HAL Id: hal-01417477
https://hal.archives-ouvertes.fr/hal-01417477
Submitted on 15 Dec 2016
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of
sci-entific research documents, whether they are
pub-lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Crustal velocity and Moho structure beneath the Gulf of
Corinth, Greece
Barry Zelt, Brian Taylor, Maria Sachpazi, Alfred Hirn
To cite this version:
Barry Zelt, Brian Taylor, Maria Sachpazi, Alfred Hirn. Crustal velocity and Moho structure beneath
the Gulf of Corinth, Greece. Geophysical Journal International, Oxford University Press (OUP), 2005,
162 (1), pp.257-268. �10.1111/j.1365-246X.2005.02640.x�. �hal-01417477�
GJI
T
ectonics
and
geo
dynamics
Crustal velocity and Moho structure beneath the Gulf of Corinth,
Greece
Barry C. Zelt,
1Brian Taylor,
1Maria Sachpazi
2and Alfred Hirn
31School of Ocean & Earth Science & Technology, Department of Geology and Geophysics, University of Hawaii, 1680 East-West Road,
Honolulu, HI 96822, USA. E-mail: taylorb@hawaii.edu
2Geodynamic Institute, National Observatory of Athens, Athens, Greece
3Laboratoire de Sismologie Exp´erimentale, D´epartement de Sismologie, Institut de Physique du Globe de Paris, Paris, France
Accepted 2005 March 14. Received 2004 December 19; in original form 2004 May 5
S U M M A R Y
The Gulf of Corinth (GOC), Greece is a continental rift with high rates of seismicity and ex-tensional strain. How this strain is accommodated in the crust and whether there are variations in the mechanism along strike remain open questions, in part because of a lack of wide-angle reflection/refraction studies that constrain crustal velocity structure. In 2001, an extensive mul-tichannel seismic survey was conducted within the GOC, one component of which included the wide-angle recording of sources from within the gulf at stations on land surrounding the gulf. In this paper we use wide-angle data in two separate, but allied, studies to constrain crustal velocities and depth to the Moho. A 2-D inversion of refraction and reflection traveltimes along an axial profile through the GOC constrains the shallow crustal velocity structure, images the Moho at 29 km depth in the east, dipping to 39 km in the west, and images the eastward subducting African slab beneath the western GOC at a depth of 74 km. The 1-D average of the 2-D velocity model was used in a tomographic inversion of PmP reflection times to solve for depth to the Moho throughout the Corinth region. This model shows generally thick, isostati-cally compensated crust (≥37 km) beneath the Hellenide Mountains, except immediately south of the GOC, and a singular region of thin crust (<30 km) beneath the Perahora Peninsula at the eastern end of the gulf. A comparison with Moho depths derived from gravity inversion shows a general agreement with crust thickening from east to west, but a number of differences in detail. The 3-D crustal thickness variations are more complex than those predicted by either pure shear or simple shear models of continental extension and suggest significant pre-rift structural variability.
Key words: crustal structure, Greece, Gulf of Corinth, Moho, tomography, velocity.
I N T R O D U C T I O N
The Gulf of Corinth (GOC; Fig. 1), which separates the Pelopon-nisos from mainland Greece, is an active continental rift on the western edge of the Aegean, which itself is one of the most active extensional continental regions in the world (Armijo et al. 1996). Extension throughout the Aegean started in the late Oligocene (Jolivet 2001); however, the Corinth rift is primarily a Quaternary feature (Armijo et al. 1996). Earthquake focal mechanisms through the GOC region show a pattern of east–west-trending normal fault-ing, consistent with geodetic measurements which show that exten-sion is directed north–south across, and is mainly localized beneath, the gulf (Billiris et al. 1991; Clarke et al. 1997, 1998; Davies et al. 1997; Briole et al. 2000). There is, however, considerable debate concerning the geometry of crustal faulting that is accommodat-ing the deformation, especially in the light of focal mechanisms (Hatzfeld et al. 1996) and microseismicity studies (Rigo et al. 1996;
Rietbrock et al. 1996) which suggest that, in the western and cen-tral parts of the gulf, slip is occurring on low-angle (<30◦) normal faults.
The crustal velocity structure, including variations in crustal thickness, are important parameters that can help resolve issues relating to the mode of extension. Such issues include whether the crustal thinning occurs offset from, or beneath, surficial rifts (i.e. simple versus pure shear) and the relationship between upper crust faulting and lower crust deformation (e.g. Wernicke 1985; Ruppel 1995). To date, the only detailed map of Moho topography beneath the Corinth rift comes from the inversion of gravity data (Tiberi et al. 2001). Furthermore, there are no published reports of crustal-scale velocity variations in the immediate GOC region. Here, in addition to using wide-angle reflection/refraction data to constrain a 2-D crustal velocity model for an along-axis profile through the gulf, we present a map of Moho topography derived from tomographic inversion of wide-angle reflection traveltimes.
21˚ 22˚ 23˚ 24˚ 38˚ 39˚ 0 50 km
GOC
s m PP 17 18 19 20 36 38 48 49 21 22 64 50 21 12 21 52 117 112 16 27 15 78 70 77 88 54 7 79 51 31 68 62 11 123 31 144 22 59 0.0 0.5 1.0 1.5 2.0 2.5 Elevation (km)A
P
Hellenic Trough NGOE GOP C SGHellenides
Peloponnisos
Aegean Sea GREECE Mediterranean RidgeFigure 1. Topographic map of central Greece centred on the Gulf of Corinth (GOC). Dotted lines within the GOC denote locations of the ship track and shot
points. The thick east–west line within the gulf represents combined shot lines 8 and 23 used in 2-D modelling. White and black dots are land stations. Black dots are stations used in 2-D modelling. Both black and white dots (except stations 36 and 48) were used in 3-D modelling. Large numbers are station numbers referred to in the text. Small numbers are the number of PmP data used in the 3-D modelling.+ symbols between stations 38 and 49 show approximate bounce points for Ps phase (reflections from subducting African Plate). Triangles are centres of volcanic activity: s, Susaki solfataric field, m, Methana arc volcano. A, Athens; C, Cephalonia; GOP, Gulf of Patras; NGOE, North Gulf of Evia; P, Patras; PP, Perahora Peninsula; SG, Saronic Gulf. The inset map shows region of the larger map (rectangle) relative to the Greece/Aegean region.
