Extension rates across the northern Shanxi Grabens, China, from Quaternary geology, seismicity and geodesy Free

Abstract

1 INTRODUCTION

Discrepancies between geological, seismic and geodetic rates of deformation are a cause of controversy in many regions of active faulting (Thatcher 2009). Yet these discrepancies are important for our understanding of how both crustal deformation and seismic hazard may vary through time. In California and western New Guinea, for example, seismic strain rates appear to match relative plate velocities, whilst in the Aegean, Iran and parts of Asia the rates of seismic deformation are substantially lower than those predicted from plate velocities (Jackson & McKenzie 1988; Ekström & England 1989). In the New Madrid seismic zone, present-day geodetic observations are not able to detect any significant strain, and yet major earthquakes took place in 1811 and 1812 and the palaeoseismic record indicates that large earthquakes have occurred every 500 yr (Stein et al. 2009; Craig & Calais 2014). On the scale of individual faults, there is significant disagreement between geological and geodetic slip rate estimates on faults such as the San Andreas, Altyn Tagh, Kunlun and Haiyuan (Searle et al. 2011; Wang et al. 2011; He et al. 2013; Tong et al. 2014). Are these differences because fault slip rates vary substantially through time? Can off-fault deformation explain the discrepancies (e.g. Milliner et al. 2015)? Could large fractions of strain build-up be accommodated by aseismic creep on major faults (e.g. Jolivet et al. 2012)? Or is tectonic loading on the continents distributed across complex systems of interacting faults (e.g. Stein et al. 2009; Liu et al. 2011)?

The northern Shanxi Grabens lie at the northeastern corner of the Ordos Plateau, a piece of non-deforming continental lithosphere in northern China. Extension in the northern Shanxi Grabens and the associated seismicity is of significant concern because the adjacent North China Plain is one of the most densely populated regions on Earth. The cities of Beijing, Datong, Shijiazhuang and Taiyuan, with a total population in excess of 35 million people, all lie in or next to the northern Shanxi Grabens. Yet previous studies have disagreed significantly about the magnitude and pattern of strain release in this region (see Section 2 for details).

2 PREVIOUS WORK IN THE NORTHERN SHANXI GRABENS

Estimates of Quaternary slip rates exist for many of the major faults in the northern Shanxi Grabens (Cheng et al. 1995; Xu & Deng 1996; Xu et al. 1997; Cheng & Yang 1998; Pavlides et al. 1999; Xu et al. 2002; Jiang et al. 2003; Yu 2004; Deng et al. 2007; Ding et al. 2009; Wang 2012). In general, the slip rates are small and the calculated rates of extension do not exceed 1 mm a⁻¹ for any of the faults in the region. A cross-sectional area balance suggests a total Pliocene-Quaternary extension rate of between 0.2 and 0.5 mm a⁻¹ across the whole northern Shanxi region (Zhang et al. 1998). Zhang et al. (1998) also estimate a principal extension direction of 165 ± 15° from the maximum height of triangular facets, whilst Zhang et al. (2003) estimate an extension direction of 120° to 150° from fault striation measurements.

Earthquake records indicate that the northern Shanxi Grabens have been seismically active over the last few thousand years, though there are relatively few recorded events in the last ≈250 yr (Lee et al. 1976, 1978; Liu et al. 2011). A Kostrov summation of seismic moment tensors for the four largest historical earthquakes in the Shanxi Grabens gives a north–south extension rate of 0.4 mm a⁻¹ (although this is not necessarily along the direction of the maximum extension) (Wesnousky et al. 1984). Xue & Yan (1984) infer a mean extensional direction of 153° from T axes of focal mechanisms, whilst Li et al. (2015a) conclude from another catalogue of focal mechanisms that the minimum compressive stress (σ₃) is at an azimuth of 139° (though orientations of principal stress need not coincide with orientations of principal strain). Wang et al. (2012) even claim from their analysis of P axes of focal mechanisms that prior to 2001 it was the maximum compressive stress (σ₁) that was oriented northwest-southeast.

Geodetic constraints on current crustal surface motion in northeast China are provided by numerous GPS studies (Shen et al. 2000; Wang et al. 2001; He et al. 2003, 2004; Gan et al. 2007; Wang 2009; Zhang & Gan 2008; Li et al. 2012; Zhao et al. 2015). Shen et al. (2000) conclude that 4 ± 2 mm a⁻¹ of extension occurs across the northern Shanxi Grabens at an azimuth of 105°. They also note that their extension direction is ≈45° anticlockwise from geological estimates, and postulate that this is because the geological estimates do not take account of hidden faulting occurring within the grabens at a different azimuth. Meanwhile, He et al. (2003) and He et al. (2004) find no evidence of any extension across the northern Shanxi Grabens at all, which they attribute to timescale dependent fault behaviour. Wang et al. (2011) fit a block model to the GPS data of Zhang & Gan (2008) and obtain an extension rate in the Shanxi Grabens of 0.8 ± 0.3 mm a⁻¹. Similarly, Qu et al. (2014) fit a block model to the GPS data of Wang (2009) and obtain a strain rate of 28 nanostrain a⁻¹ (roughly equivalent to an extension rate of 2.8 mm a⁻¹) at an azimuth of 95°.

