Precise relocation of low-frequency earthquakes in Northeast Japan: new insight into arc magma and fluids

Summary

High-resolution (∼10–20 km) P- and S-wave velocity (Vp, Vs) tomography of the crust and uppermost mantle is determined to relocate precisely a large number of low-frequency earthquakes (LFEs) which occurred in Hokkaido and Tohoku during 2002–2016. The LFEs and seismic tomography are combined to study the arc magma and fluids in the study region. We divide the 4036 LFEs in Hokkaido and 4946 LFEs in Tohoku into 43 groups. Most of the LFEs are located in or around low-Vp, low-Vs and high Poisson's ratio anomalies beneath active arc volcanoes, which indicate the existence of abundant fluids and magmatic activities in the crust and uppermost mantle beneath the volcanoes. Our results also reveal the influence of large crustal earthquakes on the spatial and temporal distributions of the LFEs. Many of the LFEs occurred at edges of the low-Vp and low-Vs zones within ∼15 km of the active volcanoes, indicating transportation and/or cooling of the arc magmas.

Seismicity and tectonics, Seismic tomography, Volcano seismology, Subduction zone processes

1 INTRODUCTION

Low-frequency earthquakes (LFEs) occur actively beneath active arc volcanoes in Hokkaido and Tohoku in Northeast Japan (Fig. 1). The LFEs are small events (M ≤ 2.3) occurring at depths of ∼10–45 km in the crust and uppermost mantle and radiate low-frequency seismic waves (1–8 Hz; e.g. Hasegawa & Yamamoto 1994; Ide et al.2007; Aso & Tsai 2014; Fig. 2). Previous studies of various volcanic areas in the world have shown that most of the LFEs take place at depths of a few kilometres within or close to magma reservoirs (e.g. Aki et al.1977; Riuscetti et al.1977; Ferrick et al.1982; Chouet 1985; Crosson & Bame 1985; Nakamichi et al.2003; Neuberg et al.2006; Furukawa 2009), suggesting that fluid and magmatic activity play an important role in generating the LFEs. Many researchers have investigated 3-D seismic velocity and attenuation (Q) structures of the crust and upper mantle beneath Hokkaido and Tohoku, and found that the LFEs are generally located in or around low-velocity (low-V) and high-attenuation (low-Q) zones beneath active arc volcanoes in the region (e.g. Zhao et al.1992; Wang & Zhao 2009; Huang et al.2011a; Liu et al.2013, 2014; Niu et al.2016; Wang et al.2017). However, all the previous studies used hypocentral parameters of the LFEs determined by the Japan Meteorological Agency (JMA), which are less accurate because a simple 1-D velocity model is used in the JMA routine earthquake location.

Figure 1.

Tectonic settings of the Hokkaido and Tohoku subduction zones (modified from Liu et al.2013). The red circles denote LFEs. The black triangles represent active arc volcanoes. The red solid sawtooth line and the red dashed line represent the plate boundaries (Bird 2003). The thin red lines denote depth contours of the upper boundary of the subducting Pacific slab. The white stars denote large earthquakes (M_JMA ≥ 6.0) that occurred during 1901–2011 (Usami 2003; JMA Unified Earthquake Catalogue), whose magnitude scale is shown at the bottom. The topography data are derived from the GEBCO08 Grid (version20100927, http://www.gebco.net). The red box in the inset map shows the location of the present study region.

Figure 2.

Three-component velocity seismograms of a low-frequency event (a1–a3) and a normal crustal earthquake (b1–b3) recorded at a same Hi-net station (N.HMTH). The two events have comparable hypocentral distances (32.4 and 30.4 km). Their epicentral distances are 10.9 and 30.3 km, whereas their focal depths are 30.5 and 2.5 km, respectively. The vertical blue lines denote first P-wave arrivals. (a4, b4) Power spectrum densities of the east–west component seismograms shown in (a1) and (b1); the grey dotted lines show the background spectra.

Precise hypocentral locations are necessary to understand the generating mechanism of the LFEs, which is very important for clarifying the fluid and magmatic processes in subduction zones (e.g. Aso & Tsai 2014). In this work, we first determine high-resolution (∼10 km) models of P- and S-wave velocity (Vp, Vs) tomography of the crust and upper mantle beneath Hokkaido and Tohoku. Then we use the obtained 3-D Vp and Vs models to relocate precisely 4036 LFEs in Hokkaido and 4946 LFEs in Tohoku (Fig. 1). Our high-resolution tomographic images and precise hypocentral locations of the LFEs shed new light on the arc magmatism, fluid migration and subduction dynamics beneath Northeast Japan.

