Published as: Calvert, A., Sandvol, E., Seber, D., Barazangi, M., Roecker, S., Mourabit, T., Vidal, F., Alguacil, G., and Jabour, N., 2000. Geodynamic evolution of the lithosphere and upper mantle beneath the Alboran region of the western Mediterranean: Constraints from travel time tomography, J. Geophys. Res., 105, 10871-10898.
Geodynamic evolution of the lithosphere and upper mantle beneath the Alboran region of the western Mediterranean: Constraints from travel time tomography Alexander Calvert, Eric Sandvol, Dogan Seber, and Muawia Barazangi Institute for the Study of the Continents and Department of Geological Sciences, Cornell University, Ithaca, New York
Steven Roecker Department of Earth and Environmental Sciences, Rensselaer Polytechnic Institute, Troy, New York
Taoufik Mourabit Tetuoan University, Tetuoan, Morocco
Francisco Vidal Instituto Geografico Nacional, Madrid, Spain and Instituto Andaluz de Geofisica, Granada, Spain
Gerardo Alguacil Instituto Andaluz de Geofisica, Granada, Spain
Nacer Jabour Centre National de Coordination et de Planification de la Recherche Scientifique et Technique, Rabat, Morocco
2
Abstract. A number of different geodynamic models have been proposed to explain the extension that occurred during the Miocene in the Alboran Sea region of the western Mediterranean despite the continued convergence and shortening of northern Africa and southern Iberia. In an effort to provide additional geophysical constraints on these models, we performed a local, regional, and teleseismic tomographic travel time inversion for the lithospheric and upper mantle velocity structure and earthquake locations beneath the Alboran region in an area of 800 x 800 km2. We picked P and S arrival times from digital and analog seismograms recorded by 96 seismic stations in Morocco and Spain between 1989 and 1996 and combined them with arrivals carefully selected from local and global catalogs (1964-1998) to generate a starting data set containing over 100,000 arrival times. Our results indicate that a N-S line of intermediate-depth earthquakes extending from crustal depths significantly inland from the southern Iberian coast to depths of over 100 km beneath the center of the Alboran Sea coincides with a W to E transition from high to low velocities imaged in the uppermost mantle. A high velocity body, striking approximately NE-SW, is imaged to dip southeastwards from lithospheric depths beneath the low velocity region to depths of ~350 km. Between 350 and 500 km the imaged velocity anomalies become more diffuse. However, pronounced high velocity anomalies are again imaged at 600 km near an isolated cluster of deep earthquakes. In addition to standard tomographic methods of error assessment, the effects of systematic and random errors were assessed using block shifting and bootstrap resampling techniques, respectively. We interpret the upper mantle high velocity anomalies as regions of colder mantle that originate from lithospheric depths. These observations, when combined with results from other studies,
3 suggest that delamination of a continental lithosphere played an important role in the Neogene and Quaternary evolution of the region.
1. Introduction The processes responsible for recycling of lithospheric material back into the mantle are usually described in terms of subduction. The contribution of other processes such as lithospheric delamination (sensu strictu, s.s.) [Bird, 1979] or convective removal [Houseman et al., 1981] continues to be controversial.
Both delamination and
subduction models have been proposed to explain the formation of many of the Neogene extensional basins located along the Alpine convergent zone between Africa and Eurasia. This study focuses on the Alboran Sea and the surrounding regions, located at the westernmost limit of this zone (Figure 1a). A variety of models have been proposed to explain the synchronous extension and contraction that occurred in the Alboran region and its surroundings during the Early Miocene despite the continued convergence of Africa and Iberia. Presently, delamination (s.s.) [e.g., Seber et al., 1996a; Mezcua and Rueda, 1997; Platt et al., 1998] or convective removal models [e.g., Platt and Vissers, 1989; Vissers et al., 1995] are gaining popularity but subduction models [e.g., Lonergan and White, 1997] are also still being proposed. Previous tomography studies that have imaged mantle structure beneath the western Mediterranean [Blanco and Spakman, 1993; Plomerová et al., 1993; Spakman et al., 1993; Seber et al., 1996b; Piromallo and Morelli, 1997; Bijwaard et al., 1999] have found a pronounced high velocity anomaly located beneath the Alboran Sea and Southern
4 Spain. However, the possible depth and lateral extent of the anomaly were not well defined with results interpreted both in terms of a subduction [e.g., Blanco and Spakman, 1993] and delamination models [e.g., Seber et al., 1996b]. Blanco and Spakman [1993], used both local and teleseismic events extracted from the International Seismological Centre’s (ISC) catalog up to 1986. Seber et al. [1996b] used more recent data (19891994) from the new Moroccan national network but did not use any local events or readings from Spanish stations. In this tomographic study, we take advantage of the expansion of seismic networks in Spain and Morocco and the large volume of new data that have been collected since these earlier studies and use all available data (1964-1998) to investigate and accurately map the velocity structure of the region. The resulting velocity model is carefully evaluated using not only the standard synthetic tests and examination of the resolution matrix but also bootstrap resampling of the travel time data and shifting of model block boundaries to eliminate the dependence of the derived model on a particular parameterization. The implications of our results for previously proposed models are discussed and synthesized with observations from other recent studies into a model for the Neogene evolution of the Alboran Sea region.