According to these studies, therefore, extension occurs in the northern Shanxi Grabens at anywhere between 0 and 6 mm a⁻¹ at an azimuth of between 95° and 180° (see later Fig. 15). In this study we compile fault extension rates from ages of offset geomorphological features and estimates of fault dip; we construct a catalogue of historical earthquakes with M_s > 5 for the period 1315 to 2015 and apply Kostrov’s relation to this dataset; and we analyse the latest campaign GPS data from Zhao et al. (2015). We then compare these geological, seismic and geodetic observations and seek to ascertain whether or not there are any real discrepancies between these different estimates of the magnitude and pattern of strain release in the northern Shanxi Grabens.

3 TECTONIC SETTING

The Shanxi Grabens lie to the east of the Ordos Plateau in northern China, where they form an S-shaped series of en echelon basins some 1200 km long (see Fig. 1). The grabens can be divided into three broad zones, two of which are wide, ENE trending, extensional zones (in the north and the south), separated by a third, narrow, NNE trending, right-lateral transtensional zone in the centre (Xu & Ma 1992; Xu et al. 1993). Although extension in the Shanxi Grabens only started in the Pliocene or the very end of the Miocene (Ye et al. 1987; Zhang et al. 1998, 2003), it is likely that the location and geometry of the grabens are in part determined by the pre-existing, northeast striking, fold-and-thrust belt structures of the Late Triassic (235 to 200 Ma old) Indosinian orogeny and the Middle Jurassic-Cretaceous (150 to 65 Ma old) Yanshanian orogeny (Xu & Ma 1992; Xu et al. 1993; Zhang et al. 1998). Manifestations of these older orogenies are still evident today as thickened crust (40–44 km) and a corresponding gravity anomaly over the region (Xu & Ma 1992). Tapponnier & Molnar (1977) note that the width and length of individual grabens decrease from south to north, whilst the mean elevation of the graben floors increases (from 380 m in the south to 1010 m in the north (Li et al. 2015a)), suggesting a northward propagation of the onset of rifting through time.

Figure 1.

SRTM topography (Farr et al. 2007) of the Ordos Plateau in northeastern China. Focal mechanisms from the Global CMT catalogue (Ekström et al. 2012) are shown for recent earthquakes with magnitudes greater than M_w 5.0. Earthquakes of M_w 2.0–5.0 from the ISC catalogue (Storchak et al. 2013) are shown as pink dots. Rivers are indicated in dark blue and inset globe shows the location of Fig. 1 within Asia. The blue polygon indicates the region shown in Fig. 2.

Open in new tabDownload slide

The northern Shanxi Grabens form a zone around 390 km long and 160 km wide (see Fig. 2), and are characterized by ENE striking normal faults and asymmetric half-grabens with footwall blocks tilted to the SSE, creating typical basin-and-range topography. Xu & Ma (1992) estimate a total of 8 km of crustal extension in the southwest of the zone and 1.5 km of crustal extension in the northeast of the zone, based upon the tilting of Pliocene-Quaternary sediments within the grabens. Initial sedimentation in the northern Shanxi Grabens commenced in the Miocene, but this earliest period of extension was confined to the Huaizhuo and Yanfan Grabens in the far northeast (see Fig. 2)—an exception to the general northward propagation of rifting. The main phase of rifting in the northern Shanxi Grabens occurred during the Pliocene and Quaternary (Xu & Ma 1992). Most of the grabens are also blanketed by a thick cover of yellow, Late Pleistocene loess (Li et al. 1998; Zhang et al. 2003).

Figure 2.

SRTM topography (30 m resolution) of the northern Shanxi Grabens (Farr et al. 2007). Faults, marked as thin black lines, have been mapped from satellite imagery (source: http://earth.google.com) based upon earlier fault maps from Tapponnier & Molnar (1977), Xu & Ma (1992) and Yu (2004). Cities are marked by blue squares and black stars show the positions of volcanic cones. Red outlines indicate the locations of later figures. Faults are numbered as follows: 1 Yunzhongshan Fault, 2 Wutaishan Fault, 3 Hengshan Fault, 4 Kouquan-Emokou Fault, 5 Laiyuan Graben Fault, 6 Lingqiu Graben Fault, 7 Yuguang Graben Fault, 8 Liulengshan Fault, 9 Yangyuan Graben Fault, 10 Yanggao-Tianzhen Graben Fault, 11 Nankou Fault, 12 Yanqing Fault, 13 Huailai Fault, 14 Huai’an Graben Fault, 15 Zhangjiakou-Bohai Fault, 16 Daihai Graben Faults (see also Table 3).

Open in new tabDownload slide

This paper is primarily concerned with the major normal faults in the northern Shanxi Grabens that accommodate the approximately northwest-southeast extension (see Fig. 2). The Wutaishan and Hengshan Faults dip to the northwest and border the northern sides of the Wutaishan (Wutai mountains) and Hengshan (Heng mountains) respectively. The Hengshan Fault also forms part of the southern boundary of the Datong Graben. The Kouquan-Emokou Fault dips southeast and runs along the northwestern side of the Datong Graben, behind the city of Datong. Further east, the Yuguang Graben Fault runs along the southeastern side of the Yuguang Graben and dips to the northwest. The Liulengshan Fault borders the northern and western sides of the Liulengshan (Liuleng mountains), running along the eastern side of the Datong Graben and the southern side of the subsidiary Yangyuan Graben, also dipping to the northwest. In the northeast of the northern Shanxi Grabens, the Yanqing and Huailai Faults dip southeast on the northwestern sides of the Huaizhuo and Yanfan Grabens. The northern boundary of the whole zone is the WNW striking, left-lateral Zhangjiakou-Bohai Fault.

