Earth and Planetary Science Letters 209 (2003) 275^290 www.elsevier.

com/locate/epsl

The lithosphere^asthenosphere boundary in the western Mediterranean from 3D joint gravity and geoid modeling: tectonic implications
Conxi Ayala a; , Montse Torne a , Jaume Pous b
a

Department of Earth Sciences, Institute of Earth Sciences Jaume Almera-CSIC, Llu|s Sole i Sabar|s s/n, 08028 Barcelona, Spain b ' Departament de Geodinamica i Geof|sica, Universitat de Barcelona, Mart| Franques s/n, 08028 Barcelona, Spain Received 19 September 2002; received in revised form 5 February 2003; accepted 13 February 2003

Abstract A new integrated three-dimensional (3D) gravity and geoid forward modeling was used to investigate the lithospheric structure of the northwestern Mediterranean basins, namely the Gulf of Lyons, Liguro-Provencal Basin, Valencia Trough, and South Balearic Basin. Available geological and geophysical data were used to construct preliminary maps of the thickness of the sediments and the base of the crust and lithosphere, which were further refined by 3D modeling. Our model results show that the base of the crust thins from 22 km in the SW areas of the Valencia Trough to less than 12 km at its NE transition to the oceanic domain of the Provencal Basin and from 28 km in the shelf area of the Gulf of Lyons to about 12 km in the deep oceanic part of the basin. Upper mantle densities vary from 3240 kg/m3 beneath the Valencia Trough (where P-wave velocities are anomalously low, suggesting the presence of an anomalous uppermost mantle), to 3320 kg/m3 beneath the Gulf of Lyons and the Provencal Basin, where normal P-wave velocities are recorded. The base of the lithosphere rises from V75^80 km at the shoreline of the Valencia Trough and Gulf of Lyons to 60 km in the northern part of the Valencia Trough and 55^60 km at the center of the Provencal Basin. The structure of the crust and lithospheric mantle obtained from the 3D modeling indicates differential thinning. Moreover, other geophysical observations, e.g., the uneven distribution of heat flow values, suggest an asymmetric extensional model. 2003 Elsevier Science B.V. All rights reserved.
Keywords: geoid; gravity; lithospheric structure; Western Mediterranean; 3D modeling

1. Introduction The western Mediterranean Sea is formed by a series of Neogene extensional basins that were developed in a region of overall convergence between the European and African plates [1,2]. Late Cretaceous^Cenozoic convergence between the two plates gave rise to the present-day western Mediterranean conguration, characterized by a

* Corresponding author. Tel.: +34-409-54-10; Fax: +34-93-411-00-12. E-mail addresses: [email protected] (C. Ayala), [email protected] (M. Torne), [email protected] (J. Pous).

0012-821X / 03 / $ ^ see front matter 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0012-821X(03)00093-1

276

C. Ayala et al. / Earth and Planetary Science Letters 209 (2003) 275^290

system of orogens which bound the extensional basins (Fig. 1). In this convergent scenario there is no consensus on the mechanism capable of triggering the opening of the small Neogene basins surrounded by the Alpine belts. Some of the basins are of continental nature, e.g., Gulf of Lyons, Valencia Trough and Alboran Sea, whereas others are thought to be partially formed by oceanic crust, e.g., Provencal and South Balearic ba sins (Fig. 1). Earlier studies carried out in the area have resulted in dierent geodynamic hypotheses. Some authors support the view that intracontinental rifting is the mechanism triggering basin formation [1,2], while others believe that the basins are a consequence of roll-back of the Tethyan lithosphere slab [3] or back-arc extension induced by subduction (e.g., [4]) and largely inuenced by several episodes of slab detachment [5]. Other hypotheses like mantle delamination [6] or orogenic collapse [7], which exerts its inuence mainly at a crustal level [8] apply only to the Alboran^Betic area. Moreover, there is no consensus on the detailed kinematics of the whole region despite the large amount of quality geophysical and geological data (see [9], [10] and references therein) currently available from the area (seismic data is summarized in Fig. 2).

The wealth of geophysical data, including heat ow, gravity, near-vertical reection and refraction/wide-angle reection data have provided a detailed picture of the crustal conguration of the Gulf of Lyons (e.g., [11,12]) and the Valencia Trough (see [9] for further details). However, little is known of the crustal conguration of the Provencal and South Balearic basins and of the pre cise location and crustal geometry of the ocean^ continent boundary or the ocean^continent transition zone. Furthermore, although a number of studies have been carried out to investigate the topography of the base of the lithosphere (see for instance [13,14]), they have concentrated on 2D proles cutting across dierent areas of the Valencia Trough and the Gulf of Lyons. To date no attempt has been made to study the 3D lithospheric structure of the western Mediterranean. In this paper we present an integrated threedimensional gravity and geoid forward model derived for the northwestern Mediterranean basins to obtain a 3D lithosphere image of the area and the distribution at depth of the dierent crustal and lithosphere density boundaries. Gravity data, when constrained with seismic data, provide accurate information about the crustal structure, whereas geoid data can provide reasonably good information on the topography of the lithosphere^asthenosphere boundary. Therefore, the inherent uncertainties associated with the modeling of each of these potential elds alone are signicantly reduced. The 3D modeling has allowed us to construct a map of the depth to the Moho and to the base of the lithosphere for the NW Mediterranean region.

2. Input data 2.1. Elevation, gravity and geoid data Elevation, gravity and geoid data used in the modeling are shown in Figs. 2, 3 and 4. Gravity and geoid are the two observables used to constrain the density model, whereas elevation (topography and bathymetry) was used to dene the geometry of the uppermost layer of the den-

Fig. 1. Tectonic summary map of the northwestern Mediterranean (modied from Roca, 2001 [10]). The study area is outlined in black. EB, Ebro Basin; CCR, Catalan Coastal Ranges; CI, Columbretes Islands; VT, Valencia Trough; SBB, South Balearic Basin; C, Corsica; S, Sardinia; Ka, Kabilias; O, Olot.