2. Tectonic Setting The entire western Mediterranean region has undergone significant extension despite continuous convergence of Africa and Eurasia since the Cretaceous [Dewey et al., 1989; Srivastava et al., 1990]. The most striking evidence for this extension is the Neogene age of the oceanic crust beneath the majority of the western Mediterranean
5 (Figure 1a). The original Mesozoic oceanic crust of the Neo-Tethys that lay between Africa and Europe was almost completely recycled into the mantle during the Late Oligocene and Miocene as a subduction zone rolled back to the south towards Africa and southeast towards Calabria (See Figures 1b and 1c) [Rehault et al., 1985; Bouillin et al., 1986; Dewey et al., 1989; Yilmaz et al., 1996; Carminati et al., 1998a; Carminati et al., 1998b]. A narrow NE-SW trending mountain belt generated by Cretaceous-Paleogene convergence that is believed to have existed along the present day coasts of SE France and E Spain [Alvarez et al., 1974] collapsed behind the retreating subduction zone and was dispersed across the western Mediterranean. Fragments of this mountain belt now form the Internal zones of the Betics, Rif, Tell, and Calabria [Guerrera et al., 1993]. Situated at the western edge of the extended region, the Alboran region suffered the least extension. It has also experienced the least amount of convergence relative to the rest of the Alpine belt (~200 km since the Cretaceous [Dewey et al., 1989; Srivastava et al., 1990]). Convergence was initially NNE-SSW to N-S before shifting to NW-SE in the Tortonian (~10-7 Ma) [Dewey et al., 1989; Mazzoli and Helman, 1994]. The plate boundary between Africa and Iberia is diffuse with Neogene continental deformation distributed over a broad zone more than 500 km wide stretching from the High Atlas in Morocco to the Betic Cordillera in Spain (Figure 2). Marine studies in the Gulf of Cadiz and off the Atlantic coast of Morocco demonstrate that the diffuse character of this deformation extends offshore with no clear plate boundary visible to the west until the Gloria Fault (Not on map, ~20°W) [Hayward, 1997]. The Alboran Sea is bounded by the arcuate Betic and Rif mountain belts (See Figures 1a and 2). These mountains are typically subdivided into Internal and External
6 zones [e.g., Choubert and Faure-Muret, 1974] (Figure 2) owing to a pronounced difference in timing, style of deformation, and lithology. A region of highly deformed flysch separates these zones in the western Betics and Rif. The Internal zones are fragments of a collisional mountain belt composed of Mesozoic marine sediments shortened during Cretaceous and Paleogene convergence. The crustal thickness beneath this belt may have been in excess of 50 km in places [Platt and Vissers, 1989]. The flysch was deposited on the southern and western margins of the belt on thinned continental or possibly oceanic crust [Comas et al., 1999]. During formation of the belt, the External zones were relatively stable continental margin areas subject to marine deposition. During the Late Oligocene-Early Miocene (~25 Ma), the belt collapsed along with limited injection of basaltic dikes beneath the Betic (northern) portion of the belt (Figure 2) [Torres-Roldan et al., 1986]. The extension seems to have occurred initially along NNW dipping faults followed by more substantial extension along WSW dipping low angle faults [Martínez-Martínez and Azañón, 1997]. Deep crustal rocks were rapidly exhumed from depths of 40 km at a minimum of 4 km/Ma [Platt et al., 1998]. Peridotites emplaced at the base of the extending crust in the Early Miocene were rapidly exhumed beneath extensional faults to form the Spanish Ronda and Moroccan Beni Boussera peridotite massifs [Van der Wal and Vissers, 1993] (Figure 2).
Extension in this
convergent setting was accommodated by thrusting in the bounding flysch and External zones with a primarily WNW, W, and WSW directed transport direction in the Betics, Gibraltar, and Rif regions, respectively [Frizon de Lamotte et al., 1991; Kirker and Platt, 1998]. This coeval extension of the Internal zones and shortening in the External zones
7 was accompanied by distributed calc-alkaline volcanism predominantly in the Internal zones but also on the perimeter of the External zones (Figure 2). By Burdigalian times (~21-16 Ma) the western portions of the belt (now the West Alboran Basin (WAB) (Figure 2)) had subsided below sea level and were subject to marine deposition [Comas et al., 1992].
Active extension of the Alboran sea ceased in the Middle Miocene
(Serravallian/Tortonian, ~10 Ma) [Watts et al., 1993; Chalouan et al., 1997] but subsidence continued in the WAB and to a limited extent in the east Alboran basin (EAB) [Watts et al., 1993]. Extension ceased in the Betics in the Tortonian [Johnson et al., 1997] (perhaps as early as the Serravallian [Andriessen and Zeck, 1996]). Predominantly west directed transport continued in the Rif and Betics until the end of the Miocene [Frizon de Lamotte et al., 1991]. The total amount of westerly directed shortening is significant but not well determined with estimates ranging between 200 and 500 km [Platt and Vissers, 1989; Balanyá et al., 1997; Martínez-Martínez and Azañón, 1997]. In the Pliocene and Quaternary, gross plate motions reasserted control of the deformation in the region with primarily NNW-SSE shortening and strike-slip faulting [Comas et al., 1999]. Uplift of the margins of the extended region differentiated the Rif and Betic Internal zones from the Alboran Sea. Limited potassic volcanism (shoshonites and lamprophyres) [Hernandez and Bellon, 1985] occurred in the eastern External zones during the Pliocene and was superseded by alkaline basaltic magmatism in the Quaternary.