2 DATA AND METHOD

To determine the 3-D Vp and Vs models, we used a great number of high-quality P- and S-wave arrival-time data recorded at 331 permanent stations of the High-Sensitivity Seismic Network (Hi-net; Okada et al.2004) in Hokkaido and 322 Hi-net stations in Tohoku during June 2002 to December 2016 (Fig. 3). Following the previous studies (e.g. Wang & Zhao 2009; Liu et al.2013; Niu et al.2016), we selected local earthquakes (without LFEs, Figs 4 and 5) carefully based on the following criteria: (1) all the selected events were recorded at more than 10 seismic stations; (2) they are located beneath the seismic network and so they have reliable hypocentral locations (with uncertainties <3 km); and (3) they are declustered to pursue a uniform distribution of hypocentres in the study area. Although an effort was made to keep a uniform distribution of hypocentres, the selected events are actually not uniformly distributed, because no earthquake occurs in the mantle wedge beneath the study area (Figs 4b and c, and 5b and c). The picking accuracy of the arrival times (for both the LFEs and non-LFEs) is estimated to be 0.05–0.15 s for P-wave data and 0.1–0.2 s for S-wave data (http://www.hinet.bosai.go.jp).

Figure 3.

Distribution of 331 seismic stations in Hokkaido (a) and 322 stations in Tohoku (b), which are used in earthquake location and tomographic inversions in this study. The names of volcanoes are labelled. The other labelling is the same as that in Fig. 1.

Figure 4.

(a) Epicentral distribution of the selected 13 812 earthquakes in Hokkaido. (b) East–west and (c) north–south vertical cross-sections of the earthquakes shown in (a). The colours denote the focal depths, whose scale is shown at the lower-right corner.

Figure 5.

(a) Epicentral distribution of the selected 5039 earthquakes in Tohoku. (b) East–west and (c) north–south vertical cross-sections of the earthquakes shown in (a). The colours denote the focal depths, whose scale is shown at the lower-right corner.

As a result, our Hokkaido data set contains 312 667 P-wave and 198 990 S-wave arrival times from 13 812 local earthquakes which occurred in the crust and the subducting Pacific slab (Fig. 4). Our Tohoku data set contains 148 401 P-wave and 133 354 S-wave arrival times from 5039 shallow and intermediate-depth earthquakes (Fig. 5). From the JMA LFEs catalogue, we selected 4036 LFEs in Hokkaido and 4946 LFEs in Tohoku which occurred during June 2002 to December 2016. Each of the LFEs was recorded at four Hi-net stations or more. Fig. 2 shows three-component seismograms and spectra of a typical LFE and a normal crustal earthquake, which occurred and were recorded at a same station (N.HMTH) in Tohoku.

We used the tomographic method of Zhao et al. (1992, 2015) to determine the 3-D Vp and Vs models. An efficient 3-D ray-tracing scheme (Zhao et al.1992) is exploited to compute theoretical travel times and ray paths. Station elevations are taken into account in the 3-D ray tracing. The local earthquakes used are relocated in the inversion process. The iterative inversion algorithm (LSQR, Paige & Saunders 1982) with damping and smoothing regularizations is adopted to solve the large but sparse system of observation equations (Zhao et al.1992, 2012). More details of the tomographic inversions are shown in the Supporting Information. The obtained 3-D Vp and Vs models are then used to relocate the LFEs precisely using the method described by Zhao et al. (1992). The LFEs with a mislocation uncertainty <5 km are selected in the following analysis.

3 ANALYSIS AND RESULTS

3.1 Model evaluation

For the Hokkaido region, the P-wave root-mean-square (RMS) travel-time residuals before (A) and after (B) the tomographic inversion are 0.379 and 0.270 s, respectively, and the corresponding S-wave RMS residuals are 0.467 and 0.340 s, respectively. The variance reductions (VRs) of the P- and S-wave data are 49 and 47 per cent, respectively, which are obtained using the relation, VR = (A² − B²)/A². For the Tohoku district, the P-wave RMS travel-time residuals before and after the tomographic inversion are 0.226 and 0.173 s, respectively, and the corresponding S-wave RMS residuals are 0.298 and 0.232 s, respectively. The VRs of the P- and S-wave data are 41 and 39 per cent, respectively. For the tomographic inversion in each region, the optimal value of the damping parameter is obtained from a trade-off curve which is drawn using the results of many tomographic inversions we conducted (Supporting Information Fig. S1).

