Introduction

The Lower Rhine Graben (LRG) is one of the most seismically active regions in central Europe. Yet, the Holocene and historical surface deformation history for most of the faults in this low-strain region remains enigmatic, limiting our understanding of the seismogenic potential of these structures and associated seismic hazards.

In addition to the spatiotemporal distribution of paleoearthquakes (e.g. Stein and Friedrich 2014), a long-standing debate has centred on whether active faults in central Europe deform by coseismic rupture or slow aseismic slip (Ahorner 2001; Camelbeeck and Meghraoui 1996, 1998; Houtgast et al. 2003; Vanneste and Verbeeck 2001b). In contrast to arid regions (e.g. Friedrich et al. 2004), differentiating coseismic from aseismic fault motion in the sedimentary record is difficult in populated and humid low-strain regions, because fault-slip rates are often <0.1 mm/year, whereas surface degradation processes and anthropogenic overprint occur at equal or significantly greater rates (Bork and Lang 2003; Cerdan et al. 2010). In the LRG, mining-related subsidence and anthropogenic activities associated with the high population density have severely modified the landscape. As a result, fault scarps are rarely preserved.

In this study, we present results from a paleoseismological excavation across the Schafberg fault—a minor normal fault in the SW-portion of the LRG (Figs. 1, 2)—and the observed sediment deformation exposed in the trench. We focus on deformed coarse clastic layers characterized by fractured and rotated clasts in the damage zone. The main objectives of the study were to map the deformation features across the fault zone, place timing constraints on the latest paleoearthquake(s) and to identify features indicative of coseismic versus aseismic deformation.

Fig. 1
figure 1

Regional setting and seismotectonic framework of a the Rhine Graben rift system and b the Lower Rhine Graben. Instrumental and historical earthquakes from (Leydecker 2011; Hinzen and Reamer 2007; Meidow 1994) fault data from (Vanneste et al. 2013). BG Bresse Graben, HG Hessian Graben, LRG Lower Rhine Graben, URG Upper Rhine Graben; BEE Beegden fault, BEL Belfeld fault, BEN Benzenrade fault, BIR Birgel fault, BOC Bocholt fault, ECK Eckelrade fault, ELS Elsloo fault, ERF Erft fault, ERP Erp fault, FEL Feldbiss fault, GEL Geleen fault, GEU Geulle fault, GRO Grote Brogel fault, KAL Kalken fault, KAS Kaster fault, KIR Kirspenpicher fault, KON Koningsboscher fault, KOT Kotenforst fault, LAU Laurensberg fault, LUE Luelkener fault, MUE Muenstergewand fault, NEE Neerbeek fault, PEE Peelrand fault, REP Reppel fault, ROE Roettgener fault, RUR Rurrand fault, SAN Sandgewand fault, SCH Schafberg fault, STE Steinstrass fault, STO Stockheimer fault, TEG Tegelen fault, VEG Veghel fault, VEL Veldhovenfault, VIE Viersener fault, WEG Wegberger fault, WIS Wissersheimer fault, ZAN Zandberg fault. Map projection: UTM Zone 31, Datum: WGS84. Digital elevation model derived from a SRTM data (90 m resolution) and b AsterDEM data, 30 m resolution

Full size image
Fig. 2
figure 2

Geormorphological overview of the Schafberg and Birgel faults. Shaded relief map derived from 1 m LiDAR data provided by Geodatenzentrum NRW (Ⓒ Geobasis NRW 2010)

Full size image

Regional setting

The LRG is the northwestern branch of the European Cenozoic Rift System (Ahorner 1962; Illies 1975; Ziegler 1992), and covers an area of ~3.600 km2 encompassing northeastern Belgium, the southern Netherlands and northwestern Germany (Fig. 1a).

Extensional faulting in the LRG initiated in the Oligocene along NW–SE striking normal faults representing reactivated structures from older sedimentary basins of Carboniferous and Mesozoic age (Ahorner 1962, 1975; Hinzen and Reamer 2007).

The central and eastern sector of the LRG is characterized by large SW-dipping border faults and records of up to 100 and 180 m of vertical Quaternary displacement along the Erft and Rurrand faults, respectively, corresponding to Quaternary fault-slip rates of <0.1 mm/year (Geluk et al. 1994; Gold et al. 2017). In contrast, the southwestern sector of the LRG consists of complex NE-dipping normal fault systems, including the Feldbiss-, Sandgewand, Münstergewand and Schafberg-Birgel fault zones (Ahorner 2001; Gold et al. 2017) recording <50 m of vertical Quaternary displacement (Ahorner 1962; Holzapfel 1904; Fliegel 1922; Quitzow and Vahlensieck 1955).

