Early Post-Seismic Deformation Revealed After the Wushi (China) Earthquake (Mw = 7.1) Occurred on 22 January 2024

Abstract

The Mw = 7.1 Wushi earthquake is the second-largest digitally recorded earthquake in the Tianshan seismic zone and provides an opportunity to explore the structural characteristics of the Tianshan seismic zone. In this study, we calculated the early (11-month) post-seismic deformation of the Wushi earthquake using Sentine-1 ascending and descending InSAR time series data. We found that the 11-month post-seismic deformation was dominated by afterslip along the up-dip continuation of the coseismic fault. The seismic moment released by the afterslip was Mw = 6.20, with 6.5% of that released by the mainshock. Moreover, we explored four slip models for the Mw = 5.7 aftershock that occurred on 29 January and found that this event primarily ruptured a thrust fault. However, determining the thrust fault type based on the current field investigations and InSAR data remains difficult. Finally, the Coulomb stress changes indicated that both the afterslip and aftershock were promoted by the Wushi earthquake.

1. Introduction

The Mw = 7.1 Wushi earthquake occurred in Wushi, Aksu Region, Xinjiang, China, at 18:09 (GMT time) on 22 January 2024. Records from the Global Centroid Moment Tensor Catalog (GCMT) showed that the epicenter was located at 41.19°N, 78.56°E and had a depth of 16 km. The Wushi earthquake occurred approximately 6 km away from the Maidan Fault [1], which belongs to the complex fold-and-thrust belts of the south Tianshan Mountains. It is the second-largest digitally recorded earthquake to have occurred in the Tianshan seismic zone [2] after the 1992 Suusamyr Ms = 7.3 earthquake. After the mainshock, the largest aftershock reached a magnitude of Mw = 5.7, which occurred on 29 January, with an epicenter of 41.12°N, 78.63°E.

Three kinds of processes can result in post-seismic deformation after an earthquake: (i) afterslip [3,4,5], (ii) viscoelastic relaxation of the lower crust or/and upper mantle [6,7], and (iii) poroelastic rebound [8,9]. Post-seismic deformation offers researchers a useful tool to explore the fault properties (velocity-strengthening/weakening frictional behavior along the fault, etc.) and the lithosphere rheological properties. As the Wushi earthquake ruptured a blind thrust fault [10], it provides an opportunity not only to probe the seismogenic fault but also to explore the coseismic fault properties and/or rheological properties of the lithosphere in the Tianshan seismic zone. In this study, while we do not consider the poroelastic rebound mechanism, we do explore the afterslip/viscoelastic relaxation mechanism for the early post-seismic deformation of the Wushi earthquake.

Interferometric synthetic aperture radar (InSAR) has been a crucial tool for calculating surface deformation [11,12,13]. In this study, we used 11-month Sentinel-1 data (ascending and descending tracks) to obtain the early post-seismic deformation after the Wushi earthquake. The afterslip characteristics of the Wushi earthquake were investigated and the coseismic slip model of the aftershock that occurred on 29 January was explored; then, a stress-triggering relationship between the mainshock and afterslip/aftershock was demonstrated by calculating the Coulomb stress change.

2. Geologic Background

The Tianshan Mountains or the Tianshan tectonic belt (TTB), with a length of 2500 km from east to west [14], separates the Tarim and Junggar basins and can be divided into north, middle, and south Tianshan [15]. The TTB is one of the largest active inland intracontinental orogenic belts in the world [16] due to extrusion, crustal shortening, and tectonic uplift [17].

The TTB is characterized by frequent and uneven seismic activity [14] (Figure 1a). Paleoseismological studies [2] indicate that since 1716, 17 earthquake events with magnitudes larger than 7 have occurred, including 4 earthquake events with magnitudes larger than 8. Moreover, 11 earthquakes with magnitudes larger than 7 have occurred since the 1900s [15]. The strongest earthquake was the 1902 Artux earthquake, with a magnitude of $8{\displaystyle \frac{1}{4}}$, but this was not recorded with digital instruments [14,15]. Therefore, the 1992 Ms = 7.3 Suusamyr earthquake and the 2024 Mw = 7.1 Wushi earthquake are the largest and second largest digitally recorded earthquakes to have occurred in the TTB, respectively [2]. The Wushi earthquake occurred approximately 6 km away [1] from the Maidan Fault (MDF), which lacks systematic studies due to its low seismic activity (Figure 1b) and sparse geodetic data [18]. The MDF has weak seismic activity, especially in the western segment, and the latest paloearthquake on this fault occurred in ~40–480 AD with a magnitude of Mw = 7.5 [19].

