1 Introduction

On September 17–18, 2022, the eastern coast of Taiwan was rocked by the Taitung earthquakes, registering magnitudes the highest magnitude of 6.8 ML (cwb.gov.tw 2022). This seismic event resulted in considerable damage to structures and infrastructure, particularly impacting the Yuli area in eastern Taiwan. With seismic intensity levels reaching 6 + and a peak ground acceleration of 607 gal, widespread destruction ensued, including the collapse of buildings and bridges (Chou et al. 2022). The first earthquake, with a magnitude of 6.5, was characterized by a strike-slip fault. Conversely, the subsequent 6.9 magnitude earthquake was identified as a reverse slip fault, believed to be westward dipping and potentially associated with the Central Range (Chou et al. 2022).

Positioned at the convergent boundary between the Philippine Sea Plate and the Eurasian Plate, Taiwan is a hotbed for tectonic and seismic activities (Barrier and Angelier 1986; Yu et al. 1997; Yu and Kuo 2001; Bird 2003). The resultant crustal deformation triggers frequent earthquakes, emphasizing the need to comprehend the geological processes fueling such seismic events for proficient disaster management and earthquake hazard assessment. The 2022 Taitung earthquakes manifested a lack of understanding about the interaction between these large-magnitude earthquakes and the surface ruptures in this geologically active region.

In several notable instances, past research has successfully utilized GNSS network measurements to investigate coseismic deformation associated with earthquakes, yielding significant insights as documented in studies by Yu and Kuo (2001), Ge et al. (2008), Ohta et al. (2012), Xu et al. (2010), and Wu et al. (2019). This paper seeks to extend this approach, utilizing the Real-Time Kinematic Global Navigation Satellite System (RTK GNSS) measurements to scrutinize the impact of the Taitung earthquakes on the Yuli area and to characterize the spatial distribution and behavior of the resultant ruptures. The seismic complexity of these events was explored by Yagi et al. (2023), who unearthed an irregular rupture process, implying a heterogeneity of stress and structure within the source region. While some remote sensing approaches are often employed to analyze surface ruptures and the extent of damage, studies by Valkaniotis (2022) divided the northing displacement of the eastern part and south direction displacement of the western part through optical satellite images, Lin (2022) demonstrated the affected area distribution through the differential synthetic aperture radar (DinSAR), Tang et al. (2023) explore the interplay between faults during Taiwan’s 2022 Chihshang earthquake sequence, revealing how the Central Range fault's seismic shifts affect stress patterns and slipping on the Longitudinal Valley fault through enhanced by satellite images and continuous GNSS data, sheds light on the dynamic responses within fault systems, underscored by the 2022 Taitung earthquake's activity. Lee et al. (2023) utilized the real-time moment tensor (RMT) collaborated with continuous GNSS during the earthquake events and showed the distribution of the affected areas with the slip direction generated during the earthquake events. While remote sensing has successfully mapped the distribution and rupture lines, the need for ground truthing to refine accuracy arises from limitations in satellite image resolution, ground conditions, and technical constraints. Concurrently, although continuous GNSS data has provided precise displacement measurements, the network's sparsity calls for denser coverage to improve the comprehensiveness of the findings. To bridge this gap, our study incorporates a network of the RTK GNSS to survey a more substantial and dense number of stations, somehow continuous GNSS has better precise (mm precision), but for the coseismic measurement that has > 10 cm movement, the RTK GNSS is the most suitable to be performed for the dense number of stations with accuracy 1–10 cm (Saghravani et al. 2009). Through the RTK GNSS measurement, we aim to offer a more granular understanding of the deformation patterns associated with the Taitung earthquakes.

By investigating the spatial distribution of the affected areas, this study aims to illuminate the underlying tectonic forces and elucidate the relationship between the ruptures and the local geological structures. Previous research into Taiwan's tectonics has yielded valuable insights into the region's intricate fault systems (Shyu et al. 2006, 2007; Yen et al. 2008, 2011; Murase et al. 2013; Huang et al. 2019; Lai et al. 2022; Yagi et al. 2023; Lee et al. 2023; Tang et al. 2023). Building upon these findings, this study seeks to delve deeper into the 0918 Taitung earthquake, examining its implications on the geological setting of the Yuli area. Specifically, the study analyzes the coseismic displacements of the Central Range block, Coastal Range block, and Longitudinal Valley block in relation to the earthquake. The findings have the potential to enhance our knowledge of seismic activities in the region, contributing to hazard assessment, disaster preparedness, and future mitigation strategies. Moreover, the research outcomes can provide valuable insights into the complex dynamics of fault systems, aiding in the advancement of seismology and tectonic studies not only in Taiwan but also in other seismically active regions worldwide.

