Large Earthquakes in Northern Europe More Likely Than Once Thought

Eos Vol. 77, No. 42, October 15, 1996, pp. 405, 409. . © 1996 American Geophysical Union.
Permission is hereby granted to journalists to use this material so long as credit is given, and to teachers to use this material in classrooms.

T. Camelbeeck and M. Meghraoui

Northern Europe would typically be classified as a relatively stable continental region where seismic hazard is low. Indeed, in comparison to active zones in the Mediterranean, seismic activity is low in northern Europe and the Lower Rhine graben region. Until recently, no one would have suspected that the normal faults related to graben structures in the Lower Rhine embayment could produce large earthquakes. Yet, on April 13, 1992, an Ms = 5.3 earthquake struck the city of Roermond in the Netherlands. This moderate earthquake is considered to be one of the major historical seismic events in this region, and it is stimulating concern among seismologists and geologists. Fundamental questions are being asked about fault behavior in the region and the earthquake distribution in time and space.

The Royal Observatory of Belgium recently launched a seismotectonic program to estimate the true seismic hazard. A major aspect of this program was to identify active faults and conduct associated paleoseismic studies along the Belgian side of the Roer Valley, which is the northwestern branch of the Rhine graben system (Figure 1). After a thorough field investigation using geologic and geomorphological methodologies coupled with geophysical prospection (radar and electric tomography), we were able to identify active faults and correlate Holocene surface faulting with historical and prehistorical earthquakes in trenches.

The Roer Valley, which crosses three countries (Belgium, The Netherlands, and Germany), is bounded by two north-northwest, south-southeast trending Quaternary normal fault systems. The eastern boundary is defined by the Peel boundary fault, along which the 1992 Roermond earthquake occurred [Camelbeeck and van Eck, 1994] , and the western boundary is defined by the Feldbiss fault zone, which is partly located in Belgium. Evidence of recent tectonic activity along the Feldbiss fault zone is visible on seismic profiles that show more than 600 m of offset in Neogene deposits [Demyttenaere and Laga, 1988] . Although Ahorner [1975] demonstrated the existence of the " ...Rhenish seismoactive zones... " and recommended a comprehensive analysis of Quaternary structures and background seismicity, coseismic movements were considered to be improbable, and active faults remain largely unidentified.

Fig. 1. Major Quaternary faults along the lower Rhine embayment. The Roer graben is bounded on the northeast by the Peel fault, which produced the 1992 Roermond earthquake (Ms = 5.3), and on the southwest by the Feldbiss fault, along which the paleoseismological investigations were conducted near the city of Bree. Two focal mechanisms correspond to the largest seismic events that have occurred in the area during the last 20 years [Camelbeeck, 1993]. Note the northwest-southeast striking seismicity and the normal-faulting solution related to the Roermond earthquake.

Active Faulting and Evidence of Coseismic Displacements

Are the Quaternary normal faults along the Roer graben capable of producing an earthquake larger than the Roermond event? Paleoseismology is a fundamental tool for addressing this critical question and relating tectonic activity and seismicity in this area. On the basis of previous seismotectonic studies [Camelbeeck, 1993] , we conducted detailed geological and geomorphological investigations of fault scarps in the Roer graben, searching for the slightest evidence of late Pleistocene and Holocene movement along the main Quaternary faults.

Near the town of Bree (Belgian Limburg, Figure 1), and along the Feldbiss fault zone, a prominent north-northwest, south-southeast trending fault scarp separates the Campine Plateau to the west from the Roer Valley to the east [Paulissen et al., 1985] . The geomorphic expression of the scarp consists of ~10-km-long escarpment that has 15 20 m of vertical topographic relief (Figure 2). In detail, the scarp is multiple, and the frontal fault trace consists of an en echelon geometry that suggests a component of left-lateral slip. The fault dips 75 NE and offsets young deposits and alluvial terraces in the flat valley. Leveling profiles along the frontal fault scarp yield an average vertical displacement of 0.6 m, which may correspond to the last coseismic event along the Bree fault (Figures 2 and 3a).

Two 80-m trenches were excavated across the Bree fault scarp to characterize the recent faulting and related paleoearthquakes ( Figures 3b and 3c). The two trenches revealed outstanding coseismic features including faulted and flexured Holocene alluvium, evidence of at least three historical and prehistorical seismic events, liquefaction structures, and minor normal faulting. Despite the 2.5-m-deep water table in trench I, two well-sorted gravel horizons intercalated with sandy-clay deposits were excavated that have 0.5 m of vertical net slip ( Figures 3a and 3b); the alluvial deposits are flexured, and the maximum displacement is 0.80 m. This offset agrees with the geomorphic measurements of the fault scarps, indicating the presence of at least two tilted and displaced surfaces with a maximum vertical offset of 1.70 m at this site that are likely due to previous coseismic displacements (Figure 3a). Paleosols and sedimentary deposits containing organic matter, charcoal and peat deposits suggest the last coseismic event occurred between 432 and 941 A.D.

