1 Introduction

The documentation of coseismic surface ruptures associated with major earthquakes (e.g., Chen et al. 2001; Clark 1972; Goto et al. 2017; Huang et al. 2019; Litchfield et al. 2018) provides fundamental knowledge to understand the characteristics of active fault systems precipitating in seismic events. Such knowledge not only provides important insights to understand the earthquake itself, but also adds fundamental information for seismic hazard and risk mitigation analyses. Furthermore, detailed investigations of coseismic surface rupture would also improve the interpretation of tectonic-related geomorphic features, especially for areas where features related to active faults are covered or modified by later natural or anthropogenic processes.

For the island of Taiwan which experiences rapid and prominent tectonic deformations, the records of fault surface rupture also provide irreplaceable information for the understanding of its recent tectonic activities, especially for strike-slip and oblique faults in plain areas (e.g., Hsu 1962; Omori 1907; Otuka 1936). In the central Longitudinal Valley where the topography is largely dominated by alluvial and fluvial processes, the recent 2022 Guanshan and Chihshang earthquakes provided a unique opportunity to understand and examine the active fault distribution in the valley, where our understanding of active faults was largely built upon the documentation of earthquake surface ruptures in 1951 (Hsu 1962; Shyu et al. 2007). The 2022 Guanshan-Chihshang earthquake sequence is considered the continuation of a seismic sequence since 2018 that extended from the northern part of the Longitudinal Valley (Huang and Wang 2022; Jian and Wang 2022). This sequence includes the Mw 6.4 Guanshan foreshock and the surface rupturing Mw 6.8 Chihshang mainshock on 17 and 18 September 2022 (Fig. 1a), and the Mw 6.8 Chihshang mainshock represents the largest earthquake in the Longitudinal Valley since the 1951 Longitudinal Valley earthquake sequence (e.g. Chen et al. 2008; Chung et al. 2008; Lee et al. 2008; Shyu et al. 2007).

Fig. 1
figure 1

a The neotectonics setting of Taiwan showing the geometry of the collision between the Philippine Sea Plate and the Eurasian Plate, modified from Shyu et al. (2005a). The 2022 Mw 6.4 Guanshan and Mw 6.8 Chihshang earthquake occurred within the Longitudinal Valley Suture (LVS), which is composed by the Central Range fault system at the eastern base of the Central Range, and the Longitudinal Valley Fault system along the western flank of the Coastal Range. The red lines are major active faults from Shyu et al. (2005a). Black lines are boundaries of major geological provinces of Taiwan. Colored dots are seismicity between 17 September 2022 and 16 December 2022 obtained from the Central Weather Administration (Central Weather Administration 2012). Focal mechanism data of the Mw 6.4 and Mw 6.8 earthquakes were acquired from GCMT catalog (Dziewonski et al. 1981). Most of the seismicity after the Guanshan and the Chihshang earthquakes occurred in the central and southern part of the Longitudinal Valley, along the Central Range fault. b Map of major active structures along the central and southern part of the Longitudinal Valley, between the Central and Coastal Ranges (Shyu et al. 2016, 2007, 2006b, 2020). The east-dipping Longitudinal Valley fault system (LVFs; in orange) and the west-dipping Central Range fault system (CRFs; purple dashed line) are the two major active fault systems along the valley. Yuli fault (YLF) is the active fault that ruptured during the 2022 Chihshang mainshock. The rupture was marked in dark red colored line (solid and dashed). Yellow colored solid and dashed bold lines are the surface ruptures during the November 1951 earthquake (Shyu et al. 2007). The blue colored vector shows the relative motion between the volcanic arcs on the Philippine Sea Plate and the Eurasian continental margin (Yu et al. 1997)

Full size image

Immediately following the Mw 6.8 mainshock, photos and videos shared on social media showed building and road damages, indicating that the mainshock was accompanied by surface fault ruptures in the Longitudinal Valley. The affected areas included the town of Yuli, which had also experienced surface ruptures and damages during the Nov-1951 earthquake (Bonilla 1975; Hsu 1962; Shyu et al. 2007). Thus, to document the co-seismic fault surface ruptures associated with the 2022 earthquake and their relationship with the active faults in the central part of the Longitudinal Valley, we initiated a series of post-earthquake field surveys from the second morning after the mainshock. The first survey was conducted 18 h after the mainshock, with subsequent surveys aimed at monitoring ground displacement following the coseismic deformations. The primary objective of our survey is to document the near-fault ground deformations and their changes, with a particular focus on the Yuli fault, which ruptured during both the 1951 and 2022 earthquakes.

In this study, we summarize our field observations and report the characteristics of the coseismic fault surface ruptures along the Yuli fault. We complement the coseismic cross-fault displacement measurements with subsequent measurements made within a few months after the mainshock to document post-seismic cross-fault displacements. We then discuss the significance of these observations with respect to the shallow geological properties along the Yuli fault, and the role Yuli fault plays in the active tectonic system within the central Longitudinal Valley.

1.1 Geological background

The relative motion between the Philippine Sea and Eurasian plates is accommodated through a double suture system across the island of Taiwan (Fig. 1a). This system features primary active faults situated along both the island’s western fold-and-thrust belt and its eastern deformation belt (e.g., Angelier et al. 1997; Hsu et al. 2003; Shyu et al. 2005a). Within these two deformation belts, more than one-third of the present-day convergence, approximately 3 to 4 cm/year (Hsu et al. (2003); Hsu et al. (2009b); Yu et al. (1997); Yu et al. (1990); Yu and Kuo (2001)) is accommodated by the eastern deformation belt, which is sometimes referred as the Longitudinal Valley Suture (LVS), located between Taiwan’s backbone Central Range and the arc-dominated Coastal Range (Fig. 1a) (Shyu et al. 2005a).