1.1 Tectonic settings and geological backgrounds

Taiwan is a young and intricate geodynamic body situated between the Chinese part of the Eurasian continent and the northwestern part of the Philippine Sea, covering an area of 36,000 km2. As a result of compression and shearing, Taiwan has a maximum elevation of nearly 4 km, making it the highest fold mountain range on the northwest coast of the Pacific Ocean. The island has an arcuate shape, with its shorter arm extending to the Ryukyu islands in the east and its long arm stretching southward to the Philippines. The Central Range is the backbone of this mountainous island, which the Western Foothill Zone flanks to the west and the Coastal Range to the east. The Longitudinal Valley separates the Central and Coastal Ranges (Biq et al. 1985; Chemenda et al. 2000). In the neotectonics setting, the study area located in the Eastern Taiwan domain, located between the metamorphic rocks of the Central Range and the volcanic rocks of the Coastal Range, is characterized by a long, narrow valley and a steep offshore escarpment (Shyu et al. 2005). The predominant neotectonic component is the east-dipping Longitudinal Valley fault (LVF), known for its high rates of sinistral reverse motion (Lee et al. 1998; Hu et al. 2001). There's evidence of slip partitioning in the southern section of the Taitung Domain (Angelier et al. 1997). The fault, confirmed to be active and slipping rapidly through geodetic measurements, caused two significant earthquakes in 1951 (Cheng et al. 1996). East-facing scarps on the seafloor east of the Longitudinal Valley indicate active thrust faulting (Hsu 1962; Hsieh et al. 2020). Despite suggestions that the LVF is a back thrust rising from the principal fault of the Coastal Range block, its high rate of slip suggests that it might be the primary fault (Carena et al. 2002). Lastly, evidence indicates the possible existence of a west-dipping reverse fault beneath the eastern flank of the Central Range, which has been named the CRF. This fault has potentially contributed to significant uplift and the distinct characteristics observed in the region (Yu and Kuo 2001) that have a role in the tectonic implication (Liu et al. 1998).

Shyu et al. (2006) have calculated the minimum millennial rate of dip-slip of the LVF using the ages of the Hsiukuluan River terraces. With this information, they can start to evaluate preliminary earthquake recurrence intervals of the fault, given their limited knowledge of historical earthquakes along the fault. The most recent earthquake with surface ruptures occurred on the LVF in 1951, with a maximum coseismic offset of up to 2.1 m and an average coseismic uplift of approximately 0.5–1 m (Shyu et al. 2007). The presence of rapid aseismic creep in the southern part of the LVF near Chihshang is well-documented (Angelier et al. 1997; Lee et al. 2001; Lee et al. 2003). However, current data are insufficient to determine if significant aseismic creep occurs around the Hsiukuluan Canyon (Yu and Liu 1989). Shyu et al. (2006). From the river terraces, the long-term uplift rate near the fault is about 20 mm/yr, if the 0.5–1 m event in 1951 is a characteristic event on the fault, the average recurrence interval would be around 50 years. The authors note that the segment of the LVF near Hsiukuluan Canyon did not rupture during the 1951 earthquakes, and it may behave differently, perhaps because it is actually a separate structure (Shyu et al. 2006).

Shyu et al. (2007) stated that the 1951 earthquakes in the Longitudinal Valley, eastern Taiwan were the most destructive in recorded history, but information on their source parameters is limited. To better comprehend the relationship between the valley's neotectonic architecture and the ruptures, a thorough re-evaluation was undertaken, referencing the study by Cheng et al. (1996). The revised rupture map divides the surface ruptures into the Chihshang, Yuli, and Rueisuei sections. The Chihshang and Rueisuei ruptures are along segments of the LVF, while the Yuli rupture is part of a separate fault. The complex behavior of these structures and their interaction are essential in understanding the future seismic hazard of the area (Shyu et al. 2007). Yen et al. (2008) investigated Eastern Taiwan excavations to show LVF ruptured at least three times with magnitudes of 7.0–7.2 up to 390 years prior to the 1951 earthquake. Two pre-1951 earthquakes occurred between 1564–1680 and 1736–1898. The average time between earthquakes is about 150 years, with a minimum uplift rate of 8.5–12.2 mm/yr. Murase et al. (2013) stated that leveling surveys were done in eastern Taiwan to study the deformation between stable fault creep and the asperity area. Their results indicate that the Yuli area, which marks the geological boundary of the Lichi Melange Formation, had a steady uplift rate of 30 mm/yr, while the Chikeshan route experienced different rates of deformation. A two-dimensional single-fault model indicated a creeping slip episodically accelerated in the transition zone between stable and episodic creep areas, which aligns with the geological boundary of the Lichi Melange Formation. Lai et al. (2022) stated that the high-resolution magneto-biostratigraphy and back-stripping analysis of marine deposits in an active arc-continent suture zone of eastern Taiwan have revealed short cycles of vertical crustal oscillations. The arc crust has undergone rapid uplift to form the modern Coastal Range during transpressional strain, with rates of ~ 9.0–14.4 mm/yr since ~ 0.8–0.5 Ma. Huang and Wang (2022) reanalyzed seismic structures under the northern Longitudinal Valley using a 3D crustal model and recent earthquake data, identifying a westward-tilting fault that could facilitate the northwest subduction of the Philippine Sea Plate and influence the reactivation of the Offshore Eastern Taiwan Thrust Belt. Tang et al. (2023) examined the fault dynamics during Taiwan's 2022 Chihshang earthquake sequence, revealing the Central Range fault's impact on the Longitudinal Valley fault's behavior, as evidenced by satellite and continuous GNSS data during the Taitung earthquake's occurrence. These transient vertical crustal motions are likely characteristic of continental growth by arc accretion over geologic timescales.