To evaluate the adequacy of ray coverage and robustness of the obtained Vp and Vs models, we carried out checkerboard resolution tests (CRTs) following the approach of Zhao et al. (1992). In the CRTs for Vp and Vs tomography (Supporting Information Figs S2–S9), positive and negative velocity perturbations are assigned alternatively to the 3-D grid nodes to make an input checkerboard model. Theoretical travel times are calculated for the checkerboard model using the same set of source-receiver paths as in our real data set. To simulate the picking errors in the arrival-time data, Gaussian noise with a standard deviation of 0.1 s is added to the synthetic travel times before conducting the tomographic inversion.

Supporting Information Figs S2 and S4 show the CRT results for the Vp and Vs tomography in Hokkaido with a lateral grid interval of 0.1°. Supporting Information Figs S6 and S8 show the CRT results for the Vp and Vs tomography in Tohoku with the same grid interval. We also conducted extensive CRTs with a lateral grid interval of 0.2° (Supporting Information Figs S3, S5, S7 and S9). The results of these tests indicate that our 3-D Vp and Vs models have a resolution of ∼0.2° in the horizontal direction and ∼10 km in depth, whereas in the well-resolved areas, the lateral resolution is ∼0.1°–0.2°.

We plot the normal distribution of the LFE focal depth errors before and after the 3-D relocation to evaluate the LFE relocation accuracy (Supporting Information Fig. S10). In Hokkaido, because many LFEs are located at the edge of the 3-D velocity model and also the seismic network, the overall focal depth errors of the LFEs have little change after the relocation (Supporting Information Fig. S10a), though the LFE locations in central Hokkaido are certainly improved. In Tohoku, however, the LFE focal depth errors are reduced significantly after the relocation (Supporting Information Fig. S10b), because almost all the LFEs occurred in central Tohoku where the tomographic resolution is very good. Supporting data files LFEs-Hokkaido and LFEs-Tohoku show the relocated hypocentral parameters for the LFEs in Hokkaido and Tohoku, respectively.

3.2 Velocity heterogeneity and LFE distribution

Main features of our Vp and Vs tomography are similar to those of the previous models (e.g. Zhao et al.1992, 2015; Huang et al.2011a; Niu et al.2016), but because of the better data sets and the smaller grid intervals adopted in this study, fine-scaled structures are revealed in the crust and uppermost mantle beneath Hokkaido and Tohoku.

We selected 43 LFE groups to show their distribution with the velocity and Poisson's ratio (PR) images (Supporting Information Table S1). The LFE groups No. 1 to No. 10 are located in Hokkaido, whereas those of No. 11 to No. 43 are located in Tohoku. Figs 6–9, and Supporting Information Figs S13, S16, S19, S22, S25, S32 and S35 show vertical cross-sections of the Vp, Vs and PR images with the LFEs, whereas Supporting Information Figs S11 and S12, S14 and S15, S17 and S18, S20 and S21, S23 and S24, S26–S31 and S33 and S34 show map views of the Vp, Vs and PR images with the LFEs. Detailed information on the 43 LFE groups is shown in Table 1 and Supporting Information Table S1. All the LFEs with the relocated hypocentral parameters are presented in the Supporting Information.

Figure 6.

(a) Epicentral distribution of the group Nos 1–4 LFEs close to the Takachidake and Taisetsuzan volcanoes whose locations are shown in map (d). (b) East–west and (c) north–south vertical cross-sections of the LFEs shown in (a). The colours denote the focal depths, whose scale is shown at the lower-right corner. The red triangles denote active arc volcanoes. The red lines in (a) show locations of the vertical cross-sections of Vp, Vs and Poisson's ratio images at the bottom. In the tomographic images, the surface topography along each profile is shown above each cross-section, the black lines mark the Conrad and Moho discontinuities, the white dots show the background seismicity (M_JMA > 1.0, during June 2002 to December 2016) within a 5 km width of each profile, the open circles with vertical bars (i.e. focal depth errors) represent LFEs within a 15 km width of each profile, and the scales for the velocity and Poisson's ratio perturbations (in per cent) are shown on the right. The group Nos 1–4 LFEs are labelled in (a). The temporal variations in the LFE number and focal depth are shown on the upper-right (e), where the red squares atop the No. 2 LFEs mark the active periods of the Tokachidake volcano; the red circles, green triangles and blue diamonds denote the LFE magnitudes of 0–1, 1–2, and 2–2.3, respectively; the vertical bars denote the focal depth errors; and the black lines represent the LFE cumulative numbers.

Figure 7.

The same as Fig. 6 but for the group Nos 23–27 LFEs around the Kurikoma and Naruko volcanoes. The star denotes the Iwate-Miyagi earthquake (M 7.2) which occurred on 2008 June 14.