S E I S M I C E X P E R I M E N T
In 2001 July the University of Hawaii, National Observatory of Athens, and Institut de Physique du Globe de Paris conducted a seismic survey of the Gulf of Corinth, Greece. A dense grid of about 50 multichannel seismic (MCS) lines was recorded within the gulf (Fig. 1) using the R/V Maurice Ewing (Sachpazi et al. 2002; Goodliffe et al. 2003; Taylor et al. 2003; Weiss et al. 2003; Pi Alperin
et al. 2004; Zelt et al. 2004). The seismic source was a 138 l, 20-gun,
tuned air gun array. Shot spacing was 50 m with an average time between shots of 21 s. The water depth beneath the shots ranged from 40 to 870 m. Bathymetry along each line was measured using Hydrosweep DS2 multibeam sonar.
In addition to the MCS component of the experiment, a network of 40 temporary stations was deployed across southern Greece (Fig. 1) specifically to record the marine shots. These stations were deployed and operated under a contract to Landtech Enterprises. All of the stations that provided usable data comprised an Earth Data PR2400 portable field recorder and a GeoSIG VE-53DH three-component seismometer. The maximum shot–receiver offset was 180 km. In addition, we also use data recorded at two stations of the Corinth Rift Laboratory Project (Lyon-Caen et al. 2001). In general we found that the seismic signals were strongly attenuated, suggesting that, ideally, for an experiment of this scale in this region a larger seismic source would be very beneficial.
P R E V I O U S V E L O C I T Y M O D E L S A N D E S T I M AT E S O F M O H O D E P T H
There have been few published reports of crustal velocity and Moho structure in the vicinity of the GOC. Cl´ement (2000) and Cl´ement
et al. (2004) used wide-angle reflections recorded at two stations
from a shot profile within the gulf to derive Moho depth estimates of 40 km beneath the western end of the gulf (near our station 49) and 32 km along the northern shore at the eastern end of the gulf. About 100 km south of the GOC, Makris (1978) interpreted refraction data from a profile crossing the Peloponnisos. He found that the velocity at the top of the crystalline basement was 6 km s−1, the velocity at the base of the crust was 6.5–7.0 km s−1, the Moho increases in depth from 26 km in the east to 46 km in the west before decreasing rapidly near the western shore of the Peloponnisos and the upper mantle velocity was∼7.7 km s−1. He also noted the presence of late arrivals interpreted as reflections from a subcrustal reflector at 65 km depth beneath the western Peloponnisos. At about the same latitude, Cl´ement et al. (2000) looked at data recorded along an onshore–offshore profile and noticed a reflective layer, interpreted as the lower crust, which terminated at the upper mantle at an inferred depth of∼20 km. At the latitude of the GOC, but further west, Hirn
et al. (1996) studied the crustal structure of the Ionian Islands region
using marine MCS reflection profiles combined with recordings at land stations. Their MCS data imaged a 12 km deep reflector from outboard of Cephalonia to the island’s southeastern tip, where it dips strongly eastwards. This was interpreted as the top of the subducting African slab which, at this location, is still overlain by upper Aegean Plate crust. From wide-angle reflection data recorded at stations of both sides of the Gulf of Patras, they inferred a depth to Moho of 24 km under the western Gulf of Patras at 21.25◦, SSE of our station 36.
There have been a number of earthquake tomography studies that include the Corinth region (Papazachos et al. 1995; Drakotas et al. 1997; Papazachos & Nolet 1997); however, cell sizes are typically
∼50–100 km horizontally and 10 km vertically, and thus both the horizontal and vertical resolution of these studies is poor. A tomo-graphic image of the subducting African slab (Papazachos & Nolet 1997) shows a depth of 60–65 km in the western Peloponnisos, with a shallow eastward dip of∼ 10◦, increasing rapidly to 35◦ in the central Peloponnisos.
Tiberi et al. (2001) inverted gravity data to create a map of Moho depth in the vicinity of the GOC. They found northwest–southeast-trending features superimposed on a general crustal thickening from ∼25 km in the east to ∼40 km in the west. They interpret an offset between relatively thin crust near the northern margin of the gulf with the main axis of the gulf as support for the hypothesis that extension has been accommodated by a low-angle normal fault at 12 km depth, as suggested by Rietbrock et al. (1996) and Rigo et al. (1996), and a ductile lower crust.
2 - D T R AV E L T I M E I N V E R S I O N A L O N G A N E A S T – W E S T P R O F I L E
For the 2-D study, we use shots from two east–west MCS profiles, 8 and 23 which, combined, form the longest (85 km) continuous profile through the Gulf (Fig. 1). We use data recorded at eight sta-tions: 17–20 to the east of the gulf, and 36, 38, 48 and 49 to the west of our shots. The latter two stations are Corinth Rift Labora-tory stations TEM and TRI. This east–west transect of the gulf is vastly superior to other potential 2-D profiles that could be created from our shot/receiver distribution: it is the longest possible transect (250 km), the overall data quality is the best and it has the most in-line shots. For these reasons, we concentrate on this single profile.
Our ability to constrain 2-D structure is limited by two factors: (1) the acquisition geometry, which is less than ideal because there is no overlap of shots and receivers and (2) the quality of data, which is limited by the size of the seismic source and apparent strong atten-uation in this region. We confidently identify three seismic phases: refractions through the upper crust (Pg), reflections from the Moho (PmP) and a later phase (Ps) which, based on its arrival time,
am-Figure 2. Receiver gather for station 36 and shot line 8+23. Data are bandpass filtered between 3 and 8 Hz and reduced at 8 km s−1. The horizontal axis is shot–receiver offset. The strong, late phase between 10.6 and 13.0 s is Ps, interpreted as a reflection from the subducting African slab. The weaker phase around 7 s is probably PmP. The inset nomogram shows the location of the receiver (black dot) and shot line (black line) in relation to other stations and shots.
plitude and moveout characteristics, corresponds to reflections from a deep, sub-Moho reflector. The maximum offset observed for the
Pg phase varies from 43–72 km; Pg is not observed at station 36
and is only tentatively observed at station 17. Pn (refractions through the uppermost mantle) is not observed. PmP is generally promi-nent except at station 48, where it is very weak, and at station 36, where its identification is indefinite (Fig. 2). Ps is a prominent, late phase recorded only at station 36 (Fig. 2). Pick uncertainties between 50 and 250 ms were calculated using an automated scheme based on the signal-to-noise ratio in a short (250 ms) time window around the pick (e.g. Zelt & Forsyth 1994). Picks were binned such that the average spacing between picks was 250 m. Table 1 gives the statistics of the data.
