Stepovers have been widely suggested to be important structural boundaries that control earthquake rupture extent and therefore the size of earthquakes. Previous studies suggested that ~4-5 km wide stepovers are likely to arrest fault rupture. However, recent earthquake cases show that even much wider stepovers (e.g., ≥7-8 km wide) sometimes may not effectively impede seismic rupture propagation, which requires us to further explore cascading rupture mechanism of large earthquakes at wider stepovers. Here, we constrained slip rates and paleoseismic earthquakes of two fault sections that bound a constraining stepover with width of approximately 7-8 km along the southern segment of the Daliangshan fault along the southeastern margin of the Tibetan Plateau. Multiple landform offset and radiocarbon dating results constrained that the two fault sections show a moderate slip rate at approximately 5 mm/yr. Moreover, three or four paleoseismic events Z through W, in 1489 AD, 620-515 BC, 4475-3700 BC, and 6265-4510 BC, were revealed on the Jiaojihe fault section. Based on the aforementioned results, we suggest that the most recent seismic event might exhibit a jump over the restraining stepover, whereas ruptures of the older events might be arrested by the stepover. Furthermore, we suggest that moderate slip-rate faults might have similar potential with that of high slip-rate faults to rupture through wider stepovers, which increases us in understanding the generation of cascading ruptures on strike-slip faults and is helpful for evaluating seismic hazards.
The magnitudes and recurrence intervals of large earthquakes have important implications for understanding fault-zone mechanics and seismic hazard analyses . Geometrical complexities such as stepovers can significantly affect the faulting process of strike-slip faults and govern the earthquake rupture extent and therefore the size of earthquakes. Specifically, field mapping has suggested that stepovers wider than 4-5 km are likely endpoints for coseismic ruptures and may be barriers that separate faults into different fault segments [2, 3], which was also evidenced by numerical simulations of dynamic ruptures [4, 5].
However, increasing earthquake cases show that even much wider stepovers (e.g., ≥7-8 km wide) sometimes may not effectively arrest seismic rupture propagation, such as the 2016 Mw 7.8 Kaikoura earthquake ruptured at least 12 single fault segments (including stepovers of 15-20 km wide), which defy many conventional assumptions of the width of stepovers in the aforementioned studies and might underestimate evaluation of seismic hazard models . Therefore, revealing seismic rupture behaviour at wider stepovers has great significance in understanding cascading rupture mechanism of large earthquakes. One issue is whether wider stepovers are always jumped over during each seismic rupture? In other words, whether wider stepovers temporally control earthquake ruptures during multicycle earthquakes? Another issue is that recent earthquakes that ruptured wider stepovers appear to be primarily related to high slip-rate faults. For example, the Marlborough fault system associated with the 2016 Mw 7.8 Kaikoura earthquake shows slip rates varying from 18 mm to 25 mm/yr . Similarly, the 2001 Mw 7.8 Kunlunshan earthquake  that ruptured through a 10 km wide stepover shows slip rate at 10.0 ± 1.5 mm/yr . Whether moderate (e.g., 3-7 mm/yr) strike-slip faults can rupture wider stepovers? To answer the aforementioned scientific issues, geologic evidence of more earthquake cases about seismic ruptures through wider stepovers is needed.
Here, we constrained slip rates and paleoseismic earthquakes on the Butuo and Jiaojihe fault sections that bound a constraining stepover with width of approximately 7-8 km along the southern segment of the Daliangshan Fault (DF). We suggest that the most recent seismic event might exhibit a jump over the restraining stepover, whereas ruptures of the older events might be arrested by the stepover. Moreover, the two fault sections that showed a moderate slip rate (~5 mm/yr) provide insights that seismic rupture behaviour at wider stepovers might be temporally variable, which increases us in understanding the generation of cascading ruptures on strike-slip faults and is helpful for evaluating seismic hazards.