Figure 8.

The same as Fig. 6 but for the group Nos 28–31 LFEs around the Zao volcano. The red squares atop the group No. 30 LFEs mark the active periods of the Zao volcano.

Figure 9.

The same as Fig. 6 but for the group Nos 40–43 LFEs around the Nasudake, Takahara and Nikko-Shirane volcanoes. The red squares atop the group No. 41 LFEs mark the active periods of the Nikko-Shirane volcano.

Table 1.

Features of the LFE groups in Hokkaido and Tohoku.

		Seismic velocity anomaly
		HVP, HVS, HPR/LPR/NPR	LVP, LVS, HPR	NVP, NVS, HPR/LPR/NPR
Hokkaido	Upper crust	8	8	10
	Conrad	1, 6	4, 5, 6, 8
	Lower Crust		1, 2, 3, 4, 5, 6, 9	8, 10
	Moho		3, 7, 9, 10
Tohoku	Upper crust		21, 35	34
	Conrad	25, 37, 42	13, 15, 23, 26, 35, 36	19
	Lower Crust	16, 19, 26, 27, 33, 38	13, 14, 18, 20, 21, 22, 23, 24, 28, 29, 30, 31, 32, 34, 35, 36, 40, 41, 43	11, 20, 30, 39, 42
	Moho	12, 17, 41	19, 25, 33, 34, 39, 42

Notes: The LFE group numbers in red and black denote that the distance from an LFE group to its closest volcano is less than 5.0 km and greater than 5.0 km, respectively. Moho, the Moho discontinuity; Conrad, the Conrad discontinuity. HVP: high P-wave velocity (Vp) anomaly; LVP: low Vp anomaly; NVP: edge of Vp anomaly; HVS: high S-wave velocity (Vs) anomaly; LVS: low Vs anomaly; NVS: edge of Vs anomaly; HPR: high Poisson's ratio (PR) anomaly; LPR: low PR anomaly; NPR: edge of PR anomaly.

Most LFEs occurred in or around the low-V and high-PR anomalies in the lower crust beneath Hokkaido and Tohoku (e.g. Figs 6–11), a small part of the LFE groups are located around the Conrad discontinuity (e.g. group No. 1 in Fig. 6) or the Moho discontinuity (e.g. group No. 25 in Fig. 7), whereas a few LFE groups are located in the upper crust (e.g. group No. 23 in Fig. 7).

Figure 10.

Vertical cross-sections of the LFEs beneath the active volcanoes in Hokkaido. The inset map shows the location of the vertical sections. The other labelling is the same as that in Fig. 6.

Figure 11.

Vertical cross-sections of the LFEs beneath the active volcanoes in Tohoku. The inset map shows the location of the vertical sections. The other labelling is the same as that in Fig. 6.

There are another two significant features. One is that the number of LFEs increased during the active periods of the active volcanoes, e.g. the number of No. 2 LFEs increased during the active periods in 2002–2016 (http://www.jma.go.jp/en/volcano/) of the Tokachidake volcano (Fig. 6). The other feature is that the number of LFEs changed before and after the 2008 June 14 Iwate-Miyagi earthquake (M 7.2) which occurred in the upper crust (e.g. the No. 25 group LFEs in Fig. 7 and Supporting Information Fig. S40) and the 2011 March 11 Tohoku-oki megathrust earthquake (M_w 9.0; e.g. the LFEs of No. 27, No. 30, No. 36 and No. 43 groups shown in Figs 7–9, Supporting Information Figs S32 and S40).

We investigate the relationship between the LFE depth distribution and the distance between the LFEs and volcanoes (hereafter we call it the E-V distance; Fig. 12, Supporting Information Figs S41 and S42). In Hokkaido, most of the LFEs are located in the lower crust within an E-V distance of 5.0 km. In Tohoku, there are distinct gaps in the LFE depth distribution when the E-V distance increases. When the E-V distance is less than ∼2.0 km, the gap between the LFEs in the upper crust and those in the lower crust becomes large (about 10.0 km in depth). When the E-V distance is 2.0∼5.0 km, the gap occurs between the LFEs near the Conrad discontinuity and those in the lower crust. No clear gap is visible in the lower crust when the E-V distance is about 5.0∼6.5 km. However, when the E-V distance is greater than ∼6.5 km, the gap occurs between the LFEs in the lower crust and those near the Moho discontinuity.

Figure 12.

