Geomorphological signatures of the evolution of active normal faults along the Langshan Mountains, North China

Abstract Segmentation, propagation, and linkage of normal faults often occur in regions of active extension, and observations of the distribution and structural properties of segment boundaries can provide important insights for seismic hazard assessment. In this study, we carry out quantitative geomorphological analysis to evaluate the relative tectonic activity along the Langshan Piedmont Fault (LPF), which bounds the NW margin of the Hetao Graben, North China. On the basis of obtained morphometric indices (HI, BS, Smf, VF, SLK, and χ), tectonic knickpoint heights, footwall topography, and small unmanned aerial vehicles (sUAV)-based field observations, we demonstrate that: (i) The Langshan landscape is in a state of disequilibrium in response to active rock uplift and channel incision; (ii) The LPF consists of two major fault segments with lengths of 65 and 95 km, respectively, which likely have been linked with each other; (iii) Rupturing of the whole of one segment can generate an earthquake of M w ~7.3–7.5, and earthquake magnitude may reach M w ~7.8 if the entire fault trace of ~160 km is ruptured, posing a significant seismic risk in the western Hetao Graben. These findings would further our understanding of normal fault evolution through space and time in actively extending regions.

The continental rifts characterized by active normal faults around the tectonically stable Ordos Block    as a whole, or consists of several segments with distinct rupture histories. To date, only few studies on individual geomorphic indices have been carried out (e.g. Dong et al., 2018a;Liu et al., 2016), systematic morphotectonic analysis indicative of the level of tectonic activity in the Langshan region is still lacking.
In this study, we integrate the morphometric characteristics of drainage basins, channel long-profiles and range-fronts with the features of tectonic landforms mapped from small unmanned aerial vehicles (sUAV)-acquired DEMs during field surveys, to evaluate the relative tectonic activity along the LPF. As a result, we demonstrate the segmentation and likely linkage of this major fault, and discuss seismic potentials in the study area. Meanwhile, the results of this study would be helpful in better understanding structural Zhang, 2018;Jiang, Xiao, & Xie, 2000;Rao, Chen, Hu, Yu, & Qiu, 2016;Rao, Lin, & Yan, 2015;Zhang, Jia, Wang, & Zheng, 1985;Zhang, Yang, Zhong, & Mi, 1995;Rao et al., 2014Rao et al., , 2017Rao et al., , 2018State Seismological Bureau [SSB], 1988;Xu, Ma, & Deng, 1993;Zhang, Mercier, & Vergély, 1998). In contrast to other rift-border faults with earthquakes of M ≥7, there is an obvious lack of historical records about strong earthquakes along the Langshan Piedmont Fault (LPF) bounding the northwest margin of the Hetao Graben (Figures 1 and 2). However, existing paleoseismic data demonstrate that the ~160-km-long LPF has been active during the Holocene and is capable of generating ground-rupturing earthquakes (Dong et al., 2018b;Li, Ran, Chen, Wu, & Lei, 2015;Rao et al., 2016Rao et al., , 2018SSB, 1988). But, it remains unclear whether this major basin-bounding fault moves simultaneously   bedrocks and Quaternary sediments, and within the sediments with an average dip of 60° (SSB, 1988;Li et al., 2015;Rao et al., 2018). The estimated vertical slip rates (throw rates) of ~0.5-2.2 mm/yr based on relative chronology of tectonic landforms are with significant uncertainties due to the general lack of precise age constraints (Deng, Cheng, Min, Yang, & Ren, 1999;Huang, Zhang, Li, Liu, & Feng, 2012;Rao et al., 2018). The fault strike changes near its center from ~N35°E to the southwest to ~N50°E to the northeast (Figures 1(c)) and 2). Even if there is no recorded earthquake of M ≥ 7 in the western Hetao Graben (Figure 1(a) and (c); China Earthquake Networks Center [CENC], 2017; Deng, 2007;SSB, 1988;Wen, 2014), recent paleoseismological investigations demonstrated that ground-rupturing earthquakes have occurred on this fault during the Holocene with average recurrence interval of ~2450-2500 years (Dong et al., 2018b;Li et al., 2015;Rao et al., 2016Rao et al., , 2018, and some of them likely had ruptured the whole fault trace (Dong et al., 2018b). Topographically, the high relief and steepness on the southeastern flank of the Langshan Mountains are in contrast to the gentler slopes on its northwestern flank as revealed by the 5-km wide ridge-perpendicular swath profiles ( Figure 3). The range height, relief, and evolution of active normal faults through space and time.