2. Geologic Setting
The DF fault is a left-lateral strike-slip fault approximately 250 km long that connects to the Ganzi-Yushu-Xianshuihe fault (GYXF) and Anninghe fault (AF) in the north and terminates at the Xiaojiang fault (XF) and Zemuhe fault (ZF) in the south (Figure 1(a)). The above-mentioned faults collectively accommodate crustal deformation due to the southeastward extrusion of the Tibetan Plateau . The DF fault has a complex fault geometry characterized by fault stepovers and can be divided into the northern and southern segments consisting of six fault sections (Figure 1) [11, 12]. Moreover, paleoseismic results suggest that the DF fault has been active in the Holocene with surface-rupturing earthquakes [11–13]. The late Quaternary left-lateral slip rate of the DF fault was determined to be approximately 3-4 mm/yr by measuring offsets of several alluvial fans and terraces along the fault and mainly using optical-stimulated luminescence dating .
For the southern segment of the DF, the nearly N-S-striking Butuo fault section is not continuous with the southern Jiaojihe fault section; the two sections are separated by a restraining stepover with a width of approximately 7-8 km . In this region, several paleoseismic investigations have been conducted; e.g., Song et al.  and He et al.  conducted paleoseismic trenching at Cizijiao and revealed several Holocene paleoearthquakes. However, Sun et al.  suggested that the previous paleoseismic results were not well constrained due to poor trenching and opened more trenches at Wuke and Cizijiao on the Jiaojihe section and Daziri and Luose on the Butuo section (Figure 1(b)). To reveal seismic rupture behaviour of the stepover, we chose the two sites at Yeer village on the Butuo fault section and Damuluo village on the Jiaojihe fault section, respectively. Moreover, we conducted paleoseismic trenching at Buji site on the Jiaojihe fault section (Figure 1(b)).
3.1. Interpretation of Fault Traces and Measurement of Fault Offsets
Firstly, we used satellite imagery (Google Earth) and field observations to map the fault geometry and displaced morphology along the fault. The primary geomorphic evidence for identifying strike-slip faults include linear ridges, displaced rivers, gullies, terraces, and fans, linear-distributed sag ponds. Secondly, we used an unmanned aerial vehicle to survey offset landforms at selected sites. Specifically, the Dajiang Phantom 4 RTK was used for taking photos with an average overlap of 70% along the fault traces. Using structure-from-motion techniques, these photos were then processed to produce high-resolution (centimeter scale) digital elevation models (DEMs) and orthomosaic photos. Considering that local topography at some sites might be affected by vegetation cover, we used Trimble RTK GPS as a supplementary tool to further survey the displaced markers. Thirdly, we used ArcGIS software to produce hillshade of the DEMs and measure horizontal offsets of displaced landforms along fault trend. Due to dense vegetation, some automatic software such as LaDiCaoz_v2 is not very helpful . Therefore, we mainly used multiple repetitive measurements according to the restoration of the displaced landforms at each site to obtain average offsets, which could lower uncertainty of measurement.
3.2. Identification of Paleoearthquakes
We chose the trenching site that favours for continuous deposition with fine-grain deposits, which is helpful for preserving evidence of complete paleoseismic events. Among various indicators for recognizing paleoearthquakes in unconsolidated sediments in a strike-slip fault setting, the best and clearest evidence for fault rupturing of the ground surface includes scarp formation, scarp-derived colluvium, infilled and void fissures, and sand blows .
3.3. Dating Method
We used radiocarbon dating method to constrain ages of stratigraphic units. The collected samples are charcoal, wood, and organic sediment that have been suggested to be good types for representing sedimentation age of stratigraphic units. All samples were sent to Beta Analytic, Inc., USA, for accelerometer mass spectrometer dating. Moreover, we used the OxCal 4.4 program to calibrate all conventional ages with two sigma (95.4% confidence limits) .