Focal depth distribution of the relocated LFEs in Hokkaido (a, b) and Tohoku (c, d). The horizontal axis represents the distance (0–10 km for Hokkaido and 0–13 km for Tohoku) between the LFEs and the closest active volcano. The colour shows the velocity perturbation (in per cent) at the LFE hypocentre, whose scale is shown at the bottom. The black and red lines represent the average Conrad and Moho depths, respectively.

4 DISCUSSION

Low-frequency volcanic earthquakes are indicators of magma transport and activity within shallow conduit systems (Green & Neuberg 2006). The source models proposed so far suggest that most of the LFEs originate at depths of a few kilometres within or near the magma reservoirs (e.g. Aki et al.1977; Riuscetti et al.1977; Ferrick et al.1982; Chouet 1985; Crosson & Bame 1985; Furukawa 2009), only a small number of magma transport channels with strong barriers generate LFEs when magma is forced through them (Aki & Koyanagi 1981). Therefore, the LFE hypocentres may represent the boundary of the magma transport channels or magma reservoirs.

4.1 Volcanic LFEs

LFEs are regarded as critically important because they often precede and accompany volcanic eruptions (Chouet et al.1994). There are 25 LFE groups beneath active arc volcanoes in Northeast Japan (group Nos 1–2, 5, 6, 10, 13–14, 16, 19–22, 25–27, 29–30, 33–36, 38, and 41–43). Most of the LFEs are located in low-V and high-PR anomalies, suggesting that there are strong connections between the LFEs, volcanoes, and high-temperature and fluid content in the lower crust and uppermost mantle. In addition, the LFEs (group Nos 2, 5, 6, 14, 19–22, 30, 36, and 41) occurred at the active periods of the volcanoes (in 2002–2016, http://www.jma.go.jp/en/volcano/). However, some LFEs beneath the volcanoes, such as Taisetsuzan (No. 1), Atosanupuri and Mashu (No. 6), Towada (No. 16), Naruko (No. 27), Numazawa (No. 38), Nikko-Shirane (No. 41), and Takahara (No. 42), are located in high-V anomalies, which may reflect that strong barriers exist and play an important role in generating the LFEs (Aki & Koyanagi 1981), especially for the LFEs (group Nos 1, 6 and 41–42) around the Conrad or the Moho, where the seismic discontinuity may become a cold and strong barrier for the magma transport (Aki & Koyanagi 1981; Aso & Tsai 2014). Further detailed studies should be made to better understand the LFEs occurring in the high-V anomalies.

The LFE depth distribution in Tohoku shows distinct gaps with the increasing E-V distance (Fig. 12, Supporting Information Figs S41 and S42), suggesting that the gaps may represent strong barriers in the crust (Aki & Koyanagi 1981). In addition, the LFEs become deeper with the E-V distance, which may directly reflect the magma focus mode and magma reservoirs beneath the volcano (Aso et al.2013). Many studies have shown that fluid migration is the most possible cause of the volcanic LFEs (e.g. Aki & Koyanagi 1981; Hasegawa & Yamamoto 1994; Zhao et al.2011; Aso et al.2013), and the cause of the low-V and high-PR anomalies may be also related to the crustal fluid or magma (e.g. Zhao et al.1992, 2011; Xia et al.2007). Therefore, we may use the LFE distribution together with the seismic velocity and PR images to reveal the fluid and magma beneath the active volcanoes. Because most of the LFEs are located at edges of the low-V and high-PR anomalies beneath active arc volcanoes (Figs 10 and 11), the low-V anomalies coincident well with the LFEs may reflect the magma reservoirs (Fig. 13). Previous studies found similar features in SW Japan (e.g. Zhao et al.2011) and the Naruko volcanic area in Tohoku (Okada et al.2014).

Figure 13.

A schematic showing magma reservoirs and low-frequency events (crosses) beneath an active arc volcano in the study region. The red and blue ellipses denote hot and cooling magma reservoirs, respectively.

4.2 Isolated intraplate LFEs

The LFEs that occurred far away from plate boundaries or active volcanoes are called isolated intraplate LFEs (e.g. Katsumata & Kamaya 2004). Although such LFEs are not located beneath active volcanoes, most of them occurred near dormant Quaternary volcanoes within a distance of 50 km (e.g. Ohmi et al.2004). Because these isolated LFEs may be generated by fluid movements, and they have quite similar waveforms to those of the volcanic LFEs, they are also called semi-volcanic LFEs (Aso et al.2013).

