Gravity constraints on the structure of the northern margin of Tunisia: implications on the nature of the northern African Plate boundary

Summary

Bouguer gravity data were analysed to determine the general crustal and upper-mantle structure in northern Tunisia. Residual gravity anomalies were determined by removing the gravitational effect of crustal thickness variations imaged by regional seismic experiments. Residual gravity anomalies contain short-wavelength anomalies superimposed on a long-wavelength component that decreases in amplitude northward towards the Tunisian coastline. An edge-enhancement analysis (e.g. enhanced analytic signals) of the short-wavelength anomalies suggests a previously unknown east–west-trending gravity anomaly south of 37°N with source depths of between 3 and 7 km. Modelling of residual and Bouguer gravity anomalies indicate that there are two possible solutions for the residual gravity decrease in northern Tunisia: (1) thickening of Cenozoic and Mesozoic sediments north of a strike-slip fault or (2) a crustal and upper-mantle low-density zone interpreted as being crustal material of the remnant subducted African Plate. The latter result is favoured based on seismic tomographic images of the Mediterranean region, which implies that subducting material exists under the African coast, geological interpretations suggesting that the Tell Atlas may be a thrust wedge accreted by underplating of the African continental crust and seismic refraction models, indicating a thinning of sediments in northern Tunisia. The east–west-trending gravity anomalies south of 37°N correspond to an important structural feature that may be related to either a structural boundary (e.g. a transform fault) or subduction beneath the African Plate.

continental margin, gravity anomalies, tectonics, Tunisia

Introduction

The crustal structure of the Mediterranean region has been affected by tectonic interaction between the Eurasian and African plates. The northern margin of the African Plate has been studied extensively by numerous researchers including Cohen et al. (1980), Scandone & Patacca (1984), Marillier & Mueller (1985), Turki (1985), Zargouni (1985), Philip (1987), Dewey et al. (1989), Mueller (1989), Ben Ayed (1993), Boukadi (1994), DeJong et al. (1994), Chihi (1995), Dlala (1995), Morel & Meghraoui (1996), Seber et al. (1996), Morgan et al. (1998) and Piqué et al. (1998). Even though there is agreement on the general crustal structure, there is considerable disagreement on the specfic crustal structure and the type of the plate boundary in northern Africa (Philip 1987; Dlala 1995; Morel & Meghraoui 1996; Piqué et al. 1998). In Tunisia, geological (Cohen et al. 1980; Rebai 1993; Chihi 1995; Dlala 1995; Morgan et al. 1998; Piqué et al. 1998; Zaïer et al. 1998; Carminati et al. 1998; Doglioni et al. 1999; Zeck 1999; Frizon de Lamotte et al. 2000; Vergés & Sàbat 1999) and seismotectonic (Udias 1982; Dlala 1992; Gueddiche et al. 1998) studies have shown that the present plate boundary is either compressional along a subduction zone, sinistral along a transform boundary or a combination of the two. In this study, we will perform an analysis of gravity data in northern Tunisia to add an additional constraint on the structure and nature of the North African margin.

Most workers (see the above references) agree on the general sequence of tectonics events affecting northern Tunisia. This sequence starts with Early Mesozoic rifting with the subsequent formation of a passive margin with deposition of Triassic evaporites, and Jurassic carbonates and turbidites. After this relatively quiet tectonic period, tectonic activity increased with active strike-slip faulting (?) from the Triassic to Early Cretaceous and convergent tectonic activity between the Eurasian and African plates from the Late Cretaceous to the present (Scandone & Patacca 1984; Dewey et al. 1989; Rebai 1993). This convergent tectonic period, included possible interactions with various microplates and the Tunisian continental margin. The convergence culminated with the mid-Miocene development of the Tell Atlas. Plate tectonic reconstructions that include Tunisia (Scandone & Patacca 1984; Dewey et al. 1989; Rebai 1993) are based on a wide range of data including sedimentological, structural, paleomagnetic, deep seismic refraction and tomography, ocean bathymetry, and marine magnetic anomalies. However, a reconstruction of several continental margins within the Mediterranean region (e.g. northwest Africa) could not be accomplished owing to a lack of subsurface data and the unknown original size of the ocean basins.

In order to determine a structural model of the northern margin of Tunisia, we analysed existing gravity data by constructing residual and regional gravity anomalies based on current crustal seismic models (Buness et al. 1992). The residual gravity anomalies, analysed using edge-enhancement techniques (Hsu et al. 1996) and 2-D forward models, were used as a first-order approximation on the nature of the plate boundary margin. The results of the analysis are compared with previous geological (Ben Ayed 1993; Dlala 1995; Morgan et al. 1998; Piqué et al. 1998) and seismic models (Mueller 1989; Buness et al. 1992; DeJong et al. 1994) to determine a deep structural scheme and possible geodynamic model of northern Tunisia.