C. Ayala et al. / Earth and Planetary Science Letters 209 (2003) 275^290

277

Fig. 2. Bathymetry and topography of the study area (m, positive upwards). Contours every 250 m. A synthesis of the seismic experiments in the northwestern Mediterranean that are relevant for this paper are also shown. VT, Valencia Trough. Nums: expanded spread proles (ESP). Dashed lines: Wide-angle seismics. Dotted-dashed line: common depth point (CDP), ESCI pro gram (Estudios S|smicos de la Corteza Iberica). Continuous lines at the VT: COP+CDP, VALSIS project (COP stands for common oset point; VALSIS is the name of the project of long listening multichannel seismic proles in the Valencia Trough). Dotted line: ANNA project. VT transect shown in Fig. 9A has the same location as PI.

sity model. Moreover, topography was used in geoid calculations to compute the eect of the masses above sea level. Topography was taken from the ETOPO5 data set [15] with a resolution of 5U5 min. As shown in Fig. 2, topography values range from V0^250 m at the central parts of the Ebro Basin to maximum values of 2000^ 2500 m at the Pyrenees. Note that gridding of data resulted in ltering the short wavelength component of the topography corresponding to local peaks. However, this is not relevant to our results since they do not make signicant contribution to the geoid. Bathymetry was obtained from the International Bathymetric Chart of the Mediterranean (IBCM [16]). The bathymetric contours shown in Fig. 2 were computed from tracks whose spacing varies between 3 and 10 km in the western Mediterranean. The tracks were gridded to a constant interval of 5U5 km. The Valencia Trough is

characterized by a central channel, with depths ranging from about 1000 m at its SE end to 2000 m at the NW. Its NW margin is characterized by the presence of a 50 km wide continental shelf that extends from its southern region to the north of the Ebro Delta (located at approximately 41N, Fig. 2). Further north, the shelf narrows to values of less than 5 km. To the SE, the trough is bounded by the Balearic rise that extends along the Balearic Promontory towards the NE and it is characterized by a gentle slope, showing a clear asymmetry with the NW margin of the trough (see Fig. 2). The continental rise o Menorca is approximately 30 km wide in its northwestern part, becoming narrower to the northeast. To the SE, a narrow shelf and a rather abrupt slope (the Emile Baudot Escarpment) mark the transition from the Balearic Promontory to the at abyssal plain of the South Balearic Basin, where water depths reach values of more than 2500 m.

278

C. Ayala et al. / Earth and Planetary Science Letters 209 (2003) 275^290

Fig. 3. (A) Free air anomaly (mGal) at sea and Bouguer anomaly (mGal) on land, contours every 10 mGal. (B) Bouguer anomaly, contours every 20 mGal.

The Gulf of Lyons is characterized by a wide continental shelf (more than 50 km) whose transition to the abyssal plain of the Provencal Basin is marked by a fairly abrupt slope. The water

depths increase towards the center of the Provencal Basin, where bathymetry reaches values of more than 3000 m. For the gravity modeling we used free air

C. Ayala et al. / Earth and Planetary Science Letters 209 (2003) 275^290

279

Fig. 4. Residual geoid anomaly (m), contours every 0.5 m. Heat ow values (in mW/m2 ) are superimposed: circles: 6 80; diamonds: 80^100; squares: s 100.

anomalies at sea and Bouguer anomalies on land (Fig. 3A). Free air anomalies were obtained from the Defence Mapping Agency (DMA-Aerospace Center) with coverage of at least one point every 2 km2 . The data are referred to the International Gravity Standardization Net 1971. Free air anomalies in the study area are characterized by an intermediate to long wavelength component dominated by values of about 0 mGal in the axial and central parts of the basins, whereas their anks are characterized by elliptical to circular anomalies of variable length and amplitude that oscillate from 20 to 50 mGal. This pattern suggests that the regional features of the area are in local isostatic equilibrium, which is in accordance with previous studies that show that the lithosphere has acquired little or no strength since rifting (e.g., [17]). Bouguer anomalies on land were taken from [18]. Bouguer gravity data show that the NE corner of the Iberian Peninsula is characterized by regional relative gravity lows of variable length and amplitude that delineate the Pyrenees and Iberian chain, attaining minimum values of 3120 mGal and 390 mGal, respec-

tively. The Ebro Basin is also marked by a NNW^SSE relative gravity low, which ranges from 320 mGal nearby the shoreline to more than 350 mGal in the NW corner of the basin. Although the free air anomaly at sea was the data used in the modeling, Fig. 3B shows the Bouguer anomaly gravity map to facilitate an initial qualitative approach of the crustal thickness variations of the region. The 3D gravity eect of the water layer was computed assuming a constant density of 2670 kg/m3 . Bouguer gravity data show that the Gulf of Lyons and the Provencal Basin are associated with positive gravity values that range from 10 mGal at the shoreline of the Gulf of Lyons to more than 240 mGal in the central areas of the Provencal Basin, whereas the Valencia Trough is associated with an enlongated SW^NE trending regional anomaly high of about 100^150 mGal, which clearly delineates the axial parts of the trough. The southern part of the Balearic Islands is characterized by strong gradients that rapidly lead to the South Balearic Basin, where Bouguer anomalies range from 150 to 220 mGal. The pattern of the Bouguer gravity