Regional setting
The NE-trending Langshan Mountains in North China are bounded by the Inner Mongolian Plateau to the northwest and the Hetao Graben to the southeast, respectively ( Figure 1). The latter is one of the seismically active graben systems around the tectonically stable Ordos Block consisting of pre-Mesozoic basements (Deng, 2007;SSB, 1988). In the Cenozoic, crustal extension prevails and has rebuilt the Late Mesozoic Langshan orogenic belt, resulting in the development of a series of normal faults along the range-front and piedmont, as well as the subsidence and thick sediments of the Hetao Graben (SSB, 1988;Sun, Hu, & Niu, 1990;Zhang et al., 1998). In the study area, the exposed strata include the Pre-Archean, Archean and Proterozoic granite, diorite, amphibolite, plagiogneiss and granulite, Jurassic breccias and Cretaceous gravels, and Paleogene to Quaternary deposits composed of diluvial, alluvial, and lacustrine sediments along the piedmont and valleys ( Figure 2).
The LPF is a Holocene active fault dominated by normal slip, which develops along the contacts between values Willett, McCoy, Perron, Goren, & Chen, 2014;Yang, Willett, & Goren, 2015), we demonstrated the tectonic activity and segmentation behaviors of the LPF. In this study, swath topographic profiles were built using the SwathProfiler, an ArcGIS add-in shared by Pérez-Peña et al. (2017).

Hypsometric curves and hypsometric integral (HI) values
The shapes of hypsometric curves usually reflect distinctive stages of topographic development of drainage basins in terms of equilibrium between relief building forces and erosion processes (Figure 4(a)): convex, sigmoidal, and concave shapes for youthful, mature, and older basins, respectively. The HI index describes the relative volume that has not been eroded (Schumm, 1956;Strahler, 1952), and is defined as the area below a hypsometric curve (Figure 4(b); Alipoor, Poorkermani, Zare, & El Hamdouni, 2011;Mayer, 1990;Pike & Wilson, 1971): In general, higher HI values often correlate with a younger topography that might be affected by active tectonics, and lower HI values reflect an older landscape that have been eroded for a long time and hence seldom influenced by recent tectonism. Accordingly, it has been widely used to evaluate the level of tectonic activity in a specific region (e.g. Alipoor et al., 2011;Mahmood & Gloaguen, 2012;Ntokos, Lykoudi, & Rondoyanni, 2016;Wu et al., 2014).
mean slope increase from SW to NE, reach the maximum values on the profile C-C' , and slightly decrease adjacent to the Seertengshan Mountains ( Figure 3). Based on the optically stimulated luminescence (OSL) dating of river terraces in the Langshan area, Jia et al. (2015) demonstrated that uplift was uneven between ~58 and 42 ka, being faster in the central and northeastern portions, whereas it became almost constant in space after ~32 ka.

Study methods
DEM-based morphotectonic analysis was carried out in order to identify tectonic signals from landscape topographic features of the Langshan Mountains. Firstly, channel networks were extracted from 90-m spatial resolution Shuttle Radar Topography Mission (SRTM) DEMs using ArcGIS ArcHydro Tools. As a result, 27 drainage basins of >28 km 2 and their main channels have been acquired. Then, morphometric indices and their distribution features were analyzed to explore first-order alongstrike variations of resulting structures. In particular, we utilized the CalHypso Tools developed by Pérez-Peña, Azañón, and Azor (2009a) to obtain hypsometric curves and hypsometric integral (HI) values. Geological information was also compiled to decipher the effect of lithological contrasts on shapes of channel longitudinal profiles. Consequently, major lithological and tectonic knickpoints are mapped and differentiated. In addition, we carried out field surveys to verify the results of our geomorphological analysis. Combining with the obtained SLK values interpolated using kriging method (Pérez-Peña, Azañón, Azor, Delgado, & González-Lodeiro, 2009b), the Langshan topography (relief, slope, and width) and drainage Chi (χ) HI lack of active structures or the decrease of deformation rate (Bull & Mcfadden, 1977). The BS index describes the elongation ratio of a drainage basin and is expressed as follows (Figure 4(c); Cannon, 1976):

Drainage basin shape (BS)
The shape of a drainage basin may also record the influence of tectonic processes. In general, significant elongation in basin shape characterizes active tectonic regions, while sub-circular basins are indicative of the notes: the lithology on the profile is extracted from the geological map in Figure 2. the knickpoints at Sites 1 and 2 marked on the profile R3 are shown in Figure 11.