We model the 2-D data using the traveltime inversion method of Zelt & Smith (1992) in which a layered velocity model is parametrized by boundary (depth) nodes and velocity nodes. We use a layer-stripping approach, modelling successively deeper structures while fixing shallower structures (e.g. Zelt et al. 1993). Our exigu-ous data and less than optimal recording geometry lead us to seek the simplest model, i.e. with the fewest parameters, that fits the data. To this end, for example, except where necessary, we do not allow lateral velocity variations.
Ray coverage and a comparison between observed and calculated traveltimes are shown in Fig. 3. The preferred final model is shown in Fig. 4. Estimates of velocity and reflector depth uncertainty are given in Table 2. The model origin corresponds to station 36 at 21.11176◦E, 38.48812◦N. The model comprises seven layers. Layer 1 is water. Layer 2 comprises syn-rift sediments. Our data do not constrain velocities in this layer; we assigned a constant velocity of 2.5 km s−1, which is the average velocity for the sediments as found in a shallow tomography study of the gulf (Zelt et al. 2004). The base of this layer is taken from MCS interpretations of lines 8 and 23 (Weiss 2004). Its thickness varies from 0 to 2.3 km, and it causes significant delays and adds structure to the observed phases. It is essential that this layer is accurately represented to avoid errors in deeper parts of the model.
Table 1. Traveltime data for 2-D modelling and the goodness-of-fit statistics for the final model. Table
columns: Station, station number (Fig. 1); X , model distance of receiver (distance from station 36); Pg,
PmP and Ps, number/average uncertainty (ms) of Pg, PmP and Ps picks, respectively; Total, total number
of picks/average uncertainty (ms) for each station;χ2, normalized chi-squared misfit for each station; dtrms, rms traveltime residual for each station. Last three rows: Total, number of picks/average uncertainty
for each phase;χ2, normalized chi-squared misfit for each phase; dt
rms, rms traveltime residual for each
phase. Station X (km) Pg PmP Ps Total χ2 dtrms(ms) 36 0.0 — 48/224 145/218 193/219 0.7 142 38 44.8 20/148 90/217 — 110/205 1.0 175 49 85.0 65/228 63/205 — 128/217 1.0 178 48 92.4 87/197 22/234 — 109/205 1.5 214 20 189.0 196/142 93/201 — 289/161 1.0 135 19 215.6 72/156 162/208 — 234/192 0.6 101 18 228.3 68/190 106/215 — 174/206 0.6 123 17 252.7 9/200 106/206 — 115/206 0.8 154 Total 517/172 690/211 145/218 1352/197 0.9 147 χ2 1.1 0.7 0.7 0.9 dtrms(ms) 158 139 146 147
b
Ps PmP Pg 38 38 38 49 20 19 18 18 18 19 20 17 17 17 19 20 49 49 49 48 48 48 49 36 36 36a
Moho W E 36 38 49 48 20 19 18 17Figure 3. (a) Rays traced through the 2-D model. Every third ray is plotted. Three phases, coded by colour and identified in panel (b), were used in the
modelling: Pg, PmP and Ps. Numbered triangles are station locations. No vertical exaggeration. (b) Observed data (short, coloured vertical lines with height equalling twice the pick uncertainty) and calculated traveltimes (black dots). Numbers indicate the stations at which the data were recorded.
Layers 3 and 4, which extend to a maximum depth of∼6.5 km, are constrained by Pg. Beneath the gulf, velocities in layer 3, which represents the basement to the current rifting phase, are 5.3 and 5.9 km s−1at the top and bottom, respectively. Layer 4 is included to affect a decrease in velocity gradient. The velocity at the base of this layer is 6.0 km s−1beneath the gulf and further east. In the east, beneath stations 18 and 17, we have reduced the velocities in layer 3 to 5.0–5.5 km s−1 to affect a traveltime delay to data from these stations. The source of the causative velocity reduction,
however, cannot be constrained; it could, for example, be equally likely to arise from the presence of a thin, surficial low-velocity region. Similarly, faster velocities are required in the west to satisfy the early Pg arrival times at station 38. To this end we increase the shallow crustal velocities west of station 49 to 6.1–6.2 km s−1. Whether these higher velocities are present west of station 38 we cannot say with our data; they are not inconsistent with our data. For simplicity we extend the high velocities to the west end of our model.
Figure 4. 2-D velocity model with no vertical exaggeration. Numbered diamonds are stations used in the modelling. Numbers within the model, and inside
boxes above the model, are P-wave velocities at the top or bottom of the indicated layer. These velocities represent averages over regions that can be distinguished by similarity of colour. Velocities in layer 2 and within the upper mantle (white region) are not directly constrained by the modelling; these layers have no vertical velocity gradient. Heavy black lines denote portions of boundaries that are sampled by reflections. Broken lines on the west side of the model represent shallowing of the Moho to 24 km beneath station 36 and the approximately 30◦dip of the African Plate required to reach a depth of 12 km beneath Cephalonia; both based on Hirn et al. (1996). The thick grey line shows the position of the Moho (M) obtained from the 3-D inversion of PmP data.
Table 2. Estimated uncertainties for velocities and depths that were solved
for in the 2-D inversion. The value of velocity or depth at a velocity or boundary node was perturbed through a range of values (both negative and positive). The range which provides a statistically equivalent fit (using an
F-test) is an estimate of parameter uncertainty. Estimated uncertainties for
Moho and slab depth assume that the overlaying velocity structure is accu-rate.
Parameter/region of model Estimated uncertainty Velocity in layer 3 ± 0.1 km s−1 Velocity in layer 4 west of X= 85 km ± 0.15 km s−1 Velocity in layer 4 east of X= 85 km ± 0.1 km s−1 Depth to Moho near X= 40 km ± 1.1 km Depth to Moho between X= 60 and 185 km ± 0.7 km Depth to Moho near X= 220 km ± 2.0 km Depth to slab between X= 50 and 75 km ± 1.0 km
The remainder of the crust is represented by layer 5. We have no refractions through this layer to constrain velocity. Velocities at the top of this layer were set to the same values as at the base of layer 4 (6.0–6.2 km s−1). PmP times were inverted for depth to Moho using various fixed and constant values for the velocity at the base of the crust. Although the PmP data were consistent with velocities between 6.5 and 7.0 km s−1, we consider the optimal value to be 6.7 km s−1 because, as detailed below, this value also provided the best overall fit to our inversion of PmP for 3-D Moho topography. It is possible to fit the PmP times at all stations with-out resorting to lateral velocity variations in layer 5; however, the converse is not true. We could not fit the PmP data by assuming a flat (or flattish) Moho and allowing reasonable lateral velocity vari-ations in layer 5. PmP times from all stvari-ations except 36 constrain a Moho which dips to the west from 28.5 km beneath station 19 to 38 km beneath station 49. Using a faster or slower velocity at the base of the crust does not affect this trend: depths would vary by about±0.5 km and ±1.0 km beneath stations 49 and 19, respec-tively, for variations in the average velocity at the base of layer 5 of ±0.2 km s−1. Our identification of PmP recorded at station 36
re-quires that the Moho shallows significantly in the west beneath sta-tion 38 to 32 km.