Historical earthquakes and paleoseismicity in the Lower Rhine Graben

Historical records for the LRG document the occurrence of at least 21 M > 5 events (Leydecker 2011). The most significant historical earthquakes of the LRG include the ML 6.2 Düren earthquake of February 18, 1756 (Fig. 1b) and the M W 5.4 Roermond earthquake of April 23, 1992. The 1756 event is the strongest documented historical earthquake of Germany and—in line with the 1356 Basel earthquake (M L 6.0–6.5, Meghraoui et al. 2001)—one of the most destructive earthquakes of Central Europe. The 1992 event occurred in the central LRG and is the largest instrumentally recorded earthquake in central Europe (Leydecker 2011). This earthquake motivated regional studies in the LRG focused on constraining the frequency of large, surface-rupturing earthquakes (Haak et al. 1994). Paleoseismic studies showed that the instrumentally and historically documented earthquakes do not represent the largest possible events for the LRG and that earthquakes with M w 6.5–6.9 are likely to have repeatedly occurred since the Mid-Pleistocene (Camelbeeck and Meghraoui 1996, 1998; Camelbeeck et al. 2007; Houtgast et al. 2003, 2005; Lehmann et al. 2001; Vanneste et al. 1999, 2001; Grützner et al. 2016). Estimates for the maximum earthquake magnitude of the LRG of are on the order of M w 7.1 ± 0.3 (Ahorner 2001, Hinzen and Fleischer 2007; Vanneste et al. 2013).

Setting of the Untermaubach trench site

This investigation focuses on the 16-km-long Schafberg fault in the southwest LRG at the transition between the LRG to the east, and the northern Eifel—the NW part of the Rhenish Shield—to the west (Fig. 1b). The fault is part of the Schafberg–Birgel fault system—a pair of sub-parallel NE-dipping normal faults (Fig. 2). Ahorner (1962) and Quitzow and Vahlensieck (1955) reported combined vertical offset of several hundred metres of the Devonian basement and Triassic cover rocks of the Eifel across these faults. Between the villages of Bogheim and Untermaubach (Fig. 2) offset measurements and drilling surveys reveal Lower Pleistocene gravels (main terrace) with vertical offsets ranging from 10 to 15 m (Quitzow and Vahlensieck 1955). The geomorphic expression of the Schafberg fault is most obvious south of the village of Straß (Fig. 2) exhibiting a vertical surface offset of several tens of metres (Fig. 2). South of Untermaubach, at a small right step of the NW–SE striking normal fault, the Rur river valley traverses the Schafberg fault where it cuts through Devonian rocks of the Rhenish Shield (Quitzow and Vahlensieck 1955; Richter 1962; Ahorner 1962). On the valley floor, two, sub-parallel, 0.8–1.2-m-high east-facing scarps cross the Holocene floodplain surface (Fig. 3b) coincident with the projection of the Schafberg fault (Kübler et al. 2017). As we describe below, the eastern scarp is interpreted as a fault scarp and the western scarp is a fluvial escarpment. To the south, the scarps disappear in the active river bed; to the north, the scarps are overprinted by adjacent road construction and dense vegetation cover. The trend of the two scarps differs: the southwestern scarp strikes 125° and is curvilinear, whereas the northeastern scarp is almost perfectly straight, striking 155° (Fig. 3c). The subtle scarps affect the hydrologic conditions at the study site. In dry summers, plants growing in areas across the scarp wilt, whereas in the spring the water table is high east of the scarps (Kübler et al. 2017). Satellite images illustrate contrasts in vegetation coincident with the scarps (Fig. 3c). This observation is further supported by shallow geophysical measurements carried out prior to our trenching study (Kübler et al. 2017). The deposits at the site comprise terrace gravel of the youngest terrace sequence (lower Terrace) of the Rur valley (Klostermann 1992) overlain by Holocene and present-day flood deposits. The modern soil on the Holocene floodplain consists of silty sand with clay-rich layers and abundant organic material (Fig. 3). The study site is situated on farmland (Fig. 3d) and is 200 m northwest of the modern channel of the Rur river; therefore, the morphology of the fluvial terrace and any possible tectonic imprint on the landscape are likely overprinted by both fluvial and anthropogenic processes.