The TTB reflects the collision between the Tarim and the Siberia cratons from the late Paleozoic [20], but it was reactivated to accommodate the convergence between the Indian and Eurasian plates by crustal shortening from the early Cenozoic [15,20]. GPS observations indicate that the crustal shortening rate at the TTB western segment is ~20 mm/a which is close to half the rate of subduction between the Indian and Eurasian plates, while the shortening rate at the eastern segment is only ~4 mm/a [15,21,22]. The southern TTB, with an ENE trending orientation, is experiencing the northward subducts of the Tarim Basin and the southward subducts of the Kazakh Platform [23]. A group of ENE-trending active faults, which have not been well studied, have formed in the interior mountain range, such as the Toshgan Fault (TSF), the Maidan Fault (MDF), the Kukesale Fault (KKSF), and the Dashixia Fault (DSXF). The MDF was active during the Holocene and late Pleistocene [18,23,24] and serves as a boundary between the south Tianshan Mountains and the Tarim Basin. The MDF is a sinistral reverse fault with a length of more than 400 km [18]. The shortening rate and the sinistral slip rate of the MDF are about 1.19 ± 0.25 mm/a and 1.56 ± 0.64 mm/a, respectively [23]. Despite high activity, no earthquake with a magnitude larger than 6 has occurred within a 50 km radius of the Wushi earthquake since the 20th century according to historical records [12,25].

Therefore, when the Wushi earthquake occurred, researchers [26,27] analyzed the stress evolution surrounding the MDF by calculating the Coulomb stress changes caused by historically strong earthquakes and the interseismic tectonic loading. The historical earthquakes considered included the 1716 M = 8.0 Zhaosu earthquake, the 1889 M = 8.3 Chilik earthquake, the 1902 M = 8.2 Artux earthquake, the 1911 M = 8.0 Kemin earthquake, the 1978 M = 6.9 Alma earthquake, the 1987 M = 6.1 Wushi earthquake, the 1991 M = 6.0 Kalpin earthquake, and the 2020 Mw = 6.0 Jiashi earthquake. The results showed that the total Coulomb stress change (from 1716 to 2024) driven by the interseismic tectonic loading was 385 kPa, and the total Coulomb stress change induced by the historical earthquakes on the epicenter region of the 2024 Mw = 7.1 Wushi earthquake was −37 KPa. Thus, the results indicate that the historical earthquakes may have delayed the occurrence of the 2024 Mw = 7.1 Wushi event, and it is more plausible that sustained tectonic loading is the main factor for the occurrence of the 2024 Mw = 7.1 Wushi earthquake.

Several studies [1,10,28,29,30,31] using geodetic and/or seismic data concluded that the Wushi earthquake ruptured a blind thrust fault within the thick-skinned tectonics of the South Tian Shan. According to Yu’s study, the coseismic fault (Figure 1b) has a northwest-dipping angle of 62°, a strike of approximately 229°, and a slip rake of 49.9° (with a 94% highest density interval (HDI) of [42.09°, 58.65°]) [1]. Both the InSAR coseismic deformation [1,10] and field investigations [2] revealed no significant surface rupture for the Wushi earthquake, which is consistent with the coseismic slip models, indicating that the slip concentrated at depths ranging from ~4 to 26 km along the dip. Therefore, the potential rupture of the causative fault did not reach the surface, which is the main reason for the lack of a significant surface rupture.

3. InSAR Data and Processing

We used 28 ascending and 22 descending Sentinel-1 images from Paths 56 and 136, respectively. The ascending data ranged from 26 January 2024 to 15 December 2024, and the descending data ranged from 25 February 2024 to 9 December 2024. As Path 136 did not have an image pair covering 29 January, we used two descending images (25 January 2024 and 6 February 2024) from Path 34 to calculate the coseismic deformation of the aftershock.

