12th International Conference on Ground Penetrating Radar, June 16-19, 2008, Birmingham, UK

GPR on Active FaultsGruetzner, C. & Reicherter, K.

Neotectonics and Natural Hazards Group Lochnerstrasse 4-20, 52056 Aachen, Germany

email: c.gruetzner@nug.rwth-aachen.de

Abstract - Several active faults in the Mediterranean have been analyzed with Ground Penetrating Radar: In Greece, the Dionysos Fault near Athens, in southern Spain the Ventas de Zafarraya Fault in the western Granada Basin, Province of Granada, and the La Laja and Cabo de Gracia Faults near Tarifa, Province of Cádiz. All of them are high-angle active faults with at least the footwall consisting of hard rock. The work aimed to visualize sedimentary structures re-lated to recent deformation, to image fault geometries in the hanging walls, and to estimate slip rates. Normally, such fault planes are not easily to be imaged by GPR as their steep dip leads to the loss of reflections. In the cases studied, it was possible to identify the fault planes based on the change of reflection patterns of footwall and hanging wall. Sedimentary features like colluvial wedges, distorted depos-its, tilted layers, fallen boulders, and syn-sedimentary fault-ing could be observed in the hanging walls consisting of soft rock. The analysis of these patterns allowed to image the in-ternal structure of the faults and to draw conclusions on the fault activation history. In case of the Dionysos Fault, both footwall and hanging wall consist of marble. Here, the sedi-ment-filled gap between the two units provides information on the fault geometry. It can be shown that GPR not only al-lows to find and to map active faults, but also is a useful tool for the detailed imaging of fault geometries.

Keywords - GPR, active faults, paleoseismology, earthquake

I. INTRODUCTION

The Mediterranean is the zone with the highest seismicity in Europe. Due to the Afro-European convergence, the majority of all European earthquakes occur in this area. Greece, Italy, and western Turkey are known as the most endangered countries, but several destructive events took place in Spain also. In our study, we applied GPR on ac-tive faults that mostly have the potential to not only pro-duce strong earthquakes, but also have to be considered as hazardous to the people living in their surroundings.

1.1 Dionysos Fault

The Dionysos Fault in Greece is not classified as active, but may be re-activated, as there are several hints for Holocene movements. Situated only a few kilometers northeast of Athens in the Penteli mountain range, this fault is a potential danger and it is important to understand fault mechanics and activation history (Figure 1). The steep marble hills littered with boulders and spiny vegeta-tion do unfortunately not allow achieving several parallel profiles perpendicular to the fault. Therefore, GPR was performed on the only gravel road that crosses the fault obliquely.

Figure 1: The Dionysos Fault in Greece

1.2 Ventas de Zafarraya FaultThe Ventas de Zafarraya Fault (VZF) in the western Granada basin in southern Spain is an E-W trending nor-mal fault and known for the major Andalusian earthquake of December 25, 1884 (Figure 2). Several hundred people died and three villages have been destroyed almost com-pletely. This event is considered to be > M 6.5 and its up to three meters high fault scarp is still visible in the field. Recent paleoseismological investigations have shown that there have been three events in the last 9,000 years [1]. The steep footwall of the VZF did not allow achieving GPR profiles across the scarp. Therefore, the radar mea-

12th International Conference on Ground Penetrating Radar, June 16-19, 2008, Birmingham, UK

surements concentrated on the hanging wall sediments next to the fault, for imaging coseismic features like collu-vial wedges and fallen boulders.

Figure 2: The Ventas de Zafarraya Fault (VZF) and the Cabo de Gracia (CdG) and La Laja (LL) Faults in southern

Spain.