Close to the time of the September 17–18 Taitung earthquake, Yagi et al. (2023) conducted an in-depth analysis using the Potency Density Tensor Inversion method. Their study revealed that the foreshock gravitated toward shallow depths, moving in a south-southwest direction. Interestingly, they identified two distinct episodes of rupture during the mainshock, one in a north-northeast and the other in a south-southwest direction. The stress concentration from both earthquake events triggered a second rupture episode. On a parallel note, Lee et al. (2023) performed joint source inversions, demonstrating that the fault triggering the earthquake had a westward dip. They also concluded that the main shock originated from the hypocenter and proceeded northward at a velocity of 2.5 km/s. Furthermore, the rupture affected the east-dipping LVF, possibly due to the significant impact on the west-dipping fault. Collectively, these findings, in combination with historical seismic events, provide compelling evidence for the existence of the CRF.

Figure 1 portrays a map filled with geological information about the surrounding area. The distribution of rock formations within the Coastal Range block is redrawn based on the work of Lai et al. (2021). In contrast, the Central Range is a unified depiction derived from various rock types. The structural line is also adopted from Lai et al. (2021), with the exception of the active fault, which has been redrawn according to Shyu et al. (2006). Specific designations for these faults include the Central Range Fault (CRF), Longitudinal Valley Fault (LVF), and Chimei Fault (CF). The 0917-0918 focal mechanism illustrations are sourced from Chou et al. (2022). The map also identifies the aftermath of the 0917-0918 Taitung earthquake, highlighted by a yellow rhombus marking the reported point of the affected area that we obtained during the post-seismic field survey. A black circle indicates our GNSS measurement point, while our focus of the investigation is outlined in a white rectangle.

Fig. 1
figure 1

The surrounding area of the geological information map

Full size image

The LVF in eastern Taiwan is a seismically active structure that poses a significant earthquake hazard. Studies have revealed that the fault has a minimum millennial rate of dip-slip, and the most recent earthquake occurred in 1951, with a maximum coseismic offset of up to 2.1 m and an average coseismic uplift of ~ 0.5–1 m (Shyu et al. 2006). While rapid seismic creep is present in the southern part of the fault near Chihshang, it remains unclear whether significant aseismic creep occurs around the Hsiukuluan Canyon (Lai et al. 2022). The fault is divided into three sections: the Chihshang, Yuli, and Rueisuei sections, with the latter two ruptures occurring in separate faults (Shyu et al. 2007). Excavations have revealed that the fault ruptured at least three times with magnitudes of 7.0–7.2 up to 390 years prior to the 1951 earthquake, and the average time between earthquakes is about 150 years (Yen et al. 2008). Leveling surveys have shown that the Yuli area experiences a steady uplift rate of 30 mm/yr, while the Chikeshan route has different rates of deformation (Murase et al. 2013). The fault responsible for the 2022 Taitung earthquake was west-dipping, and the main shock originated from the hypocenter and moved northward at a speed of 2.5 km/s (Yagi et al. 2023; Lee et al. 2023). This model, along with past earthquakes, provides strong evidence for the existence of the CRF.

1.2 The RTK GNSS measurement and results

In this study, we employed network RTK GNSS to measure the benchmarks (stations, Fig. 2, Appendix 1) established near Yuli downtown, providing a comprehensive understanding of the deformation patterns associated with the 0918 Taitung earthquake. These benchmarks were set up by multiple agencies, including Township and Hualien County, the River Management Office, and the Ministry of Interior, for various reasons, reflecting the importance of monitoring tectonic activity in the region. The benchmarks have been surveyed on several occasions over the past few years, offering a valuable dataset for comparative analysis. The benchmarks that have been installed for several years, for the most part, have not been measured periodically and only provide data from when they were initially installed. In reality, even though there has been a geodetic measurement of a plate shift of 82 mm/year (Yu et al. 1997; Takahashi et al. 2019; Yu and Kuo 2001), from the time of benchmark installation (between 2008 and 2013) until the occurrence of the 0917–0918 Taitung earthquake in 2022, various minor earthquakes did not result in significant surface deformation, therefore we consider measurements of benchmarks are capable of representing the coseismic deformation resulted from 0918 earthquakes. Essentially, in the case of this earthquake which resulted in a shift of more than 100 mm, we are confident that our measurements are representative of the coseismic shift.

Fig. 2
figure 2

The horizontal coseismic displacements. The reference point during RTK GNSS measurement is designated as the DOLI station, denoted by a black circle, while the GE18x station is utilized as a local reference during analysis and represented by a black rectangular shape. The "T" shape with rectangular yellow transparent symbolizes the start–end points with 1 km width of the profile sections x − x' and y − y', which display the vertical displacement on Fig. 4