The LVS primarily comprises the well-known east-dipping Longitudinal Valley fault system (LVFs) at the western edge of the Coastal Range (Angelier et al. 1997; Shyu et al. 2006a, 2008, 2006b; Wang and Chen 1993) and the less-recognized west-dipping Central Range fault system (CRFs), which approximately aligns with the eastern margin of the Central Range (Fig. 1b) (Shyu et al. 2006b, 2005b; Wu et al. 2006). Of these two fault systems, approximately 2 to 3 cm/year of plate convergence is accommodated by the east-dipping LVFs (Hsu et al. 2003, 2009b; Shyu et al. 2006a; Yu et al. 1997; Yu and Kuo 2001). The high slip rate of the LVF has resulted in prominent surface ruptures and earthquakes in paleoseismological and historical records (Bonilla 1975; Chen et al. 2007; Shyu et al. 2007; Yen et al. 2008), and aseismic surface creeps in geodetic and creepmeter analyses (e.g., Hsu and Bürgmann 2006; Lee et al. 2003; Mu et al. 2011; Murase et al. 2013). In contrast, the less-acknowledged CRFs is likely to accommodate roughly ~ 1 cm/year mean fault slip rate on the western side of the valley (Shyu et al. 2016, 2006b, 2020). No solid evidence for fault surface rupturing events has been reported, except for blind earthquakes occurred at the northern and the southern part of the fault (Chen et al. 2009; Chuang et al. 2014; Lee et al. 2014; Wu et al. 2006).

While no conclusive evidence of fault surface rupturing event occurred in the central section of the CRFs in the past century, surface ruptures associated with the November 1951 earthquake drew our attention to the existence of active faults in the central part of the Longitudinal Valley, west of the well-known LVFs (Fig. 1b). In the heart of Yuli Town’s residential area, co-seismic surface ruptures, later been referred as the Yuli fault (Hsu 1962) were reported after the November 1951 earthquake. This surface rupture occurred in conjunction with other co-seismic surface ruptures associated with the southern and the central LVFs (Bonilla 1975; Hsu 1962; Shyu et al. 2007) (Fig. 1b).

Photos taken shortly after the Nov-1951 earthquake, as well as subsequent field investigations conducted decades later, all indicate that the Yuli fault had approximately 20 to 40 cm of co-seismic left-lateral offset and negligible vertical displacement observed in the town of Yuli (Bonilla 1975; Hsu 1962). No clear evidence of accumulated vertical deformation has been identified along its fault trace based on geomorphic analysis, suggesting that the Yuli fault may not have a significant component of reverse motion. Thus, Shyu et al. (2007) suggested that the Yuli fault is a sub-vertical fault situated between the CRFs and LVFs, with an unclear structural relationship with the east-dipping LVFs and west-dipping CRFs.

The Mw 6.4 and Mw 6.8 Guanshan-Chihshang earthquakes were the first major earthquake sequence struck the central part of the Longitudinal Valley since the Nov-1951 earthquakes. Unlike the 1951 earthquakes, the epicenters of the 2022 events were situated on the western side of the valley, at the base of the Central Range (Fig. 1a). Focal mechanisms of the foreshock and mainshock, along with the distribution of their aftershocks, all suggest that both earthquakes were caused by a steep west-dipping fault underneath the Central Range, with a total sub-surface fault rupture length of more than 40 km extending from the area between Yuli and Rueisuei in the north, to the region south of Guanshan and Chihshang (Lee et al. 2023). This interpretation is further supported by geodetic and remote sensing analyses, showing that the west-dipping fault (CRFs) lies close to the western margin of the Longitudinal Valley is the main source of the foreshock and the mainshock (Tang et al. 2023). Their model also suggested that the west-dipping ruptured fault is connecting to the Yuli fault on the surface, implying the Yuli fault is part of the west-dipping CRFs at the central part of the Longitudinal Valley.

1.2 Methods

Since the epicenters of the 2022 earthquake sequence were primarily situated near the base of the Central Range in the western part of the Longitudinal Valley, our documentation efforts were concentrated on the central and western portions of the valley where primary fault ruptures were expected to be present at the surface. Our field investigations revealed extensive ground deformations along the Yuli fault described by Shyu et al. (2007). Many of the surface ruptures were identified through various indicators, including offset road pavement and markers, damaged paddy-field boundaries and ditches, minor fault scarps, moletracks, and en échelon cracks. In addition to the surface ruptures roughly aligned with the previously mapped Yuli fault, we also observed ground deformations along the eastern margin of the valley, predominantly along the mapped Longitudinal Valley Fault (LVF) by Shyu et al. (2007) and Shyu et al. (2020). The LVF-related ground deformations often exhibited ground fractures and offsets that were not readily measurable, and typically displayed an east-side-up geometry. We decided not to report these LVF-related ground failures in this paper, as the data we collected along the LVF were not as detailed and timely as those collected along the Yuli fault rupture.

Our survey started in the town of Yuli, and progressively extended to areas north and south of Yuli where surface ruptures were less continuous than in the Yuli area. Most of the cross-fault displacements found along the surface rupture were photographed, and measured using tape measures and compasses in the field. In places where tectonic warping near surface ruptures was evident, we documented the near-fault warping (i.e. 2 to 3 m across the fault rupture) together with the cross-fault displacements using linear features such as road markers, pathway tiles, and field boundaries that were offset and deformed by the fault. These measurements were considered as cross-fault displacements. Several offset features within the Yuli area were surveyed multiple times with tape measures from September 2022 to January 2023 to see if there are post-seismic cross-fault displacements. We verified our field measurements with photos taken during our survey to enhance the accuracy and reliability of our results.

To complement our tape measure results, several total station surveys were conducted between the date after the earthquake (19th September 2022) and about 2 months after the mainshock (November 2022). The aim of the total station survey was to estimate both the vertical and horizontal displacements across the Yuli fault rupture. These surveys also provided additional constraints on the horizontal fault displacements to our tape measure results. It is worth noting that the displacement estimated from the total station survey includes both nonbrittle off-rupture displacement, such as bending and warping close to the fault trace, and cross-fault displacements. Consequently, the estimated offset is generally larger than the offset measured with tape measures and compasses. The comparison between these two sets of results allowed us to gain insights into the extent of surface warping off the main fault ruptures.

2 Field observations

In this section, we describe the cross-fault displacements and fault rupture patterns observed along the surface rupture since the day after the Mw 6.8 mainshock on 18 September 2022 (Fig. 2). We begin with the sites that exhibited the clearest evidence of co-seismic offset: the Yuli town center area (Figs. 2 and 3), and then describe the displacements north and south of Yuli, respectively.