280

C. Ayala et al. / Earth and Planetary Science Letters 209 (2003) 275^290

anomalies can be explained by the combined effect of mantle and crustal thickness variations. In the western Mediterranean, the regional gravity highs correlate with areas of crustal thinning, as suggested by available seismic data. The geoid used in the 3D modeling is the residual geoid (Fig. 4) derived by subtracting from the geoid of GEOMED (Geoid of the Mediterranean [19]) a regional geoid obtained from the coecients of the EGM96 geopotential model [20] developed in spherical harmonics up to degree and order 12. This is a standard procedure that allows to remove from the geoid data the contribution of very deep mantle density sources, which correspond to the very long wavelength anomalies (e.g., [14], [21] and references therein). As observed in Fig. 4, the Valencia Trough is associated with a wide SW^NE trending anomaly ranging from 2 m in its SW corner to 30.5 m in the northeast. The anks of the trough are characterized by residual anomalies of up to 2.5 m in the Balearic Promontory whereas they reach more than 3 m in the Pyrenees and the Iberian chain. In the Gulf of Lyons the residual geoid ranges from 1 m at the shoreline to 0 m at the continental slope. A gentle gradient characterizes the transition to the Provencal Basin where values range from 31.5 m in its northern area to 0 m at its southern part. Geoid lows correspond to a mass decit whereas highs are associated with a mass excess. In the western Mediterranean, earlier 2D gravity and geoid modeling [14] suggests that the residual geoid is largely inuenced by variations in the lithospheric structures due to the rifting episodes that shaped the basins from at least Tertiary times. Thus, an initial qualitative interpretation suggests thinning of the lithosphere from the mountain ranges to the axial region of the Neogene basins, which is in agreement with available geophysical data, on land and at sea, and with the regional pattern of the Bouger gravity anomalies. The residual geoid will henceforth be referred to as the geoid. 2.2. Sediment thickness The thickness of the Plio-Quaternary sediments was digitized from the 1:1 000 000 scale IOC maps

(International Oceanic Commission [22]), contoured every 160 m and used to dene the geometry of the upper part of the sedimentary layer. The thickness of the Miocene sediments was compiled from dierent publications. At the Gulf of Lyons and the Provencal Basin, data come from [11,12,23]. In the Valencia Trough, information about the thickness of the Miocene sediments was obtained from [24^26]. The thickness of the Miocene sediments was also rened through modeling in areas where data coverage is poor or absent (e.g., the southern parts of the Provencal Ba sin). As shown in plate 1 ([22], p. 248), the distribution of Plio-Quaternary sediments is fairly uneven, particularly in the SW half of the study area (South Balearic Basin and Valencia Trough). In this area, the thickness of the Plio-Quaternary cover ranges from 0.2 to 0.4 km, except at the Iberian margin of the Valencia Trough. The largest accumulation of Plio-Quaternary sediments is found in front of the Ebro Delta (with thickness up to 2.8 km), in front of the Rhone deep sea fan and to the NE of the Provencal Basin, where elongated depocenters with thicknesses between 0.8 and 2.2 km are recorded. Local accumulations of Plio-Quaternary sediments are also observed in the southern areas of the Provencal Basin, with thicknesses up to 1.4 km. Miocene sediments are widespread all along the study area with highly variable thickness distribution. Circular to elliptical trending depocenters are observed in the axial parts of the basins and in the continental shelves. This pattern is also observed along the axis of the Valencia Trough where the thickness of the Miocene locally varies from 0.8 to 2.4 km. The isopach maps (Fig. 5), together with the bathymetric map (Fig. 2), indicate that the Valencia Trough is partly sediment starved (see also [9]). In the Gulf of Lyons, the thickness of Miocene sediments increases from the shelf to the axial zones of the basin, where local depocenters reach values of more than 4 km. Larger accumulations occur in the northern areas of the Provencal Basin, where values of more than 6 km are recorded, whereas the values oscillate around 4 km or less in the central and southern regions of the basin.

C. Ayala et al. / Earth and Planetary Science Letters 209 (2003) 275^290

281

Fig. 5. Thickness of Miocene sediments (km, positive downwards), contours every 0.4 km.

2.3. Moho depth and lithospheric thickness Seismic data, when available (see Fig. 2 for location of the seismic experiments), were used to constrain the thickness of the crust (e.g., [11,12,27,28]). The results of the seismic experiments show that the Valencia Trough is characterized by a strongly attenuated continental crust which is underlain by an anomalous low density velocity uppermost lithosphere mantle (V7.6 km/ s), except in its northeasternmost region where transitional or even oceanic crust is probably present (see [27]). A thin continental crust is also observed to the SW of the Gulf of Lyons and in the northern part of the Provencal Basin, where the Moho is located at 12^14 km. Towards the center of the Provencal Basin, the interpreta tion of seismic data suggests that the crust has reached oceanic stage with a total thickness of 2^4 km. The Moho deepens towards the coast of Iberia and France, reaching a depth of V20^ 24 km beneath the shoreline, and towards the Balearic Promontory, with depths of V23^28 km. Information on the conguration of the lithosphere^asthenosphere boundary comes primarily

from modeling results. 2D lithosphere modeling along selected transects cutting across the Valencia Trough and the Provencal Basin (e.g., [13,14]) shows that the base of the lithosphere remains at about 65^60 km in the central regions of both basins and rises towards the emerged areas, where it is found at V70^75 km at the shoreline of the Iberian Peninsula and France. Moreover, other regional works, such as surface-wave studies (e.g., [29]) also favor thinning of the lithosphere towards the central regions of the basins.

3. Three-dimensional modeling The study area (Fig. 1) has an extension of 650U600 km between 38.5 and 43.5N, and 1W and 7E. The three-dimensional density distribution was modeled as a set of layers formed by an assemblage of rectangular prisms, each with a constant density. The model consists of six layers: elevation (topography and bathymetry), Plio-Quaternary sediments, Tertiary (Miocene) sediments, crystalline crust, lithospheric mantle and the uppermost part of the asthenosphere.