In the Yuguang Graben a maximum sediment thickness of 600 m is reached on its southeastern side, adjacent to the Yuguang Graben Fault; in the Yangyuan Graben the maximum sediment thickness is less than 500 m; and in the Datong Graben sediment thicknesses reach 1500 m adjacent to the Kouquan-Emokou Fault (Xu & Ma 1992; Li et al. 1998). The Datong Graben also hosts tholeiitic and alkaline flood basalts and volcanic cinder cones (see Fig. 2). These rocks are largely Early Pleistocene in age and are thought to result from local mantle decompression during the ongoing crustal extension (Xu & Ma 1992; Xu et al. 1993).

4 METHODS

4.1 DEMs and offset measurements

We measure vertical offsets across fault traces by drawing swath profiles through digital elevation models (DEMs) at the relevant sites. At Guangling, on the Yuguang Graben Fault (see Section 5.1.2), we construct an ≈1 m horizontal resolution DEM from a stereo pair of 0.5 m Worldview-1 images, acquired on 2008 July 25 and 2011 September 21 (see Supporting Information for methodology). Elsewhere, we use 30 m horizontal resolution data from the Panchromatic Remote-sensing Instrument for Stereo Mapping (PRISM) of the Japan Aerospace Exploration Agency (JAXA) Advanced Land Observing Satellite (ALOS; Takaku et al. 2014; Tadono et al. 2014) or the Shuttle Radar Topography Mission (SRTM; Farr et al. 2007). At each site, all DEM points within a 100–300 m wide swath are projected onto the profile, and straight lines are fitted to the data above and below the scarp. The quoted errors on our vertical offset measurements refer to the root mean square (RMS) residuals from this line fitting process. Full details of our methodology are given in the Supporting Information.

4.2 Quaternary dating of offset features

In addition to the existing radiocarbon (¹⁴C), thermoluminescence (TL), optically stimulated luminescence (OSL) and K-Ar dates, we use infrared stimulated luminescence (IRSL) dating of sedimentary feldspar grains and Ar–Ar dating of basaltic feldspars to determine the ages of offset features. See the Supporting Information for methodologies. Unfortunately, we were unable to find further samples for ¹⁴C dating during our field visit. The age data collected in this study are summarized in Table 1.

4.3 Kostrov summations

where |$\dot{\epsilon}_{ij}$| is the ijth component of the strain rate tensor, μ is the rigidity modulus, τ is the time period of observation, and |$M_{ij}^{n}$| is the ijth component of the seismic moment tensor of the nth earthquake. We use Kostrov’s relation to calculate strain rate tensors for the northern Shanxi Grabens for different earthquake catalogues, different time periods, and different magnitude ranges.

On the basis of population density maps and Gutenberg-Richter statistics, the earthquake catalogues in northern Shanxi are thought to be complete for earthquakes with M_s ≥ 6 since 1291 (i.e. soon after the beginning of the Yuan Dynasty in 1271) and M_s ≥ 4.75 since 1484 (Huang et al. 1994). We therefore construct a catalogue of 25 earthquakes with M_s > 5 for the 700 yr period from 1315 to 2015 based on Lee et al. (1976), Lee et al. (1978), Liu et al. (2011) and the Global Centroid Moment Tensor (GCMT) catalogue (Ekström et al. 2012). We note that our catalogue is unlikely to be complete for events with 5 < M_s < 6 between 1315 and 1484, but we choose to include those earthquakes that are known about; in any case, their contribution to the total moment release will be relatively small.

Focal parameters for some of these earthquakes are available in the literature (Wesnousky et al. 1984; Pavlides et al. 1999; Liu et al. 2007) or the GCMT catalogue (Ekström et al. 2012). For the remaining historical earthquakes, we estimate focal parameters from nearby instrumental earthquakes or fault striation measurements on adjacent faults. In the second case, we assume that the sense of slip during the earthquake was the same as the Quaternary motion on the nearest adjacent fault. All 25 earthquakes and their associated parameters are listed in Table 2. We then perform a variety of sensitivity tests to check how variations in fault strike, fault dip, rake, magnitude, seismogenic thickness and time period affect our results. We also conduct Kostrov summations with the modern instrumental catalogue of Li et al. (2015a), which contains 36 earthquakes in the region of interest from between 1965 and 2014, with magnitudes (M_w) in the range 3.1 to 5.8—though this catalogue only extends as far as 114.5°E.

4.4 Strain rate field from GPS

The Crustal Motion Observation Network of China (CMONOC) was established in 1998 by the State Seismological Bureau of China (now the China Earthquake Administration) (Liu et al. 2007). Today, the network includes 260 continuous GPS stations and a total of 2056 temporary survey stations, which were visited most recently during field campaigns in 2009, 2011 and 2013 (Li et al. 2012; Zhao et al. 2015). We calculate a continuously varying strain rate field from the latest campaign GPS measurements (Zhao et al. 2015), comprising these three occupations, according to the method of Shen et al. (1996) and Floyd et al. (2010). The motivation for considering a continuous strain rate field is that a number of theories of continental deformation postulate that the whole lithosphere in a deforming region responds to stresses that are either applied at its edges or result from gradients in gravitational potential energy (Bird & Piper 1980; England & McKenzie 1982). It is well known that at small length-scales the crust must be divided by individual faults and that the earthquakes occurring on these faults represent discontinuous deformation (England & Molnar 2005; Floyd et al. 2010). As a result, an earthquake, and any of its subsequent post-seismic motion, that occurs during the time period of GPS observation will significantly perturb the calculated strain rate field. However, no large-magnitude events (M_s > 5) are recorded between 2009 and 2013; we also remove small discontinuities from our field by comparing the GPS data to a smoothed field as described below.