Geological Setting

Northern Tunisia is located at the northern edge of Africa along the Mediterranean Sea (Fig. 1) and its uppermost crust is composed of Mesozoic and Cenozoic sedimentary rocks with scattered igneous lithologies that have been deformed by the Cenozoic Alpine orogeny. The main structural features include northeast-trending thrust faults and fold axes, northwest-trending grabens and dextral strike-slip faults, and minor dextral east-trending strike-slip faults (Rouvier 1977; Turki 1985; Ben Ferjani et al. 1990; Burollet 1991; Ben Ayed 1994). These features can be divided into two structural zones: the Tell Atlas and the northern Atlas (Fig. 2). In addition, offshore tectonic features include the North African or Algerian Basin (Alvarez et al. 1974).

Figure 1.

Regional tectonic features of the Mediterranean region and location of the study area. The lines with triangles represent regional thrust faults (Vergés & Sàbat 1999; Calvert 2000).

Figure 2.

Principle geological features in northern Tunisia. Stippled pattern represents the Tell Atlas.

The Tell Atlas is a major fold and thrust belt consisting of a series of nappes composed of Oligo-Miocene-aged Numidian flysch. The upper sections of the Numidian flysch were severely folded with fold axes basically trending northeast (Rouvier 1977; Cohen et al. 1980; Ben Ferjani et al. 1990). The Tell Atlas was formed during the collision of Tunisia with a continental plate to the north (e.g. the Corsica–Sardinia–Petite Kabylie continental plate) (Cohen et al. 1980). This collision also resulted in strike-slip motion on transfer faults to the east and west of the thrust belt, which may have originated in a phase of rifting (Morgan et al. 1998).

The northern Atlas is characterized by numerous northeast-trending exposures of Triassic rocks situated within younger strata. The emplacement of the Triassic rocks has been the subject of debate (El Ouardi & Turki 1995; Zaïer et al. 1998; Vila et al. 1998, 1999; Perthuisot et al. 1999) as the Triassic outcrop patterns were thought to have formed by diapirism (Perthuisot et al. 1981; Burollet 1991; Perthuisot et al. 1999). However, recent work (Morgan et al. 1998) suggests that the origin of the northeast-trending outcrops were formed during thrusting and inversion tectonics related to the Alpine orogeny. Alternatively, Vila et al. (1999) suggested that the Triassic rocks were emplaced as a ‘salt glacier’ (chaotic Triassic material interbedded within Late Cretaceous layers).

The northern Atlas also contains northeast-trending fault inversion folds, dissected by northwest-trending normal faults that formed Upper Miocene to recent half-grabens and grabens (Turki 1985; Ben Ferjani et al. 1990; Chihi 1995; Morgan et al. 1998). These structures were formed during the collision with a continental plate (e.g. Corsica–Sardinia–Petite Kabylie continental plate) with the Tunisian continental margin (Cohen et al. 1980). Additionally, metamorphic rocks within the Triassic and Lower Cretaceous layers have been noted in the region (Bolze 1954; Rouvier 1977; Alouani 1991; Soussi 2000). No detailed studies have been performed on these rocks.

The Algerian Basin (Fig. 1), which is comparable to active backarc basins, was formed in the Lower to Middle Miocene (Dewey et al. 1989; Oldow et al. 1990). The basin contains numerous volcanic layers located on steep bathymetric gradients (Girod & Girod 1977) that were formed during the opening of the basin.

Igneous rocks in northern Africa and Tunisia provide an important constraint on possible tectonic settings in northern Tunisia. In North Africa, Miocene volcanism in Morocco, Algeria and Tunisia is essentially calc-alkaline with locally shoshonitic lavas and are commonly associated with rhyolitic lavas and granodioritic plutons (Bajanik 1971; Bellon & Brousse 1977; Girod & Girod 1977; Rouvier 1977; Mauduit 1978; Halloul 1989; Laridhi Ouazaa 1994). These Miocene igneous rocks have led some workers (Auzende et al. 1975; Girod & Girod 1977) to postulate the existence of one (or more) subduction zones in the southern Mediterranean Sea contemporary with the opening of the Algerian Basin. Based on structural arguments, this subduction is thought to be north-dipping with the African continent being under the Eurasian Plate (Philip 1987; Chihi 1995; Chihi & Philip 1999).