The Ps phase recorded at station 36 constrains a deep, subcrustal reflector at 74 km depth between stations 38 and 49. This depth assumes that the average upper mantle velocity, which is uncon-strained in our model, is 7.9 km s−1. The reflector depth would be 71 or 76 km for upper mantle velocities of 7.7 and 8.1 km s−1, re-spectively. As shown in Fig. 4, the reflector, where sampled by rays, is flat; however, an eastward dip of no more than∼5◦is allowed by the data.
3 - D T O M O G R A P H I C I N V E R S I O N F O R M O H O D E P T H
We confidently identify PmP arrivals at 28 stations. PmP, where present, is generally not difficult to recognize because intracrustal reflectivity is typically very low and the only coherent, relatively strong phase is probably PmP. The prominent PmP and lack of intracrustal reflectivity in the vicinity of the GOC have also been noted by Cl´ement et al. (2004). To reduce the size of the 3-D volume we do not use PmP data from station 36 in the inversion. Including data from this station has a predictable effect: the Moho would be imaged at∼32 km approximately beneath station 38.
PmP is observed between offsets of 50–140 km. Arrival times
between different stations vary quite significantly; e.g. for offsets in the range of 90–100 km there is a difference of 3 s in PmP arrival times across the network. Exemplary recordings of PmP at stations on the west and east side of the array are shown in Fig. 5. Note the large difference in arrival times for station 38 in the west (Fig. 5a; 8–8.6 s) compared with station 21 in the east (Fig. 5b; 5.5–6.2 s), suggestive of significant Moho topography (and/or lateral velocity variation).
To provide the best areal coverage of PmP reflection points, for each station, where possible, PmP was picked on a number of shot lines. On average, shots from five lines were picked for PmP reflec-tions at each station. Picks were made on undecimated data, but the data were binned prior to inversion to give an effective shot spacing of 1.5 km. After initial inversion tests, data extremely inconsistent with nearby PmP picks, possibly due to misidentification of the PmP phase on noisy shot lines, were culled, leaving a total of 1534 PmP arrival times with an average uncertainty of 198 ms. Fig. 1 shows the number of PmP data for each station.
Figure 5. (a) Receiver gather for station 38 and shot line 30. Data are bandpass filtered between 3 and 8 Hz and reduced at 8 km s−1. The horizontal axis is the azimuth measured clockwise from receiver to shot. Offsets range from 68 km in the north to 72 km in the south. Note the strong, clear Moho reflection (PmP) at about 8 s. (b) Receiver gather for station 21 and shot line 8+23. Same plotting parameters as (a) except the horizontal axis is shot–receiver offset. Note that
PmP arrives much earlier (at about 6 s) compared with (a), suggesting a significantly thinner crust in the east. The inset nomogram in both panels shows the
location of the receiver (black dot) and shot line (black line) in relation to other stations and shots.
To solve for depth to Moho we use the tomography method of Zelt & Barton (1998), extended to include reflections (Zelt et al. 1999). This method, as applied to the inversion of reflection data to constrain reflector depth, penalizes the roughness and size of the reflector with respect to a starting model. The amount of regular-ization is automatically chosen so that the appropriate data misfit is achieved on the final iteration. Traveltimes are calculated by finite difference (Hole & Zelt 1995) on a uniform grid with a node spac-ing of 1.5 km. The inversion is calculated on a grid with a lateral node spacing of 3 km and vertical node spacing of 1.5 km. The reflector is free to assume any depth at each grid node location. The forward and inverse grid sizes were chosen to be small enough to allow accurate traveltime calculations and to adequately
repre-sent lateral Moho depth and velocity variations while maximizing computational efficiency.
In general, PmP traveltime residuals could be mapped into Moho depth variations and/or crustal velocity variations. Using only PmP data, it is not advisable to solve simultaneously for depth and velocity because the data are not capable of determining how the residuals should be partitioned. Thus, we start by using a fixed 1-D velocity model, based on an average of the 2-D velocity model for the east– west profile, and solve only for Moho depth. We stripped off the water and sedimentary layers and moved the shots to the base of the rift-onset sedimentary sequence determined by Weiss (2004). Traveltime through the water and sediments was removed from the observed data. This traveltime was estimated by ray tracing through
Table 3. 1-D velocity model
used in 3-D inversion for Moho depth. Velocities are linearly interpolated between each specified point. 0 km depth corresponds to sea level. Depth Velocity (km) (km s−1) −3.0 5.3 3.0 5.9 4.5 6.0 6.0 6.1 33.0 6.7 51.0 6.7
a simple water–sediment–crust model using the true shot–receiver offset to give the proper take-off angle from the shot and the correct ray path from the shot to the base of the sediments. The preferred 1-D velocity model (Table 3) increases from 5.3 km s−1at the top of the model (3 km above sea level) to 6.7 km s−1at 33 km depth, with an average velocity to 33 km depth of 6.25 km s−1. This model produced the best fit to the PmP data. 1-D models with average velocities within±0.1 km s−1of the preferred model resulted in Moho models which statistically fit the data equally well and which look similar, except depths are shifted by about±1.5 km, respectively.
The final Moho model, shown in Fig. 6, has a normalized χ2
misfit of 1.7 and a root mean square (rms) traveltime residual of
dtrms= 250 ms. The crust is thickest in the west beneath the
Hel-lenides, with a Moho depth up to 40 km, and relatively thin in the east, with a zone of very thin crust (25 km) beneath the Perahora Peninsula. A comparison of Moho depths along the 2-D profile is shown in Fig. 4 (grey line): depths are similar in the west, but the Moho of the 3-D model is 1–2 km deeper east of X= 120 km. The differences are not significant given the combined uncertainties of the two methods. Because a fixed, 1-D velocity model was used in the inversion, the topographic variations in this model are probably maximal, assuming that the 1-D model is an accurate representa-tion of the average velocity structure. Allowing traveltime residuals to map into both velocity and depth would produce a model with less topography. With a fixed 1-D velocity model we find the data are underfitted (χ2 > 1), which suggests that the crustal velocity
structure is not laterally homogeneous.
