Tectonic Implications

In this section, I will discuss the tectonic implications for the clustering result by comparing cluster boundaries with active structures published by CGS and the TEM. For the convenience of the discussion, I will separate the Taiwan region into four domains including (I) northern Taiwan, (II) western Taiwan, (III) eastern Taiwan, and (IV) southwestern Taiwan (Figure 16) , and discuss the inferred structures from the clustering results in details.

6.2.1 Northern Taiwan

The Shanchiao fault is the principal fault of northern Taiwan. Both CGS and TEM maps have the Shanchiao fault. The clustering result identified the clusters separated by the Shanchiao fault as k = 26 (Figure 50). The fact that it is distinguished lately suggests the slip rate of the Shanchiao fault is low and it has the inconspicuous velocity change (Shyu et al., 2005b; Shyu et al., 2016). However, the Shanchiao fault is the principal fault of the Taipei Basin where is the most populous city in Taiwan, the potential hazard of the earthquake still cannot be ignored.

The Taoyuan and Hsinchu areas are at the transition between the waning collision and the post-collision collapse (Shyu et al., 2005b). The Shuanglienpo structure, the Yangmei structure, and the Fengshan hills frontal structure are the suspicious structures

in this region. In Figure 35, the clusters are divided along the Fengshan hills frontal structure and it represents the characteristic of the crustal movement of this region with low slip rates, which is similar to the Taipei domain. Because of low geological slip rates of these structures (Shyu et al., 2016), we cannot clearly distinguish the clusters of these structures until k = 26 (Figure 50), the clusters are been divided into the Taoyuan-Hsinchu area and the Miaoli area along with the Touhuanping structure.

Figure 61 Cluster result when k = 34 in northern Taiwan.

6.2.2 Western Taiwan

In the Miaoli domain, the activity of the active structures is higher than the Taoyuan domain. The geomorphic feature of this region is presented as anticline-syncline pairs.

The geological survey suggests that these anticline-syncline pairs are the geomorphic expression of the underlying detachment in this region (Shyu et al., 2005b). In the clustering result, there is no cluster has been distinguished in the Miaoli domain, which is consistent with the pattern of folding deformation, which is more gradual than the pattern derived from faulting.

There are three subparallel thrust faults in the center of the western Taiwan from west to east including the Changhua fault, the Chelungpu fault, and the Tamaopu-Shuangtang fault, from west to east. These structures are merged at the southern Taichung domain (Shyu et al., 2005b). The Changhua fault is the western limit of the active deformation in the western Taiwan and forms several tablelands in Taichung. The Chelungpu fault is the main fault that causes the 1999 Chi-Chi earthquake. The most internal fault in this region is the Tamaopu-Shuangtang fault, the northern section is called the Tamaopu fault and the southern section is the Shuangtung fault. In Figure 34 and Figure 35, the Western Foothills can be divided into two sections along the Chelungpu fault and the Tamaopu-Shuangtung fault. After that, Figure 38 and Figure 41 are be separated again, although the clustering result cannot fit very well but still can see the parallel trend. The discrepancy between the surface faults line and cluster boundaries may result from smooth velocity gradient due to fault locking and fault-related fold geometry (Yue et al., 2005).

South of central Taiwan in the Chiayi domain, the geomorphic expression of the fluvial terraces in the west of the Chiayi hills suggests that there is the blind thrust fault system under the foothills and coastal plain. Despite the lack of the geomorphic expression of the active structures, the geodetic survey still can reveal that the shortening rate of the Chiayi domain is high. However, the Chiahyi hills and the Chianan Plain are filled with the thick fluvial sediments that cause the structures in this area to have some controversy (Suppe, 1980; Hung et al., 1999; Shyu et al., 2005b).

In addition, more clusters were calculated in southwestern Taiwan. There is a linear cluster been distinguished very early, k = 5 (Figure 29), which starts from the Chiuchiungkeng fault, the Muchiliao fault, the Liuchia fault, the Chungchou structure, the Hsinhua fault, and end at the Hsiaokangshan fault. The GPS observations also indicate that there are significant shortening of the Chiuchiungkeng-Muchiliao-Liuchia fault system with high potential for large earthquakes (Tsai et al., 2012).

Figure 62 Cluster result when k = 34 in western Taiwan.

6.2.3 Eastern Taiwan

In eastern Taiwan, the northern Ilan structure and the southern Ilan structure are able to be distinguished early when k = 9 (Figure 33). Due to the back-arc extension of the Okinawa Trough, the Ilan Plain is rapidly sinking and bounded by the two normal fault systems, which can be identified from geomorphic expression (Shyu et al., 2005b).

In addition, there is an interesting discovery that there are two stations are been separated from the surrounding stations in the Nan’ao area. Coincidentally, Takahashi et al. (2019) also found that the movements of these stations are different and can be isolated from the surrounding stations through the hierarchical cluster analysis. Affected by the back-arc spreading of the Okinawa Trough, the northernmost Central Range is bending clockwise and the local velocity fields in this area is moves southeastward to the Philippines sea plate (Hu et al., 1996).

