Spatiotemporal variations in the stress field in the northeasternmost part of the NE Japan arc: constraints from microearthquakes

We determined focal mechanism solutions of microearthquakes and examined the stress field in the low-seismicity region from southern Hokkaido to eastern Aomori, NE Japan. The stress fields determined in this study comprise (1) a reverse faulting stress regime in southern Hokkaido with the axis of maximum compressional stress (σ1) being sub-horizontal and trending WNW–ESE, and (2) a stress regime in eastern Aomori to Tsugaru Strait that shows temporal variations and differential stress of less than tens of MPa. The spatiotemporal variation in stress from eastern Aomori to Tsugaru Strait might reflect the effects of the upper-plate bending and the 2011 Mw 9.0 Tohoku-Oki earthquake. It also indicates that the compressional stress caused by the descending Pacific plate is relatively weak, which is similar to other areas in eastern parts of the NE Japan arc.


Introduction
The stress field in inland areas of NE Japan is generally recognized as being associated with a WNW-ESE compressional reverse-faulting stress regime (e.g., Townend and Zoback 2006;Terakawa and Matsu'ura 2010;Yukutake et al. 2015) related to subduction of the Pacific plate. However, recent studies have revealed that there is a localized stress field in these areas which is different from that expected from the subduction of the Pacific Plate and that the stress field is spatially inhomogeneous (e.g., Terakawa and Matsu'ura 2010;Imanishi et al. 2012Imanishi et al. , 2013Yoshida et al. 2012Yoshida et al. , 2015Yoshida et al. , 2019. Given that the stress distribution is commonly estimated using focal mechanism data, it is difficult to estimate the stress field in low-seismicity areas. However, to properly understand earthquake generation mechanisms, it is necessary to investigate the stress field in low-seismicity areas as well as in areas with high seismicity. In Japan, the High-Sensitivity Seismograph Network (Hi-net) has been deployed nationwide by the National Research Institute for Earth Science and Disaster Resilience (NIED) since the 1995 M j 7.2 Hyogo-ken-Nanbu (Kobe) earthquake, which has been greatly improved the ability for detecting microearthquakes (Obara et al. 2005). Applying the inversion method developed by Michael (1984Michael ( , 1987 to micro-and small-earthquake focal mechanism data, Yoshida et al. (2015) defined the stress regime in the forearc region of the NE Japan, where the stress field had not previously been defined in detail owing to the low seismicity in the region.
Here, we focus on the low-seismicity area from southern Hokkaido (region N) to eastern Aomori (region S), Japan (Fig. 1), where the stress field was not previously estimated by Yoshida et al. (2015Yoshida et al. ( , 2019. This

Open Access
*Correspondence: maeda.sumire@aist.go.jp 1 Geological Survey of Japan, AIST Tsukuba Central, National Institute of Advanced Industrial Science and Technology, 7, Higashi-1-1-1, Tsukuba, Ibaraki 305-8567, Japan Full list of author information is available at the end of the article region corresponds to the northeasternmost part of the NE Japan arc (Fig. 1a), which is colliding with the Kurile arc. The stress field in Iwate Prefecture, which is located to the south of region S, has been determined by Yoshida et al. (2015Yoshida et al. ( , 2019. Although the detailed stress distribution has not been determined because of the low seismicity in Hokkaido, including region N, the orientation of maximum horizontal compressive stress (σ Hmax ) has been estimated to be roughly parallel to the relative motion of the Pacific plate (e.g., Ghimire et al. 2005;Terakawa and Matsu'ura 2010;Yukutake et al. 2015). Previous estimates of the stress field are also consistent with the regional stress field estimated from the strikes of ore veins (Watanabe 1986). However, because the seismicity in the study area is very low and no large events have occurred since the deployment of Hi-net, the stress field in this area has not been determined in detail. Since the seismic waveform data have accumulated over time since the deployment of the Hi-net, it is possible to estimate the stress field in this area now.
In this paper, we first determine the focal mechanisms of micro and small earthquakes from southern Hokkaido to eastern Aomori. We then examine the spatiotemporal variation in the stress field in the study area. Finally, we discuss the factors affecting the stress field in the northeasternmost part of the NE Japan arc.