Full size image

For greater precision, our data is cross-referenced with continuous GNSS stations, JPIN and YUL1. JPIN (belonging to the Institute of the Earth Sciences) provides data from 2001 to 2023 (TGM 2023a), and YUL1 (belonging to the Central Weather Bureau) from 2010 to 2023 (TGM 2023b). A significant focus was on the events of 2022-09-18, using readings from both stations. JPIN is situated on the Coastal Range block (CoR), and YUL1 on the Central Range block (CeR). JPIN's data, modeled linearly since its installation on 2001–12-20 up to the day before the 2022-09-18 earthquakes, indicates annual displacements of east component − 11.13 ± 0.02 mm/yr, north component 32.29 ± 0.01 mm/yr, and height 5.70 ± 0.01 mm/yr (TGM 2023b). For the 2022-09-18 earthquake, JPIN reported displacements of east component − 438.76 ± 2.37 mm, north component 490.36 ± 2.38 mm, and height − 351.45 ± 11.05 mm (TGM 2022a). YUL1's data, modeled since its inception on 2010-12-19 up to the aforementioned earthquake, shows annual displacements of east component 2.84 ± 0.56 mm/yr, north component 11.43 ± 0.36 mm/yr, and height − 12.12 ± 1.77 mm/yr (TGM 2023b). However, for the same earthquake, YUL1 reported displacement of the east component − 330.54 ± 12.07 mm, north component -840.81 ± 9.76 mm, and height 905.68 ± 57.11 mm (TGM 2022b).

Numerous benchmarks were established between 2008 and 2013. To calibrate our measurements, we obtained the coseismic values by subtracting our measurement data from the total displacement from the installation year until just before the earthquake occurred. Benchmarks on the CoR relied on JPIN data for adjustments, while those on the CeR used YUL1 data. While the preferred reference for benchmarks on the Longitudinal Valley (LV) plain is the JULI station, it was not available. Hence, adjustments were made using the southern station at Chihshang, known as the TTCS station (belonging to Central Weather Bureau), which is geologically situated on the LV.

Between October 17, 2022, and January 16, 2023, we measured a substantial number of benchmarks using network RTK GNSS for data collection. This approach allowed for more detailed spatial coverage and enhanced the accuracy of the observed coseismic deformation patterns. Although RTK is less accurate than campaign GNSS and continuous GNSS, with an accuracy range of 1–10 cm (Saghravani et al. 2009), it is still sufficient for determining the location of coseismic deformation in the context of this study. The core RTK network station utilized in this study is the Continuously Operating Reference Station (CORS) located at Dongli (DOLI) as depicted in Fig. 2, which operates under the auspices of Taiwan Bai Yue Technology (TBYT) services. Additionally, for validation purposes, we employed a station managed by the National Land Surveying and Mapping Center (NLSC) under the Ministry of the Interior (MOI) to cross-verify our measurements. We used local Taiwan Datum 1997 (TWD97) or EPSG:3826 as the projection for the entire process, from measurement to analysis. Comprehensive records of the benchmarks can be found in the supplementary data.

The measurement results reveal that there are 199 stations with measured longitude, latitude, and altitude out of a total of 266 stations obtained from the local geodetic agency (see Appendix 1). The remaining 67 stations have no measured altitude and are marked with N/A data in Appendix 1. Owing to the implementation of the RTK GNSS system in our measurements, we achieve a high level of precision, confidently reaching the scale of centimeters. However, it's important to note that our Horizontal Root Mean Square (HRMS) and Vertical Root Mean Square (VRMS) values indicate an accuracy range spanning from millimeters to centimeters. For analysis to enable the integration of GNSS measurements with various dates of measurements, the GE18x station was selected as the local reference due to its proximity to the DOLI station and roughly equal distance to both blocks. The average horizontal error (HRMS) value of 0.009 m with a maximum value of 0.06 m was recorded at station GH55. Similarly, the average vertical error (VRMS) value of 0.02 m with a maximum value of 0.09 m was measured at station GL24. The Northing displacement (∆N) has noted maximum movement to the north of 1.89 m from the L29-1 station, while the maximum towards the south has been noted at 2.09 m from the U216 station. The Easting displacement (∆E) has noted maximum movement to the east of 0.06 m from GN36 station, while the maximum movement to the west has noted 1.31 m from GL35. The Height component displacement shows a maximum uplift of about 2.40 m from GE21y, while the maximum subsidence has been noted number about 1.34 m from GK02 station. The Length component (calculated from Northing and Easting) shows the maximum number of about 2.31 m towards southwest from U216 station, the GE18x station becomes the least number of each component displacement due to its assignment as the local reference. These measurements provide valuable insights into the deformation patterns associated with the 0918 Taitung earthquake. Please refer to Appendix 1 for a detailed list of the stations and their measurements.

To validate our data, within a 1 km radius of station YUL1, there are seven benchmarks. We calculated the average difference in values from station YUL1 relative to these benchmarks, resulting in the following discrepancies 0.24 m in the northward direction, 0.32 m in the eastward direction, 0.24 m in height, and an overall length difference of 0.36 m. These numbers are critical for ensuring the reliability and accuracy of the measurements recorded at YUL1, and they provide a quantifiable method to assess the precision of our data against established reference points.