Fig. 3
figure 3

Setting of the Untermaubach trench site. a Shaded relief map (1 m LiDAR data) of the site showing the projection of the Schafberg fault into the Holocene flood plain. The colour gradient emphasizes subtle two subtle topographic scarps; b satellite image (SPOT/CNES, RGB, image courtesy Google Maps©) of the trench site. Note the difference in colour of the pasture corresponding to variations in subsurface hydrology. c Photo of the Untermaubach trench, view to west-southwest, showing the final excavation stage and the drainage system connecting six wells around the trench installed to lower the groundwater table below the level of the trench basis at −4.5 m

Full size image

Methods

We excavated an 80-m-long trench with a maximum depth of 4.5 m and a maximum width of 9 m (Fig. 3d) across the suspected fault scarps with the goal of determining offset across the fault, and to characterize the age and nature of deformation related to scarp formation (Fig. 3c). As requested by the landowner we first removed and separately deposited the upper humus-rich soil horizon. The groundwater table at the trench site is ~1.5 m below the ground surface. We lowered the water table to 5 m below the surface using six 8-m-deep wells equipped with wet pit pumps. We use these well logs to determine the top of the Devonian basement. At some locations, the trench was additionally excavated by hand to a depth of ~5 m. The trench walls were installed at an inclination angle of 45° to avoid trench wall collapse. All trench logs are corrected to vertical from the 45° inclination angle to facilitate offset measurements. For further details on trench excavation, see (Kübler 2013; Kübler et al. 2017).

For logging purposes, we produced photo mosaics of both trench walls. We flagged contacts and sample locations and mapped the sedimentary deposits by describing the composition, lithology and sedimentary fabric. We further mapped the vertical displacements of sedimentary contacts across the damage zone as well as deformation features including fissures, joints and rotated and fractured clasts exposed in the gravel deposits. We mapped the location of the 237 fractured pebbles as well as the orientation of the fracture planes on both trench walls. We further mapped the long-axis orientation of all clasts larger than 5 cm (approximately 10.000 clasts per trench wall) in the field using a geological compass and on the photomosaics. From orientation data we produced rose plots and stereonet plots using stereo32©.

We collected samples for radiocarbon dating including charcoal, plant remnants and wood from critical sediment units and contacts. Samples were prepared and measured via Accelerator mass spectrometry by Leibniz-Laboratory for Radiometric Dating and Isotope Research, Kiel, Germany (samples 1–6), and by Gliwice Radiocarbon Laboratory, Gliwice, Poland (samples 7–21).

To estimate the age of surface deformation we modelled the depositional ages of sedimentary units below and above identified event horizons. We applied Bayesian statistical analysis using OxCal to integrate radiocarbon dates and stratigraphic and structural information to constrain earthquake timing (Bronk Ramsey 2013; Reimer et al. 2013).

Results and interpretation

Excavation stratigraphy

The excavation exposed a 5- to 6-m-thick cover of unconsolidated fluvial deposits overlying Devonian sedimentary rocks (Fig. 4a, b). The exposed sediments are overprinted by secondary precipitation of iron and manganese oxides, visible as horizontal-to-subhorizontal orange and black layers in the trench mosaic (Fig. 4a). The distinction between individual sedimentary units using the photo mosaic is thus very difficult without additional field observations. The basement rocks on both sides of the fault consist of platy mud and sandstone of the Heimbach Formation (Upper Lower Devonian; Knapp and Hager 1980). We subdivided the unconsolidated sediments into four units: A–D, from oldest to youngest, respectively. The lowermost deposit (unit A) consists of a ~1.5-m-thick layer of clast-supported coarse gravel with predominately prolate clasts up to 60 cm in diameter in a silt- and clay-rich matrix (Fig. 4b). Clasts primarily consist of Triassic sandstone (Buntsandstein) and Devonian shale, both of the Heimbach Formation. Few clasts consist of Devonian quartzite, presumably of the Ems Formation (Upper Lower Devonian). The top of unit A contains interspersed clay lenses with abundant organic material, which we interpret as a reworked soil horizon (unit A′, Fig. 4b). This horizon is overlain by unit B, a 1- to 2.5-m-thick inhomogeneous clast and matrix-supported coarse gravel layer with interbedded sand and gravel lenses. The lithological composition of the clasts does not significantly differ from that of unit A, but here the clasts are smaller and more rounded. Maximum clast sizes are <30 cm. Unit B is overlain by a matrix-supported clayey-and sandy-to-fine gravel layer (unit C, Fig. 4b) with well-developed zones of cross-bedded sand lenses. The overlying unit C contains predominately sandstone gravels and few mudstone gravels, and entirely lacks quartzite gravels, in contrast to units A, and B, respectively. Furthermore, the trench exposes two channels, which coincide with the position of the observed surface scarps. The channel fill consists of sandy silt- and clay-rich layers (unit D).