Fig. 2
figure 2

Map of the 2022 earthquake surface ruptures and the measured displacements. a and b show our mapped surface rupture traces and field measurement locations of the 2022 earthquake. c and d are their measured left-lateral cross-fault displacements. Yellow dots in (c) and (d) are left‐lateral cross-fault displacement measured by tape measures and compasses. Orange dots are left-lateral cross-fault displacement estimated from the total station surveys (TTS) within 3 months after the mainshock. The survey dates were annotated on the graph. Orange line in the map shows the Yuli fault trace from Shyu et al. (2007). Bold red lines are the rupture trace we mapped in the field. Thin red lines at the base of the Central Range and the Coastal Ranges are the trace of Central Range Fault (CRF) and the Longitudinal Valley Fault (LVF) from Shyu et al. (2020). Y.H. is the location of Yuli Hospital close to the northern end of Yuli Town residential area

Full size image
Fig. 3
figure 3

Locations of field measurements (blue dots) and mapped surface rupture trace (red lines) in the Yuli town center area, between the Yuli Hospital and the southern margin of the Yuli Town residential area. Green labels are locations mentioned in the maintext. Black arrow shows the viewing direction of photo in Fig. 4. The right panel chart shows the cross-fault displacement and their uncertainty measured in the field using tape measures and the total station. Background aerial photos are from National Land Surveying and Mapping Center (NLSC), Ministry of the Interior, Taiwan

Full size image

2.1 Surface rupture within the Yuli town area

Our field investigations began in the Yuli town center area at approximately 8 am on 19 September 2022, about 18 h after the mainshock. The damages of roads and pathways in the town revealed clear evidence of localized ground failures, which could be traced continuously over a 3-km-long section of the fault, extending from the Yuli Hospital in the north to the southern margin of the Yuli Town residential area in the south (Figs. 2 and 3). Both north and south of this section, the fault ruptures extended into cultivated fields which quickly turned into water saturated paddy fields, making it challenging to map a continuous trace of surface ruptures. The general orientation of the surface rupture zone is about N15E to N20E within the town of Yuli, and the rupture exhibits classic left-lateral slip features such as right-stepping en échelon Riedel shears alternating with pressure bulges (moletracks). These features offset pathways and roads left-laterally, with minimal vertical displacements across the fault (Figs. 3 and 4). The development of moletrack features was particularly pronounced on the paved and hardened surface (e.g., paved roads and residential building floors), but was less clear in open field areas (e.g., grasslands and athletic fields).

Fig. 4
figure 4

The 2022 surface ruptures in the Yuli town center area. a The foundation separation associated to the surface rupture (marked by orange colored dashed lines) at the northern side of the Yuli Hospital. The detailed rupture map of Yuli Hospital is shown in Fig. 6. b The surface rupture trace at the Yuli roundabout. c The rupture trace in front of the Hsiehtian Temple, showing left-lateral separation of the ground, with very minor (~ 5 cm) east-side uplift across the rupture. d 7.3 cm offset measured at the northern edge of the old campus of Yuli Elementary School. e Right‐stepping Riedel shears and moletracks at the Minguo Road Sect. 2. f The en échelon Riedel shear and moletracks developed in the baseball field of the Yuli High School, taken in the afternoon of 19 September 2022

Full size image

Figure 3 shows the distribution of surface ruptures and the locations of cross-fault displacement measurements within the Yuli town area. Most of the measured left-lateral cross-fault displacements were recorded on the second day after the mainshock, between 8 am and 12 pm on 19 September 2022. The majority of these measurements indicate left-lateral cross-fault displacements of less than 15 cm, with the largest measured left-lateral displacement of around 20 cm found on the driveway of the Yuli Hospital (Figs. 3 and 4a; Table S1). In this area, our field observations indicate that many of these offset features are associated with tectonic warping within several meters off the surface ruptures. Thus, we suggest that these tape-and-compass measurements represent the minimum left-lateral cross-fault displacements in the area. Furthermore, we noticed that many residential buildings located on the surface ruptures exhibited clockwise rotation when the ruptures developed along one side of their foundation. In many cases, the ruptures followed the shape of streets and building blocks, especially around sturdy building foundations (Fig. 3). This geometry is very similar to the geometry found in the lab experiment, when the rigid foundation interacts with strike-slip faulting (e.g. Agalianos and Anastasopoulos 2021). Both the observed near-fault warping in the open ground and near buildings, and the rotation of the building foundations suggest the left-lateral displacement measured in the town area could only represent the minimum left-lateral surface offset.

The surface ruptures in the Yuli town area traversed several key landmarks, including the Yuli roundabout (Fig. 4b), the Hsiehtian Temple (Fig. 4c), and the old campus of Yuli Elementary School (Fig. 4d). These locations coincide with previously reported surface ruptures during the Nov 1951 earthquakes (Hsu 1962; Bonilla 1975; Shyu et al. 2007). Hsu (1962) and Bonilla (1975) reported left-lateral offset of 40 cm and 16 cm, respectively, along the 1951 surface ruptures at the old campus of Yuli Elementary School. The left-lateral displacement measured at this same location after the 2022 earthquake was similar, approximately 15 cm. Both the 1951 and 2022 surface ruptures exhibited negligible vertical displacement in the town of Yuli, except for the moletracks developed along the ruptures (e.g., Fig. 4e and f). These similarities suggest that the 2022 surface rupture in the Yuli town area is sourced from the same fault associated with the 1951 earthquake, even though the main rupture of Nov-1951 earthquake was not on the Yuli fault (Shyu et al. 2007).

In addition to the similarity between the fault surface ruptures of the 1951 and 2022 earthquakes, we also found evidence that both ruptures developed along an existing geologic fault at Yuli. Bonilla (1975) reported that the pond and fountain in the Yuli roundabout was originally a natural spring, and another spring was present 0.6 km south-southwest of the roundabout. Although we could not locate the second spring mentioned by Bonilla (1975), we found two additional springs and free-flowing wells situated 0.2 km and 0.4 km south-southwest of the roundabout, along the trace of the 2022 surface ruptures (Fig. 3). The alignment of these springs and wells suggests that their distribution is controlled by a geologic fault (Bryan 1919). Since no clear geomorphic evidence of vertical deformation has been reported in this area, the appearance of these springs indicates that the long-term motion of the Yuli fault is primarily dominated by left-lateral strike-slip motion, at least since the formation of the alluvial fan beneath the town of Yuli.

To further investigate the vertical deformation associated with the 2022 surface rupture, we conducted total station surveys to measure the elevation change across the rupture. The first survey was conducted in the afternoon of 19 September, about 24 h after the mainshock, at the baseball field of the Yuli High School (Figs. 3, 4f and 5). Here, the surface rupture trace was clearly defined by en échelon cracks and moletracks, forming a rupture zone of approximately 2 m in width (Fig. 5). Since no linear feature in the baseball field could be used to estimate the left-lateral offset across the fault, we surveyed six profiles across the ruptures to estimate the vertical displacement (Fig. 5). All these profiles indicate that the western side of the ruptures was the upthrown side, with uplift ranging from roughly 1 to 4 cm. The vertical uplift across the surface rupture was less than the height of the moletracks and lower than the left-lateral offset found in this area. Hence, it is challenging to identify this small elevation change without detailed surveying (e.g., Profile A2 in Fig. 5). It is noteworthy that we observed a discrepancy in the amount of co-seismic uplift between the geodetic GNSS analysis and our field survey results along this section of the surface rupture. We will discuss this in the following sections.