Gravity Data and Processing

The gravity data used were obtained from the Department of Geological Sciences at the University of Texas at El Paso and the Enterprise Tunisienne des Activités Pétrolières. The two data sets, which together consist of 803 gravity values tied to the 1971 International Gravity Standardization Net (Morelli 1976), were merged and reduced using the 1967 International Gravity formula (Morelli 1976). Free-air and Bouguer gravity corrections were made using sea level as a datum and 2.67 g cm⁻³ as a reduction density. Terrain corrections were applied using a 5 min topography grid (US National Geophysical Data Center 1995), the terrain correction method of Plouff (1977) and a density of 2.67 g cm⁻³. The merged Bouguer gravity anomaly data were gridded at a spacing of 1 min (approximately 2 km) using the minimum curvature technique (Briggs 1974) and contoured at a 5 mGal interval to produce a complete Bouguer gravity anomaly map (Fig. 3).

Figure 3.

Complete Bouguer gravity anomaly map of northern Tunisia. Contour interval is 5 mGal. Station locations are indicated by +. Relative gravity minima are shown by hatched contour lines.

Data Analysis

The complete Bouguer gravity anomaly map (Fig. 3) shows a long-wavelength north to south decrease in the gravity field that can partially be explained by a gradual increase in crustal thickness (Mickus & Jallouli 1999; Jallouli & Mickus 2000). The sources of the shorter-wavelength anomalies are not completely known owing to the complex tectonic regime of northern Tunisia but include Cenozoic grabens and molasse basins, varying thicknesses of Mesozoic salt deposits and igneous intrusions (Jallouli & Mickus 2000). However, a detailed analysis of these anomalies has not been attempted previously and such an analysis will provide insight into the tectonic regime of northern Tunisia.

Regional and residual gravity fields

The first step in our analysis was to create a residual gravity anomaly field by removing the gravitational effects of widespread density contrasts from the Bouguer gravity anomaly field. One such density contrast is the crust–mantle boundary, the depth and velocity structure of which is relatively well known in northern Tunisia from the European Geotraverse seismic refraction profiles (Buness et al. 1989, 1992). The seismic refraction results in northern Tunisia indicate that the middle to lower crust (all lithologies beneath the Cenozoic sediments) have a relatively low P-wave velocity (6.0 km s⁻¹) with no clear evidence for any high-velocity crust. However, the data indicate that the mean crustal velocity increases toward the north (Buness et al. 1989, 1992), but this observation is open to interpretation owing to the poorly resolved P-wave arrivals. Additional studies of the European Geotraverse data by Boccaletti (1990) shows that the average velocity under the Tell Atlas could be as high as 6.6 km s⁻¹. So, for our purposes, the P-wave crustal velocity could range between 6.0 and 6.6 km s⁻¹. Additionally, the P-wave velocity of the upper-mantle lithologies has been shown by the above authors to vary between 7.9 and 8.1 km s⁻¹.

Using the above range of crustal and upper-mantle P-wave velocities, density contrasts ranging between 0.34 and 0.63 g cm⁻³ were determined using velocity–density relationships (Nafe & Drake 1957). These two extremes in density variation were then used in conjunction with the crust–mantle boundary determined by the European Geotraverse along a north–south profile (Fig. 4) to calculate two regional gravity anomaly profiles and their corresponding residual gravity anomalies (Fig. 5). It is clear that the residual gravity anomaly is influenced by whatever density contrast is used in creating a regional gravity anomaly (Fig. 5). Using the two density end members, a regional gravity anomaly calculated using a 0.63 g cm⁻³ density contrast, has a high gradient and a corresponding residual gravity anomaly magnitude of approximately 120 mGal. Whereas using a 0.34 g cm⁻³ density contrast, the regional gravity anomaly has a lower gradient with a corresponding residual gravity anomaly magnitude of approximately 40 mGal.

Figure 4.

Depth to the crust–mantle boundary in northern Tunisia as derived from regional seismic refraction surveys (Buness 1989). The contour interval is 1 km. The seismic refraction profiles used in constructing the regional gravity anomaly map and the gravity model location are also shown.

Figure 5.

Regional and residual gravity anomalies along a north–south profile (Fig. 4) showing the effect of different crust–mantle density contrasts. The density contrasts are in g cm⁻³.

Since the exact crust–mantle density contrast is not known and the derived density contrasts have errors inherent in the methods of calculating them (e.g. density–velocity relationships, derived P-wave velocity values), we choose to use the lower density limit (0.34 g cm⁻³) as not to overestimate the magnitude of the residual gravity anomaly.