We then compare the velocities that we obtain from our inversion to the original GPS vectors and calculate a residual at each site. We remove sites with residual velocities greater than 1 mm a⁻¹ (i.e. that do not fit a smoothly varying strain rate field) to leave 57 sites in the region of interest (73 per cent of the original data). We then re-calculate the strain rate field on a regular 0.5° grid. In other words, we assume locally homogeneous but regionally varying velocity gradients. See Supporting Information for mathematical details of these calculations.

5 FAULT SLIP RATES FROM QUATERNARY GEOLOGY

5.1 Yuguang Graben Fault

The Yuguang Graben Fault is a large, arcuate, normal fault that borders the southeastern side of the Yuguang Graben, a basin which contains the cities of Yuxian and Guangling (see Fig. 2). The fault extends ENE-WSW for approximately 120 km (from 39.759°N, 114.005°E to 40.130°N, 115.108°E), dips to the northwest, and consists of multiple fault segments. Wang (2012) measure fault striations with a rake of −80° to −90°, indicating predominantly normal motion. Towards the western end of the fault the footwall lithologies include Cambrian and Precambrian limestones and dolomites, whilst at the eastern end the footwall is composed of Jurassic rocks (Wang 2012). The hanging wall is covered by deposits of Pleistocene and Holocene alluvial sediments.

5.1.1 Western end of the Yuguang Graben Fault

The first slip rate site is 8 km southwest of the city of Guangling towards the western end of the Yuguang Graben Fault, at 39.696°N, 114.245°E. Two fluvial terraces are preserved on the eastern side of a river exiting the range to the south, with measured heights of 9.2 and 13.2 m (Wang 2012). Wang (2012) also propose that remnants of a much higher terrace can be seen to the southwest of the river at heights of around 103.9 m above the hanging wall. However, during our fieldwork, we were not able to confirm the fluvial provenance of these sediments.

Wang (2012) obtain optically stimulated luminescence (OSL) ages for the lower two terraces of 3.93 ± 0.03 ka and 21.13 ± 1.28 ka respectively, giving throw rates of 2.34 ± 0.02 mm a⁻¹ and 0.62 ± 0.04 mm a⁻¹. For a 60° dipping fault (Li et al. 2015c), these correspond to extension rates of 1.35 ± 0.01 mm a⁻¹ and 0.36 ± 0.02 mm a⁻¹. We prefer the rate of 0.36 ± 0.02 mm a⁻¹ as it is based on a longer time period and hence more complete earthquake cycles.

5.1.2 Guangling section

The second site is 7 km south of the city of Guangling, where the most recently active fault segment cuts through the toe of an alluvial fan 1.4 km from the main range front near the village of Yixingzhuang (see Fig. 3). We measure topographic profiles across the scarp along lines A–A^΄ and B–B^΄ from the Worldview-1 DEM (see Section 4.1 and Figs 3c and d). We obtain an average vertical offset at this site of 13 ± 1 m. Fig. 3(e) is a field photograph of the triangular facets along the main range front, with the most recently active fault segment in the foreground.

Figure 3.

(a) Single-band (0.5 m resolution panchromatic) Worldview-1 image from 2011 September 21 of the Yuguang Graben Fault around 7 km south of the city of Guangling. Blue arrows mark position of the range-front normal fault, which dips to the northwest; red arrows mark the most recently active fault segment. Yellow circle marks the location of the trench opened by Wang (2012). (b) 1 m resolution DEM of the same area as (a) derived from a stereo pair of Worldview-1 images from July 2008 and September 2011. (c) 100 m wide swath profile A–A^΄. Dots show DEM points that fall within the swath. Red lines indicate best-fitting parallel lines to the upper and lower slopes. We measure a vertical offset of 12 ± 1 m. Note that the vertical exaggeration exacerbates the ≈3° difference in slope between the upper and lower surfaces—see the inset profile in the red box with reduced vertical exaggeration for comparison. (d) 100 m wide swath profile B–B^΄. Annotations the same as (c). We measure a vertical offset of 14 ± 1 m. (e) Field photograph (taken at 39.718°N, 114.312°E) looking southeast at the Yuguang Graben Fault. Blue arrows mark position of range-front normal fault; red arrows mark the most recently active fault segment in the foreground.

Open in new tabDownload slide

Wang (2012) excavated two palaeoseismic trenches at 39.710°N, 114.310°E, 1.5 km south of Yixingzhuang village. They identify two prehistoric earthquakes, the older of which occurred between 16.66 ± 1.39 and 10.62 ± 0.87 ka ago on the basis of OSL dates from the trenches. Wang (2012) also estimate a total throw in the two earthquakes of 7.0 m. Our 13 ± 1 m offset of the alluvial fan surface suggests that more than the two earthquakes visible in the trench have occurred at this site. Combining the ages with the throw of 7.0 m and a fault dip of 60° (Li et al. 2015c), we obtain a throw rate of between 0.42 ± 0.04 and 0.66 ± 0.05 mm a⁻¹ and an extension rate of between 0.24 ± 0.02 and 0.38 ± 0.03 mm a⁻¹.

