Paleoseismic investigation of the seismic gap between the seismogenic structures of the 2008 Wenchuan and 2013 Lushan earthquakes along the Longmen Shan fault zone at the eastern margin of the Tibetan Plateau

In the past 5 yr, two devastating earthquakes, the 2008 Mw 7.9 Wenchuan and the 2013 Mw 6.6 Lushan earthquakes, occurred along the Longmen Shan fault zone at the eastern margin of the Tibetan Plateau (Fig. 1), killing more than ∼80,000 people and causing severe damage. The Wenchuan event ruptured the middle segment and northern segment of the Longmen Shan fault zone (specifically, surface ruptures are along the Yingxiu-Beichuan fault [YBF] and Guanxian-Jiangyou fault [GJF] shown in Fig. 1). Seismogenic structure models for the Lushan earthquake (Chen et al., 2013, 2014; X.W. Xu et al., 2013) agree that this event was associated with the Dachuan-Shuangshi fault and a blind reverse fault system along the southern segment of the Longmen Shan fault zone. Modeling of seismogenic structures for the 2008 and 2013 earthquakes shows a 50 km seismic gap between the active segments of the fault relating to the two events (Fig. 1). Geoscientists have been debating whether this gap has the potential to produce a large earthquake in the near future. Wang et al. (2010) used the method of balancing seismic moments on faults to suggest that if all the moment energy were released by a rupture of the whole southern segment, it could produce an earthquake as large as Mw 7.7 in the next 50 yr. Liu et al. (2014) used a similar method to recalculate moment deficit and suggest the seismic gap or the region south of the Lushan rupture zone along the southern segment of the Longmen Shan fault zone could produce a M ∼7 earthquake. Other studies indicate that Coulomb stress in the gap increased considerably after the Wenchuan and Lushan earthquakes, and this represents a seismic risk (Parsons et al., 2008; Shan et al., 2013). However, Li et al. (2013) integrated methods such as seismic tomography and relocated aftershocks and found a region of anomalously slow seismic wave speeds in the gap. Their results suggest that the weak and ductile crustal materials at seismogenic depths cannot accumulate enough strain for strong earthquakes. To address these opposing views, we use mainly paleoseismologic evidence from trenching and ages from multiple radiocarbon samples to reveal a more robust rupture history along the seismic gap.

The Longmen Shan fault zone, located at the eastern margin of the Tibetan Plateau, is characterized by elevations of up to 7500 m above sea level and by topographic relief of more than 5 km over distances of less than 5 km, which is the result of an intense collision between the Indian and Eurasian plates (Kirby et al., 2002).

The active Longmen Shan fault zone marks a predominantly convergent boundary with a right-lateral strike-slip component. This fault system was reactivated during late Cenozoic time along a Mesozoic orogenic belt (Burchfiel et al., 1995, 2008; Kirby et al., 2002, 2008). To the west of the Longmen Shan, East Tibet actively deforms by both right-lateral shear parallel to and convergence perpendicular to the Longmen Shan fault (King et al., 1997; Chen et al., 2000; Zhang et al., 2004; Shen et al., 2005; Gan et al., 2007). Tectonic activity in the Sichuan Basin, east of the Longmen Shan, has been low during late Cenozoic time. Three principal, subparallel, active faults comprise the northeast-trending Longmen Shan fault zone, named the Yingxiu-Beichuan fault, Guanxian-Jiangyou fault, and Wenchuan-Maoxian fault along the middle segment and northern segment, and Gengda-Longdong fault, Yanjing-Wulong fault, and Dachuan-Shuangshi fault along the southern segment (Fig. 1; Zhang et al., 2010).