Fig. 4
figure 4

Results of the trench analysis. a Photo mosaic of the northern trench wall. The solid black line indicates the original surface before excavation and removal of the upper soil horizon. The colour change of the sedimentary deposit from orange (0–19 m) to grey (19–46 m) to orange (46–80 m) is due to uneven colour balance of the photo mosaic and does not reflect a change in the sedimentary deposits. Trench log of the main units and deformation features of b the west-northwestern part and c the east-southeastern part of the northern trench wall. d Stereonet diagrams showing the fracture plane orientations of all broken clasts (left), and of fracture planes in the fault zone (right)

Full size image

The eastern channel has a total width of 12 ± 2 m and a maximum vertical thickness of 2.5 m (Fig. 4c). The channel exhibits internal horizontal stratification characterized by variations in the content of organic matter, clay and Mn- and Fe-oxides. The base of the channel lies below the present-day groundwater table and displays features typical of reducing gley soil horizons (Gr) with a greenish-blue soil colour due to the presence of ferrous oxide (FeII) (Blume et al. 2016). This horizon contains a high concentration of organic matter, including retransported plant remains, charcoal and wood. The overlying horizon is characterized by the presence of Mn- and Fe-oxides in the form of black and reddish concretions ranging in diameter from <1 mm to up to 2 cm.

The western channel is 20 ± 1 m wide with a maximum sediment thickness of 3 m. The channel is asymmetric, with a steep western flank (45° ± 5°) and a shallower eastern flank (15° ± 3°). The base of this channel lies below the groundwater table and exhibits reducing gley horizons similar to the eastern channel. The overlying deposits predominately consist of clayey silt with a few isolated clasts exhibiting different zones of vertical and lateral clay migration caused by fluctuations of groundwater and percolating water.

Sediment deformation

Identification of fault-related deformation within the gravel deposits is hampered by the coarse-grained nature of the gravel deposits and sediment alteration from groundwater fluctuation. Detailed mapping of the gravel deposits at the scale of individual pebbles reveals a 10-m-wide central damage zone characterized by fractured and offset clasts (Fig. 4c, d) and rotated clasts (shear fabric, Fig. 5), which occurs directly below the eastern surface scarp. A less prominent, but still recognizable damage zone of fractured and rotated clasts extends 10–15 m east and west of the main fault zone (Fig. 4c).

Fig. 5
figure 5

Summary of gravel deformation observed in the trench exposure; a long-axis inclination angles of clasts >5 cm. Grey lines indicate inclination angles of 0°–60°; red lines indicate inclination angles of 61°–90°. Rose diagrams above trench log show the clustered distribution of the long-axis orientation of clasts at different locations. b Fractured and offset Buntsandstein-pebble showing pure dip-slip displacement, offset ~5 cm; fracture plane dips to the NE, fractured pebble is exposed in a zone characterized by numerous rotated clasts; c sub-vertically fractured Buntsandstein-pebble, fracturing perpendicular to internal bedding; d sub-vertically fractured quartzite pebble; e fractured Buntsandstein-pebble, fracturing sub-parallel to internal bedding, fracture plane dips tot he SW

Full size image

Most clasts record pure dip-slip fractures; however, a subset of the clasts record a left-lateral strike-slip component. We observed fracturing parallel and perpendicular to internal bedding (Fig. 5). Fractured pebbles occur only in units A and B. At the contact of units A and B, a partly degraded soil horizon (unit A′, Fig. 4) has developed on sheared and cataclasized pebbles of unit A. The majority of fractures occur in sandstone clasts (n = 214 clasts, 90.3%), the percentage of fractured quartzite and mudstone clasts are similar (mudstone: n = 11 clasts; 4.6%; quartzite: n = 12 clasts, 5.1%).

Individual fractured clasts in the fault zone are typically cut by one to five sub-parallel fractures. Over a distance of ~10 m, fractured pebbles are frequent within the unconsolidated gravel deposits and become significantly less frequent with distance away from the fault zone. Discrete fractures are absent in the surrounding matrix. However, sub-vertical zones of rotated clasts coincide with the location of fractured pebbles, representing fracture zones in the gravel deposits (Fig. 5). Fracture-parallel offsets in the clasts occur on about 15% of the fractures, offsets range from few mm to up to 6 cm.

Fractures in the gravels are generally oriented along a consistent trend. Most of the fractures dip to the northeast (mean dip direction: 41°) with mean inclination angles around 60° (Fig. 4d). This trend is even more pronounced in the fractures exposed in the ~10-m-wide fault zone below the surface scarp. A second, smaller population of fractures is oriented in the opposite direction with dip directions predominately to the SW and dip angles around 75° (Fig. 4d).

We identified liquefaction features in fine-sand and silt layers of unit C including sill and dyke structures (Dirkzwager et al. 2000; Duin et al. 2006; Vanneste et al. 2010), and small-scale folding of intact sand layers (Fig. 6a, b). Liquefaction is predominately restricted to unit C, with the exception of 15 × 5 cm small injection feature exposed at the contact between unit C and D, respectively (Fig. 6b). Whether this feature represents subsurface injection after deposition of unit D, is difficult to judge as the quality of exposure is limited.