In the Figure 14, the cluster analysis distinguished the Longitudinal Valley fault first, which is the principal active structure in the eastern Taiwan and also is a high-angle oblique thrust fault with weakly fault locking (Ho, 1986; Yu and Kuo, 2001; Shyu, 2005) and seismic rupture (Shyu et al., 2006). Due to the obvious velocity discrepancy, the Longitudinal Valley fault is identified as the boundary in the clustering result and divided the eastern Taiwan into the Coastal Range and the Central Range. Although the shortening rate of the Longitudinal Valley fault is high, the closure rate of the north sections is lower than the south section. The clustering result also can be used to corroborate the previous geodetic and geological survey. In Figure 43 and Figure 51, the Longitudinal Valley fault is been divided into the northern part and the southern part, which is consistent with the observations of higher deformation in the southern segment than northern segment. In addition, the Longitudinal Valley fault divided into three major units including the Central

Range, the valley between the Central Range and the Coastal Range, and the Coastal Range. Figure 43 reveals the northern segment of the Longitudinal Valley fault and part of the middle segment in the southern Longitudinal Valley fault. Figure 51 only shows the middle segment of the southern Longitudinal Valley fault. In Figure 53, the east side of the southern Longitudinal Valley fault is separated. These results can show that the Central Range fault is buried in the northern Longitudinal Valley and the southernmost Longitudinal Valley fault is partitioned into two branches as suggested by Shyu et al.

(2005b).

Except the Longitudinal Valley fault, the Milun fault is also an dominate the important fault in the eastern Taiwan, which ruptured during the 1951 Hualien-Taiwan earthquake sequence and the 2018 Hualien earthquake (Huang et al., 2019; Lin et al., 2019). According to the Figure 30, the cluster in eastern Taiwan is be divided into two sections including the Milun fault and the Longitudinal Valley fault. This result also can support that the movement characteristics of the Milun fault is different from the Longitudinal Valley fault.

According to the bathymetry data and the geological survey, Lutao-Lanyu domain is belong to the volcanic arc which is the incipient collision stage (Shyu et al., 2005b). There is one obvious linear fault scarp can be found in the west of Lutao which have significant to the shortening rate to offer the high rates of uplift for the Lutao-Lanyu (Shyu, 2005).

This evidence also can be proved in the clustering result that demonstrates Lutao-Lanyu be separated from Taiwan main island at k =16 (Figure 40).

Figure 63 Cluster result when k = 34 in eastern Taiwan

6.2.4 Southwestern Taiwan

South sections of the Chungchou structure, clusters are differentiated when k = 15, 18 and 33 (Figure 39, Figure 42 and Figure 57). Geologic core records indicate that the deformation rate is not consistent between the Chungchou structure and the Hsiaokangshan fault and it might be caused by a fold and reverse fault (Chen, 2009).

Although the slip rate of this structure is not well constrained, the clustering result still can identify this structure multiple times so the existence of the Chungchou structure is certain. In addition, a cluster is identified and bounded between the Longchuan structure and the Hsiaokangshan fault (Figure 31 and Figure 37). The crustal deformation of this area may be affected by the mud diapirs (Ching et al., 2016).

Figure 14 shows that the blue cluster is a distinct block bounded by the Chaochou fault and the Chishan-Youchang structure. However, the geodetic data indicates that the displacement does not have significant change across the Hengchun-Chaochou fault (Chen et al., 2005). Chen (2006) analyzed the GPS velocity fields in southwestern Taiwan and discovered that there are many boundaries of the velocity gradients, the Chishan fault is the boundary, which separated the Pingtung Plain and the Western Foothills. The expression of the velocity fields is different on the east and west sides of the Chishan fault in a way that the velocity at the east side is higher than the west side. According to the geodetic measurements, the Pingtung Plain, located between the collision zone and the

Manila Trench, is moving toward the southwest direction (Ching et al., 2007a). The area of the Pingtung Plain is considered to be moving as a block of tectonic extrusion (Lacombe et al., 2001; Liu et al., 2004; Hu et al., 2007) and is thought to be deformed associated with mud diapirs (Sung et al., 2010; Chen et al., 2014b). Although the deformation mechanism is not well resolved yet, it is certain that the Pingtung Plain is acting as a distinct tectonic block, different from the Western Foothills.

In addition, the cluster analysis identified a NW-SE trending structure, which is consistent with previously inferred Fenshan fault. Since the lack of the obvious outcrop and no historical earthquake happened along this structure, it is hard to judge the existence or activity of the Fengshan fault. According to the high-resolution Digital Elevation Model (DEM) and the geomorphological interpretation, Lacombe et al. (1999) indicate the possible location of the Fengshan transverse fault zone. The GPS observations and the PSInSAR analysis also indicate that there is a significant creeping in southwestern Taiwan and the velocity field is different on the east and west sides of the Fengshan fault (Ching et al., 2007a; Chao, 2016). There is no this fault in the CGS active fault map, and the TEM map shows a Fengshan Hills frontal structure, which is considered as a fold and its location is different from the Fengshan fault. In Figure 32, a cluster showed up and the geometry seems more similar to the version proposed by Ching et al. (2007a).

Figure 64 Cluster result when k = 34 in southwestern Taiwan.