Geological investigations suggest that fault motion on the middle segment and northern segment is dominated by reverse thrusting with a right-lateral component and a vertical slip rate no greater than 1 mm/yr for the past ∼10,000 yr (Ma et al., 2005; Li et al., 2006; Densmore et al., 2007). After the 2008 Wenchuan earthquake, numerous geoscientists used a range of methods to measure recurrence intervals of large earthquakes along these segments of the Longmen Shan fault zone. For example, modeling based on geodetic and geological slip rates from interferometric synthetic aperture radar (InSAR) and global positioning system (GPS) inversions estimated the average recurrence interval of large earthquakes along the Longmen Shan fault zone to be ∼3000–6000 yr (Zhang et al., 2008; Shen et al., 2009). Multiple trenching investigations integrated with geomorphic analysis and radiocarbon dating along the surface ruptures produced by the 2008 Wenchuan earthquake suggest an average recurrence interval for large earthquakes of ∼3000 yr (Ran et al., 2010, 2013). However, there are only a few studies of the paleoseismic behavior along the southern segment of the fault. Densmore et al. (2007) exposed a trench at Qingshiping (Fig. 2) on the Dachuan-Shuangshi fault and found evidence of weak deformation, identifying two paleoearthquakes—the younger event between 930 ± 40 yr B.P. and 860 ± 40 yr B.P. (corrected by OxCal 4.2 to A.D. 1045–1230) and the older prior to 930 ± 40 yr B.P. Recently, Chen et al. (2013) excavated a trench and uncovered a natural geologic exposure south of Qingshiping (Fig. 2). Their findings suggest that a paleoearthquake occurred around A.D. 1480–1890 (corrected by OxCal 4.2 to A.D. 1515–1885). Moreover, they used results of a previous scientific report (Institute of Geology, China Earthquake Administration, 2009) on the 2008 Wenchuan earthquake to place a time constraint between 1390 yr B.P. and 650 yr B.P. (corrected by OxCal 4.2 to A.D. 645–1340) on an earlier event. However, stratigraphic unconformities and a lack of material for radiometric age dating have resulted in incomplete or poorly dated paleoseismic records at the aforementioned trenches on the Dachuan-Shuangshi fault along the southern segment of the Longmen Shan fault; hence, it is difficult to estimate the average recurrence interval of paleoearthquakes or the vertical slip rate of the fault segment.

The Dachuan trenching site is located north of Lushan on the Dachuan-Shuangshi fault along the southern segment of the Longmen Shan fault zone (Fig. 2). Geologic mapping in this area suggests that Triassic strata were thrusted over Jurassic strata, resulting in typical fault-valley geomorphology; surface traces of the active Dachuan-Shuangshi fault are not well expressed, and the linear deformation zone is not continuous (Fig. 3). We followed the mapping of fault traces of the Dachuan-Shuangshi fault as revealed from the trenching and geomorphic interpretations of Chen et al. (2014) to constrain the active fault located within the wooded valley. To search for evidence of surface ruptures, we selected a marshy depression with less vegetation cover and opened two trenches that traversed the fault valley floor. Trenches A and B are ∼30 and 50 m in length, respectively (Fig. DR1¹). Trench B revealed deposits consisting mainly of mixed gravels and boulders, which may represent a high-energy deposition environment like a debris flow. Trench A revealed three groups of fine-grained clay units that are subhorizontal in the southeast part of the trench and warp up to the northwest in the northwest part of the trench. The lower stratum (U1) consists of maize-yellow (10YR8/1) fine clay containing some rounded gravel with diameters ranging from 5 to 10 cm. This unit can be divided into two subunits (indicated by the dashed line in Fig. 4). The upper subunit has a lighter color and less gravel with a vertical difference of at least 1.5 m (the minimum value). The middle section of the trench stratigraphy (from U2 to U6) is mainly peaty clay containing some decomposed leaves and interbedded with thin, fine white clays, representing a quiet-water environment. U2 is black and peaty with multiple rotten leaves and an apparent eastward dip to the bedding. In the northwest end of the trench, a boulder with a long axis greater than 1 m extends between the base of the upper subunit in U1 and the base of U3. The color of U3 is slightly lighter compared with U2; U3 is possibly divided into at least three subunits, and an obvious difference is that a thin white subunit consisting of clay has developed. Possibly disturbed by the boulder, this thin white layer is not clear at the northwest end of the trench. U4 is white clay with a thickness of ∼10 cm; however, it pinches out to the west. Units at the northwest end of the trench are warped upward, forming a buried scarp above an inferred buried trace of the fault. The part of U4 lying closer to the scarp shows compressed shortening characteristics. The color of U5 is similar to U3; it contains numerous rotten leaves and has a stable thickness. U6 is gray peaty clay that thins toward the deformation zone. The upper layer (U7) contains modern deposits of brown clay.