1.3 La Laja and Cabo de Gracia FaultsIn the southernmost part of Spain, close to the Strait of Gibraltar, a number of faults surround the Bolonia Bay (Figure 2). The Roman ruins of Baelo Claudia, which are situated in this area, show earthquake related damages on masonry and infrastructure. Paleoseismological and archeoseismological investigations proved two damaging major events in the 1st Century A.D. and in the 4th Century A.D. [2, 3]. Until now it is not clear whether the seismic source is a remote fault that produced a high-energy event strong enough to affect southern Spain (like in the Lisbon 1755 case), or if a local fault in the surroundings of the ru-ins has been activated. Two of the local faults that have to be taken into account are the La Laja Fault (LL) and the Cabo de Gracia Fault (CdG). At the CdG, GPR was ap-plied to check whether there are hints for recent move-ment in the hanging wall sediments. In case of the La Laja mountain front, GPR was able to reveal step like second-ary terraces and alluvial fans.

II. METHOD2.1 EquipmentIn case of the Dionysos Fault, the GSSI 270 MHz antenna was used with the SIR-3000 and the UtilityScan system including a handheld GPS. The Ventas de Zafarraya fault was investigated using the SIR-2 and a 200 MHz antenna.Measurements at the Cabo de Gracia Fault were taken out with different antenna types. 270 MHz and 300 MHz an-tennae provided high-resolution data from the uppermost 3 to 5 meters, while 45 MHz low-frequency profiles allowed to image deep-lying structures. At La Laja fault, the same equipment as in the Dionysos case was used.

2.2 Data ProcessingAll data processing was done with ReflexW by Sandmeier Software, Karlsruhe, Germany. The main processing steps applied were: Move starttime

Background removal

Gain adjustment (AGC or energy decay) Topography correctionThe GPS system did not provide very accurate elevation data, so the topographic corrections were made based on only few points measured in the field. In some cases, when signal drop-ins have been recorded, a butterworth bandpass (frequency filter) was applied with half the an-tenna frequency as lower threshold and twice the nominal frequency as upper threshold. 2.3 Geological Structures to be Identified with GPRA normal fault with both footwall and hanging wall con-sisting of hardrock (as in the case of Dionysos Fault) will theoretically appear like a joint or a crack in the radar data. In fact, there often is a gap between both blocks, filled up with sediments from above. These colluvial ma-terials often show broad grain size variations and even contain fallen boulders, which may be linked to earth-quake events. This fault gap is what can be imaged with GPR, as it differs from the surrounding hardrock in terms of humidity, consolidation, chemism, and geometry. In-creased humidity (and also higher clay contents) leads to a higher attenuation of the signals, which can be used for the detection of the faults. Broad fault gaps even allow imaging the internal structure of the sediments, while small ones (in a cm- or dm-range) appear as a sharp change in signal amplitude. Therefore, GPR on this kind of faults often only provides the mere identification/local-isation of fault zones.Normal fault hanging walls consisting of soft rock provide much more possibilities for GPR application. The internal structure of both the fault zone itself and the associated sediments allows drawing conclusions on the fault activa-tion history. Surface ruptures of major earthquakes nor-mally produce coseismic slip, which in case of normal faults can be used for magnitude estimations. Antithetical rotated layers act as sediment traps, in which colluvial wedges form out. As they develop repeatedly with each major event, their dip-angle steepens with increasing depth. Colluvial wedges often contain coarse-grained sedi-ments including gravel and boulders that fell down during the shaking. Generally, these structures differ from the surrounding sediments and therefore are detectable with GPR. The data then show a change in signal amplitudes due to the contrast in grain size, chemism and humidity. Rotated layers dipping towards the fault appear as contin-uous reflectors and fallen boulders can be identified based on reflection hyperbolae.

12th International Conference on Ground Penetrating Radar, June 16-19, 2008, Birmingham, UK

Strike-slip faults or normal faults with a distinct strike-slip component are not always easily to be imaged with GPR. Occasionally, positive or negative flower structures can be observed in surface near sediments. Then, GPR measure-ments show displaced reflectors. The contact zone between the moving units often shows a change in humid-ity, as here the sediments are scarified and the ground is weakened due to the lateral displacement. In this case, changes in the signal amplitude will be visible.

III. RESULTS

3.1 Dionysos FaultThe 270 MHz profile across the Dionysos fault clearly re-veals the fault gap. While the solid marbles show high-amplitude reflections and a very good data quality even at two-way traveltimes (TWT) greater than 100 ns (approx. 5 m), the fault zone is characterized by the sudden appear-ance of weak signals (Figure 4).