5.1.3 Jiugongkou overlap zone

The final slip rate site on the Yuguang Graben Fault is on the 7 km long Jiugongkou fault segment overlap zone, 15 km east of Yuxian city at 39.880°N, 144.766°E. Within the overlap zone, a minor antithetic fault approximately 5 km long dips to the south (see Fig. 4). On the northeasternmost fault strand, 1.3 km southeast of the village of Maizituan, we measure a vertical offset of 17 ± 4 m (see Figs 4c and d). Fig. 4(e) shows an exposure of the fault in an adjacent river cutting. Cheng & Yang (1998) obtain a TL age of 26.0 ± 2.1 ka for the alluvial sediments at this site, suggesting a throw rate of 0.65 ± 0.16 mm a⁻¹.

Figure 4.

(a) SRTM topography (30 m resolution) of the Jiugongkou overlap zone on the Yuguang Graben Fault 15 km east of Yuxian city (Farr et al. 2007). Red arrows indicate active faults, including a small antithetic fault within the overlap zone. Yellow circle indicates the dating site of Wang (2012) (at 39.880°N, 114.766°E). (b) Google Earth imagery from August 2015 (source: http://earth.google.com) of the same region as (a). Westward growth of the northern fault segment and movement on the antithetic fault have caused uplift of a central horst block that has deflected the major river exiting the mountains. The dotted yellow line marks the position of the drainage divide within this horst block. (c) Zoomed view of the SRTM topography on the northern fault segment, showing location of the swath profile C–C^΄. (d) 150 m wide swath profile C–C^΄. Dots show DEM points that fall within the swath. Red lines indicate best-fitting parallel lines to the upper and lower slopes. We measure a vertical offset of 17 ± 4 m. (e) Field photograph (taken at 39.880°N, 114.766°E) looking west at the Yuguang Graben Fault exposed in a river cutting. (f) Field photograph (taken at 39.852°N, 114.769°E) looking east along the antithetic fault, showing uplifted dacite in the footwall.

Open in new tabDownload slide

On the antithetic fault a dacitic deposit covered by Pleistocene sediments is uplifted in the footwall. Cheng & Yang (1998) measure a vertical separation across the fault of ≈24 m and obtain a TL date for the overlying fluvial deposits of 27.4 ± 2.2 ka, giving a vertical uplift rate of 0.88 ± 0.07 mm a⁻¹. This faster uplift rate on the antithetic fault is consistent with the asymmetry of the intervening horst.

Adding the rates on these parallel strands and using fault dips of 60° (Li et al. 2015c) yields an extension rate at this site of 0.88 ± 0.10 mm a⁻¹.

5.2 Liulengshan Fault

The Liulengshan Fault runs along the northwestern side of the Liulengshan, an S-shaped mountain range in the centre of the northern Shanxi Grabens (between 39.577°N, 113.276°E and 40.256°N, 114.727°E), forming part of the eastern edge of the Datong Graben and the southern edge of the subsidiary Yangyuan Graben (see Fig. 2). The fault comprises numerous segments (each typically ≈25 km long), which all dip to the northwest, and is approximately 150 km long in total. Additional, possibly inactive, fault segments continue further to the northeast. Xu et al. (1997) measure striations with rakes of −77° to −90° towards the eastern end of the Liulengshan Fault, indicating that the fault undergoes predominantly normal motion, at least on this segment. The principal footwall lithologies are Precambrian metamorphic rocks and Palaeozoic limestones, whilst the hanging wall consists primarily of alluvial and fluvial fan deposits, shed from the growing range front (Xu et al. 1997). These sediments are interbedded with, and subsequently capped by, substantial loess deposits.

Two earthquake swarms, with maximum magnitudes of M_s 6.1 and M_s 5.8, occurred on the Liulengshan Fault in October 1989 and March 1991 respectively (Xu et al. 1997). The focal depths of these earthquakes are thought to be in the range of 10 to 20 km (Xu et al. 1993) and there are no reports of surface ruptures associated with these events. The GCMT mechanisms for the largest events suggest right-lateral strike-slip motion. With the exception of one other event in 1545, there is no other recorded seismicity on the Liulengshan fault. Such inactivity is further corroborated by dating fluvial fans and soils which cover the fault trace, but have not been offset. Xu et al. (1997) find that no surface-rupturing earthquakes can have occurred on the Liulengshan Fault for at least the last 1500 yr and in some locations for as long as 23 ka.

Xu et al. (1997) measure throw rates at nine different sites along the Liulengshan Fault using TL dating of offset features and, in one location, K–Ar dating of basaltic feldspars. Using a fault dip of 65° (Li et al. 2015c), they obtain extension rates of between 0.2 and 0.4 mm a⁻¹ for different locations along strike (see later Fig. 8). Here, we seek to corroborate these results by using IRSL dating on samples from an independent site, at Qifengpocun.