Situated west of Yuli downtown, the Central Range (CeR) block, represented in purple in Fig. 1, reveals various displacement measurements. The northing component (∆N) exhibits an average displacement of approximately 0.93 m toward the south which has a standard deviation of about 0.32 m, with the GE16x station recording a maximum displacement of 1.68 m toward the south and the GF09 station noting minimum toward the south displacement of about 0.39 m. In the easting component (∆E), an average displacement of roughly 0.68 m to the west is recorded with a standard deviation of about 0.10 m, where the GN10 station presents the maximum displacement of 0.86 m toward the west, whereas the minimum displacement of 0.52 m toward the west from GL63 station. Regarding the height component (∆H), there is an average uplift of about 1.10 m which has a standard deviation of about 0.58 m, the GE16x station reports a maximum uplift of 2.31 m, while the GF09 station observes a minimum uplift of 0.21 m, signifying the least displacement in this category. The length component, combining the northing and easting components, shows an average displacement of 1.17 m with a standard deviation of about 0.29 m toward a southwest direction, the GE16x station records a maximum displacement length of 1.83 m, while a minimum displacement of 0.67 m is observed at the U011 station.

The Longitudinal Valley block, located beneath Yuli downtown and depicted as the Alluvial area in Fig. 1, necessitates the consideration of the rupture line for accurate interpretation of results. Our rupture line investigation along the Longitudinal Valley yielded various measurements that were integral to our analysis. To further enhance our understanding, we have segmented the area into two sections, demarcated by the rupture line: the western part, designated as LV1 in Fig. 5, and the eastern part, marked as LV2 in Fig. 5.

In the LV1 area, the northing component (∆N) demonstrates an average displacement of 0.53 m to the south which has a standard deviation of about 0.45 m, the U216 station records a maximum displacement of 2.09 m toward the south, while the GE62 station notes a maximum displacement of 0.80 m toward the north, and the least displacement for this component is 0.09 m toward the north at the GN39 station. The easting component (∆E) has an average displacement of 0.50 m toward the west with a standard deviation of about 0.17, with the GE62 station recording a maximum displacement of 1.17 m to the west, while the GE12y station observing a minimum displacement of 0.11 m to the west, which also signifies the least displacement in this category. The height component (∆H) presents an average uplift of 0.63 m which has a standard deviation of about 0.61 m, with the GE21y station marking the maximum uplift of 2.40 m, and the GF21 station indicating a minimum uplift of 0.003 m. The length component shows an average displacement of 0.79 m with a standard deviation of about 0.36 m towards the southwest, the maximum length displacement of 2.31 m is noted at the U216 station, while the GB07 station records a minimum length of 0.20 m.

In the LV2 area, the northing component (∆N) shows an average displacement of 0.90 m toward the north with a standard deviation of about 0.47 m, with the L29-1 station recording a maximum displacement of 1.89 m to the north, and the L50 station noting a maximum displacement of 0.20 m to the south, the least displacement in this component is observed to be 0 m at the GE18x station. The easting component (∆E) has an average displacement of 0.36 m towards the west which has a standard deviation of about 0.31 m, with the U243 station recording a maximum displacement of 1.11 m toward the west, the GN36 station observing a maximum displacement of 0.06 m toward the east, the least displacement for this component is also 0 m at the GE18x station. For the height component (∆H), the average subsidence is 0.13 m which has a standard deviation of about 0.49 m, with the GI04 station marking the maximum uplift of 1.24 m, and the GF03 station recording the maximum subsidence of 0.75 m, the least displacement is 0 m at the GE18x station. The length component reveals an average displacement of 1.02 m with a standard deviation of about 0.46 m towards the northwest, the U207 station records the maximum length displacement of 2.06 m, while the minimum length displacement of 0 m is observed at the GE18x station.

The Coastal Range (CoR) block, represented in light red in Fig. 5, is situated to the east of Yuli downtown and demonstrates a variety of displacement measurements. The northing component (∆N) has an average displacement of ~ 1.01 m toward the north which has a standard deviation of about 0.26 m, with the U031 station noting a maximum displacement of 1.43 m toward the north and the GI03 station recording a maximum displacement of 0.10 m toward south, which also marks the least displacement. The easting component (∆E) shows an average displacement of about 0.57 m toward the west with a standard deviation of about 0.17 m. The maximum displacement of 1.31 m toward the west is observed at the GL35 station, while the maximum displacement of 0.05 m toward the east is recorded at the GJ32 station and becomes the least displacement in this category. In the height component (∆H), the average subsidence is approximately 0.65 m with a standard deviation of about 0.34 m. The GH47 station reports a maximum uplift of 1.23 m, while the GK02 station notes a maximum subsidence of 1.34 m, the least vertical displacement in this category, a minuscule 0.01 m, is recorded at the GL27 station. The length component indicates an average displacement of 1.18 m with a standard deviation of about 0.22 m towards the northwest, the maximum length displacement of 1.64 m is observed at the U031 station, while the 9152 station records a minimum length of 0.24 m.

Figure 2 illustrates the pattern of horizontal coseismic displacement. The black arrow signifies the direction of these displacements, as calculated from the ∆E and ∆N values (refer to Appendix 1). The length of each arrow corresponds to the magnitude of displacement, as determined using the Length value (also found in Appendix 1). For clarity and ease of interpretation, we have incorporated a scale for the arrows in the lower right corner of the figure. The red dashed and solid line delineates the rupture line, which is based on point ruptures recorded during our post-seismic field survey. The transparent purple block was denoted as the CeR while the transparent red was denoted as CoR.