Fig. 5
figure 5

Total station survey location and results in the Yuli High School’s baseball field. The estimated vertical displacement (V.D.) is annotated in the profiles. Yellow lines are the surface ruptures mapped in the field, and the general trend of the rupture is marked by the yellow shaded bands. Red dots are the total station survey points. Orange dots are surveyed points on the surface rupture zone shown in profile A2, A4 and A5. Black dashed lines in elevation profiles are trend lines of the survey points, served as projection lines for vertical offset measurement. The survey was made on September 19, 2022. Background aerial photos are obtained from NLSC, Ministry of the Interior, Taiwan

Full size image

Similar observations of minor vertical cross-fault displacement were also found along the levee just north of the Yuli Hospital, where we conducted another total station survey about 3 weeks after the mainshock on 10 October 2022 (Figs. 3 and 6). At this location, the surface rupture trace is about N10W between the levee and the hospital (Fig. 6), slightly different from the general trend of N15E to N20E found in the town area. Our total station surveys along both the edge of the levee and the center line of the road indicate vertical displacement of ≤ 10 cm, and left-lateral offsets of 17 to 34 cm across the ruptures (Fig. 6). The amount of left-lateral offset estimated by the total station survey was higher than the left-lateral cross-fault displacement measured with tape measures and compasses (i.e., 20 cm). This difference could be attributed to off-rupture deformation (e.g., long wavelength warping on the surface) and/or shallow post-seismic motion. It is important to note that both of our surveyed profiles show an east-side-up geometry across the fault, and clear surface warping occurred east of the ruptures (Fig. 6). This east-side-up result is different from the GNSS ground deformation results by Tung et al. (2024). Since we were unable to trace the same rupture several tens meters north of the Zhou River along the projection of surface rupture, and the left-lateral cross-fault displacement decreased from 20 to 4 cm across the Zhou River (north of the Y.H. in Fig. 2), we propose that this east-side-up feature and the surface warping are associating to a step-over occurred on the ruptures at the Zhou River, where a small pressure ridge probably developed northeast of the mapped ruptures and slightly uplifted the eastern block.

Fig. 6
figure 6

Total station survey location and results between the Yuli Hospital and the Zhou River. Profile B1 is along the edge the levee offset by the ruptures, and profile B2 was surveyed along the center line of the road. The estimated vertical displacement is annotated in these profiles. Red and orange colored dots are total station survey points, and orange colored dots are survey points within the surface rupture zone (damage zone). Black dashed lines in elevation profiles are trend lines of the survey points, served as projection lines for vertical offset measurement. Yellow lines are the surface ruptures mapped in the field. Background aerial photos are obtained from NLSC, Ministry of the Interior, Taiwan

Full size image

2.2 Surface ruptures north of the Yuli town area

To the north of the Zhou River, there are much fewer roads than the Yuli town area, and the land use shifts predominantly to agriculture, with numerous water-saturated paddy fields. Between the Zhou River and Dayu, various locations within the paddy fields and riverbeds displayed clear evidence of ground failure during the mainshock (Fig. 2 and Fig. 7). However, only a few of these were definitively tectonic origin, while the others could have resulted from liquefaction or shaking-induced damages. We conducted a systemic search for evidence of ground failures along the general trend of the Yuli fault ruptures and identified features along the ruptures’ path, which we deemed as tectonic-related ground failures. This approach allowed us to map the damage zone of the ruptures, and to identify several features to estimate cross-fault displacement on the western side of the Hsiukuluan River (Fig. 2).

Fig. 7
figure 7

Surface ruptures and ground deformations north of the Yuli town center area. a The index map for field photos. Solid red colored lines are mapped surface ruptures in the field, and red dashed lines are the inferred surface rupture trace from spotty ground failures. b Left-lateral offset found along the “Horizontal 12th Road” showing three traces of surface ruptures. See text for detailed descriptions and Fig. 8 for detailed survey result. c Shaking induced surface cracks and ground subsidence (i.e. Lateral spreads and ground settlement) found on the riverbed of the Hsiukuluan River. d The left-lateral bending of the field boundary found in the “Horizontal 9th road”. e The surface ruptures across the road bent and twisted a ditch cover. See Table S1 for their locations

Full size image

Most of these identified features were in the flat field area that was previously the active Hsiukuluan riverbed several decades ago. The tectonic-related deformations were recognized by damage patterns of concrete road, drainage walls and paddy field boundaries that roughly aligned with one another. Additionally, many paddy fields experienced eastward tilting, forming a gentle monocline in the field. Although the ruptures were scattered within a broader deformation zone than those in the Yuli town area, we observed larger left-lateral cross-fault displacement in this section of the ruptures, showing at least 20 to 40 cm of left-lateral cross-fault displacement, with predominant west-side-up flexure deformation occurred in the paddy field.

It is worth noting that we identified a single ~ 120 to ~ 140 cm left-lateral offset along “Horizontal 12th Road” near Dayu, where three subparallel rupture traces were observed in the field (Fig. 7). This constitutes the largest left-lateral offset that we found north of the Yuli town area. The road has been partially modified, obscuring the original shape of the road and the ditch east of the rupture when we surveyed in November 2022 (Fig. 8). Several paddy field ridges south of the “Horizontal 12th Road” were also disrupted by the ruptures within ~ 200–300 m from this road (Fig. 8). Total station surveys along the remaining section of the ditch and road boundary showed that the estimated left-lateral offset ranges from 110 to 137 cm on the easternmost rupture trace and about 37 to 59 cm across the western two rupture traces. Vertical displacement across the rupture traces was also estimated to be between 30 and 50 cm, and most of this vertical cross-fault displacement was accommodated only by the western two traces. The accounts of local farmers also reported that several paddy fields adjacent to the surface rupture trace were tilted after the mainshock, suggesting the amount of vertical displacement we estimated through the total station survey in November only represents a fraction of the vertical deformation caused by co-seismic and post-seismic deformations.

