Fujiwara et al. (2016) demonstrated that DLs caused by the 2016 Kumamoto earthquake sequence tend to run parallel to each other with nearly constant horizontal spacing. They can therefore be classified into several groups based on their locations and features. Two regions—around the Suizenji Park and northwest of the outer rim of Aso caldera (hereinafter called “NW-Aso”)—are analyzed in detail in this study. Several parallel DLs have appeared at these places.
Figure 12 depicts epicenters of earthquakes occurring after April 14, 2016, including aftershocks, and post-seismic surface deformation detected by InSAR, respectively (Fujiwara et al. 2016). For most DLs, the observed distribution matches well with the aftershock distribution. Corresponding focal mechanisms are depicted in Fig. 12c, and as observed, the focal mechanism of aftershocks as well as the strike and displacement directions of DL agree well with the right-lateral movement, especially in the northeast and southwest regions depicted in Fig. 12c. However, there exist places, such as NW-Aso, where DL concentration occurred, despite no occurrence of aftershocks.
Post-seismic deformation (Additional file 1: Fig. S1) was observed in regions surrounding the Suizenji Park, whereas no such deformation was observed in NW-Aso. Further, several aftershocks were recorded around the Suizenji Park, whereas no aftershocks were observed in NW-Aso. Fujiwara et al. (2016) suggested that there exists a significant difference between DL causes observed at Suizenji Park and NW-Aso. DLs at both these sites demonstrate graben formation. In the Suizenji Park region, however, some deformation mechanism was observed to be active after the main shock (Additional file 1: Fig. S1; Fujiwara et al. 2016), whereas no significant post-seismic deformation or aftershocks were observed in NW-Aso. The remainder of this section examines differences between these DL groups.
Around Suizenji Park
Figure 13 depicts DLs around Suizenji Park along with a high-pass filtered up–down displacement map (Fujiwara et al. 2016) obtained using three-dimensional (3D) InSAR (Morishita et al. 2016; Morishita 2019). Figure 14 depicts the surface strain (extension and contraction) and dilatation values calculated using a 3D deformation map for co-seismic displacement (Additional file 1: Fig. S2). As can be observed, the surface strain clearly demonstrates a dominant ENE–WSW extension in this region. Therefore, large horizontal displacements along the ENE direction owing to right-lateral movement of the Futagawa fault that exists east of this region can be considered a probable reason behind DL occurrence. Because larger horizontal displacements exist in the eastern part, this region is characterized by an ENE–WSW extension field and displacements with a graben or half-graben structure. In addition, the observed DL movement coincides with increased delta CFF of main seismogenic faults, especially at shallow locations (Additional file 1: Fig. S3).
On the other hand, it is a natural predisposition to assume that faults in this region existed and that an earthquake led to their passive displacement as the DLs. In fact, some DLs can be observed to have been distributed along active faults described in the 1:25,000 active fault map “Kumamoto (revised ed.)” (Kumahara et al. 2017) published after the earthquake (Fig. 13). Subsequently, we estimated the existence of a shallow subsurface structure around Suizenji Park using borehole data and investigated the existence of faults that predispose to DLs associated with this earthquake.
Figure 13 depicts positions of borehole data acquisition and survey lines (A1–A9 and B1–B10) that connect them. Borehole data were obtained from the Geo-Station (National Research Institute for Earth Science and Disaster Resilience 2016). Figure 15a shows the vertical-displacement cross-section of InSAR, topographic cross-section, and borehole column diagrams along survey lines A1–A9 as well as the subsurface formation estimated using this information. The borehole columns, i.e., from the surface to approximately − 20 m above sea level, are divided into gravelly soil layers with N values exceeding 50 with sandy and silty soil layers above them. According to Tajiri et al. (1998), this region contains alluvium deposited from the surface to approximately 0 m above sea level and a terrace gravel layer between 0 m and − 20 m above sea level. Aso-3 or Aso-4 pyroclastic flow deposits are distributed below − 20 m above sea level. The sandy and silty soil layers depicted in Fig. 15a are considered alluvial layers and the gravelly soil layers are considered terrace gravel layers.