In the second trench (Figure 3c), the fault cuts a succession of coarse gravels with an intercalated clay horizon, a sandy unit showing a prominent flexure, and colluvial deposits. Near the surface, sandy-gravel lenses (dated at 402 206 B.C.) are cut by the fault and juxtaposed against a dark brown, fine, sandy clay that likely represents the last colluvium deposited. In the downthrown block, fine-grained sand and clay, upon which admixed coarse gravels and sandy clay form channel structures, contain minor normal faults and prominent liquefaction features. Some normal faults cut the fluvial deposits and show evidence of at least three paleoearthquakes. The paleoseismic information joining both the trench and geomorphic observations provides clear evidence of surface faulting with a historical coseismic displacement, and cumulative offsets probably due to prehistoric large earthquakes.

Fig. 2. Topographic map showing a ~10-km section of the Bree normal fault that separates the Campine Plateau to the southwest from the Roer Valley graben to the northeast. The topographic relief on the scarp ranges between 15 and 20 m, and the most recent faulting is marked by a ~ 0.5—3-m-high frontal fault scarp (also, see Figure 3b). The trenches are south of Bree.

Fig. 3. a) Plots of several leveling lines showing comparable vertical offsets of lines near the fault (~ 0.5 m, see also the surface S1): black, topography; brown, base of unit b; blue, unit e; green, base of unit g; and red, water level into trench I. S2 is the projection of an old surface that was probably uplifted by cumulative displacements from previous large earthquakes. The uplifted Holocene terrace reflects the 2 to 3 last coseismic movements on the fault. b) Central section of trench I (southeastern wall) showing ~ 0.5 m of vertical offset of units e and g (fine and coarse gravel, respectively). Paleosol c (near top center) wedges out against the fault, which is covered by unit b, reddish silty-sandy paleosol. The green unit is a brown paleosol that postdates the faulting event. Units d and f are fine, sandy-clay deposits that are internally deformed. According to the 14C dating (AMS), the offset of units below b can be correlated to the last historical seismic event. c) Southwestern section of trench II. The different fault branches x and xx and related colluvial deposits, which mainly consist of coarse gravels, indicate the intense surface deformation in this zone. Unit b is the youngest deposit affected by the fault, and colluvial wedges c, d, e, and f can be correlated with the successive fault movements during the late Pleistocene and Holocene. Units labeled g are attributed to the Cromerian (400 to 700 ka).

Large Earthquake Potential Along the Roer Graben

Subsidence in the Roer graben has increased greatly since the Pliocene, and from the study of Pleistocene terraces and late Pleistocene units in boreholes, an average long-term uplift rate of 0.06 mm/yr and 0.3 mm/yr, respectively, indicate that the deformation rate is not uniform [van den Berg, 1994; Geluk et al., 1994] . Indeed, precise geodetic data between 1875 and 1992 across the Feldbiss fault yield the most recent vertical motion of 0.8 mm/year [van den Berg et al., 1994] . Although the Roer graben has experienced at least seven earthquakes with magnitude ranging from 5.0 to 5.5 during the last 300 years, this activity alone cannot account for the current deformation rate. This is mainly because the seismic activity is widespread along the 150-km-long graben (Figure 1), and no surface faulting has been observed for either the Roermond earthquake (Ms = 5.3) or any historical seismic events. The evidence we discuss here for repeated coseismic displacements along a part of the Feldbiss fault can be compared with the geodetic deformation rate of 0.8 mm/yr (which is assumed to be representative for the Holocene). Indeed, the occurrence of three large earthquakes during the late Pleistocene and Holocene can account for a maximum of 0.3 mm/yr deformation rate. The left-lateral en echelon pattern of individual strands of the Bree active fault is consistent with the earthquake focal mechanisms showing mainly northwest-southeast faulting, with a minor left-lateral component. This fault system indicates an ongoing extension of the Roer graben in a north-northeast, south-southwest direction.

The 1992 Roermond earthquake occurred at a depth of 17 1 km, and it likely occurred near the base of the seismogenic layer [Camelbeeck, 1993] . The main shock had a seismic moment of 1.4 1017 N m, with an average vertical displacement 0.3 m, a rupture area of approximately 11 km2, and a focal mechanism solution indicating a fault plane striking 125 N and dipping 70 SW. By comparison, the last coseismic event observed along the ~10-km fault and in trenches produced an average vertical coseismic displacement of ~ 0.6 m. Assuming a fault width comparable to the thickness of the seismogenic layer at Roermond, the estimated seismic moment for the paleoearthquake is 4 1018 N m., implying a moment magnitude of ~6.4. The estimate of the potential size of the earthquake used here is low since possible activity on adjacent fault segments is not taken into account. (Also, see active faults in Figure 1.)

Implications for the Seismic Hazard

The evidence for recent surface faulting along the Bree scarp raises questions about the classical approach of developing seismic hazard assessments based solely on historical and instrumental seismicity catalogues. In addition to seismicity, comprehensive assessment of seismic hazards would require the identification of active faults, an analysis of the seismic cycle (paleoseismic investigations), and the study of the mechanical behavior of active faults (segmentation, barriers and asperities, friction, etc.).