8
3. Tomography Study 3.1. Data The data set used for this study was selected from data recorded by multiple networks in Spain and Morocco (Figure 3). Seismic stations in the region (~96) are predominantly short-period vertical component with a limited number of 3-component and intermediate band stations.
Digital waveforms from approximately 200
local/regional and teleseismic events between 1989 and 1996 were provided by Instituto Geografico Nacional (IGN) in Madrid (Spain); Instituto Andaluz de Geofisica (IAG) in Granada (Spain) and Centre National de Coordination et de Planification de la Recherche Scientifique et Technique (CNCPRST) in Rabat (Morocco).
Supplementary analog
records were also provided by CNCPRST and Physique du Globe at Mohammed V University (MOH V) in Rabat. P and S arrivals were repicked for the local events and P arrivals from the teleseismic events located between 30 and 90 degrees from the recording stations. These picks were associated and merged with catalog data from the International Seismological Centre (ISC) (1964-1994), USGS National Earthquake Information Center (NEIC) Earthquake Data Report (EDR) (1995-August 1998), and Moroccan CNCPRST bulletins to generate comprehensive databases of local and teleseismic arrivals recorded in Morocco, Spain, Portugal, and Algeria from 1964-August 1998. Errors in the central clock for CNCPRST network required that readings from this network not be merged with other networks during certain periods. The combined local/regional data set contained 7100 events with 66,200 P and 48,600 S arrivals. The S arrivals for local/regional events were included to improve constraints on the earthquake
9 locations (particularly depth). The teleseismic data set comprised 15,700 events with 47,700 P arrivals. Teleseismic arrivals reported as emergent were not extracted from the ISC and USGS catalogs, as these are likely to contain significant uncertainties. The independent repicking of phases by the authors using the waveform data allowed some quantitative assessment of the reading uncertainty (σreading) and hence 2 determination of reasonable data weights (1/σ reading) for the larger catalog dataset.
Differences between the independent picks (assuming no systematic biases) made by the authors and those reported in bulletins were calculated and the standard deviation of this difference (SDD) was determined for the different reported qualitative picking precisions (e.g., Impulsive as opposed to Emergent) for each network. The SDD of measurements was calculated after removal of any arrival time differences (outliers) that exceeded a certain threshold value. A threshold of 3 seconds was found to significantly reduce the SDD while still allowing geologically reasonable signals in the data. The standard deviations calculated after a 3 second cutoff are summarized in Table 1.
Not
surprisingly, arrivals reported as impulsive in the ISC and USGS bulletins are of better quality than those reported as emergent especially for the Spanish networks where picks are made from digital records. Teleseismic residuals were also examined and found to be consistent with clear azimuthal variations for most stations in Spain. However, some stations in Morocco showed some scatter and were assigned an increased reading uncertainty. An additional theoretical uncertainty of 0.1 seconds was also added to account for uncertainties caused by inexact ray tracing and model discretization. Local and teleseismic events recorded by 6 or more stations in the Alboran region were extracted from the database resulting in a final selected data set of 4,500 local events with
10 55,800 P arrivals and 39,500 S arrivals (Figure 4a) and 1,600 teleseismic events with 16,400 arrivals (Figure 5). Seismic source regions that provide local data are scattered, indicating the distributed nature of active deformation in the region (Figure 4). However, the majority of well constrained crustal earthquakes are concentrated in the Betic and Rif mountain belts. Intermediate-depth earthquakes occur predominantly in two regions [Buforn et al., 1995]: (1) A diffuse zone off the coast of Spain in the Gulf of Cadiz, and (2) relatively concentrated N-S line spanning the Alboran Sea and dipping S from crustal depths beneath the Betics to a depth of ~150 km beneath the center of the Alboran Sea. This dipping band of seismicity is overlain by a relatively aseismic zone [Seber et al., 1996a]. A limited number of well-constrained intermediate-depth earthquakes have also been located off the Atlantic coast of northern Morocco [Seber et al., 1996a]. No wellconstrained earthquakes have been located between depths of 200 km and 550 km in the region. However, four significant and well-constrained events have been located at depths of 550-650 km beneath southern Spain [Grimson and Chen, 1986; Buforn et al., 1995]. One of these events (Not shown in Figure 4), that occurred in 1954 with Mw = 7.8 [Chung and Kanamori, 1976], is the largest instrumentally recorded earthquake in the region. The teleseismic events are located predominantly along the subduction zones of the Americas, the Atlantic ridge system, Zagros and the India-Asia collision zone (Figure 5). Azimuthal coverage is generally good, especially from the east and west, although there is limited illumination from the south.