Ratio of valley floor width to valley height (VF)
The VF index describes the shape of a valley cross section, and is defined as the ratio of the width of a valley floor to its average height (Figure 4(e); Bull & Mcfadden, 1977;Bull, 1978). It is expressed as: where V fw is the average width of a valley floor; E ld and E rd are the elevations of the divides on the left and right sides of the valley, respectively; E sc is the average elevation of the valley floor. In general, steep and narrow (V-shape) valleys (VF ≤ 1) indicate relatively high uplift rates, and flat-floored (U-shaped) valleys (VF >1) reflect relatively low uplift rates (Bagha et al., 2014;Keller, 1986;Keller & Pinter, 2002;Özsayın, 2016). However, their shapes may significantly change if measured at different locations along the stream channels (Ramirez-Herrera, 1998). Therefore, the VF values have often been measured within a given distance upstream from catchment outlets, facilitating the comparisons among different stream channels (Mahmood & Gloaguen, 2012;Silva et al., 2003). Considering the drainage basin sizes of the study area (Table 1), the distances of ~0.5-1.0 km from catchment outlets were chosen.
where B l is the straight-line distance from headwater to mouth, and B w is the maximum width of a drainage basin. As a consequence, higher BS values reveal more elongated basins.

Mountain front sinuosity (Smf)
The Smf index reveals the sinuosity of mountain fronts, reflecting the balance between the stream erosion processes that tend to create sinuous mountain fronts, and the active vertical motions that tend to produce straight mountain fronts (Figure 4(d); Bull & Mcfadden, 1977;Keller, 1986). It was defined by Bull (1978) as: where L mf and L s are the real and straight-line lengths of a mountain front, respectively. Thus, lower Smf Smf = L mf ∕L s

Chi (χ) analysis
The parameter χ is sensitive to the loss of drainage area and disequilibrium across drainage divides Willett et al., 2014;Yang et al., 2015). Since χ removes the effect of drainage area, it can be compared among catchments of different sizes. Theoretically, water divides tend to move away from channels with low χ values towards channels with high χ values (Willett et al., 2014). Construct of χ map follows a specific protocol, and high resolution

Normalized stream-length gradient (SLK)
The gradient of a stream channel results from a long-period (10 4 -10 6 years) adjustment between the erosion and the tectonism (Demoulin, 2011;Hack, 1973;Pizzuto, 1992). Hack (1973) defined the stream-length gradient (SL) as: ΔH / ΔL is the average gradient of an evaluated channel segment, and L is the total length from the divide to its midpoint (Figure 4(f )). SL is sensitive to the variations of rock uplift rates, lithology, climate, and hydrology; high SL values often reflect high tectonic activity and/or high rock resistance (Azor, Keller, & Yeats, 2002;Brookfield, 1998;Font, Amorese, & Lagarde, 2010;Hack, 1973). In order to eliminate the influence of channel length, a normalized stream-length gradient (SLK) index is defined (Chen, Sung, & Cheng, 2003;Pérez-Peña et al., 2009b), and follows the equations: where K is a parameter representing the SL index calculated for the entire channel, H total and L total are the altitude   (Figure 7). Specifically, some of them are situated at lithological contacts, i.e. lithological knickpoints as shown in Figure 2, which are excluded in the subsequent analysis. Non-lithological knickpoints markedly change their heights along the trend of the Langshan Range, ranging from 39 m to 392 m (Figure 8).
DEMs if available have often been used (Willett et al., 2014). We firstly extracted hydrological information (flow direction, flow length, and flow accumulation) in ArcGIS Spatial Analysis Tools from 30-m spatial resolution Advanced Spaceborne Thermal Emission and Reflection Radiometer Global Digital Elevation Model (ASTER GDEM) data (the highest resolution now available for us), a joint product of the Ministry of Economy, Trade, and Industry (METI) of Japan and the United States National Aeronautics and Space Administration (NASA). Then, in combination with a Matlab script shared by Sean Willett, χ values of each streams were calculated with a reference concavity of 0.45, and a critical area of 0.5 km 2 to define the threshold for the minimum contributing upstream drainage area.