The map of Rosenhauer and Ahorner [1994] is often used to estimate the seismic hazards of the Lower Rhine Embayment and neighboring regions. For the Bree area, the map indicates a maximum intensity of 7.25 and 7.5 ( modified Mercalli scale) for a return period of 10,000 years. However, consider that the attributed intensity is 7 for the Ms = 5.3 Roermond earthquake, and according to the intensity-magnitude relationships established for the continental interior [Johnston, 1996] , the estimated epicentral intensity might be 9 for the paleoearthquake (Mw = 6.4). Then the maximum intensities proposed by Rosenhauer and Ahorner [1994] are clearly underestimates for this region and probably for others along the Rhine graben system.

The results of our study along the Bree fault scarp allow us to better estimate the potential for large earthquakes in the area and suggest that previous attributed intensities in the graben should be reevaluated. On the basis of the seismicity data (historical and instrumental), the Gutenberg-Richter relation allows the estimate of a return period of 40,000 years for an Mw = 6.5 seismic event along a single fault section, while the paleoseismic data suggest the occurrence of two to three large earthquakes during the Holocene (Figure 4). These results also reinforce our belief that paleoseismic studies are essential for estimating the true potential for large earthquakes in northern Europe. Paleoseismic studies are particularly important where deformation rates are low and where the seismic cycle on individual faults is greater than 1000 years. This has been proven by the few paleoseismic investigations performed in the so-called " stable continental regions" [Crone et al., 1996] .

Fig. 4. Frequency/magnitude distribution based on historical and instrumental data. The value of a = 1.9 is approximated for a fault length of 20 km in the Lower Rhine graben. Paleoseismic data along the Bree fault scarp do not fit with the linear trend of Gutenberg-Richter relation, and suggest a higher frequency of large earthquakes.


This project is supported by the Royal Observatory of Belgium (R.O.B.) and the European Center of Geodynamics and Seismology. We thank the Director of R.O.B., P. Pƒquet, and R. Verbeiren for support and encouragement, and A. Crone, E. Paulissen, M. Dusar, P. Alexandre, K. Vanneste, and D. Jongmans for fruitful discussions in the field. K. Sieh and S. Hough provided useful comments about the paper. Thanks to P. M. Grootes for the quick processing of our radiocarbon samples, the Engineering Geological Laboratory of the Liege University for the geophysical prospection, and the Royal Military School of Brussels for the use of the geodetic equipment. We also thank H. Martin, A. El Bouch, F. Collin, B. Bukasa, J. Krings, J. M. Delinte, and A. Somerhausen. M. Brondeel and A. Moyaert collected the geomorphic data.


Ahorner, L., Present-day stress field and seismotectonic block movements along major fault zones in Central Europe, Tectonophysics, 29, 233, 1975.

Camelbeeck, T., M‚canisme au foyer des tremblements de terre et contraintes tectoniques: le cas de la zone intraplaque belge, Ph.D. thesis, Louvain University, 1993.

Camelbeeck, T., and T. van Eck, The Roer Valley graben earthquake of 13 April 1992 and its seismotectonic setting, Terra Nova, 6, 291, 1994.

Crone, A. J., M. N. Machette, and J. R. Bowman, The episodic nature of earthquake activity in stable continental regions revealed by paleoseismicity studies of Australian and North American Quaternary faults, Aust. J. Earth Sci., in press, 1996.

Demyttenaere, R., and P. Laga, Breuken- en isohypsenkaarten van het Belgisch gedeelte van de Roerdal slenk, Belgian Geol. Surv. Prof. Pap., 1988(4), 234, 1988.

Johnston, A. C., Seismic moment assessment of earthquakes in stable continental regions, II, Historical seismicity, Geophys. J. Int., 125, 639, 1996.

Geluk, M. C., E. J. T. Duin, M. Dusar, R. H. B. Rijkers, M. W. van Den Berg, and P. van Rooijen, Stratigraphy and tectonics of the Roer Valley Graben, Geol. Mijnbouw, 73, 129, 1994.

Paulissen, E., J. Vandenberghe, and F. Gullentops, The Feldbiss fault in the Maas Valley bottom (Limburg, Belgium), Geol. Mijnbouw, 64, 79, 1985.

Rosenhauer, W., and L. Ahorner, Seismic hazard assessment for the Lower Rhine Embayment before and after the 1992 Roermond earthquake, Geol. Mijnbouw, 73, 415, 1994.

van den Berg, M.W., Neotectonics of the Roer Valley rift system. Style and rate of crustal deformation inferred from syn-tectonic sedimentation, Geol. Mijnbouw, 73, 143, 1994.

van den Berg, M.W., et al., Patterns and velocities of recent crustal movements in the Dutch part of the Roer Valley rift system, Geol. Mijnbouw, 73, 157, 1994.

Back to Seismo Home Page