Seismic imaging of active continental rifts is difficult because extensional deformations produce strong shallow heterogeneities. Marked topography, transition from land to sea and block faulting produce sharp velocity contrasts that disrupt the coherence of wavefronts and enhance multiple scattering. The determination of the geometrical relationships between high- and low-angle border faults requires careful processing [4,11], especially designed experiments [1] or additional recordings at intermediate offset ranges [9]. In this paper, we report on seismic observations made with fixed recording arrays placed on the shoulders of the Corinth Rift during the R/V Maurice Ewing July 2001 marine survey in the gulf (Fig. 1).

The rapidly extending Corinth Rift includes a subsiding basin in the gulf and the uplifted coastal ranges of northern Peloponnesus [2]. The elevation difference between the basement of the basin [6] and the southern 2.3 km-high mountain range is at least 5 km over a horizontal distance of 30 km. Sorel [13] has proposed an evolution scenario for the central rift where block faulting has progressed from south to north above a detachment fault within the basement. Large earthquakes presently occur beneath the northern part of the gulf, at depths close to 10 km. The focal mechanisms and surface deformations resulting from the recent earthquakes, such as the 1992 Galaxidi event, are compatible with low-angle normal faulting [3]. These earthquakes may occur on the down-dip extension of a northward-dipping detachment linking upper-crustal block-faulting to lower crustal ductile flow [5] and thinning [15]. The central Corinth Rift may thus be one of the few known examples of seismically active low-angle normal fault in the upper crust [16].

The rift cuts the NNW-trending external thrust belt of the Hellenides [8] at an angle of 40°. In the Galaxidi area, carbonate nappes dip gently toward the south. Tilted blocks of carbonate nappe fragments and synrift sediments are observed in the northern Peloponnesus river valleys above an inferred detachment fault in a basement schist unit [13]. In the Derveni area, the east–west-trending Valimi and Xylokastro normal faults juxtapose limestones in the footwall with up-to-a-kilometre-thick marls and conglomerates deposits in the hangingwall.

The R/V Maurice Ewing July 2001 seismic survey in the Corinth Rift was shot with a 140-l tuned airgun array and was recorded on a 240-channel 6 km-long streamer [14]. Additional land recording of these powerful shots, fired at 50 m intervals, was made by a regional network of 40 three-component stations [17]. Two linear arrays located close to the coast near Derveni and Galaxidi were also installed to probe the shallow structure of the shoulders of the rift. The Derveni array, emplaced along the road Ligia–Pyrgos–Mount Mavron, is oriented at right angle to the rift. It crosses the westward extension of the Xylokastro Fault and extends up to the foot of Mount Mavron in the hangingwall of the Valimi Fault. The Galaxidi array is located directly on the carbonates of the Helennide nappes that constitute the faulted blocks beneath the Derveni array.

The 7.2 km long Derveni array has thirty three-channel stations, each channel recording a group of six vertical geophones. The group interval is 80 m, a value chosen to provide common shot gathers without excessive spatial aliasing. The southernmost trace, at the foot of Mount Mavron, is at an altitude of 979 m. The recording was continuous during the eight-day survey, using GPS timing. Access problems due to topography or housing in the Pyrgos village and temporary station failures caused irregularities and gaps in the spatial coverage. Ten stations were emplaced with the same parameters north of Galaxidi, on the northern coast.

The source array included airguns of different volumes and bubble periods. Destructive interferences occur between individual bubble oscillations and result in an impulsive signal at short range. The attenuation of high frequencies during propagation leads to a dominance of the lowest bubble frequency (6.5 Hz) at long range. Spectra of individual traces show broad peaks of decreasing amplitude at quasi-harmonic frequencies (multiples of 6.5 Hz). Spectral balancing using a simple method of spectrum equalization [7] improves the impulsiveness and the coherence of the records, making the detection of first and later arrivals easier.

There is a clear change in the trace gathers on the Derveni array with the position of the shots. For shots within the deep basin part of the gulf (Fig. 2), first arrivals correspond to refracted waves that have penetrated through the basin sediments into basement. High-amplitude later arrivals correspond to water waves converted at the steep southern flank of the gulf. For shots located on the shallow northern shelf, the refracted waves are not seen and the first arrivals are the converted water waves. Streamer data in the Itea Bay also show that transmission from water into the high-velocity carbonates of the Hellenide nappes is ineffective [12]. The maximum inline offset on the Derveni array for the waves refracted in the basement is therefore limited to 25 km.

The first arrivals on shot gathers at the Derveni array are characterized by an average apparent velocity larger than 8 km s⁻¹. This implies the presence of a north-dipping interface at depth that reduces the angle between the conical wavefront and the surface. The decrease in the stations elevation from south to north decreases the apparent velocity. The observed variations of first arrival travel times and amplitudes across the array are quite smooth. This indicates that shallow block faulting does not perturb the propagation of the refracted waves.

