al., 2019; Hatch et al., 2020) have employed this method. Waldhauser et al. (2020) used this method for real-time hypocenter relocation in Axial Seamount. This method determines the earthquake locations using the differences in arrival times between earthquake pairs and is characterized by its simultaneous relocation of all the earthquakes in a cluster with numerous data. This method gives the most accurate relative hypocenter distribution possible from the data and can be regarded as one extreme of the traditional relative hypocenter relocation approach (e.g., Poupinet et al., 1984). However, the DD method can be computationally expensive because it solves the matrix equations for the differential arrival time data of many earthquake pairs at multiple stations. For example, Yoshida et al. (2023b) used more than 100 million differential arrival time data to relocate more than 20000 earthquakes in the Noto swarm region, and their approach of relocating all earthquakes at once makes real-time monitoring difficult. If the seismic data is further massive, this method requires so much computer memory that it may be challenging to execute in the first place. These can be weaknesses of the standard DD method when quasi-real-time processing is required.

In this study, we first developed a quasi-real-time, precise earthquake relocation algorithm based on the DD method, which can be used in the above situation of extensive data. We then applied the algorithm to the seismicity in the northeastern Noto Peninsula and evaluated its performance.

Our strategy for quasi-real-time precise earthquake relocation is to divide the target earthquakes into fine time windows and perform relocation each time new time window data is added. However, as it is, this approach does not constrain the relative locations of earthquakes in different time windows, and the shorter the time window, the larger the error in the relative locations in the overall distribution. We overcome this problem by incorporating a traditional simple relative relocation method; we constrain the relative locations of events in the new window from reference events already relocated in the previous time windows. To implement this, we first modify the equation used in the DD method (Waldhauser and Ellsworth, 2000) (Subsection 2.1) and then describe the quasi-real-time relocation algorithm (Subsection 2.2).

2.1. DD method with reference events

The arrival time ${t}_{k}^{i}$ of the seismic wave of a given event $i$ at a given station $k$ can be expressed as the integral along the path $s$ as

$$\begin{array}{c}{t}_{k}^{i}={\tau }^{i}+{\int }_{s}^{}u\left(s\right)ds,\#\left(1\right)\end{array}$$

where ${\tau }^{i}$ is the origine time, and $u\left(s\right)$ is the slowness. We first set the initial hypocentral parameters ${\varvec{m}}^{i}=({x}^{i},{y}^{i},{z}^{i},{\tau }^{i})$, Taylor expand ${t}_{k}^{i}$ around ${\varvec{m}}^{i}$ at the first order to linearize it and obtain the time difference ${r}_{k}^{i}$ for a slight change in the hypocentral parameters $\varDelta {\varvec{m}}^{i}={\left(\varDelta {x}^{i},\varDelta {y}^{i},\varDelta {z}^{i},\varDelta {\tau }^{i}\right)}^{T}$.

$$\begin{array}{c}{r}_{k}^{i}=\frac{\partial {t}_{k}^{i}}{\partial x}\varDelta {x}^{i}+\frac{\partial {t}_{k}^{i}}{\partial y}\varDelta {y}^{i}+\frac{\partial {t}_{k}^{i}}{\partial z}\varDelta {z}^{i}+\varDelta {\tau }^{i}=\frac{\partial {t}_{k}^{i}}{\partial \varvec{m}}\bullet \varDelta {\varvec{m}}^{i},\#\left(2\right)\end{array}$$

The DD method considers earthquake pairs $i$ and $j$. By using ${dr}_{k}^{ij}={r}_{k}^{i}-{r}_{k}^{j}$ at the same station $k$,

$$\begin{array}{c}{dr}_{k}^{ij}=\frac{\partial {t}_{k}^{i}}{\partial \varvec{m}}\bullet \varDelta {\varvec{m}}^{i}-\frac{\partial {t}_{k}^{j}}{\partial \varvec{m}}\bullet \varDelta {\varvec{m}}^{j}.\#\left(4\right)\end{array}$$