Continuity of most units across the trench and a lack of sharp breaks in stratigraphy suggest that the units in the trench were deformed by folding. Interpretations of paleoseismic events were therefore based on evidence of growth strata, stratigraphic warping, and unconformities (Fig. 4). U5 shows warping with a vertical displacement of ∼0.4 m, and its thickness does not change across the deformation zone, whereas U6 wedges westward against the scarp formed by the folding of U5. This indicates that U6 unconformably overlies U5 and suggests that it is a growth stratigraphic unit associated with a surface-rupture seismic event (referred to as event E1) that occurred between the deposition of U5 and U6. Similarly, the subunit within U1 is also warped, in this case with a vertical displacement of ∼1.5 m, and also does not change thickness across the deformation zone. In contrast, the units above (from U2 to U4) wedge westward against the scarp formed by the folding of U1, which indicates that they unconformably overlie the internal stratigraphy within U1 and suggests that they are growth stratigraphic units associated with another event (referred to as event E3) that occurred between the deposition of U1 and U2. The interpretation of E2 is challenging; it may or may not be associated with an earthquake. Support for the earthquake interpretation comes from the observation that U4 becomes thinner close to the deformation zone; however, away from the scarp, it shows a stable thickness. U3 is a peaty layer indicating a quiet-water depositional environment. It is much thicker and mostly deposited on the hanging wall of the deformation zone. Since the uppermost stratigraphic contact between U3 and U4 was originally horizontal, then U4 should have a continuously equal thickness; however, the actual characteristics of U4 do not support this analysis. Furthermore, as evidence of growth strata, the internal subunit above the thin white layer (its upper contact is shown with a dashed line) in U3 apparently does not change thickness across the deformation zone, even though this white layer is not clear on the hanging wall that may have been disturbed by the boulder. Unfortunately, the other wall of the trench collapsed quickly owing to the soft nature of the deposits, so no evidence is available from it that would help with the identification of paleoseismic events. Third, the vertical deformation of U3 appears to be slightly greater than that of U5. By identifying sudden changes in the stratigraphic components, it is possible to interpret that an earthquake occurred during this deposition. Conversely, an alternative interpretation is that event E3 produced a large scarp and units U2, U3, and U4 back filled against the scarp. In this case, U4 would be affected by the shape of the large scarp, but it would not represent an earthquake. The vertical displacement of U3 (0.6–0.7 m) is slightly greater than the offset of U5 (∼0.4 m), as may be explained by upward attenuation during the youngest folding.

We found numerous charcoal samples in the trenches. Twelve samples were sent for accelerometer mass spectrometer (AMS) dating at Beta Analytic, Inc., in the United States. The radiocarbon dating results are summarized in Table 1. All ages reported herein are 2σ (95.4% confidence limits) calendric ages calibrated with the OxCal 4.2 program (Bronk, 2009), using the IntCal09 atmospheric model from Reimer et al. (2009). For the calibrated age, probability density functions (PDFs) overlap between different samples. OxCal uses Bayesian statistics to reweigh 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 multipeaked distributions (Bronk, 2009; Lienkaemper and Ramsey, 2009). Most of the radiocarbon dates, except for samples LCG-T2-C6 from U1 and LCG-T2-C30 from U2, were in correct stratigraphic order. Sample LCG-T2-C6, collected as a tiny piece of rotten leaf, is possibly slightly younger than the true depositional age. LCG-T2-C30, collected as a small portion of wood, is just slightly older than a sample collected from a lower position in U2 and should be older than the true depositional age. We therefore do not use these two samples during further age constraints. Using the OxCal 4.2 program for further analysis, the ages of events E1 to E3 are constrained to the ranges A.D. 1350–1830, A.D. 590–1210, and 760–525 BC, respectively (Fig. 5). Irrespective of the paleoseismic events revealed in the Dachuan trench along the Dachuan-Shuangshi fault, unit U1 is observed to have deformed with a cumulative vertical displacement of ∼1.5 m (Fig. 4). Taking the age of this unit to be ca. 2500 yr B.P. (Table 1), we estimate the vertical slip rate to be ∼0.6 mm/yr.

Chen et al. (2013) constructed a geologic section north of Dachuan, ∼17 km away from our trench (Fig. 2), and revealed a paleoseismic event constrained between A.D. 1515 and 1885 (Fig. 6). Using the vertical offset of 0.4 m produced by the most recent event in the trench at Dachuan and the high-angle reverse fault of the Longmen Shan fault (Zhang et al., 2010), the regression analysis (reverse-fault type) of Wells and Coppersmith (1994) was applied to estimate a possible rupture length of ∼23 km—or slightly more if partial attenuation owing to folding deformation is considered. This indicates that the two trenching sites may reveal the same event. Moreover, application of the Z-statistic method (Sheppard, 1975) to the mean ages and their respective standard deviations (McCalpin, 2009; Biasi et al., 2011), i.e., A.D. 1695 ± 95 and A.D. 1580 ± 130 as calculated by the OxCal 4.2 program for the two events, gives an estimate for Z of ∼0.71. This value corresponds to a probability of contemporaneity of ∼0.5, meaning that the two events have a high possibility of being the same event. Third, overlapping distributions show that the age ranges of the two events are well matched (Fig. 6). However, there is no corresponding rupture evidence of this event from the trenching investigation of Densmore et al. (2007) or the report of Chen et al. (2013) at sites ∼18 km further to the northeast near Qingshiping (Fig. 2). This indicates that surface ruptures in the seismic gap along the Dachuan-Shuangshi fault may be segmented.