282

C. Ayala et al. / Earth and Planetary Science Letters 209 (2003) 275^290

Throughout the process, the model is assumed to be in isostatic equilibrium, a requirement necessary to avoid edge eects in the geoid calculations [14,30]. Within each layer, the vertical prisms have a variable height and a horizontal cross-section of 10U10 km, which allow us to resolve structures corresponding to intermediate and long wavelengths, such as the geometry of the Moho or the topography of the base of the lithosphere. This represents a total of 66U61 prisms for each one of the six layers. For a single prism of constant density, the analytical solution of the gravity anomaly can be found in [31]. The geoid contribution is calculated solving the integral of the gravity potential for a regular prism (Fig. 6) and substituting the potential in Bruns formula (N = T/Q, where N is the geoid and T the potential), as described as follows (Zeyen, personal communication):
Z Z Z G b x2 y2 z2 1 dxdydz NP Q x1 y1 z1 R  zy lnz R xz lny R yz lnx R3 "  !  2  2 2 2 2 2
x y yR  x p arcsin   2 y R x2 y2 x2 y2 z2    !#   2 2 2 z x z xR  p  arcsin 2 z2  2 x R x x1 y1 z1

! y x y xR p arcsin 2 x R x2 y2

p R x2 y2 z2 , where x = j3x, y = R3y, j = h3z; (j,R,h) are the coordinates of the observation point and x, y, z dene the prism coordinates ; G is the gravitational constant, 6.67U10311 m3 kg31 s32 ; b is the density of a given prism; Q is the gravity acceleration. The geoidal contribution of all prisms will be

P the sum N n N P where N is the geoidal re1 sponse of the model, NP is the geoidal contribution of each prism and n is the number of prisms. The response of the model was calculated and compared with the observables. The model parameters (densities and coordinates of the top and bottom of the prisms) were successively modied until the best tting was obtained. To avoid edge eects in the calculations the model was extended in local isostatic equilibrium far beyond the limits of the study area. Seismic velocities from available expanded spread proles (ESP) in the Valencia Trough [27] and the Gulf of Lyons (e.g., [11]) were used to assign densities to the sediments, the crust and to the lithospheric mantle using standard P-wave velocity to density empirical relationships, e.g., [32] for the sedimentary cover, [33] for the crystalline crust and [34] for the lithospheric mantle. For the oceanic crust in the Provencal Basin, the density was taken from [35]. These density values are consistent with the ones used by other authors (e.g., [13,26,36]). Density for the asthenosphere was taken from [13]. In agreement with the available geophysical data, lateral variations of density were only used for the sedimentary cover of the Gulf of Lyons^Provencal Basin transition zone, and for the uppermost lithosphere mantle of the Valencia Trough where anomalously low P-wave velocities are recorded (Table 1).

4. Results 4.1. Sedimentary cover The results obtained by the 3D modeling allowed us to quantify the gravity and geoid eect of the sedimentary cover and in turn to validate the regional trend of the thickness of the Miocene sediments in the study area. In areas of poor data coverage or where modeling results showed large mists, the thickness of the Miocene sediments was slightly modied to account for these discrepancies mainly in the southern part of the Provencal Basin and in some areas of the South Ba learic Basin. Fig. 5 shows the modied thickness of the Miocene sediments.

Fig. 6. Parameters used in the geoid formula for a prism.

C. Ayala et al. / Earth and Planetary Science Letters 209 (2003) 275^290

283

4.2. Crustal structure The obtained depth to Moho is shown in Fig. 7A. The Moho boundary from our modeling is in agreement with the regional trend of the interpretation from the seismic data. Nonetheless, in some places, we had to slightly modify the thickness of the crust to t the gravity and geoid observations. In the Valencia Trough, the Moho rises along the axis of the basin from 22 km in the SW to 12 km in the NE, at the transition to the Provencal Ba sin. Across the trough, the crust thins towards the center of the basin showing an asymmetric gradient, which is steeper from the axis to the SE, towards the Balearic Promontory, than from the axis to the NW, towards the Iberian Peninsula. In the Gulf of Lyons, the Moho rises progressively from V26^28 km along the shoreline to about 12 km at its SE end, with a gradient that becomes steeper in the direction of the Provencal Basin. In the Provencal Basin, the Moho remains at an al most constant depth of V11^12 km. The thickness of the crystalline crust (without sediments) varies signicantly between the basins. Along the axis of the Valencia Trough, the crust thickens from approximately 20 km in the SW to about 6 km in the NE. In the Balearic Promontory, crustal thickness increases from approximately 24 km beneath Ibiza to about 28 km below Mallorca, and decreases again towards Menorca, where thickness of V22 km is found. In the Gulf of Lyons, the crust thins gradually from 30^32 km oshore in France to 4^8 km beneath the central part of the Provencal Basin. Locally, in the oce anic domain in the Provencal Basin, the crust at tains a thickness of approximately 2 km. These results are in accordance with earlier studies (e.g., [37]).
Table 1 Densities used in the modeling Layer Topography Water Plio-Quaternary sediments Miocene sediments Density (kg/m3 ) 2670 1030 2220 (VT) 2400 (GL, PB) 2400^2550

Previous studies based on the interpretation of Bouguer anomalies [38] and the magnetic eld [39] and on the analysis of the P-wave velocity distribution [11] indicate that the domain of the oceanic crust is restricted only to the deeper part of the Provencal Basin. There is no agreement on the precise location of the ocean^continent boundary and on the extension of the ocean^continent transition zone. As part of the 3D modeling we tested these interpretations and conclude that the best t for the gravity and geoid anomalies is obtained using the boundary dened by Morelli [38] (Fig. 7A), who suggests that the area of oceanic crust is outlined by the bathymetric contours of 2500^2700 m. In all the models, the error tting gravity anomalies dier in general within a range of 15 mGal, but the best geoid anomalies error tting (in general 0.25 m) was obtained for the boundary proposed by Morelli [38] (Fig. 8). For [10,23,40,41], the error tting for the geoid is in places larger than 0.75 m. 4.3. Lithosphere^asthenosphere boundary The most outstanding result of our study is the 3D mapping of the lithosphere^asthenosphere boundary for the entire region. As observed in Fig. 7B, in the Valencia Trough, the base of the lithosphere shallows from land (where it is found at V75 km below the coastline), to the transition in the Provencal Basin (where it is located at V60 km). A certain asymmetry is observed across the axis of the trough, with a steeper gradient from the Balearic Promontory to the center of the trough than from the center of the basin to Iberia. This asymmetry is similar to the one observed at crustal levels. In the Gulf of Lyons, the base of the lithosphere is located at about 80 km beneath

