Seismogenic Structures Offshore Taiwan and Their Earthquake Potential from the Taiwan Earthquake Model (TEM) Project Open Access

Taiwan is an active orogenic belt with numerous seismogenic structures and frequent earthquake activities (e.g., Chen et al., 2016; Shyu et al., 2020). In this dynamic setting, the multidisciplinary team of the Taiwan Earthquake Model (TEM) project, which is funded by the National Science and Technology Council of Taiwan, integrates seismology, geology, and earthquake engineering approaches to assess future seismic hazards and risks. By continuously refining methodologies and exploring innovative research techniques, the TEM Team has achieved some leading case studies in seismic hazard research that have largely enhanced our overall understanding of earthquake hazards. For example, the seismic hazard assessment results presented by Chan et al. (2017) identified eastern and southwestern Taiwan as regions with high seismic potential, consistent with subsequent earthquake events. Overall, these studies not only enhanced our insight into regional seismic risks but also provided valuable experiences for seismic hazard assessment practices worldwide.

Although there have been many efforts to study seismic hazards in Taiwan, most of these studies have focused on onshore faults and often overlooked the impact of offshore structures (e.g., Shyu et al., 2016, 2020; Chan et al., 2020). However, there have been reports that historical major earthquakes and active structures are indeed located offshore (e.g., Deffontaines et al., 2001; Shyu et al., 2005; Wang et al., 2011).

Earthquakes generated by offshore seismogenic structures will produce significant impacts on both offshore and on‐land areas. For example, the January 2024 Noto Peninsula earthquake in Japan triggered severe ground shaking and destruction (Suppasri et al., 2024), and the April 2024 offshore Hualien earthquake in Taiwan caused road and bridge damage, utility disruptions, building collapses, and casualties (Kanwal, 2024). In both cases, the limited prior information on offshore seismic structures hindered effective earthquake impact assessments. To fill this gap, understanding the locations and seismic capacities of offshore structures is essential for disaster preparedness, as well as future exploration and development offshore.

Currently, several versions of active fault databases exist in Taiwan, including one developed by the Geological Survey and Mining Management Agency that identifies 36 active faults (Lin, Liu, et al., 2021), and another by the TEM project, which includes 45 seismogenic structures (Shyu et al., 2020). Offshore seismogenic structures, which also have the potential to generate significant earthquakes, are not included in these databases. Although Deffontaines et al. (2001) and Shyu et al. (2005) have mapped some offshore structures, their results only show structural locations, limiting their applicability for seismic hazard assessments.

Recent advances in geological and geophysical research in Taiwan’s offshore regions, including bathymetry and seismic surveys, have substantially improved our ability to identify offshore seismogenic structures. As a result, we reassessed and integrated these new results to identify potential seismogenic structures in the offshore regions of Taiwan. In addition, the geometries of these offshore structures were analyzed using pre‐existing datasets and earthquake hypocenter depths. By combining geometric parameters with empirical regression equations, we estimated potential earthquake magnitudes for these offshore faults. This study includes detailed information on the location, geometry, and earthquake potential of Taiwan’s offshore structures, which will provide important references for future earthquake hazard assessment and marine resource management.

Taiwan is located at the boundary where the Philippine Sea Plate converges with the Eurasian Plate, moving northwest at a rate about 80–90 mm/yr (e.g., Yu et al., 1997; Shyu et al., 2006; Hsu et al., 2016; Fig. 1). This convergence results in the collision between the Luzon volcanic arc and the Eurasian continental margin, forming Taiwan’s north–south‐trending orogenic belt and complex offshore topography (Yu and Song, 2000). In the Luzon subduction system south of Taiwan, the Eurasian Plate subducts beneath the Philippine Sea Plate, forming north–south‐trending features such as the Manila Trench, fore‐arc ridges, and the Luzon volcanic arc. In the Ryukyu subduction system east of Taiwan, the Philippine Sea Plate subducts beneath the Eurasian Plate, generating east–west‐trending features and the Okinawa Trough.

According to the oblique collision model of Taiwan, the orogenic process propagates from north to south (Suppe, 1981, 1984), leading to distinct orogenic stages (Teng, 1996; Shyu et al., 2005; Fig. 1). This progression, combined with varying regional stress conditions, results in different structural patterns and activities across the island. For instance, southern Taiwan, in the incipient stage of orogeny and under compressive stress, is more prone to the development of reverse faults, whereas northern Taiwan is in a post‐collision environment and tends to develop normal faults.

The area of this study spans ∼200 km offshore from Taiwan’s coast, between 20° and 26° N and 119° and 124° E (Fig. 2). In the twenty‐first century alone, this region has experienced several nearshore earthquakes that damaged buildings on‐land and caused casualties, including the 26 December 2006 Hengchun offshore event, the 31 March 2002 earthquake offshore Hualien, and the 3 April 2024 Hualien earthquake. Although both the Manila Trench and the Ryukyu subduction zone are also primary seismic sources, both interplate and intraplate earthquake sources of the subduction zones have already been considered in previous seismic hazard assessment studies (e.g., Wang et al., 2016; Chan et al., 2020). Therefore, this study only focuses on crustal structures and excludes the subduction zones.

For the identification of offshore structures, direct methods such as outcrop and trenching surveys are generally not feasible, and it is necessary to rely on geophysical methods. Based on previous reports, Taiwan’s seismogenic structures have recurrence intervals ranging from hundreds to tens of thousands of years (Shyu et al., 2020). Although earthquake data from the past few decades can help identify some potential structures (e.g., Wu et al., 2024), this timescale is too short to illustrate locked structures that have not produced recent seismicity. Therefore, this study investigates long‐term structural deformation through the analysis of seismic reflection profiles and bathymetry data.

We integrated previously published geological and geophysical data from 1980 to 2024 to efficiently locate potential offshore seismogenic structures (Fig. S1, available in the supplemental material to this article). Structure names are based on prior studies; for unnamed structures, names were assigned based on nearby geographic features. Following the convention of the TEM project, for structures that have not been reported as faults in previous studies, we refer to them as just “structures.” Different structures within the same system are distinguished by letter suffixes (e.g., A or B), and continuous segments by numbers (e.g., 1 or 2).

The credibility of these structural locations is ranked A, B, or C, from the highest to the lowest (Table S1), based on the resolution of the data using the following two criteria.

Lateral discontinuities in seismic reflections commonly indicate faulting, and the disturbance of near‐seafloor sediments is often interpreted as evidence of recent fault activity. The locations of such structures are identified based on existing interpretations of seismic reflection profiles and through the reinterpretation of original seismic data. Because of the differences in acquisition timing, methodologies, data processing, and resolution among the seismic reflection profiles, interpretations for any given structure are preferentially based on the highest‐quality profile available.