Figure 3 presents the distribution of vertical coseismic displacement. The upward-facing orange arrows indicate the vertical uplift displacement, corresponding to an ∆H value greater than 0. Conversely, the downward-facing blue arrows represent subsidence displacement, associated with a ∆H value less than 0. The length of these arrows is determined by the ∆H value (please see Appendix 1 for details). For enhanced readability, we have included a scale for the arrows in the lower right corner of the figure. The red dashed and solid line marks the fracture line, derived from fracture points identified in our post-seismic field survey. The light purple block was denoted as the CeR while the block light red was denoted as CoR. The "T" shape with rectangular yellow transparent symbolizes the start–end points with 1 km width of the profile sections x − x' and y − y', which display the vertical displacement.

Fig. 3
figure 3

The vertical coseismic displacements. The DOLI station, marked with a black circle, serves as the reference point during RTK GNSS measurements, whereas the GE18x station, indicated by a black rectangle, is used as the local reference during the analysis. The 'T' shape represents the x − x' and y − y' profile sections depicted in Fig. 4

Full size image

Figures 2 and 3 provide a clear visual representation of the rupture lines generated in the aftermath of the 0918 Taitung earthquake, based on reported points from the affected area (as shown in Fig. 1). The horizontal view (Fig. 2) and vertical view (Fig. 3) exhibit distinctive patterns that are separated by the main rupture line. Notably, the western region of Yuli experienced a southwestward displacement and vertical uplift, while the eastern region experienced a northwestward displacement and downward vertical movement. By defining the lines connected from the point of ruptures as the structure of a fault plane and considering the east and west displacement of Yuli, the fault structure type of our investigation suggests a reverse left-slip fault. Moreover, by fusing RTK GNSS measurement results with the local geological formations and current ruptures, while considering the displacement patterns, we have effectively demarcated the study area into four distinct moving blocks, a subject thoroughly investigated in the discussion section. Overall, these figures provide valuable insights into the impact of the earthquake and its fault structure classification, which can aid in future disaster preparedness and mitigation efforts.

2 Discussion

2.1 Distribution of ruptures in affected areas

The rupture points were actually investigated to understand the local fault that occurred. The delineated reported points were believed to be rupture lines that were generated during the coseismic events which show that the rupture line has a role as the boundary between two displacement patterns (see Figs. 2, 3), vertical displacement also shows the ruptures generated different vertical displacements on both sides (see Fig. 4), it is indicating that the rupture line has a role as the boundary between two displacement patterns. The previous studies show that different vertical displacement is utilized to find the ruptures between blocks (Angelier et al. 1997; Lee et al. 2003; Huang et al. 2019). The RTK measurement results, which have been shown to be useful in other studies (Ge et al. 2008; Xu et al. 2010; Ohta et al.2012; Wu et al. 2019), were particularly helpful in dividing the block movement and identifying the rupture lines. Our analysis, in line with the work of Lee et al. (2023) and Yagi et al. (2023), shows that the rupture points provide significant insights into the distribution of ruptures and their relationship with the affected regions, as previously observed in other earthquake studies. The surveyed and estimated ruptures indicate that they are primarily located in the Alluvial Longitudinal Valley, consistent with observations made by some previous studies for similar geological settings (Cheng et al. 1996; Shyu et al. 2007). Locally, the ruptures struck Yuli downtown and extended northeastwards (Shyu et al. 2006; Valkaniotis 2022; Lin 2022) which were similar to our aligned ruptures point. This observation underscores the importance of understanding the geological and geomorphological features of the area when analyzing the consequences of an earthquake (Hsu 1962; Biq et al. 1985).

Fig. 4
figure 4

The vertical displacement profile section (from Fig. 3) and the x − x’ section are located on the north yuli (top), and the y − y’ section is located across from Yuli Downtown. The purple surface line is the Central Range part, the yellow surface line is the Longitudinal Valley part, and the light-red surface line is the Coastal Range part

Full size image

2.2 Displacement of the central range block

The data shows that the Central Range block experienced horizontal movement in the southwest direction (see Fig. 5 marked on the purple block). The average displacement measured was 1.17 m with a standard deviation of about 0.29 m, with a maximum displacement of 1.83 m and a minimum of 0.67 m (see Fig. 2). In terms of vertical displacement, the average uplift was 1.10 m which has a standard deviation of about 0.58 m, with a maximum uplift of 2.31 m and a minimum uplift of 0.21 m (see Fig. 3). We also made a section on northern Yuli (x − x’ section) and a section across Yuli downtown (y − y’), which show that in only vertical terms, the Central Range has the most uplift displacement (see Fig. 4). This result shows different movement if compared to the common whole Taiwan movement of about 82 mm/yr to the northwest direction (Yu et al. 1997; Yu and Kuo 2001; Takahashi et al. 2019), this phenomenon is related to the surface movements during an earthquake can generate varying degrees of regional surface displacement (Shyu et al. 2005; Shyu et al. 2007; Wu et al. 2019; Yagi et al. 2023). These findings indicate substantial tectonic movements during the earthquake, which likely played a significant role in the observed damage and deformation patterns (Yen et al. 2008; Murase et al. 2013; Lee et al. 2023). The RTK GNSS measurements provide valuable insights into the coseismic movements that occurred during the earthquake event (Yen et al. 2019; Wu et al. 2019; Huang et al. 2019).