Hypocenter and focal mechanism data
We used waveform data obtained by Hi-net for events with Japan Meteorological Agency (JMA) magnitudes greater than 2.0 for the period from June 2002 to December 2017, for depths shallower than 25 km (Fig. 1b). We determined hypocenters and focal mechanisms via these data and the following method. First, we picked P-wave first-motion polarities and onsets from the digital waveform data of each earthquake using the WIN system (Urabe and Tsukada 1992). Next, we determined  The northern and southern parts of the study area are referred to as "region N" (southern Hokkaido) and "region S" (eastern Aomori), respectively. The region combining the southern half of region N and the northern half of region S is referred to as "region C" (Tsugaru Strait) hypocenters using the Hypomh program (Hirata and Matsu'ura 1987) with the one-dimensional velocity model of Hasegawa et al. (1978). We then used the software pick2mec (Katao and Iio 2004) to determine focal mechanisms using the phase data and the hypocenters. The pick2mec software uses a procedure proposed by Maeda (1992) to determine the focal mechanisms. This software assigns each focal mechanism a score based on the ratio of consistent polarities to all polarities. Finally, we retained only reliable solutions using the same quantitative criterion as employed by Maeda et al. (2018). Moreover, we evaluated the uncertainty of focal mechanisms by measuring the extent of these multiple fault plane solutions estimated using pick2mec with the Kagan angle (Kagan 1991). Focal mechanisms with an uncertainty in the Kagan angle of > 35° were discarded.

Stress inversion
We used the focal mechanisms obtained using the method given above to determine the stress field. We used an ordinary stress inversion procedure similar to those proposed by Michael (1984Michael ( , 1987. This technique assumes the stress to be uniform within an analysis area and minimizes the difference between observed and calculated slip directions. The method can be used to estimate the orientations of stress axes and the stress ratio R = (σ 1 − σ 2 )/(σ 1 − σ 3 ), where σ 1 , σ 2 , and σ 3 are the maximum, intermediate, and minimum compressional principal stresses, respectively. Using the above data and method, we estimated the stress field in regions N, C, and S shown in Fig. 1b. In order to smoothly visualize the spatial variation of the stress field in this area with lowseismicity, the region of C is overlapped on the regions of N and S. Then, we examined whether the stress field varied spatiotemporally. To investigate spatial variations, we compared the stress field in the three regions.
To investigate temporal variations, we divided the studied time interval into two periods, before and after the Mw 9.0 Tohoku-Oki earthquake (herein, Tohoku earthquake) that occurred on 11 March 2011, and compared the stress fields determined for these periods.

Results
We successfully determined reliable focal mechanisms for 114 events, which is 21% of all candidate events in this study. The determined focal mechanisms for the periods before and after the Tohoku earthquake are shown in Fig. 2a, b, respectively. A comparison of Figs. 1b and 2 shows that the focal mechanisms of earthquakes distributed off the coast could not be estimated because there are no offshore stations. The stress inversion results for regions N, C, and S for the periods before and after the Tohoku earthquake are shown in Fig. 3a, b, respectively.
For the period before the earthquake, the result of the stress tensor inversion in region N indicates that this region is characterized by a reverse faulting stress regime with σ 1 axis being sub-horizontal, trending between E-W and WNW-ESE. The stress ratio is < 0.5, which indicates that the magnitude of σ 2 is closer to σ 1 than to σ 3 (region N in Fig. 3a). In contrast, the results of the stress tensor inversion in regions C and S indicate that these regions are characterized by a reverse faulting stress regime with σ 1 axis being sub-horizontal and trending between N-S and NNE-SSW. The stress ratios show a wide distribution (regions C and S in Fig. 3a).
For the period after the Tohoku earthquake, region N is characterized by a reverse faulting stress regime with σ 1 axis being sub-horizontal and trending WNW-ESE; the stress ratio is generally < 0.5 (region N in Fig. 3b). Thus, the stress regime in region N is similar before and after the Tohoku earthquake. In contrast, the results of the stress tensor inversion in regions C and S indicate that these regions are characterized by a reverse faulting stress regime with σ 1 axis being sub-horizontal and trending ENE-WSW; the stress ratio is ~ 0.5 (regions C and S in Fig. 3b).
Our results indicate only minor temporal variations in the stress field in region N for the periods before and after the Tohoku earthquake. We consider this stress field to be associated with subduction of the Pacific plate, as discussed below. In contrast, the stress field in regions C and S clearly shows temporal variations though the scatter of the axis is larger before the Tohoku earthquake. The stress field differs from the generally recognized WNW-ESE compressional reverse faulting stress regime in NE Japan, regardless of the period before and after the Tohoku earthquake.