Fig. 8
figure 8

Detailed map of surface ruptures near the “Horizontal 12th road” north of the Yuli town center area. Red dots show the total station survey points along both sides of the road, orange dots are survey points within the rupture zone. Black dash lines in the elevation profile charts are trend lines of the survey points, served as projection lines for vertical offset measurement. Profile C1 was taken along the edge the ditch that partially preserved after the earthquake. C3 was surveyed along the southern boundary line of the road, and profile C4 was surveyed along the northern edge of the field boundary. Background aerial photos are obtained from NLSC, Ministry of the Interior, Taiwan

Full size image

Not all of the left-lateral separation we measured at this spot were caused by faulting itself. In addition to the three left-lateral traces we surveyed, we also found a trace with right-lateral separation near the road intersection east of these three rupture traces (Point C in Fig. 8). Our survey suggests that the preserved road boundary near the junction is in fact aligned with the road boundary west of the easternmost left-lateral trace, where the 110 to 137 cm left-lateral displacements were estimated (Fig. 8). This discrepancy led us to suspect that the measured 110 to 137 cm left-lateral separation may have been caused by strong shaking and liquefaction during the mainshock (i.e., lateral spreads). We suggest that the water-saturated block east of the fault, situated between the Farm road and the ruptures, liquefied due to the sudden and strong ground acceleration, resulting in a paired right- and left-lateral displacement very close to the ruptures. A similar but much larger-scaled instance of shaking-induced liquefaction and mudflow on a gently sloped surface was observed in the 2016 Palu earthquake (e.g. Bradley et al. 2019). In both cases, the deformation occurred in water-saturated rice cultivation regions, where the ground is less solid and coherent than alluvial fan deposits.

We also identified similar tectonic ruptures associated with the Yuli fault between Chunri and the Kuokailiang, east of the Hsiukuluan River (Fig. 2). These features roughly aligned with the pressure ridges reported by Shyu et al. (2007) west of the LVF scarp. Left-lateral cross-fault displacement of up to approximately 25 cm was observed in this section of the fault (Fig. 2 and Table S1). The rice fields also exhibited eastward tilting and surface warping along the rupture trace, showing clear left-lateral and reverse component along this section of the surface rupture. Our post-earthquake field investigation and interviews with local residents indicate that the surface rupture of the Yuli fault terminated at the southeast corner of the Kuokailiang Ridge, which is believed to be a pressure ridge associated with the strike-slip dominated Yuli fault (Shyu et al. 2007).

2.3 Surface rupture south of the Yuli town area

To the south of Yuli town center area, we found two sections of continuous surface ruptures within the Longitudinal Valley, roughly along the southward extension of the Yuli fault mapped by Shyu et al. (2007). The first section of the surface ruptures (Changliang section) emerged just north of the active riverbed of the Lele River, creating a 3-km-long surface rupture between the northern bank of the Lele River in the north and the Hsiukuluan River in the south (Figs. 2, 9a and b). The surface ruptures south of the Hsiukuluan River, in the Dongli and Jhutian area, were connected by a series of short and discontinuous ruptures close to, or even east of, the mapped LVF scarp (Fig. 9c). The other section of the ruptures (Luntian section) emerged in the Luntian area, primarily west of the Hsiukuluan River. This rupture appeared on the Luntian alluvial fans and in the old riverbed of the Hsiukuluan River (Figs. 2 and 9d). All these mapped features exhibited a similar deformation pattern, characterized by left-lateral displacement and west-side-up vertical deformations.

Fig. 9
figure 9

Surface ruptures south of the Yuli town center area. a Left-lateral offset found in the watermelon field south of the Lele River. See Fig. 10 for the detailed rupture mapping. b Offset ditch across the ruptures with 82 cm of left-lateral offset measured by the tape measure and compasses. c The sub-vertical rupture on the road pavement just west of the Dongli Junior High School. The rupture is less continued compared to the section in the north. d The surface rupture in the Luntian area, on the old riverbed of the Hsiukuluan River. See Table S1 for their locations

Full size image

For the surface ruptures just south of the Lele River, our survey conducted approximately 24 h after the earthquake revealed clear and substantial left-lateral and vertical offset along the rupture trace (Fig. 9a and b). The geomorphic characteristics of the fault scarp suggest that the ruptured fault was likely steep and west-dipping, as its surface rupture showed a relatively linear rupture trace with consistently east-facing reverse fault or flexural scarps. Our total station survey results on 21 September 2022 indicate that vertical deformation could reach up to ~ 40 cm, and the left-lateral cross-fault displacement could be as large as 86 to 140 cm in the watermelon fields (Fig. 10). This was the largest left-lateral displacement that we found along the ruptures south of Yuli. The displacement appeared to decrease southward, and the dirt road south of the watermelon fields showed only about 80 cm of left-lateral displacements.

Fig. 10
figure 10

Detailed map of surface ruptures in the watermelon fields, surveyed by the total station and unmanned aerial system (UAV). Red dots are the survey points, and orange dots are survey points between the rupture traces. Black dashed lines in elevation profiles are trend lines of the survey points, served as projection lines for vertical offset measurement. Profile D1 to D5 were surveyed along the edge of plastic covers offset by the fault, and profile D6 was surveyed along the northern edge of the dirt road between paddy fields. Background aerial photos are obtained from NLSC, Ministry of the Interior, Taiwan

Full size image

Eye witness accounts from farmers working in the watermelon fields suggest that the surface ruptures propagated from south to north during the mainshock. They observed that the plastic covers in the watermelon fields were progressively broken from south to north, accompanied by loud noises. The ground rupture in the fields occurred after they were knocked down by the mainshock’s strong shaking, indicating the S-wave and surface wave arrived earlier than the surface ruptures at Changliang. Their descriptions of rupture propagation direction are consistent with the fault slip inversion results (Lee et al. 2023; Tang et al. 2023), which showed a northward unilateral rupture along the west-dipping fault plane during the mainshock.

Farther south, we conducted another total station survey in the Luntian section, where the rupture once again appeared west of the Hsiukuluan River (Fig. 9d). The ground deformation on the Luntian alluvial fan included tectonically related left-lateral offset and east-facing flexure scarps, as well as gravitational ground failures near the toe of the alluvial fan. South of the alluvial fan, we found clear evidence of east-facing flexure scarp developed in the paddy fields on the old riverbed of the Hsiukuluan River (Fig. 11). A Total station survey conducted on 9 November 2022 in the paddy fields revealed 14 to 34 cm of left-lateral offset, and 12 to 15 cm vertical deformation across the rupture zone (Fig. 11). Like the surface ruptures in the Yuli area, the amount of vertical cross-fault displacement was nearly an order of magnitude smaller than the vertical deformation recorded by nearby GNSS stations which are less than 1 km away from the surface rupture near Luntian (Tung et al. 2024). We will discuss this in the following sections.