Figure 3: The 270 MHz GPR profile crosses the Dionysos

Fault oblique in an acute angle.

It can be shown that the Dionysos Fault consists of three parallel single faults, in the biggest of whom a colluvial wedge has been formed. The gaps seem to have widths of several meters, which is only a geometrical effect due to the acute angle between the fault prolongation and the radar profile (Figure 3). In fact, the two small gaps are about 0.5 m wide and the bigger one about 1.2 m. The high-amplitude reflections of the marble mainly appear in the first 55 m of the profile. In this zone, a number of re -flection hyperbolae give hints for the internal jointing of the hard rocks. In the western part of the profile, clear re-flections from great depths are still visible, but occur with significantly higher attenuation. Here, the gravel road was covered with a thin soil layer, which absorbed parts of the emitted energy. In the first third of the profile between 20 and 30 meters, reflections are also not as strong as in the adjacent areas, but cannot be explained by surface effects.

Figure 4: 270 MHz profile across the Dionysos Fault. The fault gaps can clearly be separated from the surrounding

hard rock and a colluvial wedge has formed in one segment.

Most likely, this feature is related to an increased jointing of the marbles close to the fault zone and the resulting in-crease of humidity.

12th International Conference on Ground Penetrating Radar, June 16-19, 2008, Birmingham, UK

3.2 Ventas de Zafarraya Fault

Figure 5: Ventas de Zafarraya Fault in the western Granada basin, trenching and GPR area. See figure 6 for details.

The Ventas de Zafarraya Fault (VZF) was object to nu-merous investigations. For a detailed analysis, paleoseis-mological trenching took place in 2004 (Figure 5). A num-ber of 200 MHz GPR profiles perpendicular to the fault have been achieved accompanying this campaign in order to find promising trenching sites, to obtain data from greater depths, and to trace the outcropped features along the fault zone (Figure 6). The GPR data of file 60 match the observations from the trench (Figure 7, upper radargram). Profile 60 was taken directly next to the top of the trench. In the figure, the out-line of the trench is drawn with a black line, observed structures have a black dotted line. Coarse-grained collu-vial wedges containing limestone debris have been found in the trenches, as well as gravel and stone layers tilted to-wards the VZF in more distant parts of the outcrop. In the radar data, these structures are clearly visible. Depth and lateral position of the reflection patterns show congruity with the observations from the field. Generally, the data quality is better in the southern parts of the profile. Due to the high clay content at the trenching site, the achieved penetration depth of the radar waves does not allow to identify deeper lying structures in file 60. Profile 57 was taken a few meters east of the outcrop with the surface being leveled at the elevation of the trench floor (Figure 7, lower radargram). Here, it is possible to image structures at a depth that could not be reached by excavation. The most important observation is the occur-rence of the fault surface in a depth of 100 ns TWT (ap-prox. 5 m). In the figure, the VZF is drawn with a bold dotted line. This fault plain is made up of lower Jurassic limestone and partly covered by fault crust breccia. It is visible in some outcrops in the surroundings. Apparently, the roughness of the fault surface and the suitable electro-magnetic properties of the overlying sediments allowed receiving reflections despite of the great depth (> 100 ns TWT) and the steep dip-angle of the structure.

Figure 6: Five GPR profiles have been achieved in the imme-diate environs of the trench.

In addition, several southward dipping elements are visi-ble in the first meters of profile 57 (black dotted lines). These are interpreted as the type of colluvial wedges that have been found in the adjacent trench and profile 60. As under interseismic sedimentation regime the layers formed by debris and colluvium from the steep hills are suspected to be inclined northwards, this observation gives a clear indication for active faulting. During the whole campaign, more than 20 GPR profiles have been taken along the Ventas de Zafarraya Fault and accompanying two more trenches. In most of the profiles, similar patterns have been found. Some profiles also showed reflection hyperbolae, which are caused by boul-ders with a diameter of at least 0.5 meters. Such big blocks in a fine-grained sediment layer might be fallen af-ter a triggering by seismic shaking. Some profiles also showed small secondary faults parallel to the main fault zone. The combination of the information of depth and spatial distribution of the observed rotated layers and col-luvial wedges achieved by GPR with the paleoseismologi-cal data provides additional information. It can be deter-mined, which segments of the fault have been activated during historical and pre-historical events. This is one im-portant input factor for paleomagnitude estimations. Fur-thermore, the occurrence of different generations of collu-vial wedges (with a progressive incline and increasing depths) allows drawing conclusions on the fault activation history. In the case of the VZF, at least three events could be identified by GPR.