5.2.1 Qifengpocun

The first site on the Liulengshan Fault is at Qifengpocun, 17 km south of the city of Yangyuan (at 39.950°N, 114.138°E). Here, a recently active fault strand cuts through alluvial fans north of the range front (see Fig. 5). Two terraces are preserved on either side of a channel exiting the Liulengshan, though their surfaces have been heavily modified for agriculture. We measure a vertical offset for the second, older terrace of 18 ± 4 m high (see Figs 5c and d). The ALOS PRISM data is not sufficiently high-resolution to resolve the height of the lower terrace, but our field observations suggest a height of 3 to 4 m.

Figure 5.

(a) 30 m resolution ALOS PRISM topography of Qifengpocun in the centre of the Liulengshan Fault (Takaku et al. 2014; Tadono et al. 2014). Blue arrows mark position of the range-front normal fault; red arrows mark the most recently active fault segment. Yellow circles indicate the locations of IRSL samples 15 and 16. (b) Google Earth imagery from March 2014 (source: http://earth.google.com) of the same area as (a). (c) 200 m wide swath profile D-D^΄. Dots show DEM points that fall within the swath. Red lines indicate best-fitting (non-parallel) lines for the upper and lower slopes. We measure a vertical offset of 18 ± 4 m. (d) 200 m wide swath profile E–E^΄. Annotations the same as (c). We measure a vertical offset of 17 ± 4 m. (e) Perspective view of the Google Earth imagery looking south at the Liulengshan Fault. Two terrace levels (T1 and T2) are visible. Blue arrows mark position of range-front normal fault; red arrows mark the most recently active fault segment. (f) and (g) Field photographs (taken at 39.948°N, 114.138°E and 39.950°N, 114.138°E) showing IRSL samples 15 and 16, taken from the sides of the T2 and T1 terraces respectively.

Open in new tabDownload slide

We collect IRSL samples for dating from both terraces (see Figs 5f and g, and b for locations). IRSL 15 yields an age of 16.3 ± 1.9 ka; IRSL 16 yields an age of 15.9 ± 1.0 ka. The age of IRSL 15 and the offset of 18 ± 4 m gives a throw rate of 1.10 ± 0.28 mm a⁻¹ and an extension rate of 0.51 ± 0.13 mm a⁻¹; the age of IRSL 16 and the offset of 4 ± 1 m gives a throw rate of 0.25 ± 0.02 mm a⁻¹ and an extension rate of 0.12 ± 0.03 mm a⁻¹ (both for a 65° dipping fault (Li et al. 2015c)). These extension rates are similar to the previous rates of 0.2 to 0.4 mm a⁻¹ obtained by Xu et al. (1997). However, we also note that the similarity in the two IRSL ages suggest that they both come from the same deposit. This could be because IRSL 16 sampled an incised wall of T2 and the T1 deposits at this site are thinner than 2.13 m (see Fig. 5g). If this is the case, then we prefer 0.51 ± 0.13 mm a⁻¹ as the more reliable extension rate.

5.2.2 Jiankoucun section

Towards the eastern end of the Liulengshan Fault, the footwall topography dies away from ≈325 m to ≈50 m, indicating that this portion of the fault has experienced considerably less slip. The ≈50 m high escarpment continues eastwards, cutting across where the main drainage leaves the Yuguang Graben, and joins an older fault that strikes northeast immediately adjacent to the Sanggan River as it flows east out of the Yangyuan Graben. It appears, therefore, that this fault segment, with uplifted former basin sediments in its footwall, is linking two older fault segments that have larger total displacements (see Fig. 2). Furthermore, the uplifted basin sediments include some of the earliest evidence for the presence of humans in northeast Asia (1.36 Ma ago) at the Xiaochangliang palaeolithic site (40.221°N, 114.662°E) (Zhu et al. 2001). Fault linkage has occurred since this time, implying a throw rate of >0.04 mm a⁻¹.

At the eastern end of the Liulengshan topography (at 40.097°N, 114.468°E) a 1.8 km long section of bedrock fault plane is exposed (see Fig. 6a). The exposure is ≈40 m high and we find an average fault striation measurement of 238/68/298 (see Figs 6b and c). (This and other fault striation measurements are discussed in more detail in Section 8.3.) Further east, within the uplifted basin sediments, fault exposures are also visible (at 40.115°N, 114.499°E—see Figs 6d and e).

Figure 6.

(a) Field photograph looking southeast at an exposed bedrock fault plane on the Liulengshan Fault. Red arrows indicate the position of the fault. (b) Field photograph (taken at 40.097°N, 114.468°E) showing a closer view of the ≈40 m high bedrock fault plane. (c) Zoomed in photograph at the same location showing the polished fault plane surface. Black arrows highlight the orientation of striations (238/68/298). (d) Field photograph (taken at 40.115°N, 114.499°E) looking east at a road-cut exposure of the Liulengshan Fault in the uplifted basin sediments towards the eastern end of the fault. (e) Zoomed in photograph at the same location showing the orientation of the fault plane (255/60) and the calcite-cemented fault gouge.

Open in new tabDownload slide

5.3 Kouquan-Emokou Fault

The 100 km long Kouquan-Emokou Fault borders the northwestern edge of the Datong Graben running from 39.668°N, 112.728°E in the southwest to 40.408°N, 113.442°E in the northeast. The fault dips to the southeast causing uplift on the northwestern side of the fault trace. A major segment boundary occurs 10 km southwest of the city of Nanjiao at 39.944°N, 113.059°E (see Fig. 2). Yu (2004) describe the Kouquan-Emokou Fault as having both vertical and right-lateral components of motion, at suggested rates of 0.5 to 0.6 mm a⁻¹ (Yu 2004) and 3.6 mm a⁻¹ (Deng et al. 2007) respectively.