11 3.2. Relocation of Local/Regional Events An initial 1-dimensional velocity model (Figure 6 and Table 2) was chosen using a combination of a reference model (PM2) determined for Europe by Spakman et al. [1993] and regional refraction surveys [Carbonell et al., 1998]. Each event was relocated using the same starting epicentral location and origin time but multiple starting hypocentral depths of 0.5, 14, 40, 90, 300, and 600 km. The depth was fixed for the early location iterations to lessen the influence of the starting epicentral location and origin time on the final hypocenter location. Residual cutoffs (used to downweight outliers) were progressively reduced from 60 seconds to 3 seconds. Despite the different starting depths, the solutions for a single event usually converged to a single hypocentral location and origin time. However, approximately 10% of events converged to two or sometimes even three locations suggesting the presence of multiple residual minima. In these cases, the location with the lowest weighted RMS residual was chosen. Events with poorly constrained locations were removed using criteria 1-8 shown in Table 3. To ensure that only events with relatively velocity insensitive locations were used for the inversion, the events were relocated in 10 different velocity models generated by adding a +/- 5% random perturbation to the reference 1-D model. Any event that moved by greater than 15 km horizontally, 20 km vertically or 25 km in total was discarded [e.g., Abers and Roecker, 1991]. Finally, to improve homogeneity of the ray coverage in the model and reduce the biasing of station corrections and 1-D inversions by arrivals from aftershock sequences, the local data set was geometrically thinned by preserving the event with the most recorded arrivals in each 5 km cube of the model [Abers, 1994]. A final local data set of 1100 events consisting of 16,000 P and 11,200 S arrivals was used for subsequent
12 1-D inversions. The three deep earthquakes with available data were not included in this data set owing to excessive coupling between the event depths and the velocity model (greater than 20 km variations).
3.3. Inversion Method The 1-D and 3-D inversions were conducted using the inversion code SPHYPIT90 developed by Roecker [1982] and further modified by numerous researchers [e.g., Roecker et al., 1987; Roecker et al., 1993; Abers, 1994; Jones et al., 1994]. The algorithm follows the nonlinear maximum likelihood formalism of Tarantola and Vallette [1982] to iterate towards the best-fit solution for the model parameters: pk+1 = pk+(GkTCdd-1Gk+Cpp-1)-1[GkTCdd-1Gk[d-g(pk)]-Cpp-1(pk-p0)]
(1)
where p0
- a priori or starting estimate of the model parameters
pk
- estimate of model parameters for the kth iteration
d
- data vector of observed travel times
g
- expected travel times (from ray tracing the model)
Gk
- matrix of partial derivatives (dg/dp)
Cdd
- diagonal a priori data covariance matrix of reading uncertainties
Cpp
- a priori model covariance matrix
The elements of Cdd are modified by two dynamic weights scaled between zero and one. Local/regional events are multiplied by a distance weight to account for a decrease in
13 picking accuracy at longer distances owing to a reduction in the amplitude and frequency content of phases with distance. Local and teleseismic readings with relative residuals greater than 3 seconds are also significantly down-weighted to reduce the effect of outliers. So the elements of Cdd are:
Cdd(i) = σ2reading/ (distance_weight * residual_weight)
(2)
Cpp is our estimated uncertainty in the slownesses assigned to the a priori model parameters p0 (1-D inversion model). In this case, Cpp is a diagonal matrix with no covariance assumed between model parameters. The elements of Cpp were chosen to correspond to a 0.2 km/s uncertainty in the a priori model.
Other values were
investigated, including allowing larger uncertainties in the crustal layers, however a constant uncertainty of 0.2 km/s was found to be optimal with a minimum of unconstrained assumptions. Travel times and partial derivatives are recalculated for each iteration by ray tracing through an average 1-D model calculated between each station-event pair and accumulating the travel time along this path in the 3-D velocity model. Although this approach is not true 3-D ray tracing, it was found to be a good approximation by Thurber and Ellsworth [1980]. Teleseismic arrivals are included using a modification of the ACH [Aki et al., 1976] method by ray tracing to the base of the model volume using the radial IASPEI91 model [Kennett and Engdahl, 1991] before tracing onto the station through the 3-D model. Parameter separation [Pavlis and Booker, 1980] allows the new slowness model and event relocations to be determined in separate steps. First, the slowness
14 inversion is conducted after the removal of the hypocenter effects then the local/regional events are relocated in the new velocity model using a separate location program. After each slowness-relocation iteration, events that fail criteria 1-10 shown in Table 3 are removed. The matrix inversions were performed using Singular Value Decomposition. Although we inverted for both P and S wave velocity structure in the lithospheric layers, we will only discuss the P inversion results below. As mentioned previously, the main reason for including local/regional S wave arrivals was to provide additional stability to the earthquake locations. 3.4. 1-D Inversion The local data set was used to invert for a best-fit 1-D model for the upper 150 km. Station corrections were calculated before each iteration step from the mean travel time residual at a station. The solution converged after three iterations. The best-fit 1-D model is shown in Table 2 and in conjunction with the station corrections results in a residual variance improvement of 30%. Mean station delays show a strong regional variation (Figure 7) suggesting that some crustal structure is being mapped into the station corrections. The most significant features are the consistently late arrivals to the stations in the Strait of Gibraltar region and the variation in the stations east of Malaga from early in the south to marginally late in the north. The substantial station delays in the Gibraltar region may be partly caused by the significant thickness (up to 10 km) of intensely deformed low velocity (2.3-5.2 km/s) sediments that were observed in the flysch regions by refraction experiments [Medialdea et al., 1986].
The significant
shortening and associated faulting that occurred in this region probably further contributes to the low velocities in this area. The high velocity signal in the station
15 corrections on the southern coast of Spain was also identified by refraction studies [Banda et al., 1993] and is discussed in more detail below. These station corrections were held constant for the 3-D inversions.
3.5. 3-D Inversion The 3-D model is parameterized by subdividing each layer into blocks using vertical interfaces.
After a number of trial inversions and examination of the ray
coverage, we chose the final model parameterization shown in Figure 8. The upper 4 layers were divided into 50 km blocks to take advantage of the dense ray coverage from the local/regional events.