Similarly, even if event E2 does exist, we suggest that a surface rupture produced by the event would be unlikely to have extended to the trenching sites of Densmore et al. (2007) and the report of Chen et al. (2013). The distance of the two sites is ∼35 km, i.e., much greater than the length of the estimated surface rupture produced by event E2 (the cumulative displacement of U3 is ∼0.6–0.7 m, which means the coseismic vertical displacement produced by event E2 is ∼0.2–0.3 m, which is less compared with that of E1). Additionally, comparing event E2 to the events from Densmore et al. (2007) and Chen et al. (2013), the Z-statistics method gives Z values of ∼1.26, which correspond to a probability of contemporaneity of ∼0.2, and the coseismic vertical displacement for the corresponding event from the trenching of Chen et al. (2013) is a small amount of ∼0.3 m, which correlates to a similar short rupture length; this more possibly suggests that the events are much less likely to be the same. Possibly, event E2 as revealed in the Dachuan trench may not represent a single earthquake, which further indicates that surface ruptures produced on the Dachuan-Shuangshi fault in the seismic gap show characteristics of segmented rupturing.

If a single earthquake could rupture the whole 50-km-long seismic gap, the coseismic vertical displacement would likely be much greater than 1 m for a reverse fault (Biasi et al., 2011). Such displacement is not consistent with the smaller stratigraphic separations revealed in the paleoseismic trenches in the gap. Therefore, based on the aforementioned analysis, we suggest that the seismic gap along the Dachuan-Shuangshi fault is unlikely to rupture completely during one earthquake, unless it ruptures together with the middle segment of the Longmen Shan fault zone.

The 2013 Mw 6.6 Lushan earthquake occurred along the southern segment of the Longmen Shan fault zone, whereas the 2008 Mw 7.9 Wenchuan earthquake ruptured the whole northern segment and middle segment. Ran et al. (2013) used trenching to reveal evidence of a paleoearthquake that occurred between 3300 and 2300 cal. yr B.P. (or 1350–350 B.C.) that was comparable in magnitude with the 2008 Wenchuan earthquake. They suggested an average recurrence interval of large earthquakes to be ∼3000 yr along the middle segment, with coseismic vertical displacements of several meters. From the Dachuan trench, assuming the existence of event E2, we suggest the average recurrence interval of surface-rupturing earthquakes to be ∼1000–1300 yr. If E2 did not occur, the average recurrence interval would be 1875–2490 yr, a much shorter recurrence interval with a smaller coseismic displacement compared with that of the middle segment. Both events E1 and E2 do not correlate to paleoearthquakes along the middle segment. However, the age of E3, constrained at 760–525 B.C. from trenching at Dachuan, overlaps with that of the penultimate event (1350–350 B.C.) revealed along the middle segment, which appears to suggest that the southern segment primarily ruptured as a separate segment from the middle segment, although rupture of the two segments together cannot be completely ruled out.

Chen et al. (2014) mapped the distributions of active faults along the southern segment and found more branch faults and secondary folding deformation along the southern segment than have been observed along the northern and middle segments of the Longmen Shan fault zone. These structural complexities along the mapped trace of the Dachuan-Shuangshi fault may retard through-going rupture on the segment; in other words, rupture of the Dachuan-Shuangshi fault may not extend across the entire southern segment. These findings are more consistent with the southern segment having a low ductile strength (Li et al., 2013) compared to models suggesting that the Dachuan-Shuangshi fault ruptures in large earthquakes.

This work was supported by the Project of Lushan M7.0 Earthquake Nucleation and its Mechanism and Influences—Study on the Tectonic Background of the Bayan Har Block and Nucleation Mechanism of the Lushan Earthquake (201408014), the National Science Foundation of China (Grant 41302160), the Project of China Earthquake Administration “Scientific Investigations on the 20 April 2013 Lushan, Sichuan Earthquake,” and the Special Projects for Basic Research Work of the Institute of Geology, China Earthquake Administration (IGCEA1304). Thanks go to Fei Han and Chenglong Liu for the field work. We are grateful to the Seismic Mitigation Bureau of Ya’an and Lushan for their support of field investigations. Great thanks go to Christopher Madden Madugo and Eric Kirby, who provided detailed and constructive suggestions on the manuscript.