This equation is matrixed for all stations and earthquake pairs as follows:

$$\begin{array}{c}\left(\begin{array}{c}d{r}_{{S}_{1}}^{{E}_{1}{E}_{2}}\\ d{r}_{{S}_{1}}^{{E}_{1}{E}_{3}}\\ ⋮\\ d{r}_{{S}_{L}}^{{E}_{N-1}{E}_{N}}\end{array}\right)=\left(\begin{array}{ccccc}\frac{\partial {t}_{{S}_{1}}^{{E}_{1}}}{\partial \varvec{m}}& -\frac{\partial {t}_{{S}_{1}}^{{E}_{2}}}{\partial \varvec{m}}& 0& \cdots & 0\\ \frac{\partial {t}_{{S}_{1}}^{{E}_{1}}}{\partial \varvec{m}}& 0& -\frac{\partial {t}_{{S}_{1}}^{{E}_{3}}}{\partial \varvec{m}}& \cdots & 0\\ ⋮& ⋮& ⋮& \ddots & ⋮\\ 0& 0& 0& \cdots & -\frac{\partial {t}_{{S}_{L}}^{{E}_{N}}}{\partial \varvec{m}}\end{array}\right)\left(\begin{array}{c}\varDelta {\varvec{m}}^{{E}_{1}}\\ \varDelta {\varvec{m}}^{{E}_{2}}\\ \varDelta {\varvec{m}}^{{E}_{3}}\\ ⋮\\ \varDelta {\varvec{m}}^{{E}_{N}}\end{array}\right), \#\left(5\right)\end{array}$$

where ${E}_{i} (i=\text{1,2},\dots ,N)$ shows the event number and ${S}_{i} (i=\text{1,2},\dots ,L)$ shows the station number. We denote the first matrix on the right-hand side by $\mathbf{G}$ and the second vector by $\mathbf{m}$.

The data of the standard hypocenter relocation method is the time residuals ${r}_{\text{o}\text{b}\text{s}}={t}^{obs}-{t}^{cal}$ between observed arrival time, ${t}^{obs}$, and calculated arrival time, ${t}^{cal}$, based on the initial hypocenter parameters as data. Unlike that, the DD method uses the difference between earthquake pairs:$\begin{array}{c}{dr}_{{k}_{obs}}^{ij}={r}_{{k}_{obs}}^{i}-{r}_{{k}_{obs}}^{j}={\left({t}^{obs}-{t}^{cal}\right)}_{k}^{i}-{\left({t}^{obs}-{t}^{cal}\right)}_{k}^{j}={\left({t}_{k}^{i}-{t}_{k}^{j}\right)}^{obs}-{\left({t}_{k}^{i}-{t}_{k}^{j}\right)}^{cal},\#\left(6\right)\end{array}$

which is called DD (Double Difference). The following matrix equation is solved to obtain hypocentral parameter $\mathbf{m}$.

$$\begin{array}{c}{\mathbf{d}}_{\text{o}\text{b}\text{s}}=Gm,\#\left(6\right)\end{array}$$

where ${\mathbf{d}}_{\text{o}\text{b}\text{s}}$ is a single-column vector including ${dr}_{{k}_{obs}}^{ij}$. $\mathbf{m}$ is a single-column vector with $\varDelta {\varvec{m}}^{i}$, and $\mathbf{G}$ is a matrix with the difference of partial differential coefficients, corresponding to the right-hand side of Eq. (5). Actual calculations are performed by applying different weights to each piece of data. The LSQR method (Paige & Saunders, 1982) is used when dealing with extensive data.

In Eq. (6), $\mathbf{G}$ is a massive matrix of [the number of components of DD] × [four times the number of events], which makes the method computationally expensive. As described above, we split the earthquake data into multiple time windows and perform sequential DD relocation for each new time window. This reduces the matrix size considerably, but, as it is, the information on the relative earthquake locations in different windows is also lost. To avoid losing this information, we use events whose hypocentral parameters have already been determined in past time windows as reference events. We fix the hypocenter parameters of these reference events but use the differential arrival time data between them and new events to relocate the new events. The residuals for the new and reference event pairs are minimized by relocating the hypocentral parameters of the new events.