Because of their limited penetration depth, seismic reflection profiles generally image only the upper few kilometers of the crust, restricting the detection of deeper seismogenic structures. Because deeper structures that do not yet reach the seafloor were excluded from this study, seismicity that occurred on them may be considered as areal sources in future probabilistic seismic hazard assessments.

Criterion 1 provides point data for a structure’s seafloor location, whereas criterion 2 defines its lateral extent. In the absence of significant erosion, active dip‐slip faults typically display cumulative deformation with lateral extensions visible at surface exposures. Structures with dominant strike‐slip components often form linear features due to topographic differences across the fault, particularly where the fault trace bends.

Faults with longer lengths tend to have a higher potential to generate large earthquakes (Wesnousky, 2008). As a result, this study focused on structures at least 20 km in length because such features are capable of generating magnitude 6.0 or greater earthquakes that may impact onshore areas.

We used ArcGIS software to analyze digital elevation models (DEMs; see Data and Resources for details), generating contour lines, shaded relief maps, and slope maps to identify seafloor lineaments. The solid lines represent structures with clear lineaments, and the dashed lines denote those inferred from seismic reflection profiles or other methods when lineaments are less distinct. DEMs for the offshore region surrounding Taiwan were obtained from the National Science and Technology Council’s Oceanography Database (200 m resolution) and the GEBCO website (450 m resolution).

Reconstructing the subsurface geometry of seismogenic structures is essential for estimating earthquake magnitudes, which is achieved by calculating the rupture area by multiplying the fault length and width (Fig. 3). The lengths were measured using ArcGIS, based on seabed lineaments. To determine fault width, we used two methods: when previous studies provided subsurface geometrical models, the published values were used. For structures without such data, the width was calculated using the fault’s dip angle and the seismogenic depth (Fig. 3). When available, we used dip angles reported in prior studies. Otherwise, we applied Anderson (1905) theory, which assumes a constant dip angle with depth: 30° for reverse faults, 60° for normal faults, and 90° for strike‐slip faults.

Seismogenic depth is defined as the vertical distance from the seabed to where brittle failure occurs in the crust (Fig. 3). We used 200 m resolution bathymetry data analyzed in ArcGIS to calculate average water depth (depth 0) along the fault lines (Appendix Excel Table). The maximum depth of historical and instrumental earthquakes, excluding the subduction zone and deeper source events (e.g., Wu et al., 2017) serves as an initial proxy for brittle failure depth. The structural geometry is then adjusted based on the subsurface intersections of faults or the geometries of nearby on‐land faults, assuming no variation in geometry along the strike direction of the structures.

For consistency with the assessments of on‐land faults, we follow the error standards defined in the TEM project (Shyu et al., 2016, 2020). Fault dip angles have an uncertainty of ±10° when estimated solely by Anderson’s theory, and ±5° when supported by geological data. Seismogenic depth has a uniform uncertainty of ±2 km (Appendix Excel Table).

This study employed three methods to estimate the potential earthquake magnitude of each structure. Two regression methods used by Shyu et al. (2020) were also used to ensure comparability with on‐land structures in Taiwan. The first method, based on Wells and Coppersmith (1994), estimates magnitudes using the structural movement type and rupture area, making it widely applicable in global and local studies (Shyu et al., 2020). The second method, adapted from Yen and Ma (2011), incorporates parameters optimized for Taiwan’s tectonic setting (Shyu et al., 2020). The third method, following Stirling et al. (2013) and Wesnousky (2008), uses the fault’s length and movement type to reduce uncertainties in rupture area estimation. Detailed results from all three approaches are compared in Table S2.

This study categorizes Taiwan’s offshore region into six neotectonic domains (Figs. 2 and 4–8) based on their geologic settings, seafloor topography, and structural characteristics. These domains are: northern offshore Taiwan (domain A), the southern Okinawa Trough (domain B), the upper plate of the Ryukyu subduction zone (domain C), eastern offshore Taiwan (domain D), southwestern offshore Taiwan (domain E), and western offshore Taiwan (domain F). However, due to transitional geological features, the boundaries between some domains are not clearly defined. The geological background and structural features of each domain will be described in subsequent sections.

Domain A is located in the northeastern offshore region of Taiwan, extending from the East China Sea Continental Shelf southeastward to the continental slope and the northern margin of the Okinawa Trough (Fig. 4a). The southwestward migration of the Taiwan orogeny (Teng, 1996) transformed the regional stress field from a compressional orogenic belt to an extensional environment during the late Pliocene to early Pleistocene. This change reactivated former reverse faults as normal faults (Emery et al., 1970; Suppe, 1984; Hsiao et al., 1998; Kong et al., 2000; Shyu et al., 2005; Shang et al., 2017; Yang et al., 2021; Nguyen et al., 2022). Seismic reflection profiles show that deeper stratigraphic layers (Yang, 2017; green line in Fig. 4b) were deformed by early orogenic compressional folding, and younger unconformities (red line in Fig. 4b) show normal‐faulting displacements (Hsiao et al., 1998).

The northwestern boundary of this structural inversion zone is near the Shanchiao fault, which trends northeastward along its strike (Hsiao et al., 1998; Nguyen et al., 2022; Fig. 4a). West of this inversion boundary, the East China Sea Continental Shelf is a flat, shallow platform with no evidence of structural inversion. Seismic reflection profiles show faults with varying dip directions buried beneath thick, undeformed sedimentary deposits, suggesting that they are inactive (Sun, 1985; Chen and Watkins, 1994; Yang, 2017).

East of the structural reversal boundary lies a collapsed orogenic zone (Nguyen et al., 2022), characterized by the Pliocene regional sequence boundary (PRSB; red line in Fig. 4b). The PRSB is offset by eleven northeast–southwest‐oriented normal faults, dipping southeast and extending to the seabed (Hsiao et al., 1998; Atomic Energy Council, 2016; Yang, 2017; Fig. 4b). The overlying Quaternary sequences thicken locally near the fault, indicating ongoing fault activity and rotation of the hanging wall during the deposition (Hsiao et al., 1998). Seabed topographic maps reveal numerous northeast–southwest linear features that facilitate the identification of fault locations (Chung, 2021).

Apart from structure 1, structures 2 to 11 exhibit similar geometric properties. Vertical offsets typically decrease from the fault center toward its ends (Cowie and Roberts, 2001), so the offset variation of the PRSB is used to determine fault length. For example, the offset of the PRSB along a fault decreases from ∼1 s in the middle of the fault to 0 s toward the southwest in the results of Yang (2017), allowing us to determine the length of the fault. This pattern also implies that normal faulting initiated in the northeast, with the reverse faults landward remaining unreactivated (Chen, 2014; Atomic Energy Council, 2016; Yang, 2017; Lai and Huang, 2023).