Layer continental crust oceanic crust lithospheric mantle asthenosphere

Density (kg/m3 ) 2850 2870 3240 (VT*) 3320 3245

284

C. Ayala et al. / Earth and Planetary Science Letters 209 (2003) 275^290

C. Ayala et al. / Earth and Planetary Science Letters 209 (2003) 275^290

285

the coastline. At the transition from the Gulf of Lyons to the Provencal Basin, the lithosphere thins towards the SE with a rather abrupt gradient in a N^S direction : the lithosphere^asthenosphere boundary rises from 75 to 60 km in V50 km of horizontal distance. One remarkable result of our modeling is that the base of the lithosphere remains at a fairly constant depth of V60 km in the Provencal Basin and in the northern part of the Gulf of Lyons and the Valencia Trough. Fig. 7B also suggests that this depth extends locally onland to the easternmost part of Iberia, in the Olot area (NE) and near Valencia and the Columbretes Islands (SE). The areas of thin lithosphere coincide with zones of recent (Neogene) volcanism [42,43]. 4.4. Error tting Fig. 8A and B show the error tting (observed data minus response of the nal model) in the gravity and geoid anomalies. The results show that, for most of the study area, the errors obtained with this model are small. These errors are in general less than 15 mGal for the gravity data and mainly 0.25 m for the geoid data, respectively. Some mists of higher amplitude and short wavelength component (like the ones located in the Valencia Trough and the Gulf of Lyons) are associated with lateral density variations at shallow crustal levels and, therefore, unrelated to variations in the lithospheric thickness. In this sense, at the Ebro platform, the positive gravity and geoid residual anomalies coincide with the thickness of Plio-Quaternary sediments locally exceeding 3.5 km. Therefore, these residual anomalies can be associated with the fact that the lithosphere has not reached local isostatic equilibrium in that area [17]. Sensitivity tests were carried out to ensure the accuracy of the model. The tests consisted of

changing the thickness of the lithosphere and the crust to evaluate the response of the model to these changes. Fig. 9 summarizes the results along two selected proles crossing the Valencia Trough (Fig. 9A) and Gulf of Lyons respectively (Fig. 9B) (see Fig. 2 for location). Changes of 2 km in the thickness of the lithosphere are within the limit of resolution for the calculated gravity and geoid anomalies (note the small dierence between these responses with respect to the solid line which corresponds to our preferred model). Variations of 5 km are clearly detected: the gravimetric response changes only few mGal, still within the accuracy of the data, whereas the geoid response increases, in general, more than 0.5 m. Variations in the thickness of the crust (depth to Moho) causes more conspicuous changes in the gravity and geoid response. Fig. 9 shows the response of the gravity and geoid anomalies when changing the thickness of the crust by 2 km which causes variations of more than 15 mGal in the calculated gravity and more than 1 m in the computed geoid. We conclude that, when the geometry of the sediments and the Moho is constrained by seismic data, the base of the lithosphere can be resolved by 2 km. In the southern part of the Provencal and South Balearic basins, where fewer seismic constraints are available, we assumed that there are no lateral density variations and used a constant value for each layer. In this area, we considered that the Moho depth is resolved with a resolution of 2 km and the lithosphere^asthenosphere boundary is determined with a resolution of 5 km. Some tests were also computed to delimit the ocean^continent boundary (OCB) in terms of lateral density variations. As we did not have enough constraints to dene a transitional zone between the oceanic and continental domains, we approximated the OCB as a line and introduced into the modeling

6 Fig. 7. (A) Depth to Moho (km, positive downwards), contours every 2 km, with the ocean^continent boundaries tested in the s, modeling according to: (1) Burru 1984 [47]; (2) Rehault et al., 1984 [23]; (3) Pasquale et al., 1996 [41]; (4) Roca, 2001 [10]; (5) present work, modied from Morelli, 1990 [38]. Grey numbers show the depth to Moho from expanded spread proles (ESP) interpretation; circles: ESP from Torne et al., 1992 [27]; crosses: ESP from Pascal et al., 1993 [11]. (B) Depth to asthenosphere (km, positive downwards), contours every 5 km.

286

C. Ayala et al. / Earth and Planetary Science Letters 209 (2003) 275^290

Fig. 8. (A) Error tting (observed3calculated) gravity anomaly (mGal), contours every 10 mGal. (B) Error tting (observed3calculated) geoid anomaly (m), contours every 0.25 m.

the boundaries proposed by a number of authors: [10,23,38,40,41]. The best t was obtained for the OCB from [38] (see outline in Fig. 7A).