From Fig. 15a, using the top elevation of the gravelly soil layer as key, vertical displacements of the stratum can be recognized on the west of boreholes A2 and A6. As observed, the position and orientation of vertical displacements exist in harmony with those of DLs detected by InSAR. In particular, an active fault can be confirmed to the southwest of borehole A6 (Fig. 13) with a high probability of fault existence. Existence of a graben structure can be confirmed between boreholes A2 and A6, and the subsidence between boreholes A2 and A6 can be observed in the subsurface region. In addition, the position and direction of faults exist in harmony with those of DLs to the west of borehole A7. The fault is likely to correspond to the northern extension of the active fault identified in the south of this borehole (Fig. 13). In addition, stratum displacements to the west and east of boreholes A3 and A4, respectively, as well as the DL locations between boreholes A8 and A9 could not be confirmed.
Figure 15b depicts the vertical-displacement cross-section, topographic cross-section, and borehole column diagrams along survey lines B1–B10 along with subsurface formation estimated using this information.Footnote 1 The borehole column diagrams reveal that although the volcanic-ash layer remains partially sandwiched from the surface to approximately − 20 m above sea level, the borehole columns remain roughly identical to those corresponding to survey lines A1–A9. However, considering the geological profile of Tajiri et al. (1998), the hard rock at B10 at a depth of more than 15 m above sea level is Aso-4 pyroclastic flow deposit. To be precise, the borehole columns largely comprise a gravelly soil layer with N value exceeding 50, capped by sand and silty soil layers at the top. Similar to the case described above, when the top elevation of the gravelly soil layer is considered key, vertical displacement of the stratum can be recognized to the west of borehole B7 and between boreholes B8 and B9. The position and orientation of these vertical displacements of the stratum were observed to be consistent with those of DLs detected by InSAR, thereby confirming the existence of a fault. In particular, an active fault can be confirmed between boreholes B8 and B9 with a high probability of fault existence because the active fault was described by Kumahara et al. (2017) and Goto et al. (2017).Footnote 2 Likewise, no displacement of the geological formation could be confirmed at DL locations detected to the east and west of boreholes B6 and B9, respectively. Although no DL was detected between boreholes B7 and B8, the existence of faults with vertical displacement was indirectly confirmed by the top elevation of the gravelly soil layer. Since there exists an active fault to the north of the gap between boreholes B7 and B8 (Fig. 13), the fault likely extends to the south as well.
In view of the above results, DLs detected by InSAR were active faults, as reported by Kumahara et al. (2017), unknown faults, or passively displaced faults caused by the Kumamoto earthquake sequence.
It is noteworthy that irrespective of DLs demonstrating similarities in length and displacement on interferograms, their respective underground structures differ significantly. It is unclear whether this is due to insufficient borehole data or if DLs move during an earthquake without necessarily being dependent on their underground structures. Therefore, further research must be performed in this regard.
Northwest of outer rim of Aso caldera (NW-Aso)
In this study, several dozen DLs were observed in the NW-Aso region. These DLs were generally aligned along the WNW–ESE direction and comprise typical dip-slip displacements. Figure 16a depicts DLs observed in NW-Aso along with a high-pass-filtered up–down displacement map, whereas Fig. 16b depicts the difference between vertical displacements to the north and south of each DL. The largest up–down displacement gap (exceeding 40 cm) was observed in the southern part of NW-Aso. Two typical cross-sections indicated in Fig. 16a are depicted in Fig. 17.
The DL group in NW-Aso was further divided into two groups—northwest and southeast—the mainly comprised dip-slip displacements comprising downward movement of the southern and northern DL sides, respectively (Fujiwara et al. 2016). The observed displacement patterns demonstrate saw-tooth shapes, directions of which differ between the two groups (Fig. 17). Suzuki et al. (2017) reported existence of up to 10 known active faults in this region. These faults (belonging to the Kuradake fault group) are characterized by graben landforms, and the positions and directions of the known active faults coincide with those of DLs (Fujiwara et al. 2016; Fig. 6).