A M_s 5 earthquake was recorded on the Kouquan-Emokou fault, just to the west of Datong, in 1582 (Liu et al. 2011). Two further earthquakes, both of M_s 6.5, occurred in 1022 and 1305 further south within the Datong Graben (Liu et al. 2011). Although both epicentres lie within the city of Huairen, and a number of smaller faults are thought to cross the centre of the Datong Graben, it seems probable that these two earthquakes also occurred on the Kouquan-Emokou Fault and were simply recorded by the inhabitants of Huairen. A further earthquake with M_s 5.4 was recorded on 1991 March 25 within the urban area of Datong itself. The epicentral location for this event indicates that it probably took place on a separate, smaller fault 12 km southeast of the main fault (see Fig. 7a); the earthquake had a right-lateral strike-slip mechanism.

Figure 7.

(a) SRTM topography (30 m resolution) of the northern end of the Datong Graben (Farr et al. 2007). Red arrows mark the positions of active faults, including the Kouquan-Emokou Fault, which bounds the northwestern edge of the graben. The yellow circle marks the location of the Ar–Ar dating sample. (b) Zoomed view of the SRTM topography at the northern end of the Kouquan-Emokou Fault, where an escarpment is preserved in a Cenozoic basalt flow. (c) 300 m wide swath profile F–F^΄. Dots show DEM points that fall within the swath. Red lines indicate best-fitting lines for the upper and lower slopes. We measure a vertical offset of 119 ± 5 m. (d) Google Earth imagery from November 2015 (source: http://earth.google.com) of the same area as (b). (e) Field photograph (taken at 40.383°N, 113.401°E) of a quarry in the footwall of the Kouquan-Emokou Fault. The yellow circle indicates the position of a basalt sample taken for Ar-Ar dating. (f) Zoomed in photograph at the same location shows the basaltic lava flow overlying a baked palaeosol.

Open in new tabDownload slide

5.3.1 Tuotuoying, Datong section

At its northern end, the Kouquan-Emokou fault passes through the western side of Datong city and offsets a Cenozoic basaltic lava flow as shown in Fig. 7(a). We measure a scarp height adjacent to the village of Tuotuoying of 119 ± 5 m (see Figs 7b and c). A sample of the basalt was taken from a quarry in the footwall of the fault 2 km northwest of Tuotuoying (see Fig. 7d for location and Figs 7e and f). We obtain an Ar-Ar age for the basalt of 8.29 ± 0.27 Ma (see Supporting Information). This gives us a minimum throw rate of 0.014 ± 0.001 mm a⁻¹ and, for a 70° dipping fault (Li et al. 2015c), a minimum extension rate of 0.005 mm a⁻¹. Using a fault length of 16.72 km within the basalt, we are also able to determine a minimum horizontal propagation rate of 2.02 mm a⁻¹. It should be emphasized that these are minimum constraints on the extension rate and propagation rate; the basaltic lava dates to the Late Miocene and hence is older than much of the graben formation in northern Shanxi. The lava could have remained undisturbed for an extended period of time before the fault cut into it. In addition, the low extension rate could also be because motion at the northern end of the Kouquan-Emokou Fault is transferred onto the short, normal faults further east (see Fig. 7a).

5.4 Summary of fault slip rates from Quaternary geology

Table 3 summarizes all of the Quaternary rates from this study and the literature. Our re-assessment of some of these sites, along with our re-measurement of some of the offsets and our additional dating on the Liulengshan and Kouquan-Emokou Faults, confirms the reliability of the previous measurements.

All of the rates originally determined are vertical throw rates across normal faults. We therefore use fault dips from Li et al. (2015c) (which are largely based on field measurements—see their Table 2 and references therein) to convert throw rates to extension rates. Where dips are not given by Li et al. (2015c), we use the mean dip from the dip ranges quoted by Yu (2004). All of the extension rates are plotted in Fig. 8. Blue dots show locations from this study; red dots show locations from other studies; and black labels are rates from Yu (2004) and Deng et al. (2007), where the location of the actual measurement site is unknown. The calculated errors for individual throw rate measurements (from propagating offset and age errors in quadrature) are typically around ±0.1 mm a⁻¹—see sites described above. We also note that there is an additional uncertainty equivalent to one earthquake offset in each of these measurements since we do not know at which point in the earthquake cycle we are making our observations. However, the majority of these throw rates come from sites where the offset feature records multiple earthquake cycles and hence the influence of this particular uncertainty is small. Furthermore, these errors are not quoted in Table 3 or Fig. 8 because they are all notably smaller than the errors introduced by uncertainties in fault dips—see below.

Figure 8.