We used the Interferometric synthetic aperture radar Scientific Computing Environment (ISCE) stack processor [32] to generate for each acquisition five nearest-neighbor interferograms with 7 looks in the azimuth direction and 21 looks in the range direction [4,33]. We used 3 arc-second SRTM [34] DEM data and precise orbit data to remove the phase error induced by the earth’s curvature and topography from the interferograms. Additionally, the statistical-cost network-flow algorithm (SNAPHU) [35] was used to unwrap the phase of the interferograms. For time series processing, we used the Miami InSAR time series software in Python (MintPy) [36]. We used the ECMWF’s ERA5 global atmospheric model to correct tropospheric delays, and we also removed the topographic residuals. A threshold for temporal coherence of 0.7 [36] was used to exclude the pixels experiencing decorrelation caused by environmental disturbances, such as seasonal snow or drought. Then, the line-of-sight (LOS) velocity was estimated based on the corrected displacement time series.

4. Modeling Approach

Considering the short post-seismic period (almost 11 months) in our study, we considered the afterslip mechanism to be the primary post-seismic process. We used the uniform rectangular dislocations in a layered half-space to estimate the aftershock and afterslip model parameters, which were the fault position, length, width, depth, dip, strike, rake, and slip. The depths and properties (p-wave velocity, s-wave velocity, etc.) of the layered half-space were obtained from the CRUST1.0 database [37], and the Green functions were calculated using the PSGRN/PSCMP software (https://github.com/pyrocko/fomosto-psgrn-pscmp, accessed on 6 April 2025) [38].

We used the quadtree down-sampling method to sample the deformation data and conducted the nonlinear inversion by calculating a misfit function ${\chi }^2$ defined as follows:

$${\chi }^2=({\mathit{d}}_{obs}-{\mathit{d}}_{sim}){\mathit{C}}^{-1}{({\mathit{d}}_{obs}-{\mathit{d}}_{sim})}^{-1},$$

(1)

where ${\mathit{d}}_{obs}$ and ${\mathit{d}}_{sim}$ are the observed and simulated deformation, respectively, and $\mathit{C}$ is the covariance matrix, calculated using the semi-variogram method to characterize the errors in InSAR data. We also used the Root Mean Square Error ($RMSE= \sqrt{\left({\mathit{d}}_{obs}-{\mathit{d}}_{sim}\right){\left({\mathit{d}}_{obs}-{\mathit{d}}_{sim}\right)}^T/n}$, where n is the number of samples) to quantify the model fit.

We used the Sequential Monte Carlo sampling algorithm implemented in the Bayesian Earthquake Analysis Tool software (BEAT v2.0.4) [39,40] to obtain the best-fitting model parameters. The 1D and 2D posterior probability density distributions and the 94% highest density interval (HDI) were used to show the uncertainties of the estimated model parameters.

5. Results

5.1. InSAR Post-Seismic Deformation

The LOS displacements, with positive values meaning movement towards the sensors and negative values meaning movement away from the sensor, from ascending Path 56 and descending Path 136 after the mainshock until December 2024 are shown in Figure 2a,b. The performance of the tropospheric delay correction using the EAR5 model is demonstrated in the Supplementary Files (Section S1), and the analysis indicates that the tropospheric delay was removed efficiently. Two lobes show up to 20 cm and 10 cm of LOS increase in the ascending data, and the descending data show two lobes of up to 10 cm and 7 cm of LOS increase. The reason for the significant deformation in the ascending data is that the time period for the ascending data (26 January 2024 to 15 December 2024) encompasses 29 January (when the Mw = 5.7 aftershock occurred), while the time period for the descending data (25 February 2024 to 9 December 2024) does not cover 29 January. In order to exclude the influence of the aftershock on the afterslip mechanism analysis, we divided the time period of ascending data into two periods, which were 26 January 2024 to 7 February 2024 (for the aftershock analysis) and 7 February 2024 to 15 December 2024 (for the afterslip analysis). Additionally, the time period for the descending data (Path 136) did not change (

Table 1. Sentinel-1 data and the time period used in this study.

Track	Time Period	Usage
AT56	26 January 2024–7 February 2024	Aftershock
	7 February 2024–15 December 2024	Afterslip
DT136	25 February 2024–9 December 2024	Afterslip
DT34	25 January 2024–6 February 2024	Aftershock

).

After excluding the aftershock’s influence, both the ascending and descending data exhibited two lobes of up to 10 cm and 7 cm of LOS increase (Figure 2e,f). Furthermore, two points were selected at the maximum lobes for the ascending and descending data, respectively. Both displacement time series show nearly linear growing patterns (Figure 2g).