4.1. Slip Rate on the Butuo Fault Section
We identified two displaced landforms at the Yeer village on the Butuo fault section of the DF fault. The first site shows that two terraces were horizontally offset by the fault (Figure 2). T1 on the northern bank of the river was left-laterally offset by 0.8 ± 0.2 m (Figure 3). However, considering that the deformation might be affected due to erosion of the river, the amount could be the minimum offset. We further cleaned the deformation exposure of T1 that shows five stratigraphic units (Figure 3). Specifically, two or three fault branches ruptured U4 and were covered by U5. Gravels along the faults show linear alignment, and U3 shows a wedge-shaped fissure within the deformation zone. Considering that U5 might be affected by human modification, we did not date the unit. However, U5 is likely to be a young deposit at the ground surface, which indicates that the offset of T1 might be associated with the most recent coseismic rupture.
Considering that survey of T2 was partly affected by several trees close to the terrace riser (Figure 2), the digital elevation model map from structure-from-motion techniques might be uncertain to measure horizontal displacement of T2. Therefore, we supplement survey using RTK GPS to constrain the offset of T2 to be 6 ± 0.2 m (Figures 2 and 4(a)). Moreover, we collected charcoal samples in the T2 and yield a conventional age at 1220 ± 30 BP (Figure 4(b)) that is calibrated to be 1145 ± 55 Cal BP using OxCal 4.4 software (Table S1), which indicates that slip rate of the fault section at the site is 5.3 ± 0.5 mm/yr.
The second site shows that three terraces were developed (Figure 5). The highest terrace T3 was the best-preserved displaced landform at the site, which was left-laterally offset by 16 ± 1 m. We cleaned a deformation exposure of T1 that shows five stratigraphic units (Figure 6). Specifically, two or three fault branches ruptured U4 and were unconformably covered by U5 that is interpreted as a scarp-related deposit. Gravels along the faults show linear alignment. One charcoal sample was collected in U5 and yielded an age of 100.37 ± 0.37 pMC, which indicates that U5 is very young and can be interpreted as a modern deposit. Moreover, the fault scarp shows a height of 1 ± 0.2 m, which might be associated with the most recent coseismic rupture. Furthermore, we collected radiocarbon dating samples in T3 and yield a conventional age at 2700 ± 30 BP (Figure 6(c)) that is calibrated to be 2805 ± 30 Cal BP using OxCal 4.4 software (Table S1). Considering that T3 was left-laterally offset by 16 ± 1 m, we suggest that slip rate of the fault section at the site is 5.7 ± 0.4 mm/yr.
4.2. Slip Rate on the Jiaojihe Fault Section
Displaced landform such as fault valley and offset gully was revealed at the Damuluo village on the Jiaojihe fault section of the DF fault (Figure 7). Specifically, a gully was cumulatively left-laterally offset by 37 ± 3 m. Moreover, a relatively young offset on the gully bank was constrained by 1.5 ± 0.2 m, and a knickpoint close to the displaced bank is developed with about 0.8 m high (Figure 7(b)), which might be associated with the most recent coseismic rupture. However, considering that the deformation might be affected due to erosion of the gully, the amount could be the minimum offset. Terrace riser of T2 was left-laterally offset by 6.5 ± 1 m (Figures 7(a) and 7(c)). Moving northward, T4 was left-laterally offset by 40 ± 4 m. However, T3 was partly modified by human, and we do not measure its offset. Two radiocarbon samples were collected in T2 and T4, respectively (Figure 8). Specifically, BDM-T2-02 in T2 yields a conventional age at 1220 ± 30 BP that is calibrated to be 1145 ± 55 Cal BP using OxCal 4.4 software (Table S1). Considering that T2 was offset by 6.5 ± 1 m, we calculate a slip rate to be 5.8 ± 1.2 mm/yr. Similarly, the offset of T4 (40 ± 4 m) and the calibrated age (9020 ± 55 Cal BP) collectively constrain a slip rate of 4.5 ± 0.4 mm/yr.