The European Geotraverse seismic model (Buness et al. 1989, 1992) indicates that the crust thins toward the north. The thickness variation along this model is 11 km with the crust–mantle boundary dipping approximately 4.2°. Since there is a high velocity contrast along this discontinuity (6–8 km s⁻¹), the refracted waves are well observed. However, the estimation of refraction arrivals and the corresponding calculation of P-wave velocities always involve errors. Any errors in the velocity calculations would be translated into miscalculations of the crustal thickness and thus our calculated regional and residual gravity anomalies. To determine a range of possible gravity anomalies, we have modified the dip of the crust–mantle boundary by up to 15 per cent. This value is well within the resolution of modern refraction surveys (Mooney 1989). Fig. 6 shows three regional gravity anomaly profiles and their corresponding residual gravity anomalies for a crust–mantle boundary with a density contrast of 0.34 g cm⁻³ using dip values of 3.5°, 4.2° and 5°. These dip values correspond to crustal thickness variations of 9.2, 11 and 12.8 km, respectively. We note that the effect of any potential error of the Moho dip is not important and will not significantly affect the magnitude of the long-wavelength residual anomaly (Fig. 6). Additionally, if a higher density contrast were used, no matter what the crust–mantle dip boundary, the residual gravity anomaly magnitude would increase and produce a potential overestimation of the residual gravity anomaly.

Figure 6.

Regional and residual gravity anomalies along a north–south profile (Fig. 4) showing the effect of varying the dip (in degrees) on the crust–mantle boundary. A crust–mantle density contrast of 0.34 g cm⁻³ was used.

Using the depth to the crust–mantle boundary from all the European Geotraverse seismic refraction profiles in Tunisia (Buness et al. 1989, 1992), a 3-D gravity model using the technique of Parker (1973) was constructed using a 0.34 g cm⁻³ density contrast between the upper mantle and the crust. The resultant gravity field (Fig. 7) represents one possible regional gravity anomaly field in northern Tunisia. This regional gravity anomaly field was then subtracted from the complete Bouguer anomaly field to obtain a residual gravity anomaly field (Fig. 7).

Figure 7.

Regional and residual gravity anomalies in northern Tunisia. The regional gravity anomaly (dashed contour lines) represents the gravitational effect of crust–mantle variations (Fig. 4) in northern Tunisia. The contour interval is 20 mGal for the regional gravity anomalies and 5 mGal for the residual gravity anomalies.

Residual gravity anomaly analysis

The residual gravity anomaly field (Fig. 7) is dominated by short-wavelength anomalies that correspond to some of the geological features mentioned above. However, a long-wavelength component still remains, as there is a general decrease in the gravity field toward the north. Its average magnitude is 40 mGal, however, it could be higher if the crust–mantle density contrast is greater than 0.34 g cm⁻³. This long-wavelength anomaly could be caused by either density variations within the crust, a regional upper crustal density feature (e.g. the large sedimentary basin in northern Tunisia), an upper-mantle feature (e.g. a subducting plate) or a combination of any of the above scenarios. However, the first alternative has been excluded since there is evidence that the mean crustal velocity increases toward the north (Buness 1989 et al., 1992; Boccaletti et al. 1990) and the regional gravity anomaly (Fig. 7) was calculated using the lowest estimated crust–mantle density contrast. Increasing the density of the crust toward the north as suggested by European Geotraverse seismic results could not explain the regional decrease in the residual gravity anomaly.

The residual gravity anomaly field (Fig. 7) was first analysed by performing edge-enhancement procedures (e.g. horizontal and vertical gradients, analytic signals), as these techniques can be used to locate the lateral boundaries of density contrasts and provide information on the location of geological units (Blakely & Simpson 1986). We used the amplitude of the horizontal gradient (Cordell & Grauch 1985) and the enhanced analytic signal (EAS) (Hsu et al. 1996) techniques to analyse the residual gravity anomaly. The results of these two techniques produced similar results, and we used the EAS technique in our analysis since it uses both the vertical and horizontal gradients of the gravitational field and may produce a better determination of the edge of the source body (Hsu et al. 1996). The use of edge-enhancement techniques provides only a first-order geological interpretation of gravity data as Grauch & Cordell (1987) showed that the results are often contaminated by nearby sources. Additionally, Dehler & Lowe (1999) showed that the interpretation of the boundaries could provide reconnaissance information concerning general structural trends but a priori information (geological or seismic reflection–refraction data) is needed in order to obtain reliable results.

