Site effects may produce large ground-motion amplification during an earthquake. The geometry of the subsoil structure, the soil types and the variation of their properties with depth, the lateral discontinuities and the surface topography modify significantly the strong ground motion and are correlated to damage pattern during a large earthquake. Based on this statement, CORSSA (CORinth Soft Soil Array) has been developed in the highly seismic area of Aegion, which was shaken by the 1995 Aegion earthquake (Ms=6.2), which caused 26 deaths and extensive damage to buildings. The aim is to develop an adequate down-hole array of 3D accelerometers and pore pressure transducers. The deepest sensor reached a stiff geological formation that consists of conglomerate and is considered as a ‘seismic bedrock’ in the well-known geological and geotechnical environment where the dynamic soil properties are well defined through field and laboratory tests. CORSSA is established by the Aristotle University of Thessalonica, the National Kapodistrian University of Athens, the ‘Institut de physique du Globe’, Paris, and the ‘Institut de radioprotection et de sûreté nucléaire’ in the framework of the European Program CORSEIS [1,2].

The Aegion city is located in the Gulf of Corinth, one of the most active seismotectonic areas in Europe (Fig. 1), with important north dipping WNW-trending active normal faulting. One of these faults is the Aegion Fault, with an average slope of 40° and an escarpment greater than 40–100 m high (Fig. 2). CORSSA has been established at the shoreline of Aegion. The selection of the site accomplishes two basic requirements:

• (a) loose-soft soil materials susceptible to liquefaction and/or significant non-linearity at shallow depths; and
• (b) relatively shallow seismic bedrock for the installation of the deepest accelerometric station.

CORSSA consists of four broadband 3D down-hole accelerometers, one surface accelerometer and two pore-pressure transducers (Fig. 3). The deepest accelerometer (used as reference station) installed within a geological formation consisting of conglomerate at a depth of −178 m, the other three are located within saturated fluvio-deltaic, marine deposits ML, CL, SC at depths of −14, −30 and −60 m, while the dynamic pore pressure probes installed at −6 and −14 m depth within saturated loose marine deposits SM and ML. These soil layers are susceptible to liquefaction; the depths of the installation of the pore pressure probes are intermediate between the accelerometers, in order to correlate their results. The accelerometers are connected with an 18-channel main recorder and the pore-pressure transducers with a 2-channel data logger. Both have common time recording.

Extensive geological, geophysical and geotechnical surveys have been performed in the site of CORSSA. Geological survey includes detailed geological mapping of the greater area of interest. The geotechnical campaign included six geotechnical boreholes (of depths: 6 m, 2 ×14, 31, 60 and 178 m), in situ (N_SPT and water table measurements) and laboratory tests on intact and disturbed soil samples. Special emphasis was given to the resonant column dynamic tests (Table 1). The geophysical campaign involved borehole (cross-hole, down-hole) as well as surface surveys (seismic reflection). These measurements have been performed in order to define the geological structure and the dynamic soil properties (P- and S-wave velocities, V_p and V_s, shear modulus, G₀ at low strains, shear modulus degradation curves and material damping vs. shear strains, $\gamma (\mathrm{G}/{\mathrm{G}}_0-\gamma -\mathrm{D}$%)), of the different geological formations.

The geological structure is described by three lithostratigraphical units: the lower one consists of very stiff clays, marls and very dense sands directly overlying the alpine substratum, which consists of Mesozoic limestones. The middle comprises typical deltaic and alluvial fan deposits mostly conglomerates and sands, overlying the previous unit. The topmost unit is characterized by alluvial fan deposits, upon which the Aegion city is located. Holocene and recent deltaic and fluvial deposits cover all the rest. Recent geological interpretation indicates the following cross section:

• (1) in the footwall: 18–27 m of alternation of silt and coarse sand (low Vs) above stiff conglomerates (high Vs);
• (2) in the hanging wall, same formations covered by loose conglomerates, silty lacustrine deposits and Holocene deposits with low velocity [4].

The proposed geotechnical profile of CORSSA is given in Fig. 4, while Fig. 5 illustrates the dynamic soil properties for all soil formations, estimated from resonant column tests.

Since the full operation of the CORSSA array (March 2002), some local as well as regional events have been recorded (M⩽4.5, R⩽200 km). An example is presented in Fig. 6, which illustrates the three components of a small event (Ms=3.5) recorded on 20 May 2002. A specific database is under construction, comprising all recordings with the description of the seismological parameters. The event presented herein is too small (PGA=0.042 g) to account for non-linear phenomena. Nevertheless, it is quite interesting to calibrate in the linear elastic range the numerical tools to be used and to study the physics of the ground motion in the vicinity of faults and topographic relief.

Preliminary 1D SH site response analyses have been conducted in CORSSA site using as input soil model the well constrained 1D profile of the site. Different input motions were used (Ricker wavelets of different central period, the deconvoluted recordings of the 1995 Aegion earthquake, as well as other recordings at outcrop conditions) scaled at different PGA values ranging from 0.1 g to 0.5 g. The input motion is introduced at the outcrop. The bedrock is considered at −178 m and maybe rigid or elastic. No major difference is found for the two assumptions. The analysis is basically equivalent linear using the software CyberQuake [3]. The resulted absolute and normalized response spectra of these motions have been compared with the corresponding elastic spectrum proposed by the National Greek Aseismic Code (Fig. 7(a) and (b)), for the soil category C (D according to EC8). The scope of these preliminary analyses is not to compare them with actual recordings, but rather to check the 1D model and identify the threshold between linear-elastic and non-linear phenomena. As it is illustrated in Fig. 8, the non-linear behaviour of soils is expected from rather low intensities (∼>0.15 g), while strong de-amplification is observed at high shaking levels independently of the input motion characteristics. Recorded ground motions are very small to identify experimentally non-linear phenomena due to soil properties. However, in the small strain range of these recordings, we have observed quite important amplifications of both horizontal (Fig. 8) and vertical components (Fig. 6).

Preliminary analyses of the liquefaction susceptibility and risk indicate that the liquefaction susceptibility is very important in the loose saturated silty sand layers found at depths 4±6 m and 12±20 m (N_SPT=5±6, D₆₀/D₁₀∼4, PI<5). The risk of liquefaction is high for acceleration values greater than 0.20 g at outcrop conditions.

CORSSA is a well-established experimental facility in Europe. With all other experiments and facilities deployed in the Corinth Gulf, it is expected to generate a set of high-quality and well-documented data (geological, geotechnical, etc.) and recordings from local and regional earthquakes that will enrich the European and World databases of strong recordings. The choice of the site is particularly successful combining high seismicity, favourable soil conditions to study the effects of non-linear and liquefaction phenomena and geological-tectonic and topographic features on ground motion. The release of the data and the accompanying studies will be very useful both for countries with high seismicity, like Greece, and low-deformation-rate areas with weak seismicity, like France.