Morphotectonic indices
In the study area, 18 of the total 27 hypsometric curves of drainage basins are convex in shape, and the other 9 are sigmoidal ( Figure 5). The obtained HI values range from 0.42 Figure 11. Field observations of knickpoints at Sites 1 and 2 along an active channel (R3 in Figure 7). the man with a 2-m tape for scale.
note: See Figures 1(c) and 2 for the locations.
heights of ~2-4 m ( Figure 11). Considering the same rock type (granite) crops out both up-and downsides of the knickpoints, their formation is attributed to tectonic uplift rather than differential rock resistances. Sites 3-9 situated along the Langshan piedmont provide evidence of active deformation along the LPF. At site 3, a NE-trending scarp can be clearly recognized from the hillshade DEM with a resolution of ~0.12 m constructed from sUAV-acquired photos using structure-for-motion techniques (Figure 12(a)). The scarp is estimated to be ~13.7 m in height from DEM-based topographic profile (Figure 12(b)). Besides the mapped fault scarp, a bedrock fault plane dipping to the southeast has also been observed in the field (Figure 12(c) and (d)). Site 4 is situated near the center of the LPF, where two identified fault traces are nicely displayed by a panoramic photo (view to the northwest) (Figure 13(a)). Near the Langshan range front, striations on bedrock fault plane demonstrate a dominant normal slip-sense (Figure 13(b) and (c)), which are consistent with previous field observations The along-strike variations are also prominent in map view of SLK values: the highest in basin Nos. 5-8 and the lowest in basin Nos. 12 and 27 ( Figure 9). The obtained χ map reveals obvious differences across the main water divides, i.e. the channels on the northwestern side are with higher values than that on the southeastern side (Figure 10), suggesting that the river networks draining the Langshan Mountains are in a disequilibrium state and need to adjust their drainage area through divide migration or river capture to bring the divide to a stable position (Willett et al., 2014).

Tectonic landforms
To validate the results of our morphotectonic analysis, field surveys were also carried out along the southeastern flank of the Langshan range, where pronounced tectonic landforms have been observed. At Sites 1 and 2 along a southeast-flowing river (R3 in Figure 7), two knickpoints identified from stream long profile have been observed in the field with throw divides tend to migrate from the steep southeastern flank towards the gentler northwestern flank. Accordingly, the drainage networks in the Langshan Mountains are in disequilibrium state probably in response to active rock uplift and channel incision. This inference is supported by the dominant V-shaped valleys in elongated drainage basins suggestive of intensified incision (Figure 6(b) and (d)). Nevertheless, the obtained major channel gradients with high SLK values are mostly concentrated near catchment outlets (Figure 9), suggesting the landscape topography is mostly affected by recent deformation along the range fronts that have low Smf values (<1.4; Figure 6(c)). The tectonic activity of this basin-bounding fault inferred from geomorphological analysis is also verified by the widespread fault scarps mapped in the present  and previous active tectonic studies (Dong et al., 2018b;SSB, 1988;Li et al., 2015;Rao et al., 2016Rao et al., , 2018. However, along-strike variations of the tectonic activity are also pronounced as revealed by the fluctuations of obtained morphometric indices ( Figure 6; Table 2). In combination with the Langshan landscape topography (Figure 17), two major fault segments (SW and NE, respectively) have been designated with a boundary approaching the drainage basin No. 12 based on the following observations: (1) changes of the range front and fault trends from ~N35°E in the southwest to ~N50°E in the northeast, serving as a geometric boundary (DuRoss et al., 2016;Machette et al., 1991;Manighetti et al., 2015;Zhang et al., 1991); (2) tectonic activity is higher in the centers of each segments (as delineated . Along the piedmont, sUAV-acquired DEMs and associated topographic profiles indicate that the fault-generated scarps at this site range in height from 3.7 to 14.9 m (Figure 14).
Further northeast, both escarpments and small scarplets generated by faulting have been observed at Site 5 (Figure 15(a)), and the latter was likely associated with single earthquake event. At site 6, three levels of river terraces (T1-T3) are formed (Figure 15(b)), where higher terraces have larger displacements, indicating slip accumulation along the fault. Meanwhile, fault scarps are also pronounced near both fault tips. Near the northeast extremity, prominent scarps were observed at Site 7 (Figure 15(c)). In the southwest, although the Langshan range progressively decreases in elevation, fault scarps can still be traced in the field (Figure 16).