4.3. Paleoseismic Events Revealed at the Buji Site
We have conducted detailed geologic and geomorphic investigations at the Buji site on the Jiaojihe fault section along the DF fault (Figure 1(b)). The trenching site (27°2940″N, 102°4733″E) is located on a large alluvial fan (approximately 0.6 km2) draining eastwards (Figure 9). Due to the strike-slip motion on the fault, a nearly N-S-striking fault valley is produced by blockage against a linear ridge (Figure 9), and we chose a depression in the fault valley as the trenching site because it favours continuous sedimentation from slope wash on the surface of the alluvial fan. To constrain fault distributions, correlate stratigraphic units, and present more evidence of paleoseismic events, three trenches were opened perpendicularly across the fault valley. One trench did not preserve deformational evidence well; therefore, we mainly analyzed the other two trenches. Specifically, trench T1 is ~25 m long, ~2 m wide, and~3 m deep. trench T3 is a re-excavation of a part of T1 that cuts off the two walls of T1 and is ~2 m wider and ~1 m deeper. The stratigraphic units revealed in the trenches are mostly composed of sandy clay and peat that are divided into 11 units, summarized from oldest to youngest in Table S2.
Three or possibly four seismic events were identified in the trenches. The oldest event W is shown in trench T3 (Figures 10 and 11). Specifically, in the northern wall of trench T3 (Figure 10), fault F1 cuts a grey liquefied sandy clay in U2 and is unconformably overlain by U3. U3 is a wedge-shaped sandy clay containing some small gravels at the bottom and is interpreted as a scarp-derived colluvial deposit. In the southern wall of trench T3 (Figure 11), several gravels close to fault F1 have been rotated nearly vertically and show linear alignment. Fault F1 offsets U2 and is unconformably overlain by U3. The aforementioned deformation indicates that the event W occurred between the depositional periods of U2 and U3. As for the second oldest event X, fault F2 offsets scarp-derived deposit U3 and is covered by U4, which is evidenced in the northern wall of trench T3 (Figure 10). Correspondingly, in the southern wall of trench T3 (Figure 11), U3 appears to be deformed with approximately 30 cm vertical folding, which is unconformably covered by U4. Furthermore, U4 is a thick black peaty layer that represents a local dammed pond depositional environment that was likely produced by blockage against a linear ridge related to a seismic event. Although these lines of evidence might have uncertainty, we are inclined to suggest that there was a seismic event with weaker deformation, which indicates that event X occurred between the depositional periods of U3 and U4. The evidence for event Y is shown in the southern walls of trenches T1 and T3 (Figures 10–12). Both walls show that fault F3 offsets U5 and is unconformably overlain by U6. U6 is wedge shaped and is interpreted as a scarp-derived colluvial deposit. In the southern wall of trench T3 (Figure 11), U1-2 wedges into black peaty layer U4 and deforms sandy clay unit U5, which is unconformably overlain by U6. Therefore, we suggest that event Y occurred between the deposition of U5 and U6. All trenches show deformational evidence for event Z. For example, in the southern wall of trench T1 (Figure 12), fault F4 consists of several branch faults and stands almost vertically, evidenced by gravel alignment and offset at the ground surface along the base of the linear ridge. Some thin infilled fissures consisting of black peat from U8 are developed in U7 close to the fault (Figure 12). Similar lines of evidence for fault deformation and infilled fissures are also revealed in the southern wall of trench T3 (Figure 11). Although the fault scarp was partly modified by humans, the scarp-derived colluvial deposit U10 at the base of the scarp can still be observed. Therefore, we suggest that event Z occurred between the depositional periods of U9 and U10.
Based on the aforementioned paleoseismic analysis, we further used OxCal 4.4 software to constrain ages of the paleoseismic events. This software uses Bayesian statistics to re-weigh the PDFs and account for stratigraphic ordering (overlying ages are younger) or historical age constraints. These statistics result in shifting or trimming the distributions to fewer peaks in multi-peaked distributions [18, 19]. Samples with close ages collected from the same stratigraphic unit were used in phase correction when establishing the sequence correction during processing with the software . For example, two samples (D-T3-03 and D-T1-32) are classified into phase U8. As previously analyzed, event Z occurred after U9 and prior to U10; although we do not have samples from these two units, we can use U8 to constrain the age of event Z to the interval from 620 AD to Present. In keeping with this scenario, the ages of the other events can be constrained by dated samples as follows: 6265 BC-4510 BC (event W), 4475 BC-3700 BC (event X), and 1150 BC-515 BC (event Y) (Figure 13).