There are 18 LFE groups which are located about 15–50 km away from the closest active volcano, but most of the LFEs (14 groups: Nos 3–4, 7–9, 15, 17–18, 23–24, 28, 32 and 39–40) are located in low-V and high-PR anomalies, suggesting that high-temperature zones or magmas are widely distributed in the crust and uppermost mantle beneath or around the active volcanoes. Fluids rising from the subducting Pacific slab through the mantle wedge may be the source of these LFEs (Vidale et al.2014), and the pathway of rising fluids may incline. The rest four LFE groups, all of which have a quite small number of LFEs, are located in high-V anomalies (Nos 12, 17 and 37) around the Conrad or the Moho and at the edge of a low-V anomaly (No. 11) in the lower crust, which may represent edges of fluid-filled matrices in the crust and uppermost mantle.

4.3 Large earthquakes and LFEs

During the period of 2002–2016 which we investigated, only one large earthquake, that is, the 2008 Iwate-Miyagi earthquake (7.8 km depth, M 7.2), occurred within 10 km of LFEs (group No. 25), which is located at the edge of high-V and high-PR anomalies (Fig. 7), similar to the previous studies (Cheng et al.2011; Huang et al.2011b; Wei & Zhao 2013). The number of the group No. 25 LFEs changed before and after the Iwate-Miyagi earthquake, suggesting that the arc magma and fluids may have ascended from the LFE source zone and contributed to the nucleation of the 2008 Iwate-Miyagi earthquake, and subsequent occurrence of the large earthquake also affected the LFE generation. Hence, the magma-earthquake interaction took place in the Kurikoma volcanic area.

Although the epicentre of the great 2011 March 11 Tohoku-oki megathrust earthquake (M_w 9.0) is located over 200 km away from the closest LFE group, the 2011 April 11 Iwaki crustal earthquake (M 7.0) occurred ∼50 km away from the closest LFE group. Most of the LFEs in Tohoku showed responses to the two big earthquakes, especially the LFEs beneath volcanoes Iwatesan (Nos 21 and 22), Kurikoma (No. 25), Naruko (No. 27), Zao (No. 30), Azuma (No. 36) and Nasudake (No. 43), as well as the group No. 4 LFEs which are located away from the active volcanoes, and a clear LFE time gap can be found during 2011 and 2012 [see Figs 8e and Supporting Information Fig. S40 (group 36)]. Because the fluid flow is one of the critical factors to generate the LFEs (e.g. Aki & Koyanagi 1981), these results suggest that the fluid flow status beneath the volcanoes changed before and/or after the large earthquakes, and reflow of the fluids may take months or a year.

4.4 LFEs around the Moho discontinuity

Few LFEs occurred beneath or around the Moho discontinuity when they are close to a volcano (the E-V distance <∼1.0 km), whereas when the E-V distance increases, more and more LFEs occurred around or beneath the Moho discontinuity with velocity anomalies from low-V to high-V (Figs 12c and d). Although we show the average Moho depth, the above-mentioned trend is clear. The cooling magma model of Aso & Tsai (2014) may explain this phenomenon (Fig. 13). Because the Moho is a chemical and density discontinuity, it is possible that ascending magma diapirs tend to stagnate there. A stagnant magma body would melt nearby rock soon after its intrusion and would subsequently cool. This cooling process occurs gradually and probably controls the in situ thermodynamical process for a long time. The low-V anomaly changing to a high-V anomaly may reflect the temperature lowering with the increasing E-V distance, which is visible in vertical cross-sections K-K’ in Supporting Information Fig. S19, and 1F–1F΄ and 1G–1G΄ in Fig. 9. This explanation is consistent with the previous studies, indicating that the Moho discontinuity is a strong barrier for the magma transport (Aki & Koyanagi 1981; Aso & Tsai 2014) or for the magmatic activity of mantle diapirs (Hasegawa & Yamamoto 1994).

We applied the method of Hardebeck & Shearer (2002) to calculate focal mechanisms of several LFEs using P-wave first-motion data recorded at over eight Hi-net stations (Fig. 7). Our results show that two LFEs around the Moho discontinuity have normal-faulting mechanisms, whereas one LFE in the upper crust exhibits a thrust-faulting mechanism, which support the cooling magma model (Aso & Tsai 2014). However, Nakamichi et al. (2003) observed various focal mechanisms for LFEs in a narrow area, which is attributed to the complex magmatic system in the source areas. This may explain our results that the LFEs gap in the E-V distance profiles occurred in Tohoku (Figs 12c and d) but not in Hokkaido (Figs 12a and b). More work on focal mechanisms is essential to understanding the physical processes causing the LFEs (Aso & Tsai 2014).