5. Discussion and conclusions The gravity and geoid modeling presented in this paper allows us to obtain a detailed 3D geometry of the crust and lithosphere in the northwestern Mediterranean basins. Since available

geophysical constraints were taken into account, a fairly accurate 3D geometry of the base of the crust and top of the asthenosphere was obtained. In this sense, we rened the work done by other authors (e.g., [44,45]) who had only depicted the gross features of the European Mediterranean lithosphere (for instance, Yegorova et al. [44] work with a grid of 1U1). In our study we used a grid interval of 10U10 km which allowed us to resolve the medium-to-long wavelength component of the topography of the base of the lithosphere. The results along the Valencia Trough section (Fig. 9A) are in agreement with ' the work from Zeyen and Fernandez [13], that model the lithosphere using as observables heat ow, elevation and gravity. At crustal levels we nd attenuated continental crust in the deep waters of the basins. Our model is compatible with the presence of oceanic crust in the central part of the Provencal Basin. Nonethe less, the modeled density structure of the crust in that area could also be interpreted as a transitional crust heavily intruded by magmatic material, which could also explain the observed high P-wave velocities. Moreover, the magnetic anomalies [39], do not provide clear evidence of the extension of the oceanic crust in the area (they can be mostly associated with volcanic bodies). Therefore, the characterization of the oceanic crust in the northwestern Mediterranean is still an open question. The most striking result is that the base of the lithosphere in the northern part of the Valencia Trough, the southern part of the Gulf of Lyons and in the Provencal Basin is found at an almost constant depth of V60 km. This lithospheric thinning reveals that the eect of the Neogene rifting exerts an inuence on all the basins and on land, towards the eastern Iberian Peninsula where it coincides with Tortonian to Recent alkaline volcanic centers (e.g., Olot, see Fig. 1 for location), indicating that the area could still be under an extensional regime. According to our 3D model results, dierential stretching occurs between the crust and the mantle. Relative stretching factors (L) were calculated for the crust and lithosphere, assuming an initial thickness of 32 km for the crust and 110 km for

C. Ayala et al. / Earth and Planetary Science Letters 209 (2003) 275^290

287

Fig. 9. (A) Cross-section in the Valencia Trough, at the same location as PI from [27] (see Fig. 2 for location); VT, Valencia Trough; MI, Mallorca Island; SBB, South Balearic Basin; PQ, Plio-Quaternary sediments; TS, Tertiary sediments. (B) Cross-section in the Gulf of Lyons^Provencal Basin coinciding with some ESP in the area (ESP 201^209, see Fig. 2 for location); PQ, Plio-Quaternary sediments; TS, Tertiary sediments. In panels (a) and (b), gray shade corresponds to the observed data 15 mGal for the gravity data and 25 cm for the geoid data; these panels show the results from our nal model and from the sensitivity tests for the lithosphere thinned and thicknened 2 and 5 km, and the crust thinned and thickened 2 km. Tk, thickness; Lth, lithosphere; C, crust. Panel (c), heat ow data projected along the proles.

288

C. Ayala et al. / Earth and Planetary Science Letters 209 (2003) 275^290

the lithosphere [46]. Except in the emerged areas and the southern Valencia Trough, maximum L for the crust is 2.9 whereas maximum L for the lithosphere is 2, indicating that the crustal thinning is greater than the lithospheric thinning. In the southern part of the Valencia Trough, the base of the lithosphere is located at 65^70 km (L = 1.6) in an area where the Moho lies between 16 and 22 km (L = 1.4). This suggests that the lithosphere could have been inuenced by inherited structures from the Mesozoic, since this zone was a rift basin [10] prior the Pyrenean orogeny. In the Gulf of Lyons, the lithosphere thins with a steeper gradient than the crust, which can also be interpreted as an indication of the decoupling between the crust and the lithospheric mantle. Moreover, dierential stretching at crustal and mantle lithosphere levels across the Valencia Trough and the Gulf of Lyons (Fig. 7) displays asymmetry in the geometry of the extension, suggesting that the extensional mechanism in the western Mediterranean could have involved simple shear deformation, in agreement with results obtained by other workers (e.g., Mauret et al. [12]). Another geophysical parameter that indicates asymmetry in the lithospheric conguration of the basins is the uneven distribution of the heat ow values (see Figs. 4, 9A and B), as investigated by several authors (e.g., [41,47]) who also favor the hypothesis of simple shear. As pointed out by [48], to the west of the Gulf of Lyons and to the NW of the Valencia Trough, there are two main directions of extension, mainly deduced from extensional faults ([10,48]): a NNE^SSW direction, Late Eocene^Oligocene in age, which can be related to the Cenozoic rift of western Europe (e.g., [1]) and a NE^SW direction, Oligocene^Middle Miocene in age, which is associated to the eastward retreat of the Tethys subduction zone. As observed in Fig. 7A, the crustal thinning pattern in the Gulf of Lyons, the western part of the Valencia Trough and the Iberia adjacent areas are fairly coincident with the regional extensional fault pattern observed at upper crustal levels (NNE^SSW) which suggests the inuence of the Cenozoic rift in these areas. By contrast, the crustal and mantle lithosphere pattern along the Balearic margin of the Valencia Trough (NE^SW)

seems to correlate better with roll-back related extension. In the northernmost part of the Valencia Trough and the western part of the Gulf of Lyons, the thinning pattern of the lithospheric structure (SSE) is compatible with the interference of the southward propagation of the Cenozoic rift of west Europe with the Tethyan back-arc extension, whereas in the Provencal Basin, the litho sphere structure could have resulted mainly from roll-back of the Tethyan slab. In summary three main hypotheses can be advanced to account for the opening and evolution of the Neogene basins of the NW Mediterranean : (a) Southern propagation of the Cenozoic European rift system, (b) eastward retreat of a subducting slab; and (c) a combination of both. Although our results alone are not suciently conclusive to fully support any of these models, the compilation of the dierent data sets used in this study combined with geological data seems to indicate that neither propagation of the Cenozoic rift nor the eastward retreating of a subducting slab can fully explain the current lithosphere conguration of the study area, suggesting rather that the NW Mediterranean basins are the result of both processes.