The deeper layers that are controlled predominantly by
teleseismic data were divided into 100 km blocks. Figure 9 shows a hit map for a 50 km block parameterization throughout the model volume, to provide more detail of illumination in the deeper layers. Four iterations were performed with the last iteration providing an improvement in fit of less than 1%. Comparison of the diagonal elements of the partial derivative matrix Gk for the first and last iterations indicated that the inversion was fairly linear with elements rarely different by more than 10% and usually by less than 3%. The final 3-D velocity model produced an overall variance reduction of 32% relative to the 1-D model. The final variance was 0.35 sec2 whereas the a priori variance was 0.19 sec2 suggesting that the model was not over parameterized. As the location of block boundaries is arbitrary and ray illumination is not homogeneous, sharp velocity contrasts that do not lie close to a block boundary may be smoothed over a block or displaced (“leaked”) to a block boundary. Furthermore, any
16 artifacts that are introduced by a particular choice of block boundaries will probably not be detected by the standard tests used to assess the quality and stability of models. In an attempt to limit these problems, we used a block shift-and-average approach [Evans and Achauer, 1993]. 100 separate inversions were conducted using the same starting 1-D model but with the block boundaries uniformly shifted by (10n+5) km in the north-south direction and (10m+5) km in the east-west direction for n = -5 to 4 and m = -5 to 4. The mean RMS residual for all these different model parameterizations was 0.349 sec2 with a 2σ of 0.006 sec2 (minimum = 0.345 sec2 and maximum = 0.368 sec2 ) suggesting that no one model parameterization provide a better (in a statistically significant way) fit than any other. All 100 models were resampled onto the same grid with a 10 km horizontal node spacing. The mean velocity at each grid node was calculated resulting in a smooth model that was not a function of a particular parameterization with any significant anomalies in the final model caused by an anomaly being imaged by many different parameterizations.
A model derived using only 16 block shifts of 25 km showed
identical features (allowing for sampling differences) so this less computationally demanding shifting scheme was used for the synthetic tests described below. Some measure of the parameterization dependent stability of the velocity model was provided by the standard deviation of the velocities. Areas of high variance will not only occur in regions of poor stability but may also indicate the presence of large but spatially limited (on the scale of blocks) velocity anomalies. 3.6. Uncertainty Analysis The influence of systematic and random errors on the model is often hard to quantify owing to the non-linear nature of seismic tomography, the difficulty of
17 measuring the amount of noise in the input data, and the effects of parameterization. A number of traditional and non-traditional methods of uncertainty estimation were used to ensure that only robust well-constrained features were interpreted. These methods are reviewed below using an example cross section located 25 km north and parallel to cross section BB' (Figure 10). Figure 10a shows a cross section through the final shift-andaverage velocity model generated by stacking the results of 100 inversions using different model parameterizations. Figure 10b shows the velocity model resulting from a single inversion demonstrating that the first order features are the same as those found in the shift-and-average model. 3.6.1.
Bootstrap resampling of residuals. The increasing speed of modern
workstations now allows the use of non-linear uncertainty estimators to evaluate inversion results. Hearn and Ni [1994] used a bootstrap resampling technique [Tichelaar and Ruff, 1989] to determine uncertainties for 2-D Pn velocity models.
The same
approach was used to evaluate the 3-D tomography model determined in this study. The bootstrap method involves repeated inversions of resampled data sets and examining the variance in the derived models. To generate a resampled data set, arrivals times are randomly sampled from the data pool. However, after a data point is sampled it is returned to the pool so it may be picked multiple times. The same number of data are picked from the pool as existed in the original data set but each data set will be different owing to the aforementioned data redundancy and resulting omission of other data points. A velocity model is determined from these resampled data using the same inversion parameterization. This model is stored and the process repeated approximately 100 times to generate robust parameter distributions. The standard deviation of a particular model
18 parameter is a measure of the uncertainty in that value in the final model. An important strength of a bootstrap method is that, in contrast to Monte-Carlo simulations, it does not require an estimate of the noise in the data to assess the uncertainty in the resulting model. The resampling technique allows the noise in the data to contribute to the variance in the derived models.
However, it should be noted that in this case the
bootstrap approach does not completely remove the need for noise estimates to measure model uncertainty because they are required by the inversion method (the elements of Cdd).
A bootstrap conducted by resampling events rather than individual phases
produced virtually identical results. The bootstrap uncertainties are invariably higher than standard errors calculated using the a posteriori covariance matrix (compare Figures 10c and 10d). The standard errors are less than 1% beneath the Betics in the lithospheric layers and less than 1% beneath the Alboran, Betics, and Rif in the deeper layers. The bootstrap uncertainties determined for layer depths of 100-670 km, which are controlled predominantly by teleseismic data, were encouraging with important features imaged with anomalies greater than twice the bootstrap error (see Figures 10b and 10d). Although the crust and uppermost mantle layers showed significant variance (sometimes approaching 4%), these regions also contained the largest perturbations with robust anomalies again above twice the bootstrap uncertainty. A useful method for interpreting our model data was to subtract between one and two times the bootstrap uncertainty from the imaged anomalies setting an anomaly to zero if the anomaly changed sign after this subtraction. Only the more robust anomalies remained for interpretation. 3.6.2. Resolution matrix. The resolution matrix provides an estimate of the relative contributions of the observations and the a priori model to the final model. The
19 diagonal elements of the resolution matrix are very close to one beneath the Betics suggesting that the ray geometry and choice of model parameterization and damping parameters allows the imaged crust and mantle structure to be controlled by the observations rather than the a priori model [Jackson and Matsu'ura, 1985] (Figure 10e). Crustal structure beneath Morocco is not well controlled owing to the limited number of local earthquakes. Off-diagonal elements of the resolution matrix were also examined to evaluate smearing. The off-diagonal elements are usually an order of magnitude less than the diagonal elements suggesting that at least in theory there is very limited smearing of anomalies or generation of compensating anomalies in adjacent blocks. 3.6.3.