SRTM topography (30 m resolution) of the northern Shanxi Grabens (Farr et al. 2007). Faults, marked as thin black lines, have been mapped from satellite imagery (source: http://earth.google.com) based upon earlier fault maps from Tapponnier & Molnar (1977), Xu & Ma (1992), Yu (2004) and Deng et al. (2007). Blue circles (and numbers) indicate rates of horizontal extension measured in this study. Red circles (and numbers) indicate rates of horizontal extension measured by Cheng et al. (1995); Xu et al. (1997); Cheng & Yang (1998); Jiang et al. (2003); Ding et al. (2009); Wang (2012). Extension rates given in black are from Yu (2004); Deng et al. (2007). Faults are numbered as follows: 1 Yunzhongshan Fault, 2 Wutaishan Fault, 3 Hengshan Fault, 4 Kouquan-Emokou Fault, 5 Laiyuan Graben Fault, 6 Lingqiu Graben Fault, 7 Yuguang Graben Fault, 8 Liulengshan Fault, 9 Yangyuan Graben Fault, 10 Yanggao-Tianzhen Graben Fault, 11 Nankou Fault, 12 Yanqing Fault, 13 Huailai Fault, 14 Huai’an Graben Fault, 15 Zhangjiakou-Bohai Fault, 16 Daihai Graben Faults (see also Table 3).

Open in new tabDownload slide

The total extension across the northern Shanxi Grabens can be found by summing extension rates across parallel faults. For example, along profile A–A^΄ in Fig. 8 the total extension rate is ≈1.8 mm a⁻¹; along profile B–B^΄ the total extension rate is ≈1.1 mm a⁻¹. In general, there is a small decrease in extension rate on moving northeast through the region. For a 165 km wide zone (see Section 6.2), these extension rates correspond to strain rates of ≈11 nanostrain a⁻¹ and ≈6.7 nanostrain a⁻¹ respectively.

However, uncertainties in fault dips have a large effect on the magnitudes of these summed extension rates. Allowing for dip uncertainties of ± 15° (approximately equivalent to the ranges quoted by Yu (2004)), we obtain extension rates of 0.7 to 3.3 mm a⁻¹ along profile A–A^΄ and 0.6 to 2.0 mm a⁻¹ along profile B–B^΄. In other words, the dip uncertainties can lead to a potential halving or doubling of the summed extension rates.

6 STRAIN RATES FROM SEISMICITY

6.1 Rate of moment release in the northern Shanxi Grabens

Fig. 9(a) shows the magnitudes of all 25 historical earthquakes with M_w > 5 in the northern Shanxi Grabens over the last 700 yr (see also Table 2). No events with M_w > 6 have occurred since 1750. Fig. 9(b) shows 100, 200 and 300 year moving averages of the logarithm of the rate of moment release through time. There is a notable decrease in the rate of moment release (of around a hundred-fold, from 6.3 × 10¹⁷ Nm a⁻¹ to 7.9 × 10¹⁵ Nm a⁻¹) in the northern Shanxi Grabens at around 1750. More than 99 per cent of the total moment release over the last 700 yr occurred prior to 1750. This decrease is larger than the variations that can be accounted for with the errors of ± 0.3 on the M_w values. The implications of this observation are discussed further in Section 8.

Figure 9.

(a) Plot of magnitude (M_w) against time for historical earthquakes with M_w > 5 in the northern Shanxi Grabens over the last 700 yr. Earthquakes with M_w ≥ 6.0 are shown in darker blue. (b) Moving averages of the log of the seismic moment rate (in Nm yr⁻¹) for the last 700 yr, calculated every 20 yr. White dots show 100 yr moving average; red dots show 200 yr moving average; black dots show 300 yr moving average. Error bars, based on the assumption that values of M_w are known to the nearest ± 0.3 magnitude units, are shown for the 200 yr moving average. Dotted red line shows the log of the average M₀ rate (log (3.9 × 10¹⁷) = 17.6) for the whole 700 yr period, from 1315 to 2015.

Open in new tabDownload slide

6.2 Kostrov summations

Fig. 10 shows all of the earthquakes used for conducting Kostrov summations according to the methodology outlined in Section 4.3 and the Supporting Information. The full results are recorded in Table 4. The area of the region highlighted in Fig. 10 is 64 076 km². We use a shear modulus (μ) of 3.3 GPa and a seismogenic thickness (T_s) of 20 km for all calculations unless otherwise stated. We choose T_s = 20 km as this was the value used by Wesnousky et al. (1984) and because the epicentres of recent earthquakes are found to have not exceeded a depth of 22 km (Li et al. 2015b).

Figure 10.

SRTM topography (30 m resolution) of the northern Shanxi Grabens (Farr et al. 2007). Faults, marked as thin black lines, have been mapped from satellite imagery (source: http://earth.google.com) based upon earlier fault maps from Tapponnier & Molnar (1977), Xu & Ma (1992), Yu (2004) and Deng et al. (2007). The black rectangle indicates the approximately 164 km × 390 km region included in the Kostrov volume. Historical earthquakes occurring in the volume are shown as white circles scaled according to their magnitude and including the year in which they occurred (Lee et al. 1976; Liu et al. 2011)—see Table 2 for details. The two earthquakes ringed in red are the ones used by Wesnousky et al. (1984) in their Kostrov summation. Focal mechanisms are from Li et al. (2015a); mechanisms for earthquakes with M > 5 are shown in red. Strain rate crosses below the map show the results of Kostrov summations for (from left to right): all of the historical earthquakes with known focal parameters; the two earthquakes used by Wesnousky et al. (1984); all of the historical earthquakes with known focal parameters except the two used by Wesnousky et al. (1984); all of the mechanisms from Li et al. (2015a); and all of the mechanisms with M > 5 from Li et al. (2015a). Note that the first two strain rate crosses are displayed at one tenth of their real size. See Table 4 for further details.

Open in new tabDownload slide