Acknowledgements We thank Prof. Dr. Hermann Zeyen for his help and constructive comments in the development of this study. We also thank R. Sabadini, J. Lorenzo and an anonymous referee for their useful comments to improve the manuscript.[BARD]

References
[1] P.A. Ziegler, European Cenozoic rift system, Tectonophysics 208 (1992) 91^111. [2] P.A. Ziegler, Cenozoic rift system of western and central Europe: an overview, Geol. Mijnb. 73 (1994) 99^127. ' ' [3] C. Doglioni, E. Gueguen, F. Sabat, M. Fernandez, The Western Mediterranean extensional basins and the Alpine orogen, Terra Nova 9 (1997) 109^112. [4] F. Horvath, H. Berckhemer, Mediterranean backarc basins, in: H. Berckhemer, K. Hsu (Eds.), Alpine Mediter ranean Geodynamics, American Geophysical Union, Geodynamic vol. 7 (a), 1982, pp. 141^173.

C. Ayala et al. / Earth and Planetary Science Letters 209 (2003) 275^290 [5] E. Carminati, M.J.R. Wortel, W. Spakman, R. Sabadini, The role of slab detachment processes in the opening of the western-central Mediterranean basins: some geological and geophysical evidence, Earth Planet. Sci. Lett. 160 (1998) 651^665. [6] J.E.T. Channel, J.C. Mareschal, Delamination and asymmetric lithosphere thickening in the development of the Tyrrhenian Rift, in: M.P. Coward, D. Dietrich, R.G. Park (Eds.), Alpine Tetonics, Geol. Soc. London Spec. Publ. vol. 45, 1989, pp. 285^302. [7] J.F. Dewey, M.L. Helman, E. Turco, D.H.W. Hutton, S.D. Knott, Kinematics of the western Mediterranean, in: M.P. Coward, D. Dietrich, R.G. Park (Eds.), Alpine Tectonics, Geol. Soc. London, Spec. Publ. vol. 45, 1989, pp. 265^283. [8] M.J.R. Wortel, W. Spakman, Subduction and slab detachment in the Mediterranean^Carpatian Region, Science 290 (2000) 1910^1917. ' [9] M. Torne, E. Banda, M. Fernandez, The Valencia Trough: geological and geophysical constraints on basin ' formation models, in: P.A. Ziegler, F. Horvath (Eds.), Peri-Tethys Memoir 2: Structure and Prospects of Alpine ' Basins and Forelands, Mem. Mus. Natl. Hist. Nat. vol. 170, 1996, pp. 47^40. [10] E. Roca, The Northwest Mediterranean basin (Valencia Trough, Gulf of Lyons and Liguro-Provencal basins): structure and geodynamic evolution, in: P.A. Ziegler, W. Cavazza, A.H.F. Robertson, S. Crasquin-Soleau (Eds.), Peri-Tethys Memoir 6: Peri-Tethyan Rift/Wrench Basins and Passive Margins, Mem. Mus. Natl. Hist. Nat. vol. 186, 2001, pp. 671^706. [11] G.P. Pascal, A. Mauret, P. Patriat, The ocean^continent boundary in the Gulf of Lyons from analysis of expanding spread proles and gravity modelling, Geophys. J. Int. 113 (1993) 701^726. [12] A. Mauret, G. Pascal, A. Maillard, C. Gorini, Tectonics and deep structure of the Northwestern Mediterranean basin, Mar. Pet. Geol. 12 (1995) 645^666. ' [13] H. Zeyen, M. Fernandez, Integrated lithospheric modelling combining thermal, gravity and local isostasy analysis: Application to the NE Spanish Geotransect, J. Geophys. Res. 99 (1994) 18089^18102. [14] C. Ayala, J. Pous, M. Torne, The lithosphere^ asthenosphere boundary of the Valencia Trough (western Mediterranean) deduced from 2D geoid and gravity modelling, Geophys. Res. Lett. 23 (1996) 3131^ 3134. [15] ETOPO5, Data Announcement 88-MGG-02, Digital relief of the Surface of the Earth, NOAA, National Geophysical Data Center, Boulder, CO, 1988. [16] C. Morelli, S. Valchuk, The International Bathymetric Chart of the Mediterranean (I.B.C.M.) published under the authority of the Intergovernmental Oceanographic Comission of UNESCO, IBCM Supporting VolumeChapter 1, J.K. Hall (Ed.), 1997. [17] A.B. Watts, M. Torne, Subsidence history, crustal structure and thermal evolution of the Valencia Trough: A

289

[18]

[19]

[20]

[21]

[22]

[23]

[24]

[25]

[26]

[27]

[28]

[29]

[30]

[31]