To do so, we change the equation to be solved from Eq. (5), as shown in Fig. 2. We first remove the rows of $\mathbf{m}$ corresponding to the reference events and include only the hypocentral parameters of the new events in $\mathbf{m}$. From ${\mathbf{d}}_{\text{o}\text{b}\text{s}}$, we removed rows containing differential arrival time data only between reference events, but retained the differential data only between new events as well as those between new and reference events. Thus, the size of $\mathbf{G}$ is significantly reduced.

2.2. Relocation Algorithm

In preparation for quasi-real-time relocation, our algorithm first relocates the hypocenters of previous events and creates a reference database. When a new time window containing new events is added, these events are relocated using the method described in Subsection 2.1, with reference events from the database. The newly relocated hypocenters are then added to the reference database, and the above procedure is repeated each time a new time window is added.

The number of reference events increases for later time windows, which becomes much larger than that of new events. Our modification removes the unknown parameters of reference events, which significantly reduces the size of the matrix $\mathbf{G}$ compared to the standard method of simultaneously solving for all the hypocentral parameters. As a result, computational efficiency is improved, and memory requirements are reduced.

3 − 1. Algorithm for Daily Relocation

We implement an algorithm that performs waveform processing, waveform correlation analysis, and hypocenter relocation sequentially for each day. The steps of the algorithm in each window are shown in Fig. 3. Step 1 is to create an earthquake list and obtain the arrival times of seismic waveforms, Steps 2 and 3 are to obtain differential arrival time data by waveform correlation analysis, and Step 4 is to perform relocation described in Section 2. This algorithm is intended to perform the relocation as soon as the JMA preliminary catalog becomes available.

As a pre-processing step for each new time window, we first use the preliminary JMA unified hypocenter catalog to create the list of earthquakes and obtain the arrival time of seismic waves at each station to cut out waveform windows in the subsequent step. If the picked arrival time is included in the catalog, we use the value. If not, we computed the arrival time from the initial hypocenter parameters based on the 1-D velocity structure used by the JMA (Ueno et al., 2002).

In waveform processing, we cut out the P- and S-wave windows for each earthquake and station and apply bandpass filters. The time window was set as 0.3 s before arrival to 2.5 s after arrival for the P wave and 4.0 s after arrival for the S wave, respectively. When the S-P time is less than 2.5 s, the P-wave window is truncated to 0.5 s before S-wave arrival. Two bandpass filters are applied: 2.0 Hz to 5.0 Hz and 5.0 Hz to 12.0 Hz.

We then compute the cross-correlation functions of waveforms of nearby earthquake pairs to derive the differential arrival times. Calculations are performed only on pairs of new events or those of new and reference events. We limit our correlation analysis to event pairs with initial hypocenters within 3 km of each other. The analyses are performed for each frequency band, and the differential arrival time data with the higher correlation coefficient is included in our dataset if the coefficient is greater than 0.8.

Finally, we relocate the hypocenters of new events using our modified DD method (Section 2) with all the reference events from the previous time windows. The initial hypocenters of the new events are set at the locations listed in the JMA unified catalog. As arrival time difference data, we use those obtained from arrival times in the JMA preliminary catalog in addition to those obtained by waveform correlation in Step 3. Hypocenters were updated using 20 iterations. We weighted the data derived by cross-correlation 50 times greater than the catalog data to delineate shorter scale (< 3 km) structures.

3 − 2. Application to Earthquake Sequence in the Noto Peninsula

We apply our algorithm to actual seismicity and evaluate its performance. The test is conducted in a retrospective setting for the seismicity in the northeastern Noto Peninsula (Fig. 1) from 19 June 2022 to 31 May 2023 (JST). We assumed a situation where the present is 19 June 2022, and daily hypocenter relocations have been performed since then, continuing for 347 days (348 windows) until 31 May 2023. During this period, an M_w 5.2 event occurred on the first day, and an M_w 6.2 event occurred on 5 May 2023. Earthquakes before the start time of the quasi-real-time monitoring (from 8 March 2003 to 18 June 2022) were relocated in advance as the first window.