Fig. 8 shows an EAS map of northern Tunisia that was further enhanced by applying the automatic gain control technique (Rajagopalan & Milligan 1995). The automatic gain control technique enhances small-amplitude, short-wavelength anomalies, while not removing the larger-scale anomalies. The short-wavelength anomalies correspond to known geological features including grabens and northwest- and northeast-trending faults and folds (Fig. 2). However, probably the most striking anomaly is an east–west-trending anomaly (EWTA) just south of 37°N. This anomaly has not been shown previously and has no obvious surface geological source, as most of the structural features trend either toward the northwest or northeast (Rebai 1993). Additionally, the Bouguer gravity (Fig. 3) and residual gravity (Fig. 7) anomaly maps do not indicate distinctive east–west-trending anomalies, so a casual analysis of the gravity field would miss a potentially important density source. A depth analysis using the EAS (Hsu et al. 1996) indicates that the source of the EWTA lies at depths between 3 and 7 km. This depth corresponds to the majority of hypocentres of the recent seismic events (Gueddiche et al. 1998) and the anomaly may be related to the source of these events. Below we will try to explain the origin of EWTA.

Figure 8.

Enhanced analytic signal of the residual gravity anomalies (Fig. 7). Areas with values greater than 1.7 mGal km⁻³ are shown in grey-scale. Contour interval is 1 mGal km⁻³. The thick bold line represents the east–west-trending gravity anomaly (EWTA) discussed in the text.

Gravity Modelling and Discussion

The above analysis of the residual gravity anomalies indicates that there is a long-wavelength residual anomaly, the source of which is either in the upper mantle or the crust. Superimposed on this long-wavelength anomaly are several short-wavelength gravity maxima, the sources of which are probably in the upper crust. These anomalies can be clearly seen on four north to south profiles (Fig. 9) taken between 9° and 9.75°E. These profiles show the same general trend of the gravity values decreasing toward the north. Additionally, the prominent EWTA (Fig. 8) does not stand out on these profiles. However, a close examination of the profiles shows that this area is the only region where each profile exhibits either a maxima or a change in gradient. All other short-wavelength anomalies are confined to one or two profiles.

Figure 9.

North–south profiles of the residual gravity anomalies (Fig. 7) along specific longitudes.

The lowest gravity values in Fig. 9 are located in a region consisting of predominantly northeast-trending features including thrust faults, nappe sheets (Numidian), evaporitic diapirs, anticlines and synclines (Burollet 1991) that were formed during the creation of the Tell Atlas. In particular, the Tell Atlas consists of thick piles (at least 6 km) of deep-water sediments emplaced on to the African continent (Morelli & Nicolich 1990). The emplacement of these thick sediments might, at first glance, be the source of the gravity minimum (Fig. 9) however, seismic refraction models (Buness et al. 1992) do not image thick sediments in this region and we will show below that the sediments are not thick enough to account for the entire anomaly unless they have an unusually low density.

To aid in explaining the potential sources of the residual gravity field, we constructed a series of gravity models along longitude 9°15′E using a 2-D forward modelling algorithm (Cady 1980). These bodies include five groups of Paleozoic to Cenozoic sedimentary rocks, the initial thickness of which was based on drillhole information (Jallouli & Mickus 2000) and seismic refraction investigations (Buness 1992). The initial densities of the bodies were estimated from previous gravity studies (Mickus & Jallouli 1999; Jallouli & Mickus 2000) and from seismic velocities (Buness 1992) converted to densities using velocity–density relationships (Nafe & Drake 1957). Additionally, the densities and thickness/geometries of the bodies were modified by up to 5 per cent of their initial value during the modelling process in order to obtain an acceptable observed/predicted anomaly match.

The main aspect of this research is to explain the long-wavelength decrease in the residual gravity anomaly toward the north (Fig. 7) in terms of the tectonic origin of northern Tunisia. The seismic refraction model of Buness et al. (1992) does not provide such a model as they have a thinned crust toward the north and additionally there is no thickening of the sedimentary units. Poor data quality prevented imaging of any low-density, deeper crustal/upper-mantle units, so we will speculate on two possible models that can explain the source of the regional residual trend. The first model (Fig. 10) is based on deeper sedimentary units north of the EWTA. The second model is based on an upper-mantle density contrast (possibly subducted crust) under northern Tunisia (Fig. 11). While both of these models can adequately fit the observed data, we will explain below why we favour the latter model.

Figure 10.

Gravity model to explain the source of the residual gravity anomalies (Fig. 5). This model is based on a thickening of sediments in northern Tunisia along a strike-slip plate boundary. The numbers represent the density of the body in g cm⁻³.