12th International Conference on Ground Penetrating Radar, June 16-19, 2008, Birmingham, UK

Figure 7: Profiles 57 and 60 were taken close to a paleoseis-mologic trench at the Ventas de Zafarraya Fault.

3.3 La Laja and Cabo de Gracia Faults

Figure 8: The La Laja mountain range shows hanging val-leys and big debris cones at their bases.

The mountains that surround the Bolonia Bay are made up of Aquitanian sandstone and distorted by a number of faults (Figure 9). The steep, nearly vertical La Laja moun-tain front in the northern part of the Bolonia Bay area shows evidence for recent tectonic activity.

12th International Conference on Ground Penetrating Radar, June 16-19, 2008, Birmingham, UK

Several hanging valleys have formed along the steep cliffs (Figure 8). Large debris sheets are situated beneath these triangular facets. Hanging valleys develop when the uplift of the footwall or the downwarped hanging wall proceeds faster than the erosion of the rocks and are therefore an in-dicator for active faults. At the base of La Laja, a lichen-free ribbon with a height of up to 40cm may also be due to recent movement. Boulders with a weight of up to 2 tons can be found even more than some 100 m away from the fault. Yet it is still under investigation whether that long run-out distances may indicate seismic shaking as well. The rock surface of the La Laja is bedding-parallel and overturned.GPR measurements concentrated on the debris cones and the area directly beneath next to the steep wall. All data presented here were achieved with the 270 MHz antenna. Profile 1 was taken from the top of a debris cone to its lower end. The sediment thickness varies from approx. 1.5m (30 ns TWT) at the top to 2.5m in the middle of the slope (Figure 10).

Figure 9: Bolonia Bay working area. 1) El Almarchal Unit (plastic clays); 2) Facinas Unit (plastic clays); 3) El Aljibe Flysch nappe (mainly sandstones): dotted lines = trace of

Betic of uprighted stratification planes; 4) Flysch-slabs acti-vated during the neotectonic period (Bolonia and Tarifa); 5) Post-collisional Pliocene and Quaternary deposits; 6) land-

slide units. After [3].

The debris consists of mainly coarse-grained sediments in-cluding larger boulders that appear as small reflection hy-perbolae in the data. It can be shown that an eroded rock surface is underlying the debris from the hanging valleys.

Figure 10: File 1 has been achieved along a debris cone that formed beneath a hanging valley. Variation in the debris

thickness are due to the existence of an underlying layer and not related to recent tectonic movement.

This structure belongs to a system of eroded strata that have been in front of the actual mountain range. At vari-ous points, the old layers are still visible in the field. As they are almost vertical and show a slight inclination to-wards the La Laja wall, they act as a sediment trap, which is responsible for the variation in the debris cone thick-ness. Differently to the Ventas de Zafarraya case, here no evidence for rotated layers caused by recent tectonic movement was observed. The colluvium shows no pro-gressive inclined wedges or tilted paleosols.

Figure 11: Files 2 and 3 from the immediate environs of the debris cone show the same rock surface, but give no evidence

for recent deformation.