Because of the safety assessments required for nearby nuclear power plants, this area has been extensively studied. The report of the Atomic Energy Council (2016) synthesized previous research, integrating multiple seismic reflection profiles and seabed topography to determine fault locations and activity for structures 1 to 3. Structure 1 in domain A serves as the master fault of a nearshore half‐graben structure, which suggests a potential offshore extension of the on‐land Shanchiao fault (Atomic Energy Council, 2016). For disaster prevention, this study compares the location of structure 1 with the Shanchiao fault based on the TEM database (Shyu et al., 2020) and defines it as an extended segment of the fault.

The locations of these structures are primarily based on the interpretations by Yang (2017, Fig. 4b), with minor adjustments in their locations using topographic maps (Chung, 2021).

The seismogenic depths of structures 1 to 11 are set at 10 km based on Atomic Energy Council (2016). This was constrained by seismic activity and studies of the on‐land Shanchiao fault. For structure 1, we follow the published structural geometry, allowing its dip angle to vary with depth (see Appendix Excel Table). The dip angles for the remaining structures are assumed to be 60°, as is typical for normal faults.

Domain B is located in the southern Okinawa Trough, stretching from west of 124° E to offshore the Lanyang Plain. The Okinawa Trough is a back‐arc extensional basin formed above the Ryukyu subduction system, extending from southern Japan to northeastern Taiwan (Kimura, 1985; Sibuet et al., 1998). The area near Taiwan features numerous submarine volcanoes and hydrothermal vents, indicating ongoing tectonic activity (Lee et al., 1980; Sibuet et al., 1998).

Since 0.1 Ma, the rifting direction in the southern Okinawa Trough has shifted to a nearly north–south orientation (Kimura, 1985; Hsiao et al., 1998; Sibuet et al., 1998; Tsai et al., 2021), resulting in the formation of several east–west‐oriented rift systems. Historical earthquake data revealed a normal‐fault stress regime, with seismicity concentrated in the trough’s center and distributed east–west (Sibuet et al., 1998; Kubo and Fukuyama, 2003; Huang et al., 2012).

Although it has been debated whether the extensional process in the southern Okinawa Trough is symmetrical at crustal scales (e.g., Park et al., 1998; Arai et al., 2017), the seismic reflection profiles show a series of east–west‐striking normal faults dipping toward the spreading center, with no significant differences in structural geometry on either side (Sibuet et al., 1998; Arai et al., 2017; Fig. 4c). In addition, the overall strata appear nearly horizontal, supporting a symmetrical extension model during the recent extensional phase. Offsets of reflectors across boundary structures away from the center are generally larger (Klingelhoefer et al., 2009; Arai et al., 2017; Tsai et al., 2021; Fig. 4c). Pronounced seabed cliffs indicate active boundary faults, when sedimentation rate variations across them indicate long‐term tectonic activity since the late Pleistocene (Fang et al., 2020).

This study focuses on structures west of the Yaeyama Graben for hazard studies and resource development in Taiwan (Fig. 4a). Six seismogenic structures have been identified in domain B: four normal faults in the Yonaguni Rift system (structures 14 to 17, Fig. 4a), and two normal faults on either side of the North Yilan Ridge (structures 12 and 13, Fig. 4a). Although normal faults may also develop on the Ilan Continental Shelf as the trough extends westward (Ku et al., 2009), no seabed lineament was identified that meet the criteria of this study, possibly due to thick terrigenous sedimentation.

These two normal faults are located on either side of the North Yilan Ridge, with the northern fault dipping northwest and the southern fault dipping southeast (Sibuet et al., 1998; Hsu, 1999; Tsai et al., 2021). Seismic reflection profiles and bathymetry maps reveal basement offsets of up to 0.4–1.0 s (∼600–1500 m) and clearly defined seabed lineaments. These structures may be influenced by both the formation of the Okinawa Trough and the orogenic collapse in northern Taiwan (Tsai et al., 2021).

These normal faults are located on the northern and southern sides of the Yonaguni Rift system, identified by the lateral extension of prominent seabed lineaments (Tsai et al., 2021; Fig. 4c).

Located east of the Yonaguni Rift system, these two normal faults dip toward the trough center and are identified by clear seabed lineaments (Arai et al., 2017; Tsai et al., 2021). The western endpoint of structure 17 is differentiated from the adjacent structure 14 by their opposite dip direction.

Seismogenic depths for structures 12 to 17 are set at 15 km based on earthquake distributions (Chen et al., 2022) in the southern Okinawa Trough. Dip angles for these structures are assumed to be 60°, as for normal faults. Structures 14 and 15 dip toward each other and intersect at 11.3 km. Thus, we adjusted their seismogenic depths to this value (Fig. 9, profile A).

Domain C is located on the overriding plate of the Ryukyu subduction system and is characterized by the northwestward subduction of the Philippine Sea Plate. This setting features the accretionary wedge, fore‐arc basins, and the volcanic arc (Lallemand et al., 1999; Fig. 5a). Because of the lack of recent high‐resolution seismic reflection profiles, seabed lineaments were primarily used to interpret seismogenic structures. Four structures were identified in this domain: one on the eastern edge of the Hoping Basin and three within the accretionary wedge.

In the fore‐arc basins south of the Ryukyu arc, the seabed lineaments of small‐scale faults are largely obscured by sedimentation (Dominguez et al., 1998; Font et al., 2001; Hsiung et al., 2017; Deng et al., 2023). Only one structure on the eastern side of the Hoping Basin (structure 18) is observed to cut through the seabed in seismic reflection profiles, which meets our criteria.

Two right‐lateral faults along the northern boundary of the accretionary wedge are defined by distinct seabed lineaments (19 and 20). In addition, several parallel seabed anticlines within the wedge suggest reverse faulting (Dominguez et al., 1998; Lallemand et al., 1999; Hsu et al., 2013; Chen et al., 2017; Fig. 5a), and structure 21 is a particularly large‐scale reverse fault. Other faults in the region are unlikely to generate significant earthquakes due to their limited length and the loose material of the accretionary wedge (Theunissen et al., 2010). As such, these smaller faults are not analyzed further in this study.

This north–south structure is identified by a scarp along the eastern Hoping Basin that is unrelated to submarine channel erosion, and exhibits a flower structure in seismic reflection profiles (Font et al., 2001; Font and Lallemand, 2009; Wang, Hsu, and Yeh, 2019; Fig. 5b,c). Previous studies identified it as a right‐lateral fault (Lallemand et al., 1999; Font et al., 2001; Font and Lallemand, 2009), with a 5.5 km offset of the shelf‐slope boundary. The northwestern extension of this fault aligns with the western end of the Okinawa Trough, suggesting its activity is related to the opening of the trough.