5 CONCLUSIONS

To better understand seismotectonics and magmatism in the Hokkaido–Tohoku subduction zone, we relocate precisely a large number of LFEs during 2002–2016 using detailed 3-D Vp and Vs models which are newly determined in this work. Our results shed new light on the close relationship between fluid migration, magmatic activity and the LFE distribution. Main findings of this work are summarized as follows.

1. Most of the LFEs in the study region are volcanic events which occurred in or around low-V and high-PR anomalies beneath active arc volcanoes, indicating that fluid and magmatic activities exist widely in Northeast Japan.

2. The depth distribution of LFEs shows gaps around the Moho discontinuity when the distance between the LFEs and a volcano increases, suggesting that magma cooling and strong barriers in the crust and uppermost mantle (e.g. the Conrad and Moho discontinuities) play an important role in generating the LFEs.

3. The LFE distribution may indicate the spatial extent of magma reservoirs beneath the active volcanoes.

4. Nearby large earthquakes affected the fluid flow and the generation of LFEs.

Acknowledgements

We thank the Hi-net data centre and the JMA unified earthquake catalogue for providing the high-quality data used in this study (http://www.hinet.bosai.go.jp). The volcanic activity data are downloaded from the JMA website (http://www.jma.go.jp/en/volcano). We thank Drs Aiguo Ruan, Xiaodong Wei, Zhiteng Yu and Xinyang Wang for thoughtful discussions. This work was partially supported by research grants from the JSPS (Kiban-S 23224012) and the MEXT (26106005) to DZ, the (QNYC201601) and NSFC (41506048) to XN, and the National Program on Global Change and Air-Sea Interaction, SOA (No. GASI-GEOGE-01) and MOST of China (No. 2016YFC0600402) to JL. Most of the figures were made by using GMT (Wessel & Smith 1998). Prof Egill Hauksson (the editor), Dr Naofumi Aso and an anonymous referee provided thoughtful review comments which have improved this paper.

SUPPORTING INFORMATION

Supplementary data are available at GJI online.

Table S1. Information on the LFE groups.

Figure S1. Trade-off curves for determining the optimal damping parameters for Vp tomography in Hokkaido (a), Vs tomography in Hokkaido (b), Vp tomography in Tohoku (c) and Vs tomography in Tohoku (d). The numbers beside the red dots denote the damping parameters adopted in tomographic inversions. The numbers with red hatches denote the optimal damping parameters which lead to the best 3-D velocity models.

Figure S2. Results of a checkerboard resolution test for P-wave tomography at six depths beneath in Hokkaido with a lateral grid interval of 0.1°. The white and black circles denote low and high Vp perturbations, respectively. The Vp perturbation scale is shown at the bottom. The red triangles represent active arc volcanoes. The blue line marks the location of the upper boundary of the subducting Pacifica slab. The sawtooth lines denote the plate boundary (Bird 2003).

Figure S3. The same as Fig. S2 but with a lateral grid interval of 0.2°.

Figure S4. The same as Fig. S2 but for S-wave tomography.

Figure S5. The same as Fig. S3 but for S-wave tomography.

Figure S6. The same as Fig. S2 but for Tohoku.

Figure S7. The same as Fig. S3 but for Tohoku.

Figure S8. The same as Fig. S4 but for Tohoku.

Figure S9. The same as Fig. S5 but for Tohoku.

Figure S10. Normal distributions for the relocated focal-depth uncertainties of LFEs beneath Hokkaido (a) and Tohoku (b). The black and red solid lines represent the normal distribution curves of the depth uncertainties before and after the 3-D relocations, respectively. The black and red dashed lines represent the average depth uncertainties before and after the 3-D relocations, respectively.

Figure S11. Map views of P-wave tomography and epicentral distribution of the group Nos 1–4 LFEs around the Takachidake and Taisetsuzan volcanoes. The layer depth is shown at the lower-right corner of each map. The black thick sawtooth lines and the dashed lines represent active and potential thrust faults, respectively (modified from Niu et al. 2016). The white solid circles show the seismicity (M_JMA > 1.0, during June 2002 to December 2016) within a 5 km depth range of each layer, and the open circles represent LFEs within a 5 km depth range of each layer. The velocity perturbation scale is shown at the bottom. The other labelling is the same as that in Fig. 6.

Figure S12. The same as Fig. S11 but for S-wave tomography.

Figure S13. The same as Fig. 6 but for the group Nos 5–6 LFEs around the Atosanupuri, Mashu and Meakan volcanoes. The red squares atop the No. 5 and No. 6 LFEs mark the active periods of the Meakan and Mashu volcanoes, respectively.