Model assessment
To test if, with ‘reasonable’ lateral velocity variations, this Moho model could fit the data to within the assigned uncertainties, we tried inverting the PmP data for crustal velocity variations while keeping the Moho of Fig. 6 fixed. The forward and inverse grid sizes were the same as described in the previous section; these sizes are small enough to adequately represent lateral velocity variations that our coarsely spaced data are capable of resolving. The data cannot discriminate the depth range over which lateral velocity variations occur. Thus, we allowed velocities throughout the entire crust to vary by up to±1 km s−1relative to the 1-D model in Table 3. The resultant model (Fig. 7) produced a misfit ofχ2= 1 (dt
rms= 189 ms) with
reasonable velocity variations. Shallow velocity variations near the stations (Fig. 7a) effectively represent station corrections. Relative to the 1-D starting model, velocities are typically faster down to about 15 km, then slower at greater depths. We stress, however, that the velocity model is non-unique and thus caution against attaching
Figure 6. Map of depth to the Moho from inversion of PmP traveltime
data using a fixed 1-D velocity model (Table 3). The contour interval is 1 km. Thick contour lines are drawn at depths of 30 and 40 km. Triangles are stations used in the inversion. Black dots are the locations of reflection points for the five non-linear iterations of the tomographic inversion. Only regions of the Moho within 5 km of a reflection point are unmasked. The misfit for this model isχ2= 1.7. The small region labelled ‘32 km’ on the
left-hand side of the model indicates depth to the Moho beneath station 38 based on 2-D modelling.
too much significance to the velocity variations. Most importantly, this model demonstrates that the Moho reflector in Fig. 6 together with this reasonable, smooth velocity model, fits the data to within the estimated uncertainties.
The shallow Moho (25 km) beneath the Perahora Peninsula on the isthmus separating the Corinth and Saronic gulfs is, topograph-ically, the most significant feature of the Moho model. To investi-gate the necessity for this feature we performed a test in which the Moho depth model of Fig. 6 was clipped to a minimum depth of 30 km. This interface was held fixed and PmP times were inverted to solve for velocity structure. The resultant model provided aχ2
misfit of 1, suggesting that the very shallow depths can be traded off against velocity. A region of significantly faster velocities centred around stations 21 and 22 is required relative to the model shown in Fig. 7. Velocities between depths of 3 and 15 km, however, are 6.7– 6.9 km s−1, perhaps somewhat faster than might reasonably be ex-pected, suggesting that while some trade-off between depth and ve-locity is permitted, a shallow Moho depth (<30 km) at this location is required by the data.
A comparison of our Moho model with the model of Tiberi et al. (2001), derived by inversion of gravity residuals, reveals major dif-ferences. For example, in the region where our tomography results constrain the Moho, the model of Tiberi et al. (2001) is, on aver-age, 5.5 km shallower. Using the 1-D velocity model of Table 3, the Tiberi et al. (2001) model produces a misfit ofχ2= 41 (dt
rms=
1180 ms). This is significantly worse than our model which pro-vides a misfit ofχ2= 1.7 (dt
rms= 250 ms) using the same velocity
model. To test if, with ‘reasonable’ lateral velocity variations, the model of Tiberi et al. (2001) could fit the data to within the assigned uncertainties, we tried inverting the PmP data for crustal veloc-ity variations while keeping the Tiberi et al. (2001) Moho fixed.
Figure 7. Velocity model derived by inverting PmP traveltimes while keeping the Moho model of Fig. 6 fixed. Horizontal slices at six depths show the
difference between final model and the 1-D model of Table 3. The contour interval is 0.05 km s−1. Thick contours are±0.2 km s−1. The velocity value indicated on each slice is the velocity of the starting 1-D model (Table 3) at that depth. Note that the velocity perturbations are generally smooth, and the resultant velocities are reasonable at all depths. This velocity model, combined with the Moho depth model of Fig. 6, produces a misfit ofχ2= 1.0.
The final velocity model, which gives aχ2misfit of 1.6 (dt rms=
234 ms), is quite unrealistic, with large regions of both unreasonably slow (<5 km s−1at 7.5 km depth) and fast (7.3 km s−1 at 18 km depth) velocities in the centre of the gulf. We reran this test after increasing the average depth of the Tiberi et al. (2001) model by 5.5 km (giving a mean depth of 34.5 km—the same as the preferred Moho model in Fig. 6) at all points to see if a static shift would make their model consistent with the PmP data. The resulting ve-locity model gave a misfit ofχ2= 1.3 (dt
rms= 220 ms), significantly
worse than our preferred Moho model (χ2 = 1, dt
rms= 189 ms;
Fig. 6) together with reasonable crustal velocity variations (Fig. 7). Furthermore the resultant velocity model has unreasonable velocity values, e.g.>7 km s−1beginning at 12 km depth beneath the GOC.
D I S C U S S I O N O F R E S U L T S
Our ability to constrain velocity structure is limited by the data, from which we have identified only three phases. Because of this, we have looked for the simplest model, i.e. with the least amount of structure, which fits the data. We were able to fit the data for the 2-D profile with a very simple model (Fig. 4). The only significant velocity variations are in the shallow crust (<6 km depth); the faster velocities in the west and slower velocities in the east are required primarily to delay or advance Pg arrival times relative to stations close to the gulf. The
PmP data could be fitted well without requiring significant lateral
velocity variation. The Moho dips to the west from about 28.5 km beneath station 19 to 38 km beneath station 49. Further west, beneath station 38, PmP data from station 36 images the Moho at 32 km. The location of this very rapid change in crustal thickness corresponds to the western edge of the NNW-trending Hellenides mountain chain, and suggests an abrupt termination of a relatively thick crustal root. Another onshore–offshore seismic experiment (Hirn et al. 1996) found Moho depth to be 24 km just south of our station 36 in the Gulf of Patras, as shown in Fig. 4. Makris (1978) documented an even more abrupt and more substantial thinning of the crust further south on the Peloponnisos at approximately the same location rel-ative to the Hellenides. The Moho map of Tiberi et al. (2001) also suggests significantly thinner crust to the west in the vicinity of our station 38.