As the descending Path 136 did not have an image pair covering 9 January, descending Path 34 (Table 1) was chosen to calculate the coseismic deformation of aftershocks that occurred on 29 January (Figure 2d). The ascending data (Figure 2c) show lobes of both LOS increase and decrease (40 cm and 13 cm, respectively), as do the descending data (38 cm and 16 cm, respectively) (Figure 2d).

5.2. Afterslip

The best-fitting uniform dislocation was estimated based on a layered half-space model. Unlike other earthquakes [4,11,41], two separate lobes occurred in the post-seismic deformation field. Therefore, we used three types of models to simulate this observation. In the first model, we assumed that the afterslip occurred on one fault segment (one-fault model). In the second model, we assumed that afterslip occurred on two adjacent fault segments (two-fault model), and in the third model, we assumed that afterslip occurred on three adjacent fault segments (three-fault model).

Figure 3 shows the observation, simulation, and residual results based on the best-fitting parameters for the three models.

Table 2. Fault parameters for the afterslip three-fault model.

	Length (km)	Width (km)	Depth (km)	Dip (°)	Strike (°)	Rake (°)	Slip (m)	Lat * (°)	Lon * (°)	Mw	RMSE (cm)
	Segment1 (South segment)									6.20	0.57
Min	1	1	0	0	180	0	0	41.133	78.561
Max	15	15	5	90	270	180	1
Optimum	9.38	12.81	4.33	48.8	228.3	41.1	0.18
94% HDI	3.52–14.99	9.56–15.00	3.18–5.00	40.0–57.7	211.5–247.6	19.3–63.9	0.06–0.35
	Segment2 (Middle segment)
Min	1	1	0	0	180	0	0	41.161	78.648
Max	15	15	5	90	270	180	1
Optimum	7.73	6.32	2.85	48.7	230.6	75.6	0.30
94% HDI	6.63–8.99	3.64–8.86	2.33–3.43	41.9–55.2	224.9–236.5	63.6–88.3	0.21–0.41
	Segment3 (North segment)
Min	1	1	0	0	180	0	0	41.201	78.76
Max	15	15	5	90	270	180	1
Optimum	10.02	7.74	2.46	53.7	255.7	68.3	0.21
94% HDI	4.90–12.51	4.25–10.89	1.80–3.20	46.3–59.9	248.7–262.9	56.4–80.5	0.12–0.35

* Lat/Lon: start point coordinate of the fault upper edge.

and Tables S1 and S2 show the best-fitting parameters, the 94% HDI uncertainty for estimated parameters, and the corresponding RMSEs, respectively. The 1D/2D marginal posterior probability densities of the estimated parameters are shown in the Supplementary Materials (Figures S2–S7). The RMSEs are 0.88 cm, 0.72 cm, and 0.57 cm for the one, two and three-fault models, respectively. Additionally, the qualitative comparisons (Figure 3) indicate that the three-fault model is the best model because the one-fault model cannot simulate the two lobes well and the two-fault model cannot explain the deformation in the south-west corner.

The best-fitting solution (Table 2 and Figure 3) shows that the post-seismic deformation 11 months after the mainshock is well explained by the three-fault model. Fault segments were located at the up-dip continuation of the coseismic fault (Figure 4), which indicates that afterslip primarily occurred on the up-dip continuation of the coseismic fault. The seismic moment released by the afterslip was $2.27\times {10}^{18} \mathrm{N}\mathrm{m}$ (Mw = 6.20), which was about 6.5% of that released by the coseismic slip.

6. Discussion

6.1. Early Afterslip of Wushi Earthquake

The afterslip fault segments of the three models are plotted with the coseismic fault segment from other studies and the MDF in Figure 4. All three models provide reasonable geological information. For the one-fault model, the afterslip occurred directly on the up-dip continuation of the coseismic fault. For the two-fault model, the two segments of afterslip were both located up-dip with respect to the mainshock hypocenter. Because basic information for the faults in south Tianshan was lacking [18], the north segment could be a splay fault or the up-dip continuation of the coseismic fault and the south segment is the up-dip continuation of the coseismic fault. For the three-fault model, the spatial position of the north and middle segments is almost the same as that of the two-fault model. Additionally, the south segment was located at the southwest extension of the coseismic fault. Therefore, considering both are reasonable in a geological sense, we used the RMSEs to determine that the three-fault model is the best afterslip model.