In the absence of shots within the array, shallow velocities directly beneath the stations cannot be determined. However, the strong water waves efficiently converted at the gulf flank can be considered as direct waves probing the sedimentary layer. They are clearly observed at all offsets across the first half of the array. Shots located close to the coast show these waves most clearly at offsets between 2 and 6 km, with a velocity of 4.3 km s⁻¹. In contrast with the refracted waves, these converted waves are strongly attenuated south of Pyrgos, in the footwall of the Xylokastro Fault.

The refracted first arrivals on the Galaxidi array have an apparent velocity close to 6 km s⁻¹ for offsets between 13 and 25 km. Because the array is short (2.2 km), it is difficult to have a precise determination of this velocity. The array is emplaced directly on the carbonates of the Hellenide nappes in a direction subparallel to their strike. One may thus assume that dip effects on the refraction velocity are small. We therefore consider that a basement beneath the nappes has a velocity of 6 km s⁻¹. Since the Derveni and Galaxidi arrays are installed on the same Hellenide unit, we assume that basement velocity is the same beneath both arrays. This implies a critical angle of 45° at the interface between the 4.3 km s⁻¹ layer and the 6 km s⁻¹ basement beneath the Derveni array.

We fit the first arrival times at the Derveni array for the shots that are located in the deepest part of the basin (a 5 km interval with water depths larger than 800 m). We assume that a continuous northward dipping plane interface separates carbonate units (4.3 km s⁻¹) from basement (6 km s⁻¹) beneath the array and the basin. We use velocities determined by prestack migration velocity analysis for the sediments in the basin [6]. We adjust the depth and the dip angle of the interface to fit a linear trend to the arrival times across the array. We find a uniform 15° northward dip along a 15 km distance and a 5.3 km depth of the interface beneath the shot indicated in Fig. 1. Half of the 15 km distance travelled by the refracted wave along the interface is under the basin.

Lateral velocity variations above the interface and depth variations of the interface beneath the array produce smooth time shifts relative to the times predicted for a plane interface. In addition, lateral velocity variations in basement outside the array would produce a shift in depth of the interface. Since it is not possible to discriminate between such velocity variations and depth variations of the interface with our data, we use time migration to obtain an image of the interface beneath the array. The migration is simply done along the refracted ray emerging from the plane interface because there is no evidence for diffraction, triplication or large slope variations in the refracted and following arrivals. Because the interface has an average dip of 15° toward the north and the critical angles is 45°, the refracted rays have an incidence angle of 30° with respect to the vertical. For a back-tilting angle of 30° for the blocks in the hangingwall of the shallow normal faults [13], the refracted rays cross the original block structure at near-normal incidence.

After migration (Fig. 3), the seismic traces are represented at the horizontal positions corresponding to the emergence of the refracted rays along the interface. The first-arrival times on the migrated section correspond to the one-way time to the interface along the vertical in a 4.3 km s⁻¹ constant velocity medium. Topography converted to time with the same velocity is also shown. The migration shifts the traces seaward by almost 2 km. To improve the signal/noise ratio, we stacked migrated traces from 10 consecutive shots. Despite stacking, the records at the stations located close to the coast remain noisier than those at the stations located further inland. Because there is also a spatial gap in the traces, first arrivals at the northernmost stations remain somewhat ambiguous.

The migrated image beneath the Derveni array shows an undulation of the interface that reflects the block structure across the Xylokastro Fault. The top of the undulation occurs at the position (3 km distance from coast) of outcrops of calcareous units observed east of the profile in the footwall of the Xylokastro Fault [2]. Thickness variations in a wedge of synrift sediments in its hangingwall and the throw on the fault combine to cause a smooth seaward delay time increase of 0.5 s in a distance of 4 km. This delay corresponds to a 2.2 km depth increase at 4.3 km s⁻¹ and to a local slope of the interface of 30°. Toward Mount Mavron, the almost constant delay time may indicate rollover anticline geometry. The image is consistent with the geometry proposed by Sorel [13] for the top surface of slivers of basement above a shallow detachment fault.

The installation of two land seismic spreads on the shoulders of the Corinth Rift during the July 2001 R/V Maurice Ewing seismic reflection campaign in the gulf has allowed to record clear refracted waves. The character of the waves indicates a smooth interface beneath the Derveni array. The arrival times can be fit with a 15° seaward dipping plane interface between a 4.3 km s⁻¹ carbonate layer and a 6 km s⁻¹ basement. This interface extends beneath the basin in the gulf. Its northward extrapolation goes through the hypocenter of the 1992 Galaxidi earthquake, a low-angle normal faulting event at 7.4 km depth (Fig. 4). The distribution of seismicity determined by local networks also shows a 15° dip in the seismically active zone [10]. Time migration of the refracted waves provides an image of the interface beneath the Derveni array. It shows an undulation that reflects the block structure across the Xylokastro Fault. Our data are consistent with the north-dipping detachment geometry proposed by Sorel [13].