We interpret the late and strong Ps phase recorded at station 36 as reflections from the subducting African Plate. We image the re-flector between stations 38 and 49 (Fig. 1) at a depth of 74 km, without significant dip, but we cannot ascertain whether the reflec-tions originate from the top or bottom (Moho) of the subducting crust. Papazachos & Nolet (1997) present a 3-D velocity model for the Hellenic subduction zone region derived from inversion of local earthquake events and show an east–west slice that is roughly paral-lel and about 25 km south of our profile. Their P-velocity tomogram
images the top of the subducting crust beneath the Peloponnisos at ∼60 km, dipping very slightly (3◦) to the east, and with a
veloc-ity anomaly thickness of 10–20 km. The depth of our Ps reflector would be consistent with a reflection from the base of the subducting crust as imaged by Papazachos & Nolet (1997). Otherwise, if the
Ps reflector originates from the top of the slab, this would suggest a
very significant decrease in depth (∼10 km) of the slab from north to south over a short distance (25 km). The slab as imaged by Pa-pazachos & Nolet (1997) does not change its dip until it reaches the eastern end of the GOC, where the dip sharply increases to∼30◦. In contrast, we observe Ps reflections only at station 36, yet if the slab were indeed (nearly) horizontal beneath most of the GOC we would expect to see similar phases on all of the western stations and possibly stations 20 and 19. The lack of any evidence of a Ps phase at these stations suggests that, at the latitude of our profile, the dip of the slab increases quite sharply in the vicinity of station 49 near the western part of the gulf as shown on Fig. 4. Again, if this is the case, the slab geometry changes dramatically over the short distance between our profile and the parallel profile of Papazachos & Nolet (1997).
Subduction of the African Plate begins at the south-ern/southwestern edge of the Mediterranean Ridge (Kastens 1991) (Fig. 1 inset). The dip of the slab is shallow until at least the Ionian island of Cephalonia where Hirn et al. (1996) image the top of the slab at a depth of 12 km. To reach a depth of 74 km where we image it between stations 38 and 49 implies an average dip of the slab of at least 30◦west of Cephalonia, as shown in Fig. 4.
The inversion of PmP reflection times recorded at 28 stations reveals that in the vicinity of the GOC Moho depth generally in-creases from east to west. Thick crust (>35 km) occurs beneath the Hellenides, both north and south of the gulf, where elevation ex-ceeds 2000 m; however, the crust is also relatively thick beneath the western end of the gulf. Thus, while the regions of high elevation appear to be isostatically compensated to some extent, regions of low elevation do not conversely correlate with thin crust, suggest-ing that the Moho has not responded in an obvious manner to the recent (Quaternary) phase of rifting. Just south of the gulf, however, between X= 100–120 km, we do see a region of relative thinning, with the Moho at∼35 km. The thinning south of the main Corinth depression suggests that a significant component of strain may be accommodated on faults south of the main rift. MCS reflection data from the western GOC, at about this longitude, show that the im-portance of faulting within the gulf diminishes here (Weiss 2004), which may be expected if extension transfers to, or is partitioned with, faults south of the gulf.
The thinnest crust occurs beneath the Perahora Peninsula, which is a feature on the isthmus separating the Corinth and Saronic gulfs. The central part of this region is dominated by the Gerania Mountains, which rise to over 1 km elevation. Moho depth here de-creases to 25 km; however, as noted previously, tests showed that the thin crust imaged here may in part be a response to higher crustal velocities in this region. Instead of the relatively slow-velocity crust associated with the rift it is conceivable that velocities in this re-gion, which lies at the northwestern end of the Hellenic Arc, are more typical of an arc setting, where mid-and lower-crustal rocks are, on average, about 0.3–0.4 km s−1 faster than typical rift set-tings (Holbrook et al. 1992). The nearest arc-related volcano is 50 km to the southeast on the island of Methana (Fig. 1); how-ever, there is evidence of volcanic activity on the isthmus at Susaki, which is a solfataric field (Simkin & Siebert 1994). Another poten-tial source for relatively high velocities are ophiolitic rocks, which crop out in this region. High velocities may trade-off somewhat
with depth, but it is unlikely that the crustal thickness here is greater than 30 km, meaning that, regardless of velocity variations, this region does indeed have the thinnest crust. The thin crust and to-pographic high may be a sign that the lithosphere is strong here and resistant to stretching and subsidence, or that the thin crust may in fact be a product of extension. For now this question is unresolved.
The region of the Moho constrained by PmP reflection points lies entirely within the map of Moho depth produced by Tiberi et al. (2001) from inversion of Bouguer gravity anomalies. They assume that the residual gravity anomaly, after removing the effect of the subducting slab, is due entirely to Moho topography. This is some-what analogous to our assumption that all variations in PmP times are due to Moho topography; however, because we include a good 1-D velocity model in our inversion, the depths we obtain are mean-ingful in terms of their absolute values. The gravity inversion maps gravity anomalies to an interface that is tied to an assumed ref-erence Moho depth, chosen as 30 km by Tiberi et al. (2001). We found that, on average, the Moho depths of Tiberi et al. (2001) are 5.5 km shallower than the depths derived by traveltime inversion, suggesting that a more appropriate value for their reference depth would have been∼35 km. They do point out that their results are stable for different reference depths, and thus we will compare our results with theirs after applying a 5.5 km shift. Within the region constrained by reflection points (non-grey region of Fig. 6) there is a general agreement about a regional increase in Moho depth from east to west; however, the correlation between high eleva-tion in the Hellenides with thick crust is stronger in our model. Both models show relatively thick crust beneath the western end of the GOC. The relatively very shallow Moho beneath the Perahora Peninsula is not present on the gravity Moho; instead the gravity Moho contains a very shallow feature about 40 km further east and other shallow features both slightly north and south of the central GOC. Tiberi et al. (2001) suggest that a zone of relatively thin crust along the northeastern margin of the GOC represents an offset be-tween extension-related crustal thinning and the main axis of the rift. Based on this they propose a model for extension which includes a north-dipping upper crustal low-angle normal fault with duc-tile deformation in the lower crust. Our Moho model shows a complicated pattern of depth variations that does not support this simple interpretation. It is possible that pre-rift Moho topogra-phy complicates signals associated with extension-related crustal thinning.