Fig. 5
figure 5

The kinetic model of the deformation during the 0917-0918 Taitung earthquakes

Full size image

Such large-scale movements have significant implications for seismic hazard assessment and risk mitigation in the region (Shyu et al. 2006; Hsu 1962; Lai et al. 2022). Understanding the extent and nature of these displacements can help inform the design and construction of infrastructure in seismically active areas, potentially reducing the impact of future earthquakes on communities (Yu and Liu 1989; Cheng et al. 1996; Kuochen et al. 2004). Furthermore, GNSS measurements can also contribute to the development of refined numerical models of the earthquake source and the surrounding geological structures (Biq et al. 1985; Bird 2003). These models can provide a more accurate representation of the complex interactions between the tectonic blocks and the associated faults, allowing researchers to understand better the mechanisms driving the observed deformation and damage. This knowledge can be used to enhance preparedness for future seismic events, guide land use planning decisions, and improve the overall resilience of communities located in earthquake-prone areas (Angelier et al. 1997; Lee et al. 2001; Lee et al. 2003).

2.3 Displacement of the coastal range block

The information indicates that there was a northwestward horizontal shift in the Coastal Range block (as seen in Fig. 5, highlighted in the light red block). In terms of horizontal northwest displacement, the Coastal Range block displayed an average of 1.18 m with a standard deviation of about 0.22 m, with values ranging from a maximum of 1.64 m to a minimum of 0.24 m (see Fig. 2). Conversely, the vertical displacement was notably smaller, with an average subsidence of only 0.65 m which has a standard deviation of about 0.34 m (see Fig. 3). The diagrams portraying the northern Yuli region (x − x' section) and the central area of Yuli downtown (y − y') demonstrate that when viewed in strictly vertical dimensions, the Coastal Range displays the mostly subsided (refer to Fig. 4). This result shows a similar movement compared to the common whole Taiwan movement of about 82 mm/yr to the northwest direction, surface movement during an earthquake can sometimes result in similar regional surface displacements, indicating a coherent deformation pattern across the affected area (Yu et al. 1997; Yu and Kuo 2001; Takahashi et al. 2019). Compared to the analyzed satellite image shows a similar distribution of the affected areas that mention the western part moved to the south and the east part moved to the north (Lin 2022; Valkaniotis 2022). This observation highlights the importance of understanding the relationship between earthquake rupture mechanisms and surface deformation patterns in order better to assess the impacts of earthquakes on regional scales (Yagi et al. 2023; Lee et al. 2023). The RTK GNSS measurements in our study reveal that the Coastal Range block experienced a significant horizontal movement toward the northwest during the coseismic event (Yen et al. 2019; Wu et al. 2019; Huang et al. 2019).

These observations not only underscore the intricate tectonic interactions that occurred during the earthquake (Kao et al. 1998; Kao and Rau 1999; Shyu et al. 2005) but also emphasize the potential for diverse degrees of surface displacement across different regions (Lai et al. 2022). The complex tectonic behaviors observed in this study highlight the importance of a comprehensive understanding of regional geodynamics in order to assess better the impacts of earthquakes on local geology, infrastructure, and hazard mitigation strategies (Hsu 1962; Bird 2003; Shyu et al. 2007). Furthermore, the varying surface displacements suggest that the interplay between the underlying tectonic structures and the regional geological settings could lead to spatially heterogeneous stress distribution and strain accumulation patterns (Kuochen et al. 2004; Biq et al. 1985).

3 Displacement of the longitudinal valley block

The data suggest that the Central Range block underwent a horizontal movement towards the northwest (refer to Fig. 5, where this is marked in a light-yellow block and texted as LV1). The western part appears to have been predominantly influenced by the Central Range block (see Fig. 4), moving horizontally in the southwest direction with an average displacement of 0.79 m with a standard deviation of about 0.36 m and experiencing an average vertical uplift of 0.63 m which has a standard deviation of about 0.61 m (see Figs. 2 and 3). This phenomenon related to the surface movements during an earthquake can generate varying degrees of regional surface displacement, as demonstrated by the heterogeneity in the distribution of coseismic slip during such events (Takahashi et al. 2019; Huang et al. 2019; Wu et al. 2019). In contrast, the eastern part (refer to Fig. 5, where this is marked in a light yellow block and texted as LV2) was likely affected by the Coastal Range block, moving horizontally in the northwest direction with an average displacement of 1.02 m with a standard deviation of about 0.46 m and recording an average vertical subsidence of 0.13 m which has a standard deviation of about 0.49 m (see Figs. 2 and 3). Surface movement during an earthquake can sometimes result in similar regional surface displacements, indicating a coherent deformation pattern across the affected area (Takahashi et al. 2019; Huang et al. 2019; Wu et al. 2019). The RTK GNSS measurements emphasize that the recent ruptures divided the Longitudinal Valley block into two distinct parts, showcasing the complex interplay between these regions and their respective tectonic movements (Yen et al. 2008; Yagi et al. 2023; Lee et al. 2023). These findings suggest that the earthquake caused significant and lasting alterations to the local geological structures, with potential implications for understanding future seismic activity in the region and developing effective mitigation strategies (Biq et al. 1985; Yu et al. 1997; Yu and Kuo 2001; Lin 2022; Valkaniotis 2022).