The magnitudes of the earthquakes analyzed were limited to Mj ≥ 1.0, and the total number is 20,840, of which 11,546 were events in the quasi-real-time relocation period. The waveforms are derived from 21 stations around the northeastern Noto Peninsula (Fig. 1a) by the JMA, Hi-net of the National Research Institute for Earth Science and Disaster Resilience (NIED), and national universities. All available components of the continuous waveform recording are used. Figure 4(a) shows the number of newly relocated events in each window.

Figure 5 shows an example of the daily relocated hypocenters for the three days from 4 May. At 14:42 on the middle day (5 May), the M_w6.2 event occurred, and the seismicity-active region expanded in one day to the north into an area where only a few events had occurred. The number of events with M_j ≥ 1.0 was ten on 4 May, 791 on 5 May and 964 on 6 May. Our relocated hypocenter distribution on 5 May shows a distinct southeast-dipping planar structure, consistent with the results of previous studies (Yoshida et al., 2023b; Kato et al., 2024).

Figures 6(a) and (b) compare the distribution of hypocenters from 5 May to 6, 2023 between our algorithm results and the JMA hypocenter catalog. Figure 6(c) shows the hypocenters relocated by Yoshida et al. (2023b), who used the standard DD method. Since the standard DD relocation should have the best spatial resolution, we here consider the hypocenters of Yoshida et al. (2023b) to be true. The hypocenter distribution in the JMA catalog (Figs. 6a) does not clearly show fault structures involved in the aftershocks and differs significantly from that in Yoshida et al. (2023b). However, our new algorithm was able to derive the aftershock distribution with an accuracy that allowed us to determine that they were concentrated on a single fault zone, similar to Yoshida et al. (2023b). Thus, our method succeeded in delineating the aftershock fault even when seismicity expands into a region where few events have occurred.

We compiled the relocated hypocenters for each day for the entire quasi-real-time relocation period (after the second window). Figures 7 and compare the distribution thus obtained (Figs. 7b and 8b) with the initial hypocenters (Figs. 7a and 8a). The initial hypocenters appear scattered in three dimensions, whereas our relocated hypocenters are concentrated in multiple plane structures (Figs. 8a and b). In some locations where the geometry of the fault planes could not be determined in the initial hypocenters, we can see subparallel fault alingnments (e.g., cross sections g and i) after the relocation. Figures 7(c) and 8(c) show the results of Yoshida et al. (2023b) in the same period and region. The overall picture generally agrees with our new algorithm's results, showing that our algorithm well constrains the relative locations of earthquakes between different time windows.

To examine the accuracy of the hypocenter relocation in our algorithm, we quantitatively compared the hypocenters in our catalog to those by Yoshida et al. (2023b). We focused on the relative locations of earthquake pairs for each catalog and compared the distances of the same earthquake pairs between the two catalogs. For comparison, the same was done between the initial hypocenters of the JMA catalog and the relocated hypocenters by Yoshida et al. (2023b). Figure 9 shows the histograms of distance differences, showing that the deviations of our relocated hypocenters are generally smaller than those of the initial hypocenters. The median deviation is 0.38 km (Fig. 9b), about half of the initial hypocenter (0.80 km; Fig. 9a).

A case of an earthquake swarm leading to a major (M_w7.5) earthquake occurred on the Noto Peninsula. A detailed study of this sequence is critical to advance our understanding of the characteristics of earthquake swarms that lead to large earthquakes (e.g., Shelly et al., 2024). To utilize the knowledge gained, we need to be able to monitor new cases with the necessary accuracy. Therefore, this study was undertaken to develop a quasi-real-time algorithm for precisely relocating earthquake hypocenters.

The DD relocation method simultaneously determines the location of all earthquakes in a cluster. While simultaneous relocation of all events provides the greatest resolution possible from the data, it can be computationally expensive, which is a drawback when real-time monitoring is required. The present algorithm aims to reduce the number of unknown parameters and data by fixing the locations of previously relocated events. An essential factor is constraining the relative locations of new and past events by the arrival time difference data between them. In some ways, the present method is intermediate between the conventional relative hypocenter relocation and DD methods.