Figure 11.

Gravity model to explain the source of the residual gravity anomalies (Fig. 7). This model is based on the continental crust being subducted in northern Tunisia beneath the European continent. The numbers represent the density of the body in g cm⁻³.

Fig. 10 shows a model in which the sedimentary units abruptly thicken toward the north. A sedimentological study of northern Tunisia by Soussi (2000) shows that the Mesozoic sediments are thicker north of the EWTA than to the south. Soussi also indicates that this region has been the site of increased subsidence and thickening of sediments since the Middle Lias (Jurassic). This increased subsidence was formed as part of a complete passive margin complex with northern Tunisia either being the site of a continental slope or abyssal plain (Soussi 2000). In this scenario, the abrupt thickening (Fig. 10) would mark the edge of the Mesozoic continental slope with the African Plate boundary being located either north or south of this region. One possible interpretation of the prominent EWTA is that it marks the edge of the continental slope.

Another interpretation of the EWTA is that the thick sediments within a pull-apart basin cause the abrupt thickening. Various authors have considered that at least part of the North African Plate boundary has been transpressive during convergence (Morel & Meghraoui 1996) or a transform fault (LeBlanc & Olivier 1984; Ben-Avraham et al. 1987). Before the Alpine convergence started in the Late Cretaceous and culminated in the Miocene, extensional regimes have existed at various times in northern Tunisia (Dewey et al. 1989; Morgan et al. 1998) since the Triassic. Thus, pull-apart basins may have been formed in northern Tunisia causing the increased sedimentary thickness and the accumulated thickness could explain the residual gravity minimum. Cenozoic rift basins exist south of the present-day Tell Atlas region (Burollet 1991) and may possibly exist beneath the Tell Atlas. However, based on gravity modelling (Jallouli & Mickus 2000), these basins are small structures that have a maximum thickness of 5 km and are represented by short-wavelength gravity anomalies.

The above model (Fig. 10) is based on sedimentary units thickening toward the Tell Atlas to explain the regional decrease in the residual gravity anomaly toward the north. However, exact sediment thickness values are not available in northernmost Tunisia and there is no information available concerning the thickness or existence of Paleozoic sediments north of the EWTA. However, a thinning of the Mesozoic sediments in the northernmost Atlas and Tell Atlas is shown by numerous studies (Rouvier 1977; Ben Ferjani et al. 1990; Alouani 1991). An isopach map of the Abiod formation (Late Campanian–Lower Maastrichtian) constrained by seismic reflection profiles and well data (Ben Ferjani et al. 1990) shows that the Abiod formation becomes thinner toward the northernmost Atlas and Tell Atlas.

The main argument against increased sedimentary thickness in northernmost Tunisia is the seismic refraction models of et al. (1992). Despite poor refraction arrivals off deeper sedimentary, lower crustal and upper-mantle layers, Buness et al. (1992) concluded that there was enough information to determine that the depth to the mantle and sediment thickness decreased under northern Tunisia. Additionally, Della Vedova et al. (1995) analysed heat flow density (HFD) values from Tunisia across the Mediterranean Sea through Sardinia. They found high HFD values (100–140 mW m⁻²) in northern Tunisia and the adjacent Sardinia Channel and attributed this to thinned crust caused by recent crustal thinning. While local high heat flow anomalies can be found over deep sedimentary basins, widespread high HFD values are more commonly associated with tectonic boundaries or continental rifting (Morgan & Gosnold 1989). However, the uncertainty in the various data does not rule possible the model shown in Fig. 10 or models similar to it.

An alternative model to the above model is a low-density crustal–mantle body, which is the source of the regional decrease towards the north of the residual gravity anomaly. Fig. 11 shows one possible model of low-density material (possibly thin continental crust) being subducted. While traditional plate tectonic theory would argue against the subduction of continental crust–lithosphere material, recent studies (Mattauer 1995; Michaud & Chopin 1995; Chemenda et al. 1996, 1997) have suggested that subduction of continental material does indeed occur and is important in the development of many mountain ranges and tectonic environments, including the Urals, the Himalayas and the evolution of the Adriatic microplate in relation to the Alps (Blundell et al. 1992). Physical modelling studies by Chemenda et al. (1996, 1997, 2000) indicate that continental material can be subducted to at least 100 km and possibly 250 km before buoyancy forces cause the possible exhumation of the subducted material. While gravity data cannot determine the exact physical properties and geometry of such deep bodies, the model in Fig. 11 shows that continental material may have been subducted in North Africa and such a model is consistent with the available data.