Figure S14. Map views of P-wave tomography and epicentral distribution of Nos 5–6 LFEs around the Atosanupuri, Mashu and Meakan volcanoes. The other labelling is the same as that in Figs S13 and S11.

Figure S15. The same as Fig. S14 but for S-wave tomography.

Figure S16. The same as Fig. 6 but for the group Nos 7–10 LFEs around the Usu and Tarumae volcanoes.

Figure S17. Map views of P-wave tomography and epicentral distribution of Nos 7–10 LFEs around the Usu and Tarumae volcanoes. The other labelling is the same as that in Figs S16 and S11.

Figure S18. The same as Fig. S17 but for S-wave tomography.

Figure S19. The same as Fig. 6 but for the group Nos 11–13 LFEs around the Iwaki volcano.

Figure S20. Map views of P-wave tomography and epicentral distribution of Nos 11–13 LFEs around the Iwakisan volcano. The other labelling is the same as that in Figs S19 and S11.

Figure S21. The same as Fig. S20 but for S-wave tomography.

Figure S22. The same as Fig. 6 but for the group Nos 14–17 LFEs around the Towada and Hakkoda volcanoes.

Figure S23. Map views of P-wave tomography and epicentral distribution of Nos 14–17 LFEs around the Towada and Hakkoda volcanoes. The other labelling is the same as that in Figs S22 and S11.

Figure S24. The same as Fig. S23 but for S-wave tomography.

Figure S25. The same as Fig. 6 but for the group Nos 18–22 LFEs around the Akita-Yakeyama, Hachimantai, Iwatesan and Akita-Komagatake volcanoes. The red squares atop No. 19 and No. 20 LFEs mark the active periods of the Akita-Yakeyama volcano and the Akita-Komagatake volcano, respectively. The red squares atop group No. 21 and group No. 22 mark the active periods of the Iwatesan volcano.

Figure S26. Map views of P-wave tomography and epicentral distribution of Nos 18–22 LFEs around the Akita-Yakeyama, Hachimantai, Iwatesan and Akita-Komagatake volcanoes. The other labelling is the same as that in Figs S25 and S11.

Figure S27. The same as Fig. S26 but for S-wave tomography.

Figure S28. Map views of P-wave tomography and epicentral distribution of Nos 23–27 LFEs around the Kurikoma and Naruko volcanoes. The other labelling is the same as that in Figs 7 and S11.

Figure S29. The same as Fig. S28 but for S-wave tomography.

Figure S30. Map views of P-wave tomography and epicentral distribution of Nos 28–31 LFEs around the Zao volcano. The other labelling is the same as that in Figs 8 and S11.

Figure S31. The same as Fig. S30 but for S-wave tomography.

Figure S32. The same as Fig. 6 but for the group Nos 32–36 LFEs around the Azuma, Adatara and Bandai volcanoes. The red squares atop the group No. 36 LFEs mark the active periods of the Azuma volcano.

Figure S33. Map views of P-wave tomography and epicentral distribution of Nos 32–36 LFEs around the Azuma, Adatara and Bandai volcanoes. The other labelling is the same as that in Figs S32 and S11.

Figure S34. The same as Fig. S33 but for S-wave tomography.

Figure S35. The same as Fig. 6 but for the group Nos 37–39 LFEs around the Numazawa volcano.

Figure S36. Map views of P-wave tomography and epicentral distribution of Nos 37–39 LFEs around the Numazawa volcano. The other labelling is the same as that in Figs S35 and S11.

Figure S37. The same as Fig. S36 but for S-wave tomography.

Figure S38. Map views of P-wave tomography and epicentral distribution of Nos 40–43 LFEs around the Nasudake, Takahara and Nikko-Shirane volcanoes. The other labelling is the same as that in Figs 9 and S11.

Figure S39. The same as Fig. S38 but for S-wave tomography.

Figure S40. The space-time plot of LFEs with time intervals in days and hours for LFE groups 33, 36 and 25. The red circles, green triangles and blue diamonds denote the LFE magnitudes of 0–1, 1–2, and 2–2.3, respectively. The other labelling is the same as that in Figs 6 and 7.

Figure S41. The same as Fig. 12 but the largest distance between the LFEs and active arc volcanoes is 50 km in Hokkaido and Tohoku.

Figure S42. Focal depth distribution of the relocated LFEs (coloured dots) in Hokkaido (a, b) and Tohoku (c, d). The colour represents vertical distance from each LFE to the Conrad (a, c) and Moho (b, d) discontinuities. The colour scale is shown in each panel.

Please note: Oxford University Press is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the paper.