Fig. 6
figure 6

Liquefaction features in sandy deposits of unit C, and the contact between unit C, and D, respectively a Dm-scale folding of fine-sand and silt deposits; b “sills-and-dykes” structure (Obermeier 1996) of fine-sand and silt layers injected into coarser deposits. Examples shown in (a) and (b) are exposed at the western head of the trench

Full size image

Vertical offset of sedimentary contacts is constrained by three markers (Fig. 4c, Online Resource 4)—the contact between units A and B (1.15 ± 0.25 m), the contact between units B and C (0.9 ± 0.2 m), and the surface scarp (0.8 ± 0.15 m). The contact between fluvial deposits (unit A) and Devonian basement provides an additional and complimentary constraint. This contact is exposed only at one location in the trench, at the eastern end of the fault zone (Fig. 4c). However, drill cores derived from the installation of wells for groundwater lowering provide basement depths at two further locations east of the fault zone of 5.8 and 5.9 m, respectively (Fig. 4c). These measurements indicate that the basement is offset ~1.2 m, with an uncertainty of 0.1–2 m caused poor drill core recovery from the screw drill device, and thus is on the same order compared to the offset measurements of displaced Holocene sediments.

Deformation is best described by diffuse step faulting with each minor fault accommodating a few mm–cm of vertical slip. Subtle warping/folding may also be an important process that occurs at a scale beyond our detection threshold. No further exposures of the Schafberg fault facilitate along-strike comparison to evaluate rupture style and fault complexity. However, the sharp and linear fault morphology north of Untermaubach suggests a simple fault geometry for most of the fault length.

Paleoearthquakes

We document evidence for one or two paleoearthquakes. In detail, seismogenic faulting at the trench site has vertically offset gravel units A and B. In unit C we identify evidence for liquefaction, but find no signs of coseismic rupture. One explanation for these relationships is that unit C was deposited prior to the surface-rupturing earthquake but did not preserve diagnostic features, because the high sand content this unit did not favour formation of liquefaction features comparable to the gravel units A and B. Other explanations are that unit C was deposited after the latest rupturing event of the Schafberg fault, and liquefaction is the result of a later event that did not rupture the surface or faulting occurred at a neighbouring source causing liquefaction in the surrounding area including the Untermaubach site. Unit D entirely lacks deformation features such as discrete fault cracks and forms an erosional contact with unit B below the eastern scarp (Fig. 4b, c). Owing to its clay-rich nature and overprinting from soil formation and groundwater fluctuation, unit D could have been affected by faulting without preserving evidence of seismic motion. Coseismic deformation postdating deposition of unit D is supported by two observations, the small injection feature between the contact of unit C and D (Fig. 6b), and the presence of a surface scarp above the fault zone affecting unit D. However, strong surface modification from present-day fluvial and anthropogenic activity makes it difficult to unambiguously connect the surface scarp with tectonic activity of the Schafberg fault. In fact, the fluvial channel at the western end of the trench also produced a surface scarp. Further, the lack of earthquake-related deformation structures makes it difficult to prove that this deposit pre-dates the most recent surface-rupturing earthquake (MRE). It is unclear whether the contact between unit C and D represents the same paleoearthquake or a younger event leading to liquefaction in unit C. We consider all scenarios in our analysis, below.

Fracture density and the number of fractured gravels in unit A is higher compared to unit B. A few transgranular fractures terminate at the contact between unit A and A′. This could indicate an older event predating the deposition of unit A′. Alternatively, these upward fault terminations could have resulted from rheological differences between these two units. We prefer this latter, simpler explanation.

Age control

We used radiocarbon dating to date 21 organic samples including detrital charcoal, wood and plant remnants from throughout the stratigraphic section exposed in the trench (Fig. 4, Online Resource 1, 2). The base of unit D was particularly rich in plant remnants, presumably deposited under reducing conditions. We found wood samples at the base of unit D at the western channel as well as within a sand lens at the top of unit B, below the eastern channel. In both cases, the branches are several dm to up to 1 m long, and up to 10 cm thick. The wood samples lacked a sufficient number of tree rings for dendrochronological analysis.

Radiocarbon analyses yielded mixed results. 9 of 21 samples yielded ages that are apparently either too young (sample 3, unit B) or too old (samples 5, 11, 13, 16, 17, 18, 19 overlaying units C and D, respectively) in their stratigraphic context (Online Resource 1, 2). Radiocarbon samples from unit D in the western fluvial channel are systematically older than those of the underlying units. Anomalously young sample age likely result from unrecognized root penetration and the old ages are attributed either to fluvial transport and redeposition of plant remnants at the base of a river channel (samples 11, 12, 16) or to recycling of detrital charcoal from deposits on the forested slopes surrounding the river valley (samples 5, 13, 17, 18, 19). These problematic samples are excluded in the subsequent analysis.