Temporally variable stress field from eastern Aomori to Tsugaru Strait
Our study results indicate that the stress field in regions C and S differs from that expected from subduction of the Pacific plate. In this section, we discuss the temporal variation in the stress field in these regions, in detail.
The stress field in regions C and S for the period before the Tohoku earthquake shows a reverse faulting stress regime with the maximum compressional principal stress (σ 1 ) axis being sub-horizontal and trending N-S. Recent studies have identified a region with sub-horizontal σ 1 axis trending N-S in the forearc region in NE Japan, and ascribed this orientation to the effect of bending of the upper-plate due to interplate coupling and/or the effect of subducting plate interface curvature (Imanishi et al. 2012(Imanishi et al. , 2013Yoshida et al. 2015Yoshida et al. , 2019. Yoshida et al. (2015) suggested that in parts of the forearc region, E-W compressional stress is low compared with N-S compressional stress because of upper-plate bending. The plate bending deformation produces E-W extensional stress, and thus N-S compressional stress is relatively larger than the E-W compressional stress. N-S trending σ 1 axis in regions C and S can be similarly explained by the effect of upper-plate bending, because the regions are located at the northern margin of the forearc region in Honshu.
However, we also found that the stress field in regions C and S for the period after the Tohoku earthquake constitutes a reverse faulting stress regime with σ 1 axis being sub-horizontal and trending ENE-WSW, suggesting that σ 1 axis was rotated by ~ 90° as a result of the Tohoku earthquake. Since there is no significant difference in hypocenter distribution for the periods before and after the Tohoku earthquake in regions C and S (Fig. 2), this stress rotation cannot be explained as an artifact caused by spatial variation in the stress field or by a change in hypocenter distribution. Rather, the rotation in σ 1 axis is explained by the effect of the Tohoku earthquake, as follows. Yoshida et al. (2012Yoshida et al. ( , 2019 reported that the stress field in inland areas of eastern Honshu, Japan, changed as a result of the Tohoku earthquake. The stress field in regions C and S is consistent with that in inland areas that  were affected by coseismic slip of the Tohoku earthquake (Yoshida et al. 2012(Yoshida et al. , 2019. Such a stress change caused by the 2011 Tohoku earthquake could have triggered some earthquakes with unfavorable fault orientation with the typical WNW-ESE compressional stress field in NE Japan. In addition, the strain distribution obtained from GEONET [GNSS Earth Observation Network System, operated by the GSI (Geospatial Information Authority of Japan)] shows that ENE-WSW contraction and NNW-SSE extension were predominant in the period after the Tohoku earthquake in northernmost Honshu, including regions C and S (e.g., Figure 10 in Geospatial Information Authority of Japan 2018a). To release such strains, the ENE-WSW compressional reverse fault earthquakes may have been caused. All of the above suggests that the stress field in regions C and S for the period after the Tohoku earthquake was influenced by coseismic static stress change and postseismic crustal deformation related to the Tohoku earthquake. Yoshida et al. (2012) reported that the Tohoku earthquake resulted in a coseismic change in differential stress of < 1 MPa in the present study area. The strain change due to postseismic deformation (for the period up to 2017) is estimated to have been the order of 10 −5 based on GNSS observations (Geospatial Information Authority of Japan 2018b), which corresponds to the order of 1 MPa change in the horizontal differential stress (σ Hmax − σ Hmin ) assuming a rigidity of 30-40 GPa. We just show an order-estimation of the strain change here because the strain change estimated from GNSS station data varies with a subjective weighting parameter (e.g., "distance-decaying constant" proposed by Shen et al. 1996); but we can safely conclude that the change in the horizontal differential stress (σ Hmax − σ Hmin ) due to the coseismic and postseismic deformation caused by the Tohoku earthquake should have been < 10 MPa. Thus, the value of (σ 1 − σ 2 ) before the occurrence of the Tohoku earthquake should have been also < 10 MPa because horizontal σ 1 and σ 2 axes in the region S exchanged their directions after the earthquake as shown in Fig. 3. As the stress ratio is estimated to have been ~ 0.3 for the period before the Tohoku earthquake, the differential stress (σ 1 − σ 3 ) is thought to have been less than tens of MPa. The obtained differential stress is consistent with the estimate of Yoshida et al. (2015Yoshida et al. ( , 2019. In regions C and S, earthquakes are generated by movement on normal and strike-slip faults as well as reverse faults. Miyauchi (1985) identified geological folding structures in these regions that strike E-W as well as N-S. Bouguer anomaly patterns reveal small-scale structures with variable strike superimposed on largescale N-S-striking structures ( Fig. 4; Yamamoto 2005). Based on the information shown above, it is considered that fault planes with various orientations exist in these regions. The occurrence of various fault types of earthquakes under such low differential stress in these regions may be due to the presence of such weak fault planes.