For comparison, we perform quasi-real-time earthquake relocation without using reference events. We relocated the hypocenters of the events at each window as independent of those at the other windows. The obtained hypocenters are shown in Figs. 7(d) and 10(b); they are scattered and do not show a clearly recognizable planar structure (Fig. 10b), which differs from our main algorithm (Figs. 8b and 10a) and relocated hypocenters by Yoshida et al. (2023b) (Fig. 8c). The relocation by this algorithm reduced the mean deviation of the hypocenter pairs by only about 10% compared to the JMA initial hypocenters (Fig. 9b). This indicates that constraints between different windows are essential for improving the spatial resolution of hypocenter distribution when using multiple time windows.

Although the relationship between matrix size and computation time is not a simple linear relationship, matrix size can be used as an indicator of computation time. In our algorithm, the median size (product of the number of columns and rows) of the matrix G of each day (1st to 348th windows) was only 0.0005% of that when all events are relocated at once using the standard DD method. Figures 4(b) and (c) show the number of the size of $G$ for each relocation. This indicates that the present algorithm can significantly reduce daily computation time and facilitate quasi-real-time monitoring.

The hypocenter relocation results obtained from our algorithm are similar to those obtained using the standard DD method (Yoshida et al., 2023b). They reveal multiple planar structures not found in the cloud-like distribution of the initial hypocenters (Fig. 8), showing that our algorithm can reveal fault structure and earthquake migration at a much higher resolution than the initial hypocenters. Still, there are multiple improvements that can be made to this algorithm. The current algorithm relocates an event only once and fixes the hypocenter afterward. As a result, errors in hypocenters relocated in past windows may subsequently and continuously affect the relocation of new events. This problem may be alleviated by introducing a process such as repeatedly relocating and updating the hypocenters of past events. For example, after each longer period of time (e.g., a week or a month), the relocation of all previous events can be made using the standard DD method, and the hypocentral parameters of the reference events can be updated. Also, our algorithm may slow down calculations in later windows as the number of reference events accumulates. In future implementations, adding an algorithm that limits the number of past reference earthquakes may be helpful. In addition, the algorithm we tested targeted events already cataloged by the JMA, but a combination with, for example, earthquake detection using AI (e.g., Zhu & Beroza, 2019) may enable faster monitoring.

The Double-Difference relocation method (Walduhaser & Ellsworth, 2000) can be regarded as one extreme of the traditional relative hypocenter relocation method and provides the highest spatial resolution of relative hypocenters possible from the data. However, the DD method can be computationally expensive because it solves the matrix equations for the differential arrival time data of many earthquake pairs at multiple stations. This study developed an algorithm for quasi-real-time hypocenter relocation based on the DD method by reducing the computation cost. This algorithm reduces the number of unknown parameters and data by fixing the location of previously relocated events when relocating new events. The key element is that our algorithm constrains the relative locations of events in different time windows using the arrival time difference data. We tested the application of this algorithm to earthquake sequence in the northeastern Noto Peninsula. Despite the very small computational cost, the relocation results were generally consistent with those obtained by the standard DD method, revealing fine fault structures from the cloud-like distribution of initial hypocenters. We hope that this algorithm contributes to future earthquake monitoring.

Availability of data and materials

Waveforms were collected and stored by the JMA, national universities, and NIED Hi-net (2019). The figures were created using GMT (Wessel and Smith, 1998). This study used hypocenter and arrival time data from the JMA　unified catalog (https://www.data.jma.go.jp/svd/eqev/data/bulletin/hypo.html).

Competing interests

We declare no conflict of interest.

Authors' contributions

RM developed the daily relocation algorithm, obtained the results of this study, and wrote the first draft. KY created the basic parts of the algorithm, modified the DD code, and reviewed and edited the draft.

Funding

This study was supported by JSPS KAKENHI (Grant Number 23K17482) and the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan under the Second Earthquake and Volcano Hazards Observation and Research Program (Earthquake and Volcano Hazard Reduction Research).