To explore further the possible relationships between crustal thickness (Moho depth plus elevation, Fig. 8a) and elevation (Fig. 8b), we cross-plot the two within the area of our tomographic inversion (Fig. 8c). Assuming pointwise Airy isostasy, the ratio of crustal thickness differences to positive elevations should equal the crustal density divided by the density difference of mantle and crust. A linear best fit to the observed data has a gradient of 4.16 and an intercept of 32.8 km. Likewise, the ratio of crustal thickness differ-ences to submarine depths should equal the crustal density contrast with water divided by the density contrast of mantle versus crust. A linear best fit to the observed data has a gradient of 2.03 and an intercept of 34.2 km. Constraining the two relationships to have the same crustal thickness at sea level (33 km at zero elevation), we plot the combined best fit curve in Fig. 8(c).
Note that a subaerial gradient of 4.16 is equivalent to a crustal density of 2.67 g cm−3given a mantle density of 3.30 g cm−3. The former is the crustal density commonly used in Bouguer gravity reductions. This value would predict a submarine gradient of 2.16 if the bathymetry represented crustal depths. We know, however, that
Figure 8. (a) Map of crustal thickness (i.e. Moho depth in Fig. 6 plus elevation). The contour interval is 1 km; heavy lines are the 30 and 40 km contours.
(b) Map of elevation. The contour interval is 0.5 km; the heavy line is the 1 km contour. (c) Plot of crustal thickness versus elevation for each point of the grid constrained by seismic data (non-grey regions in other panels). The line represents a best estimate of crustal thickness versus elevation for regions both below and above sea level. (d) Map of the difference between crustal thickness and predicted crustal thickness as predicted by the line in (c). In panels (a), (b) and (d) triangles are stations used in the 3-D inversion to derive Moho depth.
there are up to 2.5 km of sediments beneath the GOC and that, if sediments of average density 2 g cm−3filled the rift basin, they would produce a gradient of 1.16. Therefore, the observed intermediate value (2.03) of the submarine gradient also appears reasonable.
The crustal thickness versus elevation data (Fig. 8c) show sub-stantial variations from the simple gradients predicted assuming Airy isostasy, indicative that the crust/lithosphere is able to support loads (both positive and negative). Fig. 8(d) shows the spatial pat-tern of these variations. They are not random. We note the following differences relative to Airy isostatic model predictions:
(1) The crust is differentially thickest beneath the western GOC and thinnest beneath the Perahora Peninsula. The east–west gradi-ent in differences along the GOC mimics the regional gradigradi-ent in topography and Moho depths outside the gulf (i.e. high/thick in the Hellenides to low/thin towards the Aegean).
(2) As expected, short-wavelength topographic features are not locally compensated. For example, the longitudinal river valleys at
X= 95 km and at the head of the Gulf of Itea, north of the GOC,
and at X= 110 km south of the GOC, interrupt the regional crustal difference patterns with locally thicker anomalies.
(3) There are two well-resolved regions of crust thinner than pre-dicted by local Airy isostasy. The first, north of the GOC, parallels the North Gulf of Evia and is similar in location to that derived from gravity modelling by Tiberi et al. (2001). The second, immediately south of the GOC, follows the east–west normal fault trends there. Both these regions also correlate with generally positive topography (Fig. 8b), so the question arises as to the extent of undercompen-sation (or possibly lack of local compenundercompen-sation) represented by the relative crustal thinning. Resolution of this issue is beyond the scope of this study but will be investigated by joint inversion of the gravity data and crustal structure in future work.
C O N C L U S I O N S
In this study we have presented a 2-D velocity model for an axial profile through the Gulf of Corinth, and a regional map of Moho depth derived from inversion of reflection traveltimes. Our results contribute important information towards the understanding of the deep structure in the GOC region. The 2-D analysis reveals a simple velocity structure with a west-dipping Moho and a highly reflective subducting African slab at∼74 km depth beneath the western end of
the gulf. The 3-D analysis reveals thick crust beneath the Hellenides, relatively thin crust immediately south of the GOC, and a singular region of shallow crust centred on the Perahora Peninsula.
It is clear that the structural reality in the GOC (Fig. 8d) is more complex than either uniform pure shear or asymmetric simple shear models of crustal extension would predict. The region is not in local Airy isostatic equilibrium. There are substantial inherited (pre-rift) topography and crustal thickness variations that are superimposed by a 3-D pattern of crustal thinning that does not simply correlate with the location of upper crust brittle deformation. Notably, the western GOC is little thinned, whereas both the areas immediately northeast of and south of the GOC are.
A C K N O W L E D G M E N T S
Primary funding for this project was from the US NSF, with addi-tional funding from Greece (NOA) and France (CNRS). We thank the scientists and crew of the R/V Maurice Ewing cruise 0108. We thank the Hellenic Coast Guard for facilitating our operations and protecting the 6 km streamer with a chase boat. We thank Colin Zelt for providing the reflection tomography code and for guidance on its usage. We also thank H´el`ene Lyon-Caen for providing the record-ings of our shots at the Corinth Rift Laboratory seismic stations.
R E F E R E N C E S
Armijo, R., Meyer, B., King, G., Rigo, A. & Papanastassiou, D., 1996. Quaternary evolution of the Corinth rift and its implications for the late Cenozoic evolution of the Aegean, Geophys. J. Int., 126, 11– 53.
Billiris, H. et al., 1991. Geodetic determination of tectonic deformation in Central Greece from 1900 to 1988, Nature, 350, 124–129.
Briole, P. et al.,, 2000. Active deformation of the Gulf of Korinthos, Greece: results from repeated GPS surveys between 1990 and 1995, J. geophys.
Res., 105, 25 605–25 625.
Clarke, P.J. et al.,, 1997. Geodetic estimate of seismic hazard in the Gulf of Korinthos. Geophys. Res. Lett., 24, 1303–1306.
Clarke, P. et al.,, 1998. Crustal strain in central Greece from repeated GPS measurements in the interval 1989–1997, Geophys. J. Int., 135, 195– 214.
Cl´ement, C., 2000. Imagerie sismique crustale de la subduction H´ell´enique et du golfe de Corinthe, PhD thesis, Universit´e Denis Diderot, Paris.
Cl´ement, C., Hirn, A., Charvis, P., Sachpazi, M. & Marnelis, F., 2000. Seis-mic structure and the active Hellenic subduction in the Ionian islands,
Tectonophysics, 329, 141–156.
Cl´ement, C., Sachpazi, M., Charvis, P., Graindorge, D., Laigle, M., Hirn, A. & Zafiropoulos, G., 2004. Reflection-refraction seismics in the Gulf of Corinth: hints at deep structure and control of the deep marine basin,
Tectonophysics, 391, 85–95.
Davies, R., England, P., Parson, B., Billiris, H., Paradissis, D. & Veis, G., 1997. Geodetic strain of Greece in the interval 1892–1992, J. geophys.