Fig. 11
figure 11

Detailed map of surface ruptures in the Luntian area surveyed by total station. Yellow dashed line and dots are ruptures mapped in the field. Red dots show the total station survey points along both sides of the road, orange dots are survey points within the rupture zone. Black dashed lines in the elevation profiles are trend lines of the survey points, served as projection lines for vertical offset measurement. The profile E1 was taken along the northern edge of the road. The profile E2 was surveyed along the southern boundary of the road. Background aerial photos are obtained from NLSC, Ministry of the Interior, Taiwan

Full size image

2.4 Post-seismic cross-fault displacement

In addition to co-seismic surface ruptures and offsets observed in the central Longitudinal Valley area, our repeated field surveys of offset markers revealed significant shallow post-seismic deformation along the 2022 surface rupture. We repeatedly surveyed five offset markers within the Yuli town center from September 2022 to January 2023. These survey sites are situated at the Yuli roundabout, the old campus of the Yuli Elementary School, and the Yuli Hospital. The tape measure results from all these locations indicate substantial post-seismic deformation occurring within 3 to 4 months after the 2022 mainshock (Table 1 and Fig. 12).

Table 1 Post-seismic cross-fault deformation measured in the town of Yuli
Full size table
Fig. 12
figure 12

The repeated survey result in the heart of Yuli town area between 19 September 2022 and 10 January 2023. See text for detailed discussion

Full size image

Our result suggests that the left-lateral cross-fault displacement at the Yuli roundabout increased from approximately 7 cm to 13 cm between 19 September and 14 December, 2022 (Table 1). The post-seismic cross-fault movement is evident by the reappearance of surface cracks on the repaired concrete pavement at the roundabout, and the widening surface cracks in the surrounding parking spaces. The tiles of a nearby shop also displayed continued left-lateral movement on the floor within a few months after the mainshock, which required frequent repairs since the 2022 mainshock. In fact, local residents informed us that floor tile failures at the roundabout persisted even into late June of 2023, approximately 9 months after the earthquake.

Although accounts from local residents suggest a prolonged and persistent post-seismic cross-fault movement in the Yuli town center, our survey data indicate that the majority of post-seismic cross-fault movement occurred within 2 to 3 months after the mainshock. The survey results in December 2022 and January 2023 at the old campus of Yuli Elementary School showed similar amounts of left-lateral displacement, at 18 cm and 19 cm, respectively (Table 1; Figs. 12 and 13a–c). The other survey within the campus showed that left-lateral displacement increased from about 15 cm to 24 cm between September 2022 and December 2022, and remained similar in January 2023 (Fig. 13d–f). The left-lateral separation at the Yuli Hospital also exhibited no significant increase between our November and December surveys, suggesting an exponential-like decay of the post-seismic deformation 3 months after the mainshock. We also observed a similar pattern of post-seismic vertical deformation at Luntian when comparing photos taken in September 2022 and November 2022. However, we did not identify a suitable marker in the field to perform repeated measurements at the Luntian area.

Fig. 13
figure 13

The sequential photos taken in the old campus of Yuli elementary school, showing the change of fault separation from 19 September 2022 to January 2023. See text for detailed discussion. The locations of these two sites are shown in Fig. 12. The white box in the lower-right of each picture shows the date when the photo was taken in the field. The white box near the top of the image d to f shows our best estimation of the left-lateral sepration of the ditch when the picture was taken

Full size image

3 Discussion

3.1 Patterns of surface deformation along the Yuli fault within the Yuli Town

Our field observations of the 2022 surface rupture along the Yuli fault and its southward projection revealed a distinct surface rupture pattern, which varies significantly between the Yuli town center area and the adjacent low-elevation cultivated areas. Within the heart of Yuli Town, a 3-km-long section of the surface ruptures was characterized by a series of right-stepping en échelon Riedel shears alternating with pressure bulges or moletracks. The orientation of the observed Riedel shear varies from nearly N20W to approximately N8E between the Yuli roundabout and the Yuli High School (Fig. 3). Notably, near the Yuli roundabout, many observed en échelon Riedel shears exhibit orientations of up to N20W. The angle between these en échelon ground cracks and the general orientation of the Yuli fault (i.e. N15E to N20E) at this location exceeds 30 degrees. Although some of these observed cracks may belong to T-fractures in the Riedel shear system, most of the observed en échelon ground cracks’ angle are larger than the classic 15-degree angle between Riedel shears and the primary shear zone (e.g. Davis et al. 2000). However, the angle of these cracks still fall within the upper limits of the observed angle between Riedel shears and the respective faults [~ 30 degree, Petit (1987)].

The angle between the Riedel shear and the primary shear zone can be influenced by many factors, including the internal friction of the material involved. We suggest that the large angle observed between the Riedel shear and the primary fault zone found near the Yuli roundabout is a result of material heterogeneity at shallow depths within the Yuli town area, since we found the pattern of streets and building blocks had a great influence on the angle between the Riedel shears and the primary shear zone within the Yuli town center. Many of these observed rupture features within the town area appeared to follow the outline of building blocks, and the orientation of the en échelon ground cracks changed from as large as N20W to approximately N-S, or even N8E when the pattern of the street changed from N18W to N8E at the old campus of Yuli Elementary School (Fig. 3). Given that both areas are located on the surface of the same alluvial fan between the Zhou River and the Lele River and exhibit similar building and road pavement conditions, we do not expect the rheology between these two areas to be diffident at shallow depths. Therefore, we suggest that the arrangement of building blocks and material strength heterogeneity played a crucial role in controlling the distribution of fault surface ruptures within the town of Yuli.

3.2 The cross-fault displacement along the Yuli fault

Our documentation of cross-fault displacement of the 2022 surface rupture also suggests that the observed cross-fault deformations in Yuli were significantly smaller than far-field ground refererveys. Tung et al. (2024) used continuous and campaign GPS data to estimate the co-seismic ground deformation across the Yuli fault, and suggest the vertical motion across the Longitudinal Valley could reach up to approximately 1 m, whereas our observations indicate less than 10 cm of vertical cross-fault displacements in the Yuli town center area. For the area north and south of the Yuli Town, our field observations suggest vertical cross-fault displacement could exceed 40 cm across the surface rupture. Even though the significant vertical cross-fault displacement may indicate the change of the shallow fault geometry in favor of more reverse component at the surface, most of the vertical cross-fault displacement is still smaller than the far-field vertical displacement estimated by the GNSS data across the Longitudinal Valley from Widiatmoko et al. (2023) and Tung et al. (2024).