Effect on the stress field of southwestward movement of the Kuril Forearc sliver in southern Hokkaido
Our results show that the stress regime (reverse faulting with σ 1 axis oriented sub-horizontal and trending WNW-ESE) in region N was essentially unchanged between the periods before and after the Tohoku earthquake (Fig. 3). The results suggest that the effect of the subduction of the Pacific Plate is more dominant than the effect of the southwestward movement of the Kuril Forearc sliver (e.g., Kimura 1981Kimura , 1996Seno 1985;Moriya 1986;Arita et al. 2001) on the stress field in this region even after the Tohoku earthquake.
On 6 September 2018, the Mw 6.6 Hokkaido Eastern Iburi earthquake (herein, the Iburi earthquake) occurred in the northeastern part of region N. Ohtani and Imanishi (2019) estimated the stress field around the northeastern edge of region N using the focal mechanism solutions of the aftershocks that occurred in the region indicated by a rectangle in Fig. 4. They found that this area is characterized by a reverse The locations of these regions are shown in Fig. 1b. The left panel shows stress tensor inversion results. Red, green, and blue dots correspond to the 95% confidence regions of the σ 1 , σ 2 , and σ 3 axes, respectively, and red, green, and blue circles correspond to the best-fit solutions of the same axes, respectively. The right panel shows the frequency distributions of stress ratio R, where R = (σ 1 − σ 2 )/ (σ 1 − σ 3 ), for the respective regions faulting stress regime with σ 1 axis oriented sub-horizontal and trending ENE-WSW, and suggested that this stress regime has resulted from southwestward migration of the Kuril Forearc sliver and its collision with NE Japan. The focal mechanism solutions used in the present analysis for region N are located mostly in the western part of the region. Therefore, it is likely that the stress field changes between the western and eastern parts of region N. The Bouguer anomaly shown in Fig. 4 also shows the difference in the structure between the western and eastern parts there. However, the details of the spatial variation in the stress are unclear, owing to insufficient data. To overcome this limitation, it will be necessary to use data from ocean-bottom stations (e.g., NIED 2019a) in future work.

Conclusions
We determined the focal mechanisms of earthquakes in the low-seismicity region extending from southern Hokkaido to eastern Aomori (northeasternmost NE Japan arc) and investigated spatiotemporal variations in the stress field. Our results indicate that the stress regime in southern Hokkaido remained essentially unchanged after the Tohoku earthquake, whereas the regime changed after the earthquake in the region from eastern Aomori to Tsugaru Strait. The differential stress is estimated to be less than tens of MPa before the Tohoku earthquake in the latter regions. Bouguer anomaly map for the study area (modified from Geological Survey of Japan, 2013). A density of 2.67 g/cm 3 was used for calculating Bouguer anomalies. The contour interval is 2 mGal. The white line shows the coastline. The black dashed rectangle indicates the distributional range of the aftershocks used in the stress tensor inversion by Ohtani and Imanishi (2019). The big double arrows indicate the locations of large-scale N-S-striking structure (Yamamoto 2005). There are many small-scale structures, among which some are shown with small double arrows