Res., 102, 24 571–24 588.
Drakotas, G., Karantonis, G. & Stavrakakis, N., 1997. P-wave crustal to-mography of Greece with use of an accurate two-point ray tracer, Ann.
Geofis., XL, 25–36.
Goodliffe, A.M., Weiss, J.R., Taylor, B., Sachpazi, M., Hirn, A., Stefatos, A. & Laigle, M., 2003. Variations in the distribution and control of syn-rift deformation in the Gulf of Corinth, Greece, Geophys. Res. Abstr., 5, 04697.
Hatzfeld, D. et al.,, 1996. The Galaxidi earthquake of November 18, 1992: a possible asperity within the normal fault system of the Gulf of Corinth (Greece), Bull. seism. Soc. Am., 86, 1987–1991.
Hirn, A., Sachpazi, M., Siliqi, R., McBride, J., Marnelis, F., Cernobori, L. & the STREAMERS-PROFILES group, 1996. A traverse of the Ionian islands front with coincident normal incidence and wide-angle seismics,
Tectonophysics, 264, 35–49.
Holbrook, W.S., Mooney, W.D. & Christensen, N.I., 1992. The seismic ve-locity structure of the deep continental crust, in Lower Continental Crust, pp. 1–43 eds Fountain, D.M., Arculus, R. & Kay, R.W., Elsevier, New York.
Hole, J.A. & Zelt, B.C., 1995. Three-dimensional finite-difference reflection traveltimes, Geophys. J. Int., 121, 427–434.
Jolivet, L., 2001. A comparison of geodetic and finite strain pattern in the Aegean, geodynamic implications, Earth planet. Sci. Lett., 187, 95– 104.
Kastens, K.A., 1991. Rate of outward growth of the Mediterranean Ridge accretionary complex, Tectonophysics, 199, 25–50.
Lyon-Caen, H., Makropoulos, K., Papadimitriou, P., Deschamps, A., Bernard, P., Castarede, H., Cornet, F. & Patau, G., 2001. The Corinth rift laboratory seismological network: first results, Geophys. Res. Abstr., 3, 2544.
Makris, J., 1978. The crust and upper mantle of the Aegean region from deep seismic soundings, Tectonophysics, 46, 269–284.
Papazachos, C. & Nolet, G., 1997. P and S deep velocity structure of the Hellenic area obtained by robust nonlinear inversion of travel times, J.
geophys. Res., 102, 8349–8367.
Papazachos, C.B., Hatzidimitriou, P.M., Panagiotopoulos, D.G. & Tsokas, G.N., 1995. Tomography of the crust and upper mantle in southeast Eu-rope, J. geophys. Res., 100, 12 405–12 422.
Pi Alperin, J.M., Marthelot, J.M., Galv´e, A., Sachpazi, M., Taylor, B., Laigle, M. & Hirn, A., 2004. Seismic refraction imaging of the southern Corinth rift shoulder at Derveni. C. R. Geosci., 336, 251–257.
Rietbrock, A., Tiberi, C., Scherbaum, F. & Lyon-Caen, H., 1996. Seismic slip on a low angle normal fault in the Gulf of Corinth: evidence from high resolution cluster analysis of microearthquakes, Geophys. Res. Lett.
14, 1817–1820.
Rigo, A., Lyon-Caen, H., Armijo, R., Deschamps, A., Hatzfeld, D., Makropoulos, K., Papadimitriou, P. & Kassaras, I., 1996. A microseismic study in the western part of the Gulf of Corinth (Greece): implications for large-scale normal faulting mechanisms, Geophys. J. Int., 126, 663– 688.
Ruppel, C., 1995. Extensional processes in continental lithosphere, J.
geo-phys. Res., 100, 24 187–24 215.
Sachpazi, M. et al.,, 2002. Korinthos EW0108 cruise and the July–August 2001 Seismic survey of the Gulf of Corinth, Geophys. Res. Abstr. 05885. Simkin, T. & Siebert, L., 1994. Volcanoes of the World, Geoscience Press,
Tuscon, AZ.
Taylor, B., Goodliffe, A., Weiss, J., Sachpazi, M., Hirn, A., Laigle, M. & Stefatos, A., 2003. Detachment tectonics in the Gulf of Corinth rift, in EGS-AGU-EUG Joint Assembly, Nice, France, paper EAE03-A-07222
Tiberi, C., Diament, M., Lyon-Caen, H. & King, T., 2001. Moho topography beneath the Corinth Rift area (Greece) from inversion of gravity data,
Geophys. J. Int., 145, 797–808.
Weiss, J.R., 2004. A geophysical investigation of the Gulf of Corinth, Greece,
MSc thesis, University of Hawaii, Honolulu.
Weiss, J.R., Taylor, B., Goodliffe, A.M., Hirn, A., Sachpazi, M., Stefatos, A. & Laigle, M., 2003. Fault-controlled sedimentation in the Gulf of Corinth, Greece, Geophys. Res. Abstr., 5, 07036.
Wernicke, B., 1985. Uniform-sense normal simple shear of the continental lithosphere, Can. J. Earth Sci., 22, 108–125.
Zelt, B.C., Ellis, R.M. & Clowes, R.M., 1993, Crustal veloc-ity structure in the eastern Insular and southernmost Coast belts, Canadian Cordillera, Can. J. Earth Sci., 30, 1014– 1027.
Zelt, B.C., Taylor, B., Weiss, J., Goodliffe, A.M., Sachpazi, M. & Hirn, A., 2004. Streamer tomography velocity models for the Gulf of Corinth and Gulf of Itea, Greece, Geophys. J. Int., 159, 333– 346.
Zelt, C.A. & Barton, P.J., 1998. Three-dimensional seismic refraction to-mography: a comparison of two methods applied to data from the Faeroe Basin, J. geophys. Res., 103, 7187–7210.
Zelt, C.A. & Forsyth, D.A., 1994. Modeling wide-angle seismic data for crustal structure: southeastern Grenville province, J. geophys. Res., 99, 11 687–11 704.
Zelt, C.A. & Smith, R.B., 1992. Seismic traveltime inversion for 2-D crustal velocity structure, Geophys. J. Int., 108, 16–34.
Zelt, C.A., Hojkoa, A.M., Flueh, E.R. & McIntosh, K.D., 1999. 3D si-multaneous seismic refraction and reflection tomography of wide-angle data from the central Chilean margin, Geophys. Res. Lett., 26, 2577– 2580.