In term of the horizontal deformation during the 2022 earthquake sequence, the co-seismic southwestward motion recorded by stations at the base of the Central Range can reach approximately 90 cm (YUL1; Tung et al. 2024), and the fault-parallel motion recorded by stations east of the Yuli fault can reach around 43 cm along the northeastward direction (13R3; Tung et al. 2024). This suggests that the far-field left-lateral motion across the Yuli fault can exceed 130 cm, and only a small fraction of this left-lateral displacement (≤ 34 cm; mostly ≤ 20 cm) was observed right across the 2022 surface ruptures in the heart of Yuli Town. Similar discrepancies can also be found along the ruptures north of Yuli, where far-field displacements estimated from the relative motion of GNSS stations were much larger than the left-lateral cross-fault displacements. This significant difference strongly suggests that either the main ruptures stopped at a shallow depth (i.e. ≤ 1 km), or the co-seismic ground deformation is accommodated by off-fault deformations close to the surface along this segment of the fault. Similar observations have been reported in many surface rupturing earthquakes, including the Mw 6.4 Hualien earthquake in 2018 (Kuo et al. 2018), the Mw 6.0 South Napa earthquake in 2014 (Brooks et al. 2017), and the Mw 7.3 Landers earthquake (Xu et al. 2016).

In contrast to the limited left-lateral cross-fault displacements observed in the Yuli town center area, we observed significant cross-fault displacements at the Changliang site, less than 4 km south of the Yuli town center. Total station measurements indicate left-lateral cross-fault displacements ranging from 73 to 147 cm, which were comparable to nearby GNSS survey results from Tung et al. (2024). For instance, a GNSS station (ZCRS) recorded around 93 cm of co-seismic ground displacement about 2.5 km west of the ruptures, and the fault-parallel motion recorded by a station east of the Yuli fault also reached approximately 42 cm northwestward near the LVF scarp (1172; Tung et al. 2024). If the relative motion between these two stations represents the far-field left-lateral motion across the Yuli fault at Changliang, the difference between the cross-fault measurements and the far-field displacements at Changliang is much smaller than the difference observed at the Yuli area. This indicates that a significant portion (> 50% to nearly 100%) of the rupture may have propagated to the surface and accommodated by on-fault deformation at Changliang.

Significant discrepancies between cross-fault measurements and far-field displacements estimated from remote sensing or GNSS data are not uncommon, especially for strike-slip dominated events. For example, Milliner et al. (2015) re-examined displacements estimated from pixel-tracking results of aerial photos and compared them to field measurements for the Mw 7.3 Landers earthquake in 1999. They found that only a few peak offsets measured in the field were comparable to the estimates from remote sensing data in this complex rupture event. Gold et al. (2015) also noted a similar difference between near-field and far-field displacements in the Mw 7.7 Balochistan earthquake in 2013 using high-resolution satellite imagery, and found that approximately 85% of the rupture length exhibited far-field displacement values higher than on-fault displacements. For simpler and smaller rupture events, Wang et al. (2014) suggest that most field offset measurements were smaller than the estimates from ALOS pixel-tracking results for the Mw 6.8 Tarlay earthquake in 2011, where a single trace of fault surface rupture appeared along a linear valley at the westernmost Nam Ma fault. The cause of the discrepancy between cross-fault (or near-field) and far-field observations may be related to factors such as the width of the fault zone (Gold et al. 2015; Milliner et al. 2015), the property of shallow crustal materials near the rupture (Kaneko and Fialko 2011; Wang et al. 2014; Zinke et al. 2014), or even the frictional properties of the fault interface at different depths (Lapusta et al. 2000). For the 2022 earthquake event, since we did not observe a significant difference in the width of the surface rupture zone between Yuli and Changliang on the surface, and the shallow crustal materials at these two locations are likely to be similar young fluvial deposits sourced from the Central Range, which are less capable of localizing co-seismic deformation at shallow depths than consolidated bedrocks, it is difficult to conclude that the properties of shallow crustal materials within the Longitudinal Valley play an important role in controlling the changes of cross-fault displacement. Other factors, including the depth of basement rocks and the ground water conditions, may be the important factors that affect the difference in off-fault deformation between these two locations.

3.3 The Central Range fault and the Yuli fault

Even though the co-seismic deformations were not fully localized along the Yuli fault and its southward extension in the valley, our total station survey results suggest that the western side of the fault is the upthrown side in most of our surveyed locations. The vertical deformation across the Yuli fault associated with the 2022 earthquake, the mainshock focal mechanism, and the hypocenter distribution of subsequent aftershocks strongly suggest that the deeper part of the Yuli fault is linked to the west-dipping fault system and extends beneath the Central Range along its downdip direction; hence the Yuli fault is likely to be a branch of CRFs previously proposed by Shyu et al. (2006b). Based on geomorphic analysis, Shyu et al. (2006b) suggest that the main fault trace of the reverse dominated CRFs crops out at the base on the Central Range, and the Yuli fault is a left-lateral dominated sub-vertical fault east of the CRFs. The primary reason to suggest Yuli fault as a left-lateral dominated fault is the lack of geomorphic evidence for accumulated vertical deformation on the fault. The left-lateral dominated interpretation by Shyu et al. (2006b) is similar to what we observed along the Yuli fault during the 2022 earthquake in the field, except for the minor but persisted west-side uplift revealed by our total station profiles. The west-side-up character along the Yuli fault also matches the far-field ground deformation pattern in Tung et al. (2024), as well as fault slip inversion results from Lee et al. (2023) and Tang et al. (2023). The difference between cross-fault measurements and these inversion results suggests that either the majority of fault slip stopped at a shallow depth, or the slip was accommodated by significant off-fault deformations during the mainshock. The former explanation requires the shallow crustal material to be less elastic to avoid unrealistically high stress near the buried fault tip (Nevitt et al. 2020), and the latter explanation requires less cohesive materials near the surface (Kaneko and Fialko 2011). In either case, the lack of localized vertical deformation would prohibit the Yuli fault to form a prominent geomorphic scarp at the surface. Even if a scarp had formed along the Yuli fault, it could be easily buried by the large sediment fluxes from the Central Range (Fuller et al. 2003; Hovius et al. 2000) or the Coastal Range, if the long-term vertical deformation rate of the fault is smaller than the sedimentation rate in the valley.