Spike test. Spike tests are traditionally used to assess resolution in
tomographic inversions. Synthetic data are generated by ray tracing a velocity model consisting of a regular 3 dimensional pattern of perturbations to the reference model (e.g., Figure 10f). A representative amount of noise is often added and the synthetic travel times inverted following the same procedure that was used for the real data.
The
recovered velocity model is compared to the velocity model used to generate the synthetics. Regions where the recovered model closely matches the input model are considered well resolved in the real velocity model whereas a poorly recovered region is considered to be suspect in the final model. The addition of noise usually degrades the recovered data but in this case it has a limited effect (compare Figures 10g and h). The degree of recovery is also sensitive to the geometry of the synthetic test [Lévêque et al., 1993] and in particular its relationship to the model parameterization used to conduct the inversion. Figure 10h would seem to indicate that even with a representative level of noise added to the data, very good model recovery is possible.
20 However, if the model parameterization is shifted laterally 25 km with respect to the synthetic model, the amount of recovery is significantly reduced. A shift-and-average inversion demonstrates that contrary to what would be inferred from examination of the resolution matrix or from conducting a synthetic spike inversion with velocity perturbations matching block boundaries (Figure 10h), discrete blocks of 100*100*50 km are not resolvable at depths of 100-400 km (Figure 10i). A synthetic test conducted using a single +5% spike of 150x150km in the 200-260 layer showed that this was not an artifact caused by the choice of spike distribution. Synthetic anomalies spanning two layers (Figures 10i & j) are imaged suggesting that depth rather than lateral resolution is the limiting factor. The recovery in the deepest layers of even the thin anomalies is surprising. It probably results from rays becoming more horizontal with depth and hence traveling a greater horizontal distance before crossing a layer boundary. However, it should be noted that these spike inversions might be giving an optimistic assessment of resolution in the deepest layers. At these depths, the radius of the first Fresnel zone for a 1 Hz wave is ~75 km but the effects of signal frequency are not included in the synthetic travel times generated by ray tracing. Certainly, there is little to be gained from using blocks much smaller than 100 km at these depths.
3.7. Interpretation and Evaluation of Velocity Model Figures 11a-m and Figures 12a-d show horizontal and vertical slices through the block shift-and-average model. The ray segments close to the cross sections are also shown to provide an indication of the control of the travel time observations on the different portions of the model (please see our website at: atlas.geo.cornell.edu for depth
21 slices of these ray segments).
Regions that contain many crossing rays are better
controlled than regions that contain a few sub-parallel rays. The most robust feature imaged in the upper crustal layer (Layer 1) is the pronounced low velocity region seen in vicinity of the Strait of Gibraltar (Figure 11a) both in Spain and Morocco. The low velocities are obtained despite the inclusion of significant station corrections for stations in this region (see Figure 7).
Inversions
conducted without station corrections yielded low velocity perturbations of ~18% indicating that 8% is probably a minimum value and that a significant portion of the anomaly may be being absorbed by the station corrections. These low velocity anomalies are significantly larger than the bootstrap standard deviation of ~3%. Low velocities in this region were imaged by other tomography studies [Blanco and Spakman, 1993; Gurria et al., 1997; Piromallo and Morelli, 1997] and refraction studies [Medialdea et al., 1986] and, as mentioned earlier, may be caused by a combination of low velocity sediments and significant crustal deformation in the flysch regions. A high velocity anomaly is imaged in the eastern Rif to the SE of the Nekor fault (NF, Figure 2) in both crustal layers. This anomaly coincides with an aeromagnetic anomaly subparallel to the Nekor fault that was interpreted by Michard et al. [1992] as indicating the presence of peridotite. The most significant feature in the lower crustal layer (Layer 2) is the high velocity anomaly imaged off the Alboran coast of Spain in the vicinity of Malaga (Figure 11b). High velocities were also identified in this region by a refraction experiment [Banda et al., 1993]. These authors reported that ray trace modeling was unable to discriminate between a high velocity peridotite body and a thin crust as the source of this
22 anomaly. The absence of stations in the Alboran Sea results in very poor control of its shallow velocity structure with only limited rays passing through this layer beneath the Alboran Sea (See Figure 9). Gravity modeling and limited refraction profiling [Working group for deep seismic sounding in the Alboran sea, 1978; Ben Sari, 1987] provide evidence of thinned (~20 km) crust but the underlying mantle velocities reported were highly variable ranging from 7.5-8.4 km/s on intersecting profiles and, therefore, may be suspect. Low velocities are also imaged in the Rif and Betics. Layer 3 from 30-45 km contains striking anomalies, consisting of a ‘horseshoe’ of anomalously low velocities (relative to an already low mantle velocity of 7.9 km/s) beneath the Betics, west Alboran Basin (WAB) and Rif surrounding a region of high velocity beneath the Alboran Sea (Figure 11c). The low velocity region may result from a mixture of two phenomena: (1) A crustal signature resulting from crustal thicknesses of up to 38 km in places beneath the Betics, and (2) The presence of anomalously low velocity mantle [e.g., Banda et al., 1993]. The high velocity region in the center of the Alboran Sea is surprising and difficult to reconcile with our expectations of what to find beneath a region that underwent extension during the Miocene. One possible geological explanation for this anomaly is that a significant thickness of thermal lithosphere has been recovered since the termination of extension in the Middle Miocene. Although, Polyak et al. [1996], based on heatflow measurements, interpreted a lithospheric thickness in this region of only ~40 km. The central Alboran Sea region contains few crustal earthquakes and effectively no stations. Another possible explanation is that it is an artifact caused by assumption inherent in our approach. For example, we assume that a 1-D velocity model is a good approximation for the first ray tracing and inversion
23 iteration.