A compressional tectonic environment has been present in northern Tunisia since at least the Middle or Upper Miocene (Frizon de Lamotte et al. 2000) and continues to the present (H'Faiedh et al. 1985; Dlala 1992; Gueddiche et al. 1998). While the present-day northwest convergence is slow (<1 cm yr⁻¹) and there are no present-day deep earthquakes in northern Tunisia (Udias 1982; Gueddiche et al. 1998) these facts do not rule out the possibility of subducted crustal material. There are a several lines of evidence supporting the development of subduction along the North African Plate. Plate tectonic reconstructions of the entire Mediterranean region (Gealey 1988; Dewey et al. 1989; Doglioni 1999; Frizon de Lamotte et al. 2000; Vergés & Sàbat 1999) indicate that there was north-dipping subduction of (probably) oceanic lithosphere until at least 15 Ma and possibly as recently as 8–13 Ma (Vergés & Sàbat 1999). However, while these studies imply that subduction stopped along northern Tunisia, the current nature of the plate boundary was not described except that it is compressional.

The evidence for subduction in the above plate tectonic reconstructions is partially based on the distribution and type of igneous activity occurring in North Africa. Petrological analyses (Girod & Girod 1977; Wilson & Giraud 1998) of plutonic and volcanic lithologies in North Africa suggest that calc-alkaline magmatic activity began in the Early Miocene and continued to the Quaternary. Most workers (see Wilson & Giraud 1998) considered that these lithologies were formed by subduction processes either directly or post-collisional. In northern Tunisia, Miocene volcanic activity (8–14 Ma) occurred in three main regions (La Galite, Nefza and Mogod). Trace element analyses by Laridhi Ouazaa (1994) indicated that Th/Ta ratios range from 8 to 10 and are comparable to other plate convergent regions with subduction zones.

The most important evidence for recent subduction has been regional seismic studies (Mueller 1989; Spakman 1990; DeJong et al. 1994) that used P-wave delay-time tomography to analyse the crustal and mantle structure of the Mediterranean region. Clearly evident from these studies is a north-dipping zone of higher-velocity material that may represent the remnant of the Miocene subduction along North Africa. The dimensions and the amplitude of velocity anomalies beneath northern Tunisia are smaller than those seen under the Rif-Betic Mountains to the east in Morocco (Mueller 1989; DeJong et al. 1994; Seber et al. 1996). Seber et al. (1996) interpreted these velocity anomalies along with gravity data and earthquake locations in Morocco as being caused by active delamination of continental lithosphere into the upper mantle. This may also be a cause of the velocity anomalies seen in the upper mantle of northern Tunisia.

A closer examination of the regional seismic models (Mueller 1989; DeJong et al. 1994) shows a shallow (<150 km) north-dipping low-velocity zone along northern Tunisia. While the evidence is again stronger for a low-velocity zone under the Rif-Betic Mountains, it is still present under Tunisia. Seber et al. (1996) interpreted this low-velocity zone as being caused by low-density asthenospheric material replacing the delaminating lithosphere. However, Vergés & Sàbat (1999) using estimates of crustal shortening within northeastern Algeria (the Kabylies, Tell Atlas and northern Atlas units) suggested that the lower crust and thin units of the Kabylies and Tell could have also been subducted. Using the thrust mechanic models of Platt (1986), geometrical relationships between the Tell Atlas and the northern Atlas and their structural styles, Boccaletti et al. (1990) suggest that the Tell Atlas sector of Tunisia might be interpreted as a thrust wedge accreted by underplating of the North African continental crust. This is analogous to processes occurring in the Calabrian Arc (Cello et al. 1981) and is confirmed by recent seismic tomographic models (Di Stefano et al. 1999) that show a low-velocity subducted crust on top of the Adriatic lithosphere. Additionally, to explain the distribution of deformation within North Africa, Chihi (1995) and Chihi & Philip (1999) proposed a geodynamic evolution for northern Tunisia and Algeria based on a continental subduction setting from the Lower to Middle Miocene. These authors consider that this subduction has been locked in place since the Late Miocene, however, the whole domain remained controlled by a compressive regime until the Quaternary owing to the continuous northward motion of the African Plate.

The final model of the crustal and upper-mantle structure of northern Tunisia is shown in Fig. 12. Additionally, a geological interpretation of this model is shown in Fig. 13. Complete Bouguer gravity anomalies were used to construct this model as the regional and residual gravity models were combined. The geometry and thickness of the Cenozoic, Mesozoic and Paleozoic units were taken from seismic refraction (Buness et al. 1992) and previous gravity models (Jallouli & Mickus 2000). This model represents the subduction of crustal material either currently being subducted or recently subducted. However, gravity data cannot distinguish between these two possibilities.