3.4 Shallow rapid after-slip along the Yuli fault trace

In addition to the co-seismic deformations we documented, our cross-fault measurements following the earthquake indicate a noteworthy phenomenon that the amount of left-lateral displacements increased over time in the Yuli town center area, at least during the first 3 months after the mainshock. It is noteworthy that those post-earthquake displacements we observed in the Yuli town center area is likely associated with the slip at very shallow part of the fault plane, which is different to the modelled post-seismic fault slip on the CRF using GNSS post-seismic displacements during the first 34 days after the earthquake (Tang et al. 2023). The result derived from GNSS data suggests the majority of posit-seismic fault slip occurred on the down-dip side of the CRF, with insignificant post-seismic slip on the most shallow part of the ruptured fault. Detailed aftershock catalog from Sun et al. (2024) also showed very few aftershocks occurred near the area of Yuli, suggesting our observed after-slip is not associated to ground deformation caused by local aftershocks. Hence, the observed after-slip along the 2022 surface rupture is likely aseismic. Although aseismic creeping and rapid after-slip have long been reported along the LVF at the central and southern part of the valley (e.g. Cheng et al. 2009; Hsu and Bürgmann 2006; Hsu et al. 2009a; Lee et al. 2001, 2003; Peyret et al. 2011; Thomas et al. 2014), even after the 2022 earthquake (Tang et al. 2023), there has been no report for shallow after-slip or aseismic creeping on the Yuli fault before the 2022 earthquake.

One particular site that has been studied for aseismic creeping on the Yuli fault is the old campus of the Yuli Elementary School, where observations in the past half century showed insignificant fault creep during the inter-seismic period. For instance, Bonilla (1975) examined the damages of the school buildings constructed after the 1951 earthquake and suggests the amount of fault creep was not significant in this campus, as the classrooms built in 1958 showed less than 1 cm of deformation in his visit of 1973. Yu and Liu (1989) surveyed a geodetic network deployed in the same campus from 1980 to 1987, and found no significant vertical or horizontal motion associated to aseismic creeping across the Yuli fault. More recently, Chen et al. (2021) analyzed repeating leveling surveys between 2004 and 2018 and found that most of the inter-seismic vertical deformations occurred along the LVF at the Yuli bridge (see Fig. 2 for location), with no clear motion across the Yuli fault. All these lines of evidence suggest the shallow portion of the Yuli fault is locked between the 1951 and the 2022 earthquakes, and the cross-fault after-slip only appeared after the 2022 earthquake.

Similar phenomena of inter-seismic locking with shallow post-seismic slip on surface ruptures have also been reported in many other earthquake events, such as the Mw 6.0 South Napa earthquake in 2014. Delong et al. (2015) used a laser scanner to monitor post-seismic deformations along a short section of the 2014 earthquake surface rupture (i.e., the West Napa fault) and found the right-lateral displacement increased from 0.22 to 0.29 m at 0–2.5 days to 0.33–0.42 m about 2 months after the earthquake. Lienkaemper et al. (2016) suggest the post-seismic slip (after-slip) of the West Napa fault may last much longer than 1 year along its southern ruptured section, but there was no evidence to suggest the West Napa fault creeps inter-seismically before the earthquake. The 1999 Izmit earthquake on the North Anatolian fault that was deeply locked prior to the earthquake also showed shallow after-slip even 19 years after the earthquake (Aslan et al. 2019; Cakir et al. 2012). On the other hand, after-slip phenomena have also been observed in faults that show inter-seismic aseismic creeping at the surface. These examples include the 1966 Parkfield earthquake on a creeping segment of the San Andreas fault (Smith and Wyss 1968), and the 1987 Superstition Hills earthquake on the Superstition Hills fault that has been frequently induced by earthquakes nearby and experienced aseismic creeping at the surface (Hudnut and Sieh 1989). All these examples suggest after-slip could occur on the shallow part of faults no matter if the fault has inter-seismic creeping or not.

The cross-fault after-slip that we observed along the Yuli fault is likely to be very shallow, and may not be linked to post-seismic fault slip at depth, since GNSS analysis showed insignificant post-seismic slip on the up-dip side of the fault, and the deep post-seismic motion occurred well below the co-seismic ruptured patch (Tang et al. 2023). Considering the large differences between the far-field and the near-field co-seismic displacements, the significant shallow after-slip on the ruptured fault, the along-strike variation of the cross-fault displacements, and the lack of secondary faults along the Yuli fault, we suggest the rapid and significant shallow after-slip on the fault is controlled by the elasto-plastic properties of shallow crustal materials around the fault.

The shallow cross-fault after-slip also poses long-lasting hazards and concerns to buildings built on top of the fault. Due to the small cross-fault displacement of the coseismic surface ruptures, many buildings on top of the fault were only damaged but not completely destroyed by the earthquake. However, the subsequent after-slip may continue to produce tearing and damage to the structure of the buildings for months or even years after the earthquake. Most earthquake recovery activities may have not considered such effects, which would also cause secondary hazards such as seepage erosion from the leaking pipelines after the earthquake. Hence, it is important to take the cross-fault after-slip into account when reconstructing structures across the surface rupture, to prevent frequent damages and repairs after the earthquake.

4 Conclusions

In summary, our field mapping along the surface ruptures during the September 2022 Guanshan-Chihshang earthquake revealed the Mw 6.8 earthquake resulted from ruptures of the Yuli fault, which is likely a high-angle branch of the Central Range fault system, extending from below the Central Range to the Longitudinal Valley at the surface. Field observations confirmed that co-seismic fault slip was dominated by left-lateral strike-slip, with minor but persisted west-side uplift along mapped Yuli fault traces. The amount of left-lateral cross-fault displacements measured in the field were mostly much smaller than far-field ground displacements estimated by GNSS analysis, indicating either the occurrence of strong off-fault deformation or the main rupture did not reach the surface. In addition to the co-seismic offset measured soon after the earthquake, our repeated cross-fault survey results in the town of Yuli indicate large and rapid cross-fault after-slip along the previously locked Yuli fault. The large differences between the far-field and near-field co-seismic displacements along the ruptures, the shallow and rapid after-slip, the along-strike variation of cross-fault displacements, and the lack of secondary ruptures along the Yuli fault all suggest that the elasto-plastic properties of shallow crustal materials around the fault play an important role on the fault rupture behaviors during and after the earthquake event.