The afterslip occurred along the up-dip continuation of the coseismic fault and released a moment of about 6.5% of that released by the coseismic slip. The position of the afterslip segments indicates a velocity-strengthening frictional zone above a velocity-weakening zone along the fault. This up-dip afterslip behavior was also observed for many earthquakes, such as the 2003 Zemmouri [11] and the 2017 Kermanshah earthquake [4]. Compared with other earthquakes (10%, 12%, and 56% for the Kermannshah [4], Gorka [42], and Pakistan [41] earthquakes, respectively), the amount of afterslip moment release (6.5%) was relatively small.

Because an afterslip is promoted by high temperatures, elevated pore-fluid pressures (indicated by $V_p/V_s$ ratio), and clay-rich sediments/faults [3], we collect data on the heat flow (45–55$\mathrm{m}\mathrm{W}/{\mathrm{m}}^2$) [43] and $V_p/V_s$ ratio (1.85 ± 0.04) [44] in the south Tianshan region. The low heat flow and pore-fluid pressure may have induced the relatively short duration and small moment of the afterslip. However, this study cannot determine the end time of the afterslip as the time period for the post-seismic deformation was only 11 months. Thus, longer post-seismic observations should be made in the future to investigate the exact duration and released seismic moment of afterslip.

Besides the afterslip mechanism, we also explored whether the 11-month post-seismic deformation could be explained by the viscoelastic relaxation mechanism (Section S4 in Supplementary Files). The results showed that the viscoelastic relaxation mechanism is not the main factor for the 11-month post-seismic displacement. This phenomenon is consistent with the different spatial and temporal behaviors of afterslip and viscoelastic relaxation. Afterslip mainly occurs in the near-field while viscoelastic relaxation mainly occurs in the far-field [4], and afterslip always lasts for a few months to years while viscoelastic relaxation lasts for decades. Therefore, years or even decades of post-seismic deformation observation data are needed in the future to explore the crustal rheological characteristics of the south TTB region.

6.2. Coulomb Stress Triggering Relationship

What is special about the Wushi earthquake is that the mainshock did not rupture to the surface [10]. The modeling results showed that afterslip primarily occurred at a shallow depth. Therefore, we investigated the triggering relationship between the mainshock and afterslip. Coulomb stress change has been widely used to explore stress triggering and perturbation in many studies [1,5,10,45]. Therefore, we calculated the change in Coulomb failure stress [46] imparted by the mainshock in the inferred direction of slip along the inferred fault segments in an elastic model. In addition, we divided the afterslip fault segments into 2 × 2 patches to increase the spatial resolution of stress changes.

$$∆CFS=∆\tau +{\mu }^{\prime }∆{\sigma }_n$$

(2)

where the $∆\tau$ is the change in shear stress, $∆{\sigma }_n$ is the change in normal stress, and ${\mu }^{\prime }$ is the effective friction.

Figure 5 shows that 100%, 100%, and 75% of the north, middle, and south segments received positive ∆CFSs, respectively. The average ∆CFSs are 10.08 bar, 6.27 bar, and 5.77 bar for the north, middle, and south segments, respectively. Additionally, the standard deviations of ∆CFS are 13.98 bar, 5.88 bar, and 7.82 bar for the north, middle, and south segments, respectively.

6.3. Aftershock on 29 January 2024

Several researchers [1,10,29,30] investigated the coseismic model for the aftershock that occurred on 29 January, but the conclusions are not consistent with each other. Based on their studies, the aftershock fault can be divided into two categories, a back-thrust [10,30] fault (SW-dipping) and a splay thrust fault (NE-dipping) [29]. Moreover, the field investigations [2] conducted between 24 January 2024 and 30 January 2024 found a surface rupture trace, with a length of 2 km and a trend toward N60°E, which may have been induced by the aftershock, but the slip could be along the back-thrust fault or the splay thrust fault. Furthermore, these studies also have disagreements on the number of slip faults. Only one thrust event occurred on 29 January 2024 in the Global CMT Catalog, while Yu [1] and Qiu [10] argued that the coseismic deformation field (Figure 2c,d) was induced by the rupture of two faults. Thus, they used the two-fault model with the primary slip that occurred on the thrust fault to simulate their observations. However, Famiglietti [30] and Zheng [29] used a one-fault model.