The 1-D model places the Moho at depth of 30 km.
Rays connecting
earthquakes on one coast of the Alboran with stations on the opposite coast, in reality, are probably traveling at depths of ~15-20 km while they are traced through the model at depths of 30 km. This extra segment of ray path before refracting along the Moho results in the theoretical arrival time being delayed relative to the actual arrival times. However, as the rays are not passing through the 15-30 km layer above, the velocities are increased in the 30-45 km layer in the least constrained region in the center of the Alboran Sea. If this is true, then one must also consider the possibility that the surrounding lows might be enhanced by the inversion attempting to compensate for the effect of these elevated velocities on longer ray paths. Another possibility is that significant anisotropy may exist in the mantle and that inhomogeneous ray coverage might be causing this anisotropy to be mapped into the isotropic velocity model. To check these hypotheses we conducted three tests: (1) Synthetics were generated using a mantle velocity of 7.5 km/s in the 15-30 km layer beneath the Alboran. After adding noise, these data were inverted using the same starting 1-D velocity model. A significant portion of the anomaly in the 15-30 km layer was mapped into the 30-45 layer below, confirming that thinned Alboran crust might be a viable explanation for this anomaly. However, although slight compensating lows were generated beneath the Betics and Rif, they were minor compared to the very pronounced low velocity anomalies imaged in the actual model. (2) As a further check on the independence of the Betic/Rif lows from this artifact, inversions were conducted after Pn rays passing beneath the Alboran were removed. The African and Iberian phase data were split into separate data sets and all local rays that
24 intersected an E-W line passing through the Strait of Gibraltar removed. These data sets were inverted separately using the same model parameterization.
Although these
inversions used no ray paths traveling beneath the Alboran Sea significant low velocity anomalies were still imaged beneath the Betics and Rif. (3) To investigate these anomalies further, as well as the possible influence of anisotropy, the Pn arrivals extracted for various station-event separations were inverted for Pn velocity and anisotropy using the code of Hearn and Ni [1994].
Inversions were
conducted using varying amounts of Laplacian damping on the isotropic and anisotropic velocity components to evaluate any trade-offs. Even when anisotropy was very weakly damped; the low velocity beneath the Betics was imaged for all distance ranges. The Rif anomaly was also imaged but found to be less robust than the Betic anomaly perhaps owing to the sparser ray coverage. Zero or positive velocity perturbations were imaged beneath the Alboran Sea for all distance ranges and damping parameters. The use of shorter ray paths did accentuate the high velocities beneath the Alboran Sea lending support to the thin crust hypothesis; however, shorter paths might also be less affected by a thin mantle lid. Even with strong Laplacian smoothing it was not possible to extend the imaged low velocities beneath the Alboran Sea, suggesting that uppermost mantle velocities beneath the central Alboran may be higher than 7.9 km/s. The 45 - 60 km layer (Layer 4) again shows low velocity anomalies beneath the eastern Betics and Rif (Figure 11d), but these lows are now separated by a region of slightly elevated velocity beneath the western Betics and Rif. The low velocity beneath the Betics conforms well to the surface outcrop of the Internal zone and its western boundary is marked by intermediate-depth seismicity.
Higher velocities are again
25 imaged beneath the Alboran but these anomalies are relatively small and within one bootstrap error and so should not be considered robust. The 60-100 km layer (Layer 5) is arguably the most important layer for evaluating any geodynamic models, but also one of the most difficult to constrain as it marks the depth where ray coverage from local/regional data is limited to only those from intermediate-depth earthquakes and teleseismic rays are still limited to the regions close to the stations (Figure 11e). The most robust feature is the low velocity zone along the southern coast of Spain that extends eastwards from the N-S line of intermediate-depth seismicity. Despite the appearance that this anomaly extends beneath the Alboran Sea to the African coast, this feature is predominantly controlled by teleseismic rays to the Spanish coastal stations that do not sample the central Alboran. Beneath the western Betics and Rif a striking N-S high velocity anomaly is imaged to the west of the intermediate-depth seismicity.
The bootstrap uncertainties indicate that the portion
beneath the flysch unit of the Betics is probably robust. In layer 6 (100-150 km), a robust high velocity anomaly is imaged beneath the WAB and west and central Betics (Figure 11f). The SE extent of this anomaly is not well defined owing to the absence of good ray coverage (Figure 9). Assuming that positive velocity anomalies may be taken to indicate a region of cooler mantle, rather than a compositional difference, the most likely interpretation is that it indicates the presence of mantle lithosphere in view of the coherent continuation of the anomaly to depth. The high velocity anomaly beneath central Iberia was not found to be stable using the bootstrap error test.