No datable samples were identified in the coarse gravel deposits of unit A, so we lack information on the age of this deposit. We interpret unit A′ to represent an interval of subaerial exposure and soil development, which leads to the formation of a clay-rich soil horizon that was partially eroded before or during deposition of gravel unit B. Three radiocarbon ages (No. 1, 2 and 6, Online Resource 2) constrain the depositional age of unit A′ to be 7589–6568 year BP. The depositional age of unit B is constrained by six radiocarbon ages (No. 7, 9, 10, 14, 20, 21) to be 1523–975 year BP. Radiocarbon ages bracket the depositional age of unit C to be 927–800 year BP (No. 12). Radiocarbon ages in unit D range from 782 to 669 year BP (No. 4, 8).

Earthquake-event scenarios at the Untermaubach site

We model three scenarios (Figs. 7, 8, Online Resource 3) and report 95% confidence interval bounds (~2 σ equivalent): model 1 assumes two earthquakes—a surface deformation event postdating the deposition of unit B, followed by a liquefaction event (on the Schafberg fault or a neighbouring source) postdating the deposition of unit C. The surface deformation event is constrained at 1135–835 year BP (815–1115 AD) and the paleo-liquefaction event at 1114–720 year BP (836–1230 AD) (Figs. 7, 8). Model 2 assumes a single earthquake causing fracturing and rotation of clasts in unit B and liquefaction in unit C, postdating the deposition of unit C (Fig. 7) constrained at 1114–720 year BP (836–1230 AD). Model 3 assumes a single earthquake causing surface rupture and liquefaction postdating the deposition of unit D. For this model deposition of unit D constrains the maximum age of the ground-rupturing earthquake of ≤671 year BP (1279 AD, Figs. 7b, 8).

Fig. 7
figure 7

Timing constraints on most recent surface-rupturing earthquake (MRE) and paleo-liquefaction event at the Untermaubach trench site on the Schafberg fault using OxCal v4.2.4 (Bronk Ramsey 2013; Ramsey 2009); IntCal13 atmospheric curve (Reimer et al. 2013). a Complete sequence of modeled radiocarbon ages listed in stratigraphic order from bottom to top, without depth constraints. b Time constraints for three alternative event age models; detailed reconstruction of MRE shown in Fig. 8

Full size image
Fig. 8
figure 8

Reconstruction of sedimentation, erosion, and earthquake-related sediment deformation events at the exposed fault segment and three alternative models interpreting the reconstruction of the most recent faulting event

Full size image

Unfortunately, we do not find compelling evidence to rule out any of the three models. The strong contrasts of the style and intensity of sediment deformation in units A and B, compared to units C and D, particularly the lack of discrete fracturing in the finer grained units C and D, support models 1 and 2, respectively. However, the varying intensity of deformation could also have been caused by varying grain size spectra in the deposits leading to discrete shearing and clast fracturing in the coarse gravel deposits and diffuse deformation and liquefaction in the finer sand and silt deposits. The presence of a surface scarp affecting unit D (Fig. 4) above the fault zone as well as the observed injection structure at the contact of units C and D (Fig. 6b), respectively, supports model 3. However, in the temperate climate of the Lower Rhine Graben region, fault scarps are prone to be modified by fluvial activity and fault parallel river incision has been commonly observed. Evidence for subsurface injection after deposition of unit D is restricted to one poorly exposed feature (Fig. 6b). Furthermore, liquefaction in unit C could have also resulted from an earthquake on a neighbouring fault, e.g. the Birgel or Rurrand faults, ~3, and ~10 km, respectively, east of the trench site.

Discussion

We test the robustness of our results and discuss the significance of our study in light of the context of ongoing debates on spatiotemporal fault-slip patterns in low-strain regions as well as possible deformation mechanisms of coseismic deformation in gravel deposits overlying bedrock faults.