A boundary structure in the area was suggested by gravity anomalies (Doo et al., 2018), and Global Positioning System velocity data indicate right‐lateral motion across this boundary between the Yilan nearshore and Yonaguni Island (Lallemand and Liu, 1998). Although these results represent an integrated pattern over a large spatial scale, such pattern is consistent with the existence of the proposed Suao fault zone in this area.

Two right‐lateral structures near the northern boundary of the accretionary wedge were proposed and defined due to the oblique subduction (Dominguez et al., 1998; Lallemand et al., 1999; Font and Lallemand, 2009). Seismic reflection profiles show flower structures characteristic of strike‐slip motion (Font et al., 2001; Fig. 5b). The northern tip of the Gagua Ridge divides these structures into the West Yaeyama fault zone (19) and the Yaeyama fault zone (20). These faults have clear seabed lineaments, with the eastern side of structure 20 obscured by a submarine landslide (Okamura et al., 2018). As a result, we used the western boundary of the landslide as the minimum eastern extension of the fault.

A primary seabed lineament north of the Gagua Ridge represents the largest splay fault in this region (Lallemand et al., 1999; Hsu et al., 2013; Chen et al., 2017).

The Suao fault zone (18) is assigned a dip angle of 90° as a strike‐slip fault, with a depth reaching 25 km based on earthquake distributions (Wang, Hsu, and Yeh, 2019). The geometries of structures 19, 20, and 21 were constrained by the depth of their intersection with the subduction interface. Wu et al. (2009) suggested a 15° dip for the subduction interface beneath the accretionary wedge (Fig. 9, profile A; Fig. S2a). Structures 19 and 20 have dip angles of 90°, whereas structure 21 has a 30° dip, consistent with mega‐splay fault characteristics (Strasser et al., 2009; Conin et al., 2012). Based on the depths at which these structures intersect with the subduction interface, the seismogenic depths for these structures are estimated at 17 km for structures 19 and 20, and 13 km for structure 21.

Domain D lies atop the Philippine Sea Plate, where topography and geological structures reflect the transition from subduction in the south to collision in the north (Hirtzel et al., 2009; Huang et al., 2018). The offshore Hengchun Ridge extends southward from the Central Range, representing the deformation zone above the Manila subduction zone (Fig. 2). To its east, the North Luzon Trough contains thick sediments in the fore‐arc basin of the Luzon island arc (Fig. 6). Northward, plate convergence deforms these basin sediments, forming the Huatung Ridge, the South Longitudinal Trough, and the Taitung Trough (Huang et al., 1992; Fig. 6b). Farther east lies the Luzon island arc, including the volcanic islands of Lanyu and Lutao. Northward, the region transitions into Taiwan’s orogenic system, characterized by frequent seismic activity and seismogenic structures parallel to the mountain ranges (Deffontaines et al., 2001; Malavieille and Trullenque, 2009; Wang et al., 2011, 2024).

This study integrated various previous viewpoints and identified seven seismogenic structures (Deffontaines et al., 2001; Malavieille et al., 2002; Shyu et al., 2005; Malavieille and Trullenque, 2009; Huang et al., 2018; Fig. 6). Two west‐dipping reverse faults (25 and 26) are located on the eastern side of the Luzon island arc, with another on the eastern side of the Huatung Ridge (27). To the west of these uplifted terrains, only the southeastern South Longitudinal Trough shows evidence of an east‐dipping structure (28). Seismic reflection profiles do not support the presence of other westward thrusts proposed in some earlier studies (e.g., Chen et al., 2019, 2024) at the junction between the western island arc and the Taitung Trough. Off the eastern coast of the Coastal Range, three west‐dipping reverse faults (22, 23, and 24) deform the seabed. Farther east, seismic reflection profiles show no evidence of tectonic activity on buried Huatung Basin structures since at least 1.5 Ma (Hsiung et al., 2017; Hsu et al., 2021).

Seismic reflection profiles reveal two west‐dipping thrust faults deforming the strata off Taiwan’s eastern coast (Malavieille et al., 2002; Hsieh et al., 2020). The northern part of these structures may have subducted beneath the Eurasian Plate; thus, we drew their ends at the trench. The Chimei Canyon thrust (23) features more prominent fault scarps than the Takangkou High thrust (22), suggesting higher activity. Historical earthquake hypocenters align with these west‐dipping fault planes (Huang and Wang, 2022). The $ML$ 7.2 Hualien earthquake on 3 April 2024 occurred near this structural zone and may have been produced by one of these two structures.

This southeastward‐thrusting fault cuts through the seabed with a topographic scarp (Malavieille et al., 2002; Lin, 2010; Fig. 6d). Possible left‐lateral motion on this structure during the northwestward collision of the Luzon arc with Taiwan is consistent with paleomagnetic records that indicate clockwise rotations (Lee et al., 1991). However, left‐lateral motion may have ceased around 1.2 Ma (Chen et al., 2015), indicating that its current activity is primarily thrust faulting.

Seismic reflection profiles show eastward‐thrusting faults on the eastern side of the Luzon island arc (Lin, 2010; 25 and 26). Strata east of the island arc are deformed by folding, with long‐term uplift of the hanging‐wall block illustrated by marine terraces on the Lutao and Lanyu islands (Inoue et al., 2011). Structures 23, 25, and 26 may belong to the same structural system; however, this study distinguishes them based on the differences in their surface features and the trends of their fault traces.

Seismic reflection profiles indicate a major thrust on the eastern side of the Huatung Ridge, cutting through and deforming the seabed (Malavieille et al., 2002; McIntosh et al., 2005; Hirtzel et al., 2009; Malavieille and Trullenque, 2009; Chen et al., 2015; Huang, 2018). Growth strata on the eastern flank of the South Longitudinal Trough indicate continued movement of this structure (Huang, 2018). Some minor seabed lineaments atop the ridge represent small‐scale fault systems (Malavieille et al., 2002), but these were not included in our database due to their limited length.

Seismic reflection profiles from the west of the southern Huatung Ridge reveal an anticlinal structure with a small‐scale, eastward‐dipping reverse fault (Malavieille et al., 2002; Malavieille and Trullenque, 2009; Huang, 2018; Fig. 6e). Shyu et al. (2005) mapped a structure there based on submarine topography, with a distinct straight scarp. The lack of evidence for channel erosion suggests it may be a structural scarp (about 90–150 m high from ArcGIS analysis, Fig. S4).

For structures 22 and 23, two geometric interpretations are considered. Hsieh et al. (2020) proposed a 25° fault dip transitioning to nearly horizontal at the depth of 6.0–6.1 km, but Malavieille et al. (2002) suggested a continuous 25° dip until they intersect with the on‐land Longitudinal Valley fault system (Fig. 9, profile B). The coseismic coastal uplift observed during the $ML$ 7.2 earthquake near Hualien on 3 April 2024 and its deeper focal depth support the latter model. Thus, we include both geometric assumptions but prefer the second interpretation.