5.1. Slip Rates on the Southern Segment of the DF
Based on measurement of displaced landforms and radiocarbon dating, two sites at the Yeer village on the Butuo fault section constrain slip rates as 5.3 ± 0.5 mm/yr and 5.7 ± 0.4 mm/yr. Similarly, offset of the two terraces at Damuluo village on the Jiaojihe fault section shows that slip rates are 5.8 ± 1.2 mm/yr and 4.5 ± 0.4 mm/yr. The aforementioned moderate slip rates are well consistent, which indicates that the two fault sections of the southern segment of the DF show comparable fault activity. Moreover, Wei et al.  used measuring offsets of several alluvial fans along the southern segment of the DF fault and optical-stimulated luminescence dating to constrain slip rates of the fault to be approximately 3 mm/yr, which is comparably consistent with our result.
5.2. Comparison with Previous Paleoseismic Results on the Jiaojihe Fault Section
Before correlating the events between different trenching sites along the southern segment of the DF fault, it is important to evaluate the completeness of the paleoseismic record at the Buji site. Critically, a continuous depositional environment at the trenching site is a key factor in determining whether the paleoseismic record is complete. At the Buji trenching site, the linear fault ridge (Figure 9) can block slope wash from the surface of the alluvial fan and facilitate continuous deposition in the depression, which is evidenced by most stratigraphic units, such as black peat and fine-grained sandy clay.
For previous paleoseismic results in the region of the southern segment of the DF fault, the Cizijiao, Buji, and Wuke trenching sites are all located on the Jiaojihe fault section while the Luose and Daziri trenching sites are located on the Butuo fault section (Figure 1(b)). Sun et al.  summarized all the paleoseismic events inferred from trenching [11, 13]. Specifically, Sun et al.  used two walls of two single trenches to reveal three paleoseismic events at Cizijiao with ages at 150 BC-630 AD, 6500-3010 BC, and prior to 17880 BC and four events at Wuke with ages at 970-1510 AD, 620-180 BC, 12640-10320 BC, and prior to 12850 BC; these data show that only two oldest events at the two sites have possible overlaps. To explain the lack of correlations for all the other events between the Cizijiao and Wuke trenching sites, Sun et al.  suggested that depositional hiatuses resulting in missing records of paleoearthquakes might exist in the two trenches; therefore, they concluded that these seven events revealed at the two sites can represent a six-event sequence during the past 20000 years on the Jiaojihe section. Here, we compare these events using the paleoseismic results from the Buji trenching site as follows.
Firstly, the three trenching sites (Cizijiao, Buji, and Wuke) on the Jiaojihe section are closely spaced within ~18 km with continuous fault traces that show no typical geometrical complexities (Figure 1(b)), which indicates that the three trench sites along the fault section are very likely to be ruptured simultaneously during an earthquake. In other words, if age ranges of seismic events revealed at the three sites are very close, we could suggest that the events can be interpreted as the same earthquake . For example, the age of event (150 BC-630 AD) at Cizijiao is close to the age of event (620-180 BC) at Wuke . Similarly, He et al.  constrained one paleoseismic event at the Cizijiao site to 0-430 AD that is also close to the ages of the events. The slight difference among the aforementioned ages might be resulted from dating on peaty samples [11, 12] that actually represent a mixed age and might be younger or older than expected, which indicates that the three events are likely to be interpreted as the same earthquake. Secondly, in keeping with the aforementioned scenario, we can further narrow age constraints by comparing different trenching sites. For example, age of the youngest event Z at the Buji trenching site can be narrowly constrained as 970-1510 AD. Similarly, the age of event Y (1155-515 BC) revealed from Buji trenching site can be bracketed as 620-515 BC (Figure 14).