Thus, in this section, we adopt four models to explore whether distinguishing the most plausible model based on the InSAR ascending and descending data is possible. The four models are the SW-dipping one-fault model, NE-dipping one-fault model, SW-dipping two-fault model, and NE-dipping two-fault model. Regarding the SW-dipping one-fault model, we fixed its strike, dip, width, rake, length, and position according to the surface rupture trace and Famiglietti’s results [2,30]. Regarding the SW-dipping two-fault model, we fixed the geometry parameters according to Qiu’s work [10].

After multiple tests, we ultimately obtained a set of plausible model parameters that can explain the main coseismic observation characteristics (Figure 6), and the RMSEs for the four best-fitting models were also calculated (

Table 3. Best fitting parameters of the four models for aftershock.

	Length (km)	Width (km)	Depth (km)	Dip (°)	Strike (°)	Rake (°)	Slip (m)	Lat * (°)	Lon * (°)	RMSE (cm)
SW-dipping one-fault model
Optimum	9	5	1.3	41	70	90	0.43	41.144	78.565	4.82
NE-dipping one-fault model
Optimum	4.09	5.53	2.19	61.35	240.83	116.08	1.16	41.140	78.628	4.91
NE-dipping two-fault model
Optimum	5.06	5.80	1.58	60.3	232.3	107.5	0.67	41.145	78.634	4.83
	6.36	1.14	1.22	58.3	2.2	42.8	0.68	41.112	78.634
SW-dipping two-fault model
Optimum	5	8	0.12	60	61	108	6.5	41.140	78.588	6.11
	4	3.5	0.11	76	191	178	0.38	41.160	78.662

Note: Bold parameters are the fixed parameters based on other studies; * Lat/Lon: start point coordinate of the fault upper edge.

). All models indicated that the coseismic deformation was dominated by the slip along a thrust fault (named the F1 fault), and for the two-fault models, the optimal solution also reveals that a relatively small slip occurred near a south–north striking fault (named F2 fault). The aftershock slip occurred at a shallow depth which propagated to the surface and caused a surface rupture. However, none of the best-fitting models could explain the observation very well, and the RMSEs for these four models are similar. Thus, determining the best model based on the simulation-observation fitness is relatively difficult. Additionally, although a 2 km-length surface rupture was caused by the aftershock, field investigations did not find more clues to distinguish between the back-thrust fault and the splay thrust fault [2]. Moreover, ∆CFSs imparted by the mainshock were calculated for each best-fitting model (Figure 7), and the results showed that the F1 fault in the four best-fitting models received positive ∆CFSs. Therefore, no exact clues have helped us to determine the dipping direction of the F1 fault, which means that the F1 fault could be a back-thrust of the mainshock fault or a splay fault (Figure 8).

Furthermore, because the RMSEs between the one-fault and two-fault models were similar, the optimal one could not be judged based only on the information provided by the InSAR data. Moreover, no information from field investigations have supported the two-fault model assumption. Therefore, more data sources, such as GPS data, seismic profile data, geological trenching, or borehole data, are needed to provide more information to identify the dipping direction of the F1 fault, the existence of an F2 fault, and the primary-secondary relationship between F1 and F2 fault. Finally, the coseismic deformation of aftershock containing one-week post-seismic signals should be acknowledged, which could have a small influence on the inverted model. However, such an acknowledge makes no difference to the main conclusions.

7. Conclusions

We analyzed the early post-seismic deformation of the Mw = 7.1 Wushi earthquake from February 2024 to December 2024 using Sentinel-1 ascending and descending data. Based on the 11-month post-seismic deformation, the research findings are as follows:

The 11-month post-seismic deformation was induced by the afterslip along the up-dip continuation of the coseismic fault. The seismic moment released by the afterslip was $2.27\times {10}^{18} \mathrm{N}\mathrm{m}$, corresponding to a Mw = 6.2 earthquake.

The thrust fault of the Mw = 5.7 aftershock that occurred on 29 January could be a back-thrust fault or a splay fault (the branch fault of MDF developed at the front end). However, distinguishing between these two possibilities based on the current field investigation results and InSAR data is difficult. More data, such as GPS or borehole data, are urgently needed to not only determine the fault type but also determine whether another small fault also ruptured.

The Coulomb stress change in the direction of inferred slip along the afterslip fault segments imparted by the coseismic slip distribution shows that 100%, 100%, and 75% of the north, middle, and south segment received positive ∆CFSs, respectively, and the average ∆CFSs are 10.08 bar, 6.27 bar, and 4.48 bar for the north, middle, and south segments. The ∆CFS results revealed that the mainshock encouraged the afterslip.