Structure 24 shows reverse faulting with potential horizontal components. The fault dip angle thus ranges from 30° for reverse faults to 60° for faults with both reverse and strike‐slip characteristics. The seismogenic depth is set at 14 km, where it intersects with the Longitudinal Valley fault (Fig. S3). Structures 25 and 26 have dip angles of 30°, with seismogenic depths set at 25 km (Chen et al., 2016, 2018, 2019). Structures 27 and 28 are also assigned 30° dip angles, with depths of 20 and 10 km, respectively, based on their relationships with the Manila subduction system (Fig. 9, profile C).

Domain E is located on the upper plate of the Luzon subduction system, where the processes of subduction and the transition into Taiwan’s incipient collision stage dominate the tectonic framework (Shyu et al., 2005; Fig. 1). Because the Eurasian Plate subducts eastward, deformed sediments accumulate, forming north–south‐trending ridges and canyons.

Some studies related the area’s topographic highs to late Pliocene to recent mud diapirism (e.g., Sun and Liu, 1993). However, high‐gravity values suggest that compressional stress, rather than buoyant mud diapirism, plays a key role in the formation of topographic highs (Doo et al., 2015). Seismic reflection profiles show that asymmetric landforms are more likely associated with reverse faults that deform the seabed (Liu et al., 1997, 2004; Lin et al., 2008, 2022; Hsiung et al., 2018; Dirgantara et al., 2020). Moreover, recent studies indicate that some chaotic seismic facies previously attributed to diapirism may also contain stratified signals (Ko, 2022). As a result, we attribute the topographic highs in this area mainly to reverse faulting.

We also identified the Fangliao Canyon structure zone (49) and the Manila Upper Plate splay structure (50) in the eastern part of this domain.

The Small Liuchiu is the only island above sea level in this region. The East Liuchiu Island structure (47), located east of the island, accounts for the uplift that forms the marine terraces on the island (Wang et al., 2022). Other structures, identified by Lin et al. (2008), lie west of each anticlinal axis and are roughly parallel to the Manila Trench. Seismic reflection profiles show that these east‐dipping reverse faults are spaced 5–10 km apart, and that their dip angles gradually decrease toward the west. Their distribution pattern follows the typical westward propagation of thrust faults in western Taiwan’s fold‐and‐thrust belt. Although this region experiences fewer instrumental earthquakes than other parts of Taiwan (Wang et al., 2011), seismic reflection profiles show these structures deformed seabed sediments with growth strata in their footwalls, indicating their ongoing activity (Chiang et al., 2004; Lin et al., 2008).

Seismic reflection profiles reveal two structures in the submarine Fangliao Canyon (Deffontaines et al., 2016), though canyon erosion partially obscures the structural scarp. This study combined the two structures into a single fault zone located at the western edge of the west‐facing scarp. Previous studies suggest both thrust and strike‐slip components for this fault (Deffontaines et al., 2016). We interpret it as a left‐lateral fault with a thrust component, similar to the nearby Chaochou fault on land (Shyu et al., 2020).

Seismic reflection profiles show several thrust faults beneath the Hengchun Ridge off southern Taiwan (Lin et al., 2009; Zhang et al., 2022). Structure 50, a major westward‐thrusting fault, is characterized by topographic features and their lateral continuity. This structure likely branches out from the subduction system and divides the accretionary wedge into a lower slope region with fold‐and‐thrust belts and a highly deformed upper slope region.

Seabed lineaments and seismic reflection profiles suggest this fault terminates near northern Luzon Island (Fig. S5; Liu et al., 2006; Chen, 2015), where the interaction of subducting seamounts with the Manila Trench changes the geometry of the overlying accretionary complex (Zhu et al., 2013). Other westward‐thrusting structures between this fault and the trench (Lin et al., 2009) were excluded from this study due to their shorter length and long distance from Taiwan.

Structure 47 off Small Liuchiu has a fault dip ranging from 60° at shallow depths to horizontal at around 4 km based on previous results (Wang et al., 2022). For other structures, seismogenic depths are estimated based on the fault dip angles and the depth of the subduction interface (Tan, 2020; Tan et al., 2022; Fig. 9, profile C; Figs. S2b and S6). Assuming the structural evolution follows the east‐to‐west pattern as in the on‐land fold‐and‐thrust belt, the fault dip angles are set to decrease westward (Lin et al., 2008, 2022), ranging from 30° to 40° (Fig. S6; Appendix Excel Table).

The Fangliao Canyon structure zone (49) is suggested as a high‐angle strike‐slip fault (Deffontaines et al., 2016). Shyu et al. (2016) provided geometric data for the on‐land Chaochou fault, with dip angles of 70°–80°. Hence, this study assumes a dip range of 70°–90° and a seismogenic depth of 9 km (Fig. 9, profile C). For the Manila Upper Plate splay structure (50), the fault dip is assigned at 40° and a seismogenic depth of 15 km, where it intersects with the subducting plate.

Domain F is part of the Eurasian Plate, extending from the shallow Taiwan Strait, with an average water depth of 60 m, to the deeper South China Sea Basin, where depths reach 3000 m. Before the Miocene, this region experienced extensional tectonics, forming northeast–southwest‐striking normal faults and related grabens (Lin and Watts, 2002; Teng and Lin, 2004; Zhang et al., 2020; Fig. S5). Since the Miocene, the collision between the Luzon arc and the continental margin resulted in the development of a foreland basin on the eastern side of the Taiwan Strait. Although many normal faults that formed earlier have become inactive, some on the eastern side were reactivated as reverse faults (Yu and Chou, 2001; Lin and Watts, 2002; Lacombe et al., 2003; Lin et al., 2003; Yang et al., 2016, 2021).

In domain F, we mapped four seismogenic structures. Among them, two structures are related to the deformation front in western Taiwan that extends offshore in the vicinity of Hsinchu and Miaoli (Yu, 2004), where the stress conditions are favorable for the development of reverse faults (Lacombe et al., 2003; Chiang, 2016; Yang et al., 2016). Data from older oil and gas prospecting and recent offshore wind energy project assessments have enabled us to map structures 52 and 53 in this area. West of the deformation front, the fold‐and‐thrust belt is absent, and the seafloor topography is primarily shaped by terrigenous sedimentation and ocean currents (Lin and Watts, 2002; Yu, 2004; Yang, 2010; Chang et al., 2015). Only the Binhai fault zone (54) cuts through the seabed, as evidenced by seismic reflection profiles (Lin and Watts, 2002; Lin et al., 2003; Teng and Lin, 2004; Chang et al., 2015; Zhang et al., 2020; Lin, Yang, et al., 2021).