5.3. Paleoseismic Comparison of the Jiaojihe and Butuo Fault Sections
Sun et al.  used two single trenches at the Daziri and Luose sites on the Butuo fault section, which are separated by ~4 km and reveal paleoseismic events of ages 1310-1660 AD and 27800-3070 BC and five events with ages 750-400 BC, 5270-5060 BC, 7260-7060 BC, 10640-7440, and 41180-29200 BC, respectively. Similarly, if age ranges of seismic events revealed at the two sites are very close, we could suggest that the events can be interpreted as the same earthquake on the Butuo fault section. However, considering that paleoseismic events constrained on the Jiaojihe fault section are limited, we only compare these events that are younger than about 7260 Cal BP (Figure 14). Specifically, the most recent event revealed at Daziri occurred at 1310-1660 AD, which overlaps the age of the most recent event Z (970-1510 AD) at the Buji trenching site. Wen et al.  summarized all the historically recorded earthquakes along the southeastern margin of the Tibetan Plateau since ~624 AD. Although no historically recorded large earthquake was related to the DF fault , we rechecked the historical records in the adjacent regions and found some brief descriptions related to an earthquake in 1489 AD that caused strong damage in Xichang city approximately 40 km to the west of the trenching sites , but this earthquake has not been revealed by paleoseismic studies near the city [22–25]. Moreover, the other historically recorded earthquakes during the time range were related to the AF and ZF that are far away from the DF (Figure 1(b)) [22–25]. Furthermore, the young displaced landforms such as terrace T1 and the gully revealed at Yeer village and Damuluo village (Figures 3, 6, and 7) provide evidence that the youngest two events are likely interpreted to be a historical recorded earthquake. Therefore, it is likely that the two youngest events revealed at Butuo and Jiaojihe fault sections are the same event. In other words, the 1489 AD earthquake ruptured the two fault sections simultaneously. As for the second youngest events on the two fault sections (Figure 14), the event (750-400 BC) revealed at Luose site appears to be overlapped with the event Y (620-515 BC). However, there are no historical records prior to ~624 AD to clearly prove that the two events might be interpreted to be the same event. In other words, it is uncertain whether the two events might be the same earthquake. As for the older events, their age ranges do not overlap well, indicating that these events might only rupture a single fault section.
Base on the aforementioned analyses, we estimated earthquake magnitudes for different rupture modes on the Jiaojihe and Butuo fault sections. If the two fault sections ruptured simultaneously such as the youngest event, there would be surface ruptures of lengths of approximately 120 km (Figure 1) , calculated from the summarized equation for a strike-slip fault (, where denotes the magnitude of an earthquake and represents the corresponding surface rupture length) . We concluded that a magnitude M7.5 ± 0.3 earthquake was estimated for the coseismic rupture of both fault sections. If the two fault sections ruptured independently such as the older events over lengths of approximately 55 km (Butuo fault section) and 65 km (Jiaojihe fault section) (Figure 1) , the corresponding earthquake magnitudes would be M7.1 ± 0.3 and M7.2 ± 0.3, using the same method of Wells and Coppersmith . The estimated earthquake magnitudes are consistent with those of neighboring faults such as the XJ fault to the south and the ZF and AF faults to the west, which have experienced historical earthquakes of magnitudes M7 to M8 .
5.4. Implications for Rupture Behaviour at Wider Stepovers on Strike-Slip Faults
Generally, 4-5 km wide stepovers are considered to be endpoints that are likely to arrest fault rupture, which has been widely considered to be an important rule for fault segmentation and seismic hazard assessment for strike-slip faults [2, 3]. However, several recent large earthquakes show that wider stepovers (e.g., ≥7-8 km wide) were jumped over such as the 2016 Mw 7.8 Kaikoura earthquake  and the 2001 Mw 7.8 Kunlunshan earthquake , which indicates that previous analysis of fault segmentation might be underestimated. The aforementioned earthquake cases are mainly related to high slip-rate faults; few studies reported seismic rupture behaviour at wider stepovers on moderate slip-rate faults. The result on the southern segment of the DF fault indicates that wider stepovers on moderate slip-rate faults can also be jumped over. However, seismic ruptures through the 7-8 km wide constraining stepover were temporally variable. Similar seismic rupture behaviour was also revealed on other geometrical complexities. For example, paleoseismic studies along the Longmen Shan thrust fault along the eastern margin of the Tibetan Plateau suggest that all the parallel-distributed faults simultaneously ruptured during the 2008 Wenchuan Mw7.9 earthquake, whereas older events only ruptured some of the fault segments . Padilla et al.  used trenching and computational modelling to show that 20%-23% of earthquakes on the San Andreas and the San Jacinto faults are co-ruptures through Cajon Pass. Shao et al.  used paleoseismic investigation to suggest that one of four seismic events ruptured the Aksay Restraining Double Bend, which is also revealed by computational modelling that shows statistically about 10% of ruptures jump across the bend and propagate through almost the entire local fault system . Although paleoseismic studies at wider stepovers were limited, our study suggests that moderate slip-rate faults have similar potential with that of high slip-rate faults to rupture through wider stepovers.