In this study, the existence of a correlation between topography and vertical displacement can be realized. For example, the 2-km horizontal distance in Fig. 17b shows that the difference in height (topography) equals approximately 50 m, and the vertical displacement equals approximately 50 cm. If the same displacement value is recorded for each historical earthquake, it would require at least 100 such earthquakes to build the current topographic relief. This correlation between slip displacements and topography suggest that earthquakes occurring in regional tectonic stress fields tend to cause similar DLs at identical positions resulting in the formation of the observed topography (Fujiwara et al. 2000a).
There exist several parallel faults along the E–W direction in central Kyushu between the Beppu Bay and Unzen (Chida 1992). NW-Aso (the Kuradake fault group) is located near the center of this characteristic region. Figure 18a depicts the Haneyama–Kuenohirayama fault zone (located approximately 30 km north of the Kuradake fault group). To facilitate comparison between the two fault groups, Fig. 18b depicts DLs demonstrating normal fault movements with identical to those of corresponding movements observed in NW-Aso. As observed in both groups, the northern and southern parts mainly comprise dip-slip displacements with the corresponding southern and northern sides exhibiting downward movement, respectively. This suggests that fault groups in central Kyushu were formed via similar mechanisms. Based on the similarity between the Kuradake fault group and the Haneyama–Kuenohirayama fault zone, the occurrence of large earthquakes in faults such as the Futagawa fault zone (the main earthquake fault) likely triggered movements of surrounding DLs. Some of the remaining N–S extension strain in the shallow area is likely released in the normal fault group in central Kyushu. In addition, these normal fault groups exist near regions of volcanic activity, and together with wide-range stress-field characteristics, this can be considered a unique but typical geological feature of central Kyushu.
Figure 19a depicts the difference between N–S components of horizontal displacements to the north and south of each DL. This difference in displacements corresponds to N–S opening displacements at each DL. Figure 19b illustrates surface strain and dilatation values calculated using the 3D deformation map (Additional file 1: Fig. S2). As can be observed, strain distributions in the southern and eastern parts of Fig. 19b demonstrate greater variation from place to place compared to other regions owing to the southern region’s proximity to the Futagawa fault and the occurrence of local lateral spreading in the eastern region (Fujiwara et al. 2017a). In the central region, N–S extension is dominant around “Ext.” in Fig. 19b, whereas an N–S contraction exists around “Cont.” This can be attributed to the greater northward movement of the southern part of “Cont.” compared to other regions. Interestingly, DLs in “Cont.” moved to open N–S direction (Fig. 19a) despite the shortening of the strain in this direction (Fig. 19b). This is because, in this study, strain was calculated by averaging surrounding crustal deformation data, thereby resulting in an effect similar to that of low-pass-filtered displacement; the strain represents a somewhat broader force than the ones that DLs were receiving directly. (Figure 19a). In addition, DL movement observed in this study does not coincide with delta CFF caused by main seismogenic faults. In particular, at greater depths (of the order of 2 km), there exists a force that further restrains DL movement (Additional file 1: Fig. S4). Therefore, it is important that some DLs do not correspond to the strain caused by the Kumamoto earthquake sequence. In other words, the static displacement caused by the earthquake does not directly shift DLs in NW-Aso. The observed slip in each DL is likely triggered by dynamic shaking encountered during passage of seismic waves (Kaneko et al. 2015). Thus, it can be inferred that uniform DLs around the Suizenji Park are caused by the systematic and simple strain field that exists therein, while uniform DLs in NW-Aso appear to have been caused under a complex strain field caused by the Kumamoto earthquake sequence. These contrasting stress fields may consider the cause of the completely different DL movements observed at these locations.