The Tainan Basin in the south of the strait has not yet entered the compressional stress field (Huang et al., 2004). The continental margin features east‐northeast–west‐southwest‐striking normal faults formed by earlier extensional stress (Lin et al., 2005; Ding et al., 2008; Yang et al., 2016). Among these, one structure (51) shows significant seafloor lineaments (Liu et al., 1997, 2004; Lin et al., 2008; Fig. 7a).

Farther south, there is no clear evidence of active structures in the South China Sea Continental Slope, except for some local faults that do not seem to cut the seabed (e.g., Han et al., 2019, 2021; Fig. 7a). The slope transitions into the South China Sea Basin, which remains unaffected by compression with older structures covered by at least 2 km of sediments (Lin and Watts, 2002; Li et al., 2007; McIntosh et al., 2013; Han et al., 2019; Fig. 7a).

Seismic activity in domain F is relatively low compared to other regions around Taiwan. Notable events include the 1604 Quanzhou earthquake (⁠$Mw$ 7.5–8.1, Lee et al., 1976; Yen et al., 2021), likely related to the Binhai fault zone (54). Other earthquakes, such as those in 1994 (⁠$Mw$ 6.4–6.7; Kao and Wu, 1996; Huang et al., 1999; Fig. S5) and 2018 (⁠$Mw$ 5.7–6.2; Wang, Ye, et al., 2019; Wang et al., 2021; Tang et al., 2023; Fig. S5), had minimal impact onshore Taiwan due to their locations.

Domain F lacks detailed survey data compared to other offshore regions. However, new data are becoming available from recent offshore wind farm surveys led by the Geological Survey and the Mining Management Agency. Because more high‐resolution data becomes available, they will improve the identification of potential seismogenic structures off Taiwan’s western coast.

This structure cuts through the seabed, forming a clear lineament near the continental margin that corresponds to the central segment of the previously mapped Yichu fault (Liu et al., 1997, 2004; Lin et al., 2003, 2008). This fault may also be associated with seabed sediment slumping (Ding et al., 2008).

Seismic reflection profiles show that folded strata east of this structure are slightly elevated than those to the west (Chiang, 2016; Wu, 2017; Chen, 2019; Fig. 8d). This structure aligns with the onshore Hsinchu structure (6) described by Shyu et al. (2020), suggesting a combined length over 20 km, meeting the criteria of this study (Lacombe et al., 2003; Chen, 2019).

Seismic reflection profiles reveal that this fault cuts other structures, indicating that it is a relatively young structure in the area. The location of this structure is defined by the western edge of the anticline that deforms the Toukoshan Formation, the youngest late Pleistocene strata on land in western Taiwan (Huang et al., 2020; Fig. 8e).

The Binhai fault zone is located on the western side of the Taiwan Strait and is likely related to multiple earthquake events (Wen and Xu, 2005; Yen et al., 2021). Seismic reflection profiles and earthquake data reveal a 30–60 km wide structural zone cutting through the seabed, forming a right‐lateral flower structure with branches that connect at depth (Zhang et al., 2020; Fig. 8c).

Gravity, magnetic, and ambient noise tomography surveys have identified a primary structure along the proposed Binhai fault zone, marked by distinct geophysical anomalies parallel to the orientation of the Taiwan Strait (Huang and Wang, 2006; Zhang et al., 2018). Because of the lack of clear fault trace evidence, the fault’s approximate location is mapped as a dashed line, with a total length of about 500 km. Seismic reflection profiles show variations in basement depth, leading to the division of the fault into the northern Binhai fault zone (54A) and the southern Binhai fault zone (54B).

Besides strike‐slip motion, the fault likely also exhibits a reverse component, with the northwestern side as the hanging‐wall block, as evidenced by uplifted coastal features along the southeast Chinese coast (Zhang et al., 1982; Huang and Wang, 2006). Thinner Quaternary sediments on the fault’s western side also indicate ongoing relative uplift since the Quaternary (Huang and Wang, 2006).

The Tainan Basin structure (51) has a dip angle of 60° for the shallow part (0.2–4.5 km) and 20° for the deeper part (4.5–7.5 km) (Lin et al., 2008; Fig. S7, Appendix Excel Table). The Outer Hsinchu structure (52) is suggested as a reverse fault that dips 45° to the south, following the on‐land Hsinchu structure’s geometry (Shyu et al., 2020). The Toufen structure (53) is suggested as a reverse fault that dips 40° to the southeast, based on cross‐section interpretations (Huang et al., 2020).

The Binhai fault zone (54) has a seismogenic depth of 20 km, derived from earthquake depths and ambient noise tomography results (Zhang et al., 2020, 2018). The fault is divided into two parts (Fig. S8): a deep basement part with a 60° dip and a sediment layer part with dips between 50° and 90°. Seismic reflection profiles suggest sediment thicknesses of 1.5 and 2 km for structures 54A and 54B, respectively (Zhang et al., 2020).

The accuracy of our seismogenic structure identification is primarily determined by the availability and quality of seismic reflection profiles offshore. In this study, we have integrated the literature about offshore Taiwan as comprehensively as possible (Fig. S1). However, there are still many regions with only limited seismic reflection profile data, such as the East China Sea Continental Shelf, the Ryukyu arc, the Huatung Basin, and some parts of Taiwan’s coast. In these areas, it is difficult to conduct detailed structural mapping.

Another challenge comes from the resolution of the available DEMs. Because the grid resolution of the DEMs is between 200 and 450 m, it would be difficult to map the location of the offshore structural scarps with a higher accuracy than these numbers. To enhance the identification of offshore seismogenic structures in the future, it is necessary to obtain higher‐resolution bathymetry maps and more seismic survey data.

Seafloor erosion and sedimentation can obscure topographic features formed by tectonic activities. Indeed, most structures identified in this study are not parallel to offshore channels, where erosion rates are generally high. For instance, the Chimei Canyon thrust (23) was traced via scarps outside the canyon. In domain E, some structures are parallel to submarine channels but are located in the footwall of reverse faults, making them less likely to be affected by erosion. Therefore, it is possible that some structures were not identified in this study due to their topographic features being eroded.

Sedimentation rates around Taiwan vary based on the proximity to sediment sources and ocean currents (e.g., Liu et al., 2008; Huh et al., 2011; Hsiung et al., 2017). Although the slip rates of offshore structures remain highly uncertain, if we assume that offshore tectonic activity rates are comparable to those on land (∼0.5–10 mm/yr; Shyu et al., 2020), such rates would largely exceed the reported offshore sedimentation rates of 0.01–0.5 mm/yr (e.g., Hung and Chung, 1994; Chung and Chang, 1995; Dirgantara et al., 2020; Hsu et al., 2021). Therefore, we suggest that our identification of structures using topographic features is less likely affected by seafloor sedimentation.