Acknowledgements

We are deeply grateful to Professor Felix Waldhauser for making the code for the Double-Difference relocation method publicly available. We thank Makoto Naoi and Shiro Hirano for their helpful advice on speeding up the calculation of correlation functions.

Amezawa Y, Hiramatsu Y, Miyakawa A, et al (2023) Long‐Living Earthquake Swarm and Intermittent Seismicity in the Northeastern Tip of the Noto Peninsula, Japan. Geophys Res Lett 50:. doi: 10.1029/2022gl102670
De Barros L, Baques M, Godano M, et al (2019) Fluid-Induced Swarms and Coseismic Stress Transfer: A Dual Process Highlighted in the Aftershock Sequence of the 7 April 2014 Earthquake (Ml 4.8, Ubaye, France). J Geophys Res Solid Earth 124:3918–3932. doi: 10.1029/2018JB017226
Fukuyama E, Ellsworth WL, Waldhauser F, Kubo A (2003) Detailed Fault Structure of the 2000 Western Tottori, Japan, Earthquake Sequence. Bull Seismol Soc Am 93:1468–1478. doi: 10.1785/0120020123
Hatch RL, Abercrombie RE, Ruhl CJ, Smith KD (2020) Evidence of Aseismic and Fluid‐Driven Processes in a Small Complex Seismic Swarm Near Virginia City, Nevada. Geophys Res Lett 47:. doi: 10.1029/2019gl085477
Kato A (2024) Implications of Fault‐Valve Behavior From Immediate Aftershocks Following the 2023 Mj6.5 Earthquake Beneath the Noto Peninsula, Central Japan. Geophys Res Lett 51:. doi: 10.1029/2023gl106444
Nakajima J (2022) Crustal structure beneath earthquake swarm in the Noto peninsula, Japan. Earth, Planets Sp 74:. doi: 10.1186/s40623-022-01719-x
Naoi, M. et al. (2015) Unexpectedly frequent occurrence of very small repeating earthquakes (−5.1 ≤ Mw ≤ −3.6) in a South African gold mine: Implications for monitoring intraplate faults. J Geophys Res Solid Earth 120, 8478–8493.
NIED (2019) NIED Hi-net, National Research Institute for Earth Science and Disaster Resilience. doi: 10.17598/NIED.0003
Nishimura T, Hiramatsu Y, Ohta Y (2023) Episodic transient deformation revealed by the analysis of multiple GNSS networks in the Noto Peninsula, central Japan. Sci Rep 13:8381. doi: 10.1038/s41598-023-35459-z
Okada T, Savage MK, Sakai S, et al (2024) Shear wave splitting and seismic velocity structure in the focal area of the earthquake swarm and their relation with earthquake swarm activity in the Noto Peninsula, central Japan. Earth, Planets Sp 76:24. doi: 10.1186/s40623-024-01974-0
Paige, CC & Saunders, MA (1982) LSQR: An Algorithm for Sparse Linear Equations and Sparse Least Squares. Acm Transactions Math Softw Toms 8, 43–71.
Poupinet G, Ellsworth WL, Frechet J (1984) Monitoring velocity variations in the crust using earthquake doublets: An application to the Calaveras Fault, California. J Geophys Res Solid Earth 89:5719–5731. doi: 10.1029/jb089ib07p05719
Ross ZE, Rollins C, Cochran ES, et al (2017) Aftershocks driven by afterslip and fluid pressure sweeping through a fault-fracture mesh. Geophys Res Lett 8260–8267. doi: 10.1002/2017GL074634
Ruhl CJ, Abercrombie RE, Smith KD, Zaliapin I (2016) Complex spatiotemporal evolution of the 2008 Mw 4.9 Mogul earthquake swarm (Reno, Nevada): Interplay of fluid and faulting. J Geophys Res Solid Earth 121:8196–8216. doi: 10.1002/2016JB013399
Shelly DR (2024) Examining the Connections Between Earthquake Swarms, Crustal Fluids, and Large Earthquakes in the Context of the 2020–2024 Noto Peninsula, Japan, Earthquake Sequence. Geophys Res Lett 51:. doi: 10.1029/2023gl107897