The in-depth analysis of the RTK GNSS measurements provides valuable insights into the intricate tectonic interactions that occurred during the 0918 Taitung Earthquake. The boundary that divided the eastern and western parts of the Longitudinal Valley as indicated on the field from the rupture observations. The recent ruptures on the Longitudinal Valley plain we believed as Yuli fault was first introduced after the 1951 earthquake event (Hsu 1962), then detailed that the Yuli fault was related to the LVF (Shyu et al. 2007). However, the investigation location is the convergent tectonic boundary of the Eurasian (represented by the Central Range) and Philippine plates (represented by the Coastal Range) (Yu and Kuo 2001; Bird 2003; Shyu et al. 2005). Then, about after the glaciation age, the Longitudinal Valley was filled by the conglomerates (Shyu et al. 2008). The convergent boundary is believed covered under the Longitudinal Valley conglomerate deposits. The existence Yuli Fault is probably related to the convergent tectonic boundary. By understanding these movements and their impacts on the local geological structures until the tectonics settings can gain a clearer picture of the factors that contributed to the observed damage and deformation in the Yuli area.

3.1 The role of the Yuli fault during the 0918 Taitung Earthquake

The Yuli fault, first recognized by Hsu (1962) through his analysis of the 1951 earthquake ruptures, was proposed as the principal fault of the Longitudinal Valley Fault (LVF). Subsequent studies, notably by Shyu et al. (2007), the Yuli fault's steeply dipping strike-slip fault. However, they also observed that the Yuli rupture's southern part leans closer to the Central Range Fault (CRF). Recent research by Chou (2022) and Lee et al. (2023) contests these findings, asserting that the two major shocks of the 0918 Taitung earthquakes were positioned on a west-dipping fault, seemingly linked to the CRF. This discovery aligns with our rupture surveys, which found the rupture lines nearer to the CRF than the LVF. This body of research, including our own, demonstrates two discernible displacement patterns within the Longitudinal Valley plain. Intriguingly, the eastern part of the Longitudinal Valley moved comparably to the Coastal Range Block, albeit to a lesser degree (as marked in Fig. 5's yellow block denoted with LV2). Surface evidence suggests that this portion of the Longitudinal Valley plain was pulled by the Coastal Range Fault during the 2022 Taitung earthquake. Chou et al. (2022) further support this, indicating that most aftershocks occurred near an east-dipping fault likely related to the LVF. Conversely, the western part of the Longitudinal Valley moved similarly to the Central Range Block, although less extensively (see Fig. 5's yellow block denoted with LV1). Surface indications propose that this section of the Longitudinal Valley plain was influenced by the CRF during the earthquake sequence, a theory supported by Chou et al. (2022), who state the two major shocks were situated at the west-dipping fault. Therefore, based on these observations, we assert that the ruptures during the 2022 Taitung earthquakes can be ascribed to the Yuli fault (depicted in Fig. 5). The Yuli fault, characterized by its westward dip with a reverse left slip fault component, was likely set into motion by the hypocenter originating from the CRF.

Figure 5 presents a kinetic model illustrating the deformation generated by the 0918 Taitung earthquakes. The model includes the Central Range, depicted as a purple block, which has shifted towards the southwest and uplifted. The adjacent light-yellow block represents the Longitudinal Valley (LV1), which seems to have been pulled along by the movement of the Central Range. Conversely, the red block indicates the Coastal Range that has moved northwest and subsided. Another light-yellow block symbolizes a different part of the Longitudinal Valley (LV2) which appears to have been dragged by the Coastal Range's movement but slightly uplifted.

4 Conclusion

The primary ruptures identified in the post-earthquake survey, which were aligned with each point of rupture, are believed to represent the Yuli fault. The patterns of displacement linked with these fractures typically shift in multiple directions: towards the southwest, there is an uplift that includes the Central Range block and the western part of the Longitudinal Valley, whereas, towards the northwest, there is a subsiding that involves the Coastal Range block and the eastern portion of the Longitudinal Valley. The Central Range block appears to have exerted an influence on the western part of the Longitudinal Valley, causing it to shift. Conversely, the Coastal Range block has affected the eastern part of the Longitudinal Valley. The Yuli ruptures played a crucial role in the 20220918 Chihshang earthquake, serving as a fault plane. The consensus is that the mainshock was situated on the west-dipping section of the Central Range Fault, which is believed to be connected to the Yuli fault plane. The minor ruptures, located near the Coastal Range block, may potentially be related to the Longitudinal Valley Fault (LVF), their positioning is remarkably similar to where most aftershocks were recorded, indicating a potential correlation. The intricate interplay between the Yuli ruptures, the Central Range Fault, and the Longitudinal Valley Fault (LVF) highlights the complexity of the region's seismic activity. A deeper understanding of these relationships is essential for improving earthquake prediction models and developing more effective strategies for earthquake preparedness and response.