Identifying strike‐slip components in offshore faults using seismic reflection profiles is more challenging. Some faults may also exhibit horizontal movement, particularly when their orientations are oblique to the compressional stress field. In domain D, for example, we infer that some structures may have a left‐lateral component due to the oblique collision that forms the Taiwan orogen, and this is also consistent with stress patterns inferred from historical seismicity (e.g., Fuh and Liu, 1998; Kao et al., 1998; Wu et al., 2010; Hsu et al., 2016; Hui et al., 2018; Wang et al., 2024).

Although we have collected and integrated all available data to identify and characterize offshore structures in this study, the information on these structures is by no means complete. Further studies of these structures are essential for improving the structural locations and parameters. For example, estimation of fault‐slip rates and earthquake recurrence intervals would enable more comprehensive seismic hazard assessments, albeit this would require much more data and assumptions than those currently available. To accurately calculate the earthquake recurrence intervals for each fault, their long‐term slip rates need to be determined, a key objective for future offshore fault studies. For instance, direct age data of the offset seabed sediments would provide much better constraints for the long‐term slip rates of the structures.

In this study, we utilized three empirical regression equations to estimate earthquake magnitudes for rupture events, based on the subsurface geometries of reconstructed structures (Table 1, Fig. 10). Structures 1, 52, and 53 were identified as possible offshore extensions of on‐land faults, so magnitudes for scenarios involving simultaneous rupture of both on‐land and offshore segments were also calculated.

A comparison of results from the three equations revealed that most structures yield similar magnitude ranges. Discrepancies were likely attributed to geometric uncertainties in this study (seismogenic depth ±2 km, fault dip ±10°), resulting in magnitude uncertainties of ±0.11 when using Wells and Coppersmith (1994) and ±0.13 when using Yen and Ma (2011). For larger reverse fault earthquakes (e.g., structures 23, 26, 27, and 50), calculations using Wells and Coppersmith (1994) yielded lower magnitudes compared to those using Yen and Ma (2011) and Wesnousky (2008). This is likely due to differences in the regression datasets used in these studies. For such cases, we recommend using the results from Yen and Ma (2011) and Wesnousky (2008) because they have included more complete datasets in their regressions.

In domains C and E, we found that some structures may intersect with other structures at depth. To prevent this, we have set the seismogenic depths of these structures shallower, thus their widths become narrower. In these cases, length‐based magnitude estimates by Wesnousky (2008) sometimes exceed those based on fault area. Therefore, we recommend the use of Wells and Coppersmith (1994) and Yen and Ma (2011) for structures with shallow seismogenic depths.

Because of the limitation of data currently available, we modeled the fault geometry as a simple planar surface for most of the structures. Future studies may identify changes in fault dip with depth, and this would increase the estimated fault area. For example, the simplified geometry of Outer Chinshan structure 1 yielded a fault area of $261km2$⁠, corresponding to a potential earthquake magnitude of 6.4. However, data from the Atomic Energy Council (2016) suggest a dip angle variation for this structure, and the fault area could increase to $926km2$ considering such variation, raising the estimated magnitude to 7.0 (Appendix Excel Table). Further offshore investigations are needed to refine these fault geometries and to improve the magnitude estimates.

It is possible that the empirical regression relationships for the geometries of offshore faults and their associated seismic events are different from on‐land structures. However, because most previous studies have focused on on‐land events, data for offshore structures are scarce. Because more offshore earthquake data will become available in the future, it would be important to develop regression models that are more appropriate for offshore regions.

Earthquake magnitudes were estimated in this study under the assumption that the entire fault would rupture simultaneously. However, it is recognized that longer structures may rupture in segments (e.g., structure 54), as evidenced by historical seismic events (Wen and Xu, 2005). For example, the 1604 Quanzhou earthquake was attributed to the rupture of the southern segment of structure 54A (Huang and Wang, 2006; Yen et al., 2021; Fig. S5). Unfortunately, it is difficult to propose reasonable segmentation scenarios for most structures included in this study, due to the limitation of currently available data.

In addition to segmented ruptures, scenarios involving multifault ruptures were also evaluated because these can generate larger earthquakes than those occurred on a single fault, and are critical for seismic hazard assessments (e.g., Gómez‐Novell et al., 2020; Chang et al., 2023). Potential multifault rupture scenarios between offshore and on‐land structures were identified based on fault types, geometries, and the spacing of two faults that is less than 6 km (Cheng, 2023; Table 2). For instance, in domain A, many normal faults exhibit a lateral continuity and have the potential to be connected laterally or at depth (Fossen and Rotevatn, 2016; Chang et al., 2021).

In domain D, several closely spaced reverse faults on the eastern side of the island arc appear to display lateral continuity (structures 23, 25, and 26). In domain E, some structures may be offshore extensions from on‐land faults (e.g., Chiang et al., 2004; Wu et al., 2024). In addition, many faults in domain E may connect at depth to a common detachment (Fig. 9, profile C). Multifault ruptures involving the underlying plate boundary fault were also considered in this region.

The magnitudes of earthquakes resulting from these multifault rupture scenarios were estimated (Table 2). Calculations were performed using the method by Yen and Ma (2011), with the seismic moment (⁠$M0$⁠) determined with the formula $Mw=2/3logM0−10.73$⁠. The combined seismic moment from both offshore and on‐land faults, as estimated by Shyu et al. (2020), was used to calculate the total magnitude. If we consider multifault rupture scenarios, all 54 offshore structures may rupture longer than 20 km and have the potential to generate earthquakes of magnitude 6.5 or greater (Fig. 10b; Tables 1 and 2).

An understanding of historical earthquakes that may have occurred on the offshore seismogenic structures can provide additional constraints for these structures, as well as information needed for time‐dependent seismic hazard assessments (Chan et al., 2020). Among the 54 offshore seismogenic structures identified in this study, only the Binhai fault zone (54A) has been discussed as the likely source of the 1604 Quanzhou earthquake in previous studies (e.g., Lee et al., 1976; Yen et al., 2021). No other earthquakes have been proposed to have occurred on the other offshore structures.

We searched historical and instrumental earthquake events in an integrated catalog covering the time period from 1900 to 2023 to explore potential correlations between major earthquakes and the offshore faults identified in this study. This catalog combined earthquake data from multiple sources and agencies (see details in Data and Resources). In this catalog, there are 12 earthquakes with magnitudes of 7.0 or greater located offshore Taiwan (Fig. S5). We compared the location, magnitude, and focal mechanism of these events with the offshore structures and their parameters identified and proposed in this study. For the earthquake locations, we used the standard location error proposed by Wu et al. (2013), which is ∼6 km in longitude, 3 km in latitude, and 10 km in depth.