Shelly DR, Hill DP (2011) Migrating swarms of brittle-failure earthquakes in the lower crust beneath Mammoth Mountain, California. Geophys Res Lett 38:1–6. doi: 10.1029/2011GL049336
Shelly DR, Hill DP, Massin F, Farrell J, Smith RB, Taira T, (2013) A fluid‐driven earthquake swarm on the margin of the Yellowstone caldera. J Geophys Res Solid Earth 118, 4872–4886.
Sibson RH (1983) Continental fault structure and the shallow earthquake source. J Geol Soc London 140:741–767. doi: 10.1144/gsjgs.140.5.0741
Ueno H, Hatakeyama S, Aketagawa T, et al (2002) Improvement of hypocenter determination procedures in the japan meteorological agency. Q J Seism 65:123–134
Waldhauser F (2009) Near-Real-Time Double-Difference Event Location Using Long-Term Seismic Archives, with Application to Northern CaliforniaNear-Real-Time Double-Difference Event Location Using Long-Term Seismic Archives. Bull Seismol Soc Am 99:2736–2748. doi: 10.1785/0120080294
Waldhauser F, Ellsworth WL (2002) Fault structure and mechanics of the Hayward Fault, California, from double‐difference earthquake locations. J Geophys Res Solid Earth 107:ESE 3--1-ESE 3-15. doi: 10.1029/2000jb000084
Yoshida K, Hasegawa A (2018) Sendai-Okura earthquake swarm induced by the 2011 Tohoku-Oki earthquake in the stress shadow of NE Japan: Detailed fault structure and hypocenter migration. Tectonophysics 733:132–147. doi: 10.1016/j.tecto.2017.12.031
Yoshida K, Uchida N, Matsumoto Y, et al (2023) Updip Fluid Flow in the Crust of the Northeastern Noto Peninsula, Japan, Triggered the 2023 Mw 6.2 Suzu Earthquake During Swarm Activity. Geophys Res Lett 50:. doi: 10.1029/2023gl106023
Yoshida K, Uno M, Matsuzawa T, et al (2023) Upward Earthquake Swarm Migration in the Northeastern Noto Peninsula, Japan, Initiated From a Deep Ring‐Shaped Cluster: Possibility of Fluid Leakage From a Hidden Magma System. J Geophys Res Solid Earth 128:. doi: 10.1029/2022jb026047
Yoshida K, Hasegawa A, and Okada T (2016) Heterogeneous stress field in the source area of the 2003 M6.4 Northern Miyagi Prefecture, NE Japan, earthquake. Tectonophysics 206:408–419. doi: 10.1093/gji/ggw160
Yukutake Y, Ito H, Honda R, et al (2011) Fluid-induced swarm earthquake sequence revealed by precisely determined hypocenters and focal mechanisms in the 2009 activity at Hakone volcano, Japan. J Geophys Res Solid Earth 116:13. doi: 10.1029/2010JB008036
Yukutake, Y, Yoshida, K & Honda, R (2022) Interaction between aseismic slip and fluid invasion in earthquake swarms revealed by dense geodetic and seismic observations. J Geophys Res Solid Earth (2022) doi:10.1029/2021jb022933
Zhu W, Beroza GC (2018) PhaseNet: a deep-neural-network-based seismic arrival-time picking method. Geophys J Int 216:261–273. doi: 10.1093/gji/ggy423

graphicalabstract.jpg

Download PDF

Reviewers agreed at journal
08 Jul, 2024
Reviewers invited by journal
08 Jul, 2024
Editor assigned by journal
02 Jul, 2024
First submitted to journal
26 Jun, 2024

You are reading this latest preprint version

Quasi-Real-Time Hypocenter Relocation and Monitoring in the Northeastern Noto Peninsula

Status:

Version 1

Abstract

Figures

Introduction

Quasi-Real Time Relocation Algorithm

2.1. DD method with reference events

2.2. Relocation Algorithm

Application to Real Data

3 − 1. Algorithm for Daily Relocation

Results

Discussion

Conclusion

Declarations

References

Supplementary Files

Status:

Version 1