Coseismic displacement and seismogenic potential of the Schafberg fault

Compared to other potential seismic sources identified in the LRG (Vanneste et al. 2013) the Schafberg fault is a relatively small structure with a potential surface rupture length of 15 ± 2 km. Rupture of the entire Schafberg fault would produce a M Wmax ~6.5 earthquake (Wells and Coppersmith 1994), which is a reasonable magnitude for surface-rupturing earthquakes reported in other intracontinental regions (e.g. Bonilla 1988; McCalpin 2009a). For such magnitudes, vertical displacement of 1.0 ± 0.2 m measured in our study exceeds surface displacements typically expected for M W ~6.5 earthquakes (Wells and Coppersmith 1994). The phenomenon of short faults producing large offset (“short-fat-fault syndrome”) has been described in paleoseismic studies in the western USA (Taylor et al. 2010; McCalpin 2005). Another possible explanation for the large displacement observed along the Schafberg fault could be that the rupture co-ruptured with another, adjacent fault, resulting in a longer surface rupture with a larger displacement. Candidate adjacent structures include the 10-km-long Birgel fault and/or the >40-km-long Rurrand fault, both located to the west of the westward dipping Schafberg fault. The eastward dipping Rurrand fault is a particularly likely structure to interact with the Schafberg fault within seismogenic depth. Previous paleoseismic studies imply that the Rurrand fault has been active in the Holocene (Vanneste and Verbeeck 2001a; Grützner et al. 2016). However, poor age control does not allow for direct correlation of the paleoearthquakes between the Rurrand and Schafberg faults.

Association with the 1756 Düren event and other historical earthquakes

Using OxCal age modelling of radiocarbon ages we present three different scenarios I the time range between 1135 and <671 BP. Time ranges for models 1 and 2 do not overlap with the 1756 AD Düren earthquake, but suggest an older event, presumably of similar magnitude (~M 6–6.5). Historical earthquake catalogues list two candidate earthquakes in the Aachen/Düren region in 872 and 1234, both with intensities of VII (Leydecker 2011). Neither of these events has been linked to destruction comparable to the 1756 AD event; however, sparse documentation of events of that age may have resulted in underestimates of coseismic damage. Given that model 3 only provides a maximum event age of 671 year BP allows the possibility in this model that the MRE observed in the trench exposure was produced by the 1756 Düren event. However, as previously discussed not all paleoseismological observations support this model. Regardless of the exact age bracketing of the MRE our trenching results have for the first time successfully identified coseismic faulting associated with an event that occurred in historical times in the LRG.

Are fractured clasts detectors of coseismic rupture?

This study reveals detailed observations of transgranular fracturing in unconsolidated gravel deposits. Based on the systematic arrangement and orientation of fractured clasts in the exposed gravel deposits we suggest a tectonic origin for the clast fracturing. Among others, previous studies reported on tectonically fractured clasts in unconsolidated near-surface gravels in the Alhambra hill in the Eastern Granada Basin, southern Spain (Azanón et al. 2004), western Taiwan (Lee et al. 1996), the Feldbiss fault zone in northeastern Belgium (Demoulin 1996), and the Rex Hills area in the Eastern California Shear Zone (Baran et al. 2010). In addition, a recent trenching study across the 2011 Darfield earthquake rupture (Hornblow et al. 2014) reports on complex gravel deformation including fractured and rotated clasts.

Although the excavation record in our study lacks “classic” paleoseismological criteria used to identify coseismic surface rupture and displacement, such as colluvial wedges and free faces, our data provide important information on the issue of dynamic surface rupturing versus protracted, slow fault motion. This distinction is of particular importance for studies in the Lower Rhine Graben. Here, seismological and paleoseismological research has long centred on a debate, whether faults in this region are in general capable to produce large surface-rupturing earthquakes or rather slip aseismically as a result of slow fault-creep or induced by groundwater disposal from open-pit mining (Ahorner 1962; Camelbeeck and Meghraoui 1996, 1998; Görres 2008; Görres and Kuhlmann 2008; Hinzen and Reamer 2007; Houtgast et al. 2003, 2005; Vanneste and Verbeeck 2001a, b; Vanneste et al. 2001).

We interpret the fractured gravels to have resulted directly from coseismic rupture. However, here we consider alternative mechanisms that could have led to the fractures, including overburden, regional shaking, and frost weathering.

Fractured clasts occur at minimum depths of ~0.5 m below the modern ground surface, so the effect of overburden can be neglected as a cause of gravel fracturing. Our observations challenge earlier studies on fractured clasts, which hypothesized that a vertical overburden stress on the order of 10 MPa, corresponding to a sedimentary thickness of approximately 400 m, would be necessary to produce in situ fracturing of pebbles similar to those we observe in the trench (Eidelmann and Reches 1992). Alternatively, seismic shaking could generate fractures and overcome the shear resistance of the pebbles. However, seismic shaking would have caused widely distributed fractured pebbles in the gravelly layers of unit A and B. Instead, the fractured pebbles are concentrated in the immediate vicinity of the surface scarp and the offset sedimentary units (Fig. 4c). These observations suggest that the clast fractures resulted from the rupture itself and not from shaking. Second, alternative mechanisms for pebble fracturing, such as freeze–thaw cycles under permafrost conditions (e.g. Matsuoka and Murton 2008), rearrangement of pebbles during the formation of ice-wedges (e.g. van Vliet-Lanoë et al. 2004), or clast fracturing from sub-glacial shearing (Hiemstra and van den Meer 1997) are not viable because this region did not experience glacial and/or permafrost conditions during the Holocene (Davis et al. 2003; Klostermann 1992). As we are lacking constraints on the depositional age of unit A, this unit could have been affected by permafrost processes. Further, an alternative scenario leading to localized freeze–thaw processes in the Holocene, would be downward infiltration of water underneath a swale, which may fracture clasts as a result of deep freeze where the gravels are water saturated. However, the dominant trend of fractures is clearly aligned with the orientation of the fault, which strongly supports a tectonic rather than frost weathering related origin. Frost weathering would produce fractures that are controlled by water migration pathways along internal planes of weakness within the clasts such as bedding or cleavage (Matsuoka and Murton 2008) resulting in a pattern of randomly oriented fractures in the gravel deposit. In summary, we interpret the fractured and offset gravels to have resulted directly from coseismic deformation and distributed fault motion at the trench site.