Our results suggest that only two earthquake events may be related to the identified offshore structures: the 1915 M 7.2 earthquake in the hanging wall of the Ryukyu subduction system (possibly related to structure 20, or a subduction zone event) and the 1925 M 7.1 earthquake southeast of Taiwan (potentially linked to structure 27). Because these two events occurred long ago, detailed information such as their focal mechanisms is lacking, making it not feasible to further verify their seismogenic faults. The recorded magnitudes of these events are slightly lower than our estimates, possibly indicating partial ruptures of the structures. In short, only structures 20, 27, and 54 may have possible historical earthquake records, and we have included this information in the “Last event” column of the Appendix Excel Table.

Our comparison result shows that most recorded seismic events do not correlate with the structures we identified in this study. Instead, they are more likely associated with the subduction zone or deeper, subsurface seismic sources (Fig. S9). This is expected because most of the structures identified in this study may have not produced any major earthquake over the past few decades, similar to the situation for on‐land seismogenic structures. This observation illustrates that it is not appropriate to rely solely on the instrumentally recorded seismic data to identify active structures because this may largely overlook the structures that have not ruptured in the time period of the instrumental record and are still accumulating strain.

This study presents the first comprehensive offshore seismogenic structure database of Taiwan, which includes detailed information on structural locations, fault types, subsurface geometries, and potential earthquake magnitudes (Table 1, Appendix Excel Table). We envision that this database to be applied to studies of future earthquake disaster prevention, offshore infrastructure planning, resource development, and general tectonics of Taiwan.

In the event of an offshore earthquake, this database may provide some initial insights into the focal mechanism and seismogenic structure of the event, enabling rapid preliminary risk assessments. For example, all the identified structures have the potential to generate earthquakes of magnitude 6.5 or greater, which may result in the rupture of the seabed (Wu and Hu, 2024). Based on the earthquake magnitudes and structural geometries, some structures may have average displacements exceeding 3 m in an earthquake (Fig. 10a), enough to trigger tsunamis if the seabed is ruptured. Furthermore, some offshore earthquakes can cause ground shaking of intensity 6 or higher on land if they occur on structures closer to the coastline, as illustrated in historical records.

Most structures capable of generating major earthquakes greater than magnitude 7.5 are located offshore eastern Taiwan (Fig. 10). Assuming their long‐term slip rates are similar to their on‐land counterparts in eastern Taiwan (∼7–15 mm/yr; Shyu et al., 2020) and an ∼3 m slip of the fault in a magnitude 7.5 event (Table 1), the earthquake recurrence interval can be estimated in the range of 200–450 yr. Such a short interval suggests that further studies and enhanced preparedness measures are necessary for eastern Taiwan.

The complex tectonic processes of the Taiwan orogenic belt result in diverse stress fields and structural patterns in offshore regions, providing a valuable reference for similar geologic settings worldwide. For example, the structures in domain C are influenced by the oblique subduction of the Philippine Sea Plate, with both reverse faults and large right‐lateral faults developed through strain partitioning in the overriding accretionary wedge (Chen et al., 2022). This pattern resembles the prominent seismogenic right‐lateral faults observed in the Sumatra subduction system (e.g., McCaffrey, 2009; Sørensen and Atakan, 2008). In contrast, the Manila subduction system in domain D is characterized by a relatively perpendicular subduction direction and exhibits a fold‐and‐thrust belt that propagates upward from a basal detachment. This is similar to the Japan Trench subduction system, where the detachment fault is a primary seismogenic structure (e.g., Nakata et al., 2012).

Information on offshore seismogenic structures is crucial for the planning and design of offshore and coastal projects, including nuclear power plants, wind farms, submarine cables, and carbon dioxide storage facilities. Because offshore wind energy and carbon dioxide storage have become more important in Taiwan, we hope this offshore seismogenic structure database will provide important constraints for these projects. Furthermore, we also hope our results and findings will point to the directions of future offshore studies in Taiwan, to gather more high‐resolution geological and geophysical information to better understand the offshore structures of this active orogenic belt.

In this study, we collected and analyzed all available geological and geophysical data to identify and map offshore seismogenic structures. A total of 54 potential offshore seismogenic structures were identified. We also calculated structural parameters for these structures, including their fault type, geometry, and possible earthquake magnitude. Most of these structures are capable of generating earthquakes exceeding magnitude 6.5. In cases of multifault ruptures, some events could reach magnitudes above 8.0. As the very first offshore seismogenic structure database of Taiwan, we hope our results will provide additional constraints for the earthquake hazard assessment studies of the island.

All the seismic reflection profiles used in this study are from published sources listed in the references (Fig. S1). Digital elevation models (DEMs) for the offshore region surrounding Taiwan (119°–123° E, 21°–26° N) with a resolution of 200 m were obtained from the National Science and Technology Council’s Oceanography Database. DEMs beyond this range, with ∼450 m resolution, were acquired from the GEBCO website (https://download.gebco.net/, last accessed September 2021). Historical earthquake data were compiled from Broadband Array in Taiwan for Seismology (BATS), Central Weather Administration (CWA), Global Centroid Moment Tensor (Global CMT), USGS (U.S. Geological Survey), National Research Institute for Earth Science and Disaster Resilience (NIED), Japan Meteorological Agency (JMA), International Seismological Centre (ISC), and China Earthquake Networks Center (CENC) databases, covering the period from 1900 to 2023 (Taiwan Earthquake Model group, Gao et al., in preparation). The supplemental material is provided as a separate document, including a detailed example of parameter calculations (Text S1), Tables S1 and S2, and Figures S1–S9, which contain supporting data for determining the locations and geometries of the structures. In addition, an independent excel table is included, presenting the complete parameter values for 54 seismogenic structures. This table includes information such as the average strike and dip direction of the structures, the upthrown side, seafloor depth, dip angles, seismogenic depth, and magnitudes estimated using three different equations.

The authors acknowledge that there are no conflicts of interest recorded.

The authors sincerely thank C.‐H. Chan, J.‐H. Chang, L. Chen, W.‐S. Chen, R. Y. Chuang, J.‐C. Gao, W.‐C. Han, Y.‐H. Hsieh, Y.‐J. Hsu, J.‐C. Lee, L.‐F. Lin, K.‐F. Ma, H.‐Y. Tang, L. S. Teng, Y. Wang, E. Y. Yang, and E.-C. Yeh for their valuable comments and insightful discussions. Comments and suggestions from Handling Editor T.-L. Tseng and three anonymous reviewers have significantly improved this article. This research is supported by the National Science and Technology Council (NSTC) of Taiwan (Project 110‐2124-M-002-008, 111‐2124-M-002-008, 112‐2124-M-865-001, and 113‐2124-M-865-001 as the TEM project, and 110‐2116-M-002-019, 111‐2116-M-002-044, 112‐2116-M-002-030, and 113‐2116-M-002-025-MY2 to J.B.H.S.).