Young Extensional Basin in the Western Mediterranean, J. Geophys. Res. 97 (1992) 20021^20041. ' A. Casas, M. Torne, E. Banda, Mapa Gravimetric de ' Catalunya (1:500000), Servei Geologic de Catalunya, De partament de Pol|tica Territorial i Obres Publiques, 1986. n M. Sevilla, Determinacio del Geoide, Bolet|n ROA, No 6/92, Real Instituto y Observatorio de La Armada, San Fernando (Cadiz), 1992. F.G. Lemoine, D.E. Smith, L. Kunz, R. Smith, E.C. Pavlis, N.K. Pavlis, S.M. Klosko, D.S. Chinn, M.H. Torrence, R.G. Williamson, C.M. Cox, K.E. Rachlin, Y.M. Wang, S.C. Kenyon, R. Salman, R. Trimmer, R.H. Rapp, R.S. Nerem, The Development of the NASA GSFC and NIMA Joint Geopotential Model, Proceeding paper for the International Symposium of Gravity, Geoid and Marine Geodesy (GRAGEOMAR 1996), The University of Tokyo, Tokyo, September 30^October 5, 1996. E.C. Molina, N. Ussami, The geoid in southeastern Brazil and adjacent regions: new constraints on density distribution and thermal state of the lithosphere, J. Geodyn. 28 (1999) 357^374. M. Gennesseaux, P. Burollet, E. Winnock, Thickness of the Plio-Quaternary sediments (IBCM-PQ), Boll. Geos. Teor. Appl. 39 (1998) 243^284. J.P. Rehault, G. Boillot, A. Mauret, The western Mediterranean basin geological evolution, Mar. Geol. 55 (1984) 447^477. E. Clavell, X. Berastegui, Petroleum geology of the Gulf of Valencia, in: A.M. Spencer (Ed.), Generation, Accumulation and Production in Europes Hydrocarbons, Spec. Publ. EAPG, Oxford University, 1991, pp. 335^368. V. Diaz del Rio, J. Rey, R. Vegas, The Gulf of Valencia continental shelf. Extensional tectonics in Neogene and Quaternary sediments, Mar. Geol. 73 (1986) 169^179. A. Maillard, A. Mauret, A.B. Watts, M. Torne, G. Pascal, P. Buhl, A. Pinet, Tertiary sedimentary history and structure of the Valencia trough (western Mediterranean), Tectonophysics 203 (1992) 57^75. M. Torne, G. Pascal, P. Buhl, A.B. Watts, A. Mauret, Crustal and velocity structure of the Valencia Through (western Mediterranean), Part I. A combined refraction/ wide-angle reection and near-vertical reection study, Tectonophysics 203 (1992) 1^20. J.J. Danobeitia, M. Arguedas, J. Gallart, E. Banda, J. Makris, Deep crustal conguration of the Valencia trough and its Iberian and Balearic borders from extensive refraction and wide-angle reection seismic proling, Tectonophysics 203 (1992) 37^55. F. Marillier, St. Mueller, The Western Mediterranean region as and upper mantle transition zone between two lithospheric plates, Tectonophysics 118 (1985) 113^130. ' C. Ayala, M. Fernandez, M. Torne, J. Pous, Two-dimensional geoid modelling: some remarks on Chapmans algorithm, Geophys. J. Int. 127 (1996) 542^544. H. Zeyen, J. Pous, 3D joint inversion of magnetic and gravity data with a priori information, Geophys. J. Int. 112 (1994) 244^256.

290

C. Ayala et al. / Earth and Planetary Science Letters 209 (2003) 275^290 [41] V. Pasquale, M. Verdoya, P. Chiozzi, Heat ux and timing of the drifting stage in the Ligurian^Provencal basin (Northwestern Mediterranean), J. Geodyn. 21 (1996) 205^ 222. [42] J. Mart|, J. Mitjavila, E. Roca, C. Aparicio, Cenozoic magmatism of the Valencia trough (western Mediterranean): relationship between structural evolution and volcanism, Tectonophysics 203 (1992) 145^165. [43] M. Wilson, G. Bianchini, Tertiary^quaternary magmatism within the Mediterranean and surrounding regions, in: B. Durand, L. Jolivet, F. Horvath, M. Seranne (Eds.), The Mediterranean Basins: Tertiary Extension within the Alpine Orogene, Geol. Soc. London Spec. Publ. vol. 156, 1999, pp. 141^168. [44] T.P. Yegorova, V.I. Starostenko, V.G. Kozlenko, Largescale 3-D gravity analysis of the inhomogeneities in the European-Mediterranean upper mantle, Pure Appl. Geophys. 151 (1998) 549^561. [45] G.F. Panza, P. Suhaldoc, Properties of the lithosphere in collisional belts in the Mediterranean ^ A review, Tectonophysics 182 (1990) 39^46. [46] P. Morgan, M. Fernandez, Neogene vertical movements and constraints on extension in the Catalan Coastal Ranges, Iberian Peninsula and the Valencia Trough (western Mediterranean), Tectonophysics 203 (1992) 185^201. [47] J. Burrus, J.P. Foucher, Contribution to the thermal regime of the Provencal basin based on Flumed heat ow surveys and previous investigations, Tectonophysics 128 (1986) 303^354. [48] A. Mauret, C. Gorini, Structural style and geodynamic evolution of Camargue and Western Provencal Basin, southeastern France, Tectonics 15 (1996) 356^375.

[32] S.E. Nafe, C.L. Drake, Physical properties of marine sediments, in: M.N. Hill (Ed.), The Sea vol. 3, Interscience, New York, 1961, pp. 794^819. [33] N.I. Christensen, G.H. Shaw, Elasticity of mac rocks from the Mid-Atlantic Ridge, Geophys. J. R. Astron. Soc. London 20 (1970) 271^284. [34] J.P. Barton, The relationship between seismic velocity and density in the continental crust ^ a useful constraint?, Geophys. J. R. Astron. Soc. London 87 (1986) 195^208. [35] R.L. Carlson, C.N. Herrick, Densities and porosities in the oceanic crust and their variations with depth and age, J. Geophys. Res. 95 (1990) 9153^9170. [36] J. Burrus, F. Bessis, B. Doligez, Heat ow, subsidence and crustal structure of the Gulf of Lyons (NW Mediterranean); a quantitative discussion of the classic passive margin model, in: C. Beaumont, A.J. Tankard (Eds.), Sedimentary Basins and Basin-Forming Mechanisms, Canadian Society of Petroleum Geologists, Memoir 12, 1987, pp. 1^15. [37] N. Chamot-Rooke, J.M. Gaulier, F. Jestin, Constraints on Moho depth and crustal thickness in the Liguro Provencal basin from a 3D gravity inversion: geodynamic implications, in: B. Durand, L. Jolivet, F. Horvath, M. Seranne (Eds.), The Mediterranean Basins: Tertiary Extension within the Alpine Orogen, Geol. Soc. London Spec. Publ. vol. 156, 1999, pp. 37^61. [38] C. Morelli, The regional meaning of the Bouguer gravity anomalies in the Mediterranean, J. Geodyn. 12 (1990) 123^136. [39] A. Galdeano, J.C. Rossignol, Contribution de laeromag' netisme a letude du Golfe de Valence (Mediterranee occidental), Earth Planet. Sci. Lett. 34 (1977) 85^99. s, [40] J. Burru Contribution to a geodynamic synthesis of the Provencal Basin (Northwestern Mediterranean), Mar. Geol. 55 (1984) 247^270.