Distributed deformation of the Schafberg fault

The lack of localized displacement along a single fault surface, as well as the wide distribution of deformation in the trench (Fig. 4), differ from exposures in other trench studies carried out along normal faults in the Lower and Upper Rhine Graben (Camelbeeck and Meghraoui 1996, 1998; Houtgast et al. 2003, 2005; Lehmann et al. 2001; Peters et al. 2005; van den Berg et al. 2002; Vanneste and Verbeeck 2001a, b). We suggest this difference is rooted either in the local geology or in the coseismic rupture process. In the setting of the Untermaubach site, a propagating coseismic rupture and the associated seismic waves encounter a large contrast in material properties at the bedrock/fluvial deposit interface. Although we lack constraints on the geometry of the fault structure at depth below the Untermaubach site, it is reasonable to assume that the Schafberg fault localizes as a single fault strand at in the basement rocks. However, once an earthquake rupture transmits energy across the interface between bedrock and unconsolidated sediment, the rupture may disperse onto numerous small fracture branches due to increased granular deformation and low confining pressure processes in unconsolidated near-surface sediments. In contrast, in thick sedimentary deposits, as documented for the central Lower Rhine Graben, such boundary effects between bedrock and sediment are likely less significant due to higher confining pressure at greater depth, while in water-saturated sediments as observed at our study site, dispersion effects may be enhanced (e.g. Stoll 2002; Sessarego et al. 2008; Cheng 1986)

Monoclinal scarps have also been observed in cohesive gravels elsewhere (Crone et al. 1987; McCalpin 2009b). Importantly, such scarps are not typically associated with classic colluvial wedge deposits because they accommodate displacement over numerous minor faults without exposing a free face large enough to produce recognizable colluvial wedges. No colluvial wedge was observed at the Untermaubach trench. Further studies on the deeper fault structure of this and other faults in the LRG along with investigation of deformation textures in unconsolidated sediments in different regions may yield insight into these problems.

Conclusions

Paleoseismological investigations at the Untermaubach site in the LRG reveal evidence for one surface-damaging earthquake on the Schafberg fault and an additional, possibly coeval paleo-liquefaction event. Surface deformation at the trench site is characterized by broad warping of Holocene fluvial gravels over a damage zone above the Schafberg fault. The large-scale warping is the result of distributed deformation recorded by fractured, faulted, and rotated pebbles. Our detailed analysis of these fractured pebbles highlights the potential utility of these features for documenting coseismic deformation in fault zones with rheological contrasts between mechanically strong and overlying, unconsolidated weak sediments.

OxCal modelling of radiocarbon ages of sediments exposed in the trench constrain the timing of the youngest seismic event to be younger than ~1100 year BP, indicating that the Schafberg fault has produced a coseismic faulting event in historical times. Our observations on the distributions and contact relationships of gravel deformation, sediment liquefaction and surface displacement allow for three alternative models to reconstruct the latest surface-rupturing event. While the earthquake timings of models 1 and 2, respectively, suggest that the youngest surface faulting occurred 1135–720 year BP (815–1230 AD), model 3 suggests surface faulting after 671 year BP (1279 AD) and thus possibly overlaps with the 1756 AD Düren earthquake. Two older historical earthquakes are potential candidates to have been associated with the paleoearthquake(s) documented in this study. Due to sparse documentation of older historical earthquakes, the magnitude of such events may have been underestimated (Leydecker 2011; Hinzen and Reamer 2007; Meidow 1994).

The Schafberg fault is characterized by a relatively large coseismic displacement with regard to its maximum length and a large Holocene displacement compared to its total displacement. To improve constraints on coseismic rupture processes as well as spatiotemporal patterns of tectonic activity in this region, further paleoseismic and tectono-geomorphic studies in the Lower Rhine Graben are necessary.