Moreover, our paleoseismic results may provide some insights into how wider stepovers temporally control earthquakes. Specifically, the older seismic events on the two fault sections are very likely to be arrested by the constraining stepover, whereas the most recent event jumped over the stepover (Figure 14). Here, we could present two scenarios to show the rupture modes at the stepover. The first possibility is that the stepover rupture during each earthquake produced on the Butuo fault section or the Jiaojihe fault section is random. The second possibility is that arrest ability of the stepover varied during multicycle earthquakes. We suggest that the second mode is more likely than the first mode. As noted in dynamic modelling of Duan et al. , previous repeated earthquakes can have residual stress concentrations that would result in more-heterogeneous fault stress and rupture wider stepovers during subsequent earthquakes. Moreover, Finzi et al.  used 2D simulations to suggest that multicycle earthquake-related damage accumulation within stepovers facilitates rupture of wider stepovers. Although we do not know how the fault stress evolved or the damage accumulated in the stepover, the change of rupture behaviour of the stepover (Figure 14) appears to be consistent with the physics-based models [4, 5]. If the second rupture mode is true, it at least has two important implications for assessing seismic risk at wider stepovers on strike-slip faults. The first implication is that rupturing wider stepovers requires enough residual stress concentrations or accumulated damage in the stepovers by multiple previous earthquakes, which might be considered one of the reasons that field evidence of rupture through wider stepovers has not been widely developed. Second, if a wider stepover is not ruptured by the previous earthquakes revealed from historical or paleoseismic studies, it could be suddenly ruptured during the next earthquake, which enhances the uncertainty in assessing the seismic risk on strike-slip faults because it is difficult to determine which earthquake can finally rupture the stepovers. Therefore, rupture behaviour at wider stepovers can be temporally variable and more complex than previously thought, which requires us to be particularly focused in future studies .
Based on detailed field investigations, landform survey, paleoseismic trenching, and radiocarbon dating on the southern segment of the DF, which is separated by an approximately 7-8 km wide restraining stepover, we constrained slip rates and paleoseismic history of the fault. The two sites at the Yeer village on the Butuo fault section constrain slip rates as 5.3 ± 0.5 mm/yr and 5.7 ± 0.4 mm/yr, whereas offset of the two terraces at Damuluo village on the Jiaojihe fault section shows that slip rates are 5.8 ± 1.2 mm/yr and 4.5 ± 0.4 mm/yr, respectively. Combined with the results of previous paleoseismic studies, we suggest that the youngest event Z might rupture the stepover, whereas older events might be arrested, which might indicate that the rupture behaviour at the stepover may be temporally variable during multicycle earthquakes. The results have important implications for evaluating regional seismic hazards and can help us better understand the behaviour of multiple-fault ruptures on strike-slip faults.
The data used to support the findings of this study are included within the article.
Conflicts of Interest
The authors declare no conflict of interest.
Great thanks to China Scholarship Council for sponsoring the first author to study at Oregon State University for a year. Thanks are due to Jiahui Feng for field work. This study was supported by the 2nd Tibetan Plateau Scientific Expedition and Research (2019QZKK0901) and the National Science Foundation of China (Grant No. 41672207).