Discussion
All the obtained hypsometric curves are convex-and S-shaped with high HI values (average value of 0.56; Figures 5 and 6(a)), indicating that drainage basins are with low to intermediate maturity (El Hamdouni et al., 2008;Keller & Pinter, 2002;Pérez-Peña et al., 2009a). It is consistent with the observed transient rivers containing both tectonic and lithological controlled knickpoints as revealed by the channel longitudinal profiles and associated SLK values (Figures 7 and 8). The Langshan topographic features (relief and slope steepness; Figure 3) combined with χ values (Figure 10) suggest that water of the resulting faults if there is no significant change of base level on either side of the range (Densmore, Dawers, Gupta, & Guidon, 2005).
Segment boundaries may play significant roles in perturbing or arresting earthquake rupture, whereas large earthquake is still able to rupture smaller structural discontinuities (Zhang et al., 1999). Due to stress feedback between adjacent fault arrays, they might finally become linked with each other (Cowie, 1998;Cowie & Roberts, 2001). Combining with existing data, the recent paleoseismic study by Dong et al. (2018b) demonstrated that three Holocene earthquakes might have ruptured the entire fault trace, indicating the linkage of fault segments in terms of rupture behaviors. This scenario likely has been recorded by river incision processes, as the rates previously were uneven in the Langshan area but became spatially constant after ~32 ka (Jia et al., 2015). Nevertheless, due to the lack of historical records of earthquakes in the study area and the uncertainties by the high HI, BS, and SLK values, and low Smf and VF values), and bilaterally decreases towards their tips (Figures 6 and 9), demonstrating a general bell-shape in each segments; (3) distribution of major non-lithological controlled knickpoint heights that probably mirrors the trends in the total throw and footwall relief (e.g. Kent et al., 2017), demonstrates the same tendency ( Figure 8); (4) footwall local relief resembles the maximum, mean, and minimal elevations on the 5-km-wide ridge-parallel swath profiles (Figure 17(b)), which reflects different footwall landscapes (higher and steeper in the northeast segment) if considered as a proxy of fault throw (e.g. Densmore, Dawers, Gupta, Guidon, & Goldin, 2004;Kent et al., 2016;Whittaker & Walker, 2015); (5) the contrasting basin relief and mean slope indicative of differential channel incision, being higher in the northeast segment (Figure 17(c)); (6) variable range half-widths, being wider in the northeast segment (Figure 17(d)), which together with mountain reliefs reflect the geometry and spacing   As the LPF can be categorized into two major fault segments with lengths of ~65 km and ~95 km, respectively, seismic potentials of the Langshan and neighboring regions can be evaluated using the empirical relationship between the moment magnitudes (M w ) and the surface rupture lengths (SRL) of active normal faults worldwide (Wells & Coppersmith, 1994): M w = 4.86 + 1.32 * log (SRL) of age constraints, we suggest that future work is still needed to determine whether the demonstrated earthquakes occurred as temporal clustering, or could be correlated as the same event (McCalpin & Nishenko, 1996;Rockwell et al., 2000). More precisely determined fault slip rates on multiple time scales together with complete rupture histories of each segments would further our understanding of structural evolution of this fault. If rupturing the whole of one segment, earthquake magnitude would be M w ~7.3-7.5. It may reach M w ~7.8 if the entire ~160-km-long fault trace is ruptured by single event. In summary, the LPF is able to generate an earthquake of M w ~7.3-7.8, posing a significant seismic risk in the western Hetao Graben. These findings would be helpful in better understanding seismic potentials associated with this major boundary fault.

Conclusions
On the basis of GIS morphotectonic analysis and sUAVbased field observations along the Langshan Mountains, North China, we have reached the following conclusions.
(i) The Langshan landscape is in a state of disequilibrium in response to the rock uplift and channel incision on its southeastern flank affected by the active LPF; (ii) Morphometric indices together with range topography indicate that the LPF consists of two segments with fault lengths of 65 and 95 km, respectively, which likely have been linked with each other; (iii) Rupturing of the whole one segment of the LPF will generate an earthquake of M w ~7.3-7.5, and earthquake magnitude may reach M w ~7.8 if the entire fault trace of ~160 km is ruptured, posing a significant seismic risk in the western Hetao Graben.
We suggest that future investigations integrating precise constraints on the slip rates and paleoseismic histories of each segments would be helpful in better understanding the evolution of this basin-bounding normal fault through space and time.