Figure 12.

Bouguer gravity anomaly model with a subduction of a continental slab beneath northern Tunisia. The numbers represent the density of the body in g cm⁻³.

Figure 13.

Geological interpretation of the deep structure of the Bouguer gravity model shown in Fig. 12. The stippled pattern represents the Paleozoic and younger sediments. The grey-scale pattern represents basement crustal rocks.

The model in Figs 12 and 13 fits within current Mediterranean Plate tectonic models (Carminati et al. 1998; Frizon de Lamotte et al. 2000; Doglioni et al. 1999; Zeck 1999; Vergés & Sàbat 1999) that have a recent (either 8 or 15 Ma) north-dipping subduction zone off North Africa with continued compression between Europe and Africa. Current tectonic models have a northwest to west subducting African slab that started in the Alboran Sea 25–30 Ma and progessively rolled back to the east to its current position in the Apennine and Calabrian arc. The models differ in the specific mechanisms of how the subduction zone rolled back as Carminati et al. (1998), Zeck (1999), Frizon de Lamotte et al. (2000) and Vergés & Sàbat (1999) believe that the subduction zone detached or jump-backed in either two or four steps to its present position, whereas Doglioni et al. (1999) states that the subduction continuously rolled back. The detached subduction zone models all show that subduction did not occur off northern Tunisia. However, our gravity analysis indicates that subduction may have occurred. This fact would enforce the model of Doglioni (1999), however, it would not rule out the detached models, only modify them with these new results.

The EAS map (Fig. 8) shows an anomaly (EWTA) that does not correspond to any known surface feature. Forward modelling of the Bouguer gravity anomalies shows an uplift of the Paleozoic sediments and the basement in the same location as the EAS anomaly. While this solution is non-unique and the uplift could also be included in the younger sediments, it does show that some type of density anomaly is needed to explain the observed gravity maximum. A depth analysis of the EAS anomalies indicated that the source of the anomaly is between 3 and 7 km, which would be located in some type of Mesozoic sediments. The cause of this anomaly is unknown but it could be caused by uplift in the upper plate directly over the maximum curvature of a retreating (or rolling back) subduction zone (Royden 1993). In this model, extension occurs northwest of the uplift region and is represented by the Algerian Basin which has high HFD flow values indicative of a backarc basin with thinned crust (Della Vedova et al. 1995). Another explanation is the exhumation of the continental crustal material (Chemenda et al. 1997) that can occur when continental material has been subducted. However, only scattered outcrops of metamorphic rocks within Triassic and Lower Cretaceous layers have been noted in northern Tunisia (Bolze 1954; Rouvier 1977; Alouani 1991; Soussi 2000). If this scenario is correct, the EWTA, which corresponds to the short-wavelength positive residual gravity anomalies (Figs 8 and 9), may be caused by low-temperature metamorphic rocks that were emplaced somewhere in the upper crust.

Conclusions

Edge-enhancement techniques and 2-D forward modelling in conjunction with regional seismic experiments and geological studies have been used to analyse the Bouguer gravity field of northern Tunisia in order to constrain the structure of the northern margin of Tunisia. EAS analysis of residual gravity anomalies indicated a previously unknown east–west-trending feature with source depths between 3 and 7 km. Forward modelling of residual and Bouguer gravity anomalies implies that a residual gravity anomaly decreasing toward the north in northern Tunisia is caused by either thickening of sediments along a strike-slip fault or crustal material that was subducted beneath northern Tunisia. The subduction model is favoured as seismic tomographic images of the Mediterranean region indicate that subduction probably has occurred off along the African continent. Additionally, geological studies indicate that the Tell Atlas may be a thrust wedge accreted by underplating of the African continental crust as seismic refraction models show thinning of sediments in northern Tunisia. The east–west-trending EAS anomaly corresponds to an important feature in northern Tunisia. It could be associated with the location of a strike-slip/transform or subduction beneath the African Plate, however, gravity measurements alone cannot make this determination.

Acknowledgments

We would like to thank the Enterprise Tunisienne des Activités Pétrolières for releasing the gravity data that were used in this paper. We would also like to thank Allan Cogbill for assisting us with the terrain corrections. We are grateful to M. Diament who allowed one of us (CJ) to use his laboratory to initiate the project. The reviews of D. Seber and G.R. Keller greatly improved the manuscript, however, all conclusions remain our own.