12th International Conference on Ground Penetrating Radar, June 16-19, 2008, Birmingham, UK

Profile 2 and 3 show two parallel sections perpendicular to the mountain front, leading from the vertical rock surface down the slope to the valley (Figure 11). Both profiles were taken about 20 meters away from the debris cone shown in file1 and have the same direction. The afore-mentioned sandstone layer is visible in both profiles. Again, it leads to variations in the sediment thickness, which seems not to be associated with recent deformation. Profile 3 shows large reflection hyperbolae that are caused by big single boulders. These blocks have a diameter of more than 0.5 meter and are embedded in the colluvium. The lack of rotated layers and other evidence for coseis-mic slip in the uppermost sediments lead to the conclu-sion, that there was no recent major or moderate earth-quake event at this fault. West of the Roman city of Baelo Claudia, the NE-SW trending Cabo de Gracia reverse fault can be observed at various locations. Several features like striae indicate re-cent deformation. Prior to a paleoseismologic trenching campaign, GPR was applied in order to explore promising trenching sites, both considering tectonic features and a sufficient sediment cover. Several profiles were taken at a steep rock front, which was suspect to be the Cabo de Gracia fault scarp (Figure 12).

Figure 12: Reflection hyperbolae indicate sandstone layers at the Cabo de Gracia Fault zone.

File 42 (270 MHz) leads from the nearly vertical rock sur-face to the southeast. The data show the presence of sev-eral layers, buried close to the wall. These surface-near hard rock structures appear as strong hyperbolae within the weak reflections of the surrounding sediments. Any clear evidence for coseismic deformed sediments could not be derived from the GPR data and it could not be de-termined, whether the sandstone units were moved by re-cent earthquakes. Nevertheless, the results proved evi-dence for the planning of the trenching campaign. Excava-tions showed, however, that no recent major event took place at this part of the fault, although striae have been found at the beds.

Figure 13: 45 MHz profile (file 12) crossing the suspected fault trace along the spring (above), and 300 MHz profile of

the same location (file 13).

Files 12 and 13 have been collected close to profile 42. Here, the steep rock surface ends abruptly and a spring forms a small pond at the base of the rocks. This place provided an excellent trenching site on the first glance. The GPR data are of generally bad quality. File 12, which was taken using the 45 MHz antenna, does not provide any structural data from the subsurface (Figure 13). In-stead, there is a clear change in the reflection amplitudes in the middle of the profile. The same patterns occur in file 13 (300 MHz), which was taken at the same location. The GPR data were interpreted as a clayey layer that causes the high attenuation of the signals, and a change in soil humidity that causes the lateral differences in the radargrams. In addition, the position of the nearby spring matches the signal change of the radar data. Therefore, young tectonic deformation was supposed to be the reason for these effects.Later excavations showed that there is a complex system of small faults in the area that was predicted to be a promising trenching site. At least the uppermost 2 meters of the ground consist of clays. Next to the spring, there was a change in humidity and an increase of moisture that even lead the trench walls to collapse partly. Geological mapping in this case proved the GPR interpretation. The

12th International Conference on Ground Penetrating Radar, June 16-19, 2008, Birmingham, UK

tectonic structures found in the trenches show recent de-formation, but no indications for a recent major event.

IV. CONCLUSIONSGPR profiling at several faults with different moving mechanisms and various lithologies showed that this geo-physical technique can provide a lot of additional informa-tion to the classical geological methods. It is possible to draw conclusions on fault activation history, on slip rates, fault kinematics, and fault geometry from the imaging of the fault zone itself and its associated sediment structures. Sedimentary features like rotated layers, colluvial wedges, secondary faults, and filled fault gaps can be imaged in most cases.

REFERENCES[1] Reicherter K. (2001). Paleoseismologic advances in

the Granada basin (Betic Cordilleras, Southern Spain). Acta geologica hispanica, v. 36 (2001), n° 3-4, p. 267 – 281

[2] Silva P.G., Borja F., Zazo C., Goy J.L., Bardají T., De Luque L., Lario J. & Dabrio C.J. 2005. Archaeo-seismic record at the ancient Roman city of Baelo Claudia (Cádiz, South Spain). Tectonophysics, 408, 129-146.

[3] Silva P.G., Reicherter K., Grützner C., Bardají T., La-rio J., Goy J.L., Zazo C., & Becker-Heidmann P., 2008. Surface and subsurface paleoseismic records at the ancient Roman city of Baelo Claudia and the Bo-lonia Bay area, Cádiz (South Spain). Geol Soc of Lon-don Spec. Vol.: Paleoseismology: Historical and prehistorical records of earthquake ground effects for seismic hazard assessment. In press.