26 Layers 7, 8, and 9 (150-200-260-330 km) show a continuous progression of the high velocity anomaly to the southeast beneath the Alboran Sea (Figures 12g,h, and i). However, it should be noted that if the entire Alboran Sea is underlain by a high velocity region at these depths, this progression might in part be a reflection of the progressive improvement of ray coverage beneath the Alboran Sea with depth. The anomalies become relatively small and diffuse in layers 10 and 11 (330-410490 km) (Figure 11j and k). However, significant concentrated anomalies are imaged in layers 12 and 13 (490-570-650 km). Any features imaged towards the base of the model must be interpreted with great care, as these regions are the most likely to contain artifacts caused by deep mantle structure outside the model volume.
One of the
important assumptions of an ACH-type approach is that any deviations from a spherically symmetric earth outside the model volume will be sampled equally by all the teleseismic rays from a single event. Masson and Trampert [1997] showed that this assumption is sometimes not valid for a true heterogeneous earth. Examination of a recent global tomography model [Bijwaard et al., 1999] suggests that some heterogeneity may exist directly beneath the model volume but the size of imaged anomalies is limited (variations of less than 1% from mean). Cross sections AA', BB', CC', and DD' (Figure 12) provide a good summary of the major features seen in the velocity model and its relationship to seismicity. A low velocity anomaly is imaged in the uppermost mantle along the SE coast of Spain. Its is bounded on its western edge by the intermediate depth seismicity. A high velocity body appears to dip south and east from near lithospheric depths beneath Spain and the Strait of Gibraltar under the low velocity region. This high velocity anomaly is imaged to be
27 continuous to depths of 350-400 km where it reduces in amplitude and becomes more diffuse. A pronounced high velocity anomaly is again imaged at depths of 500-650 km near the base of the model in the vicinity of the deep earthquakes. 3.8. Synthetic Smearing Test and Comparison to Previous Velocity Models The depth extent and volume of the high velocity body in the upper mantle are important constraints for any geodynamic model for the region. Despite no explicit regularization between blocks, the smoothing inherent in least squares and the shift-andaverage method will result in smearing of the imaged anomalies. To examine the effect of this smearing, synthetic travel times generated using a representative velocity model of the region (Figure 13) were inverted. Gaps were placed in key regions of the model (Layers 5, 10, & 11 – Figures 13e,j & k) where smearing may have been occurring. Only limited smearing occurred into the 60-100 km layer (Figure 13e). A small positive anomaly was mapped into the region just to the west of the intermediate-depth seismicity, in the vicinity of the imaged high velocity in the real model, suggesting that smearing may have played some role in generating this anomaly.
However, low
velocities are not smeared into this layer from above suggesting that the low velocities beneath the southern coast of Spain in this layer are real. The existence of anomalous low velocity mantle beneath the coastal Betics is further supported by the low Q values found for this region by Canas et al. [1991]. Smearing clearly occurs into layers 10 and 11, indicating that the inversion is probably biased towards filling in any gap if it exists. Most previous studies have reported a relatively continuous high velocity body from depths of 200 to 650 km [Blanco and Spakman, 1993; Spakman et al., 1993; Piromallo and Morelli, 1997; Bijwaard et al., 1999].
Examination of Blanco and
28 Spakman's depth slices shows that similar to our model, a high velocity is also imaged beneath the western Betics, Gibraltar, and Rif in their 70-120 km layer before moving east in deeper layers to join with the deeper anomaly at 200 km.
The studies by
Plomerová et al. [1993] and Seber et al. [1996b] image a high velocity body beneath the Alboran and Southern Spain but owing to bottom layers beginning at 200 and 150 km, respectively, they provide limited information on the depth extent of this body. Bijwaard et al. [1999] show a section cut from their global model (BSE) in the Alboran region that also shows a pronounced reduction in the size of the anomaly beneath 400 km (although this reduction in amplitude in the transition zone may be a global feature of the BSE model [Personal communication, Spakman, 1999]). All these inversions used similar data sets and so they possess the same potential for systematic smearing errors. Establishing continuity of this body from 350-500 km is probably beyond the resolution of studies using existing station distributions. However, existence of a single coherent body extending from < 200 km to 650 km is difficult to explain given the constraints on the relatively small amount of Tertiary convergence between Africa and Iberia (~200 km). The hypothesis that the deep earthquakes are occurring in lithospheric bodies of limited size is supported by the waveforms recorded in Spain. Buforn et al. [1997] suggest that one possible explanation for puzzling arrivals from these deep earthquakes may be caused by S to P conversions near the source and propose that they are occurring in a body of limited size.
29
4. Discussion 4.1. Previously Proposed Models Recent publications have converged to two competing but not necessarily mutually exclusive hypotheses, one proposing a progressive rolling or peeling back of lithosphere, the other proposing detachment of lithosphere. Within these broad categories, there is also debate about the involvement of crust in any recycling of lithosphere. Figure 14 shows a simple categorization and schematic cartoons of the most current models taken from their original publications.
4.1.1. Alboran block and retreating subduction. The predominant westerly transport direction in the Betics and Rif lead to the proposal of the Alboran block model [Andrieux et al., 1971]. The Internal zones were considered to have behaved as rigid block that escaped to the west during N-S convergence. Subsequent observations that the Internal zones underwent extension during westerly migration of the Alboran block necessitated modification of the model. Migration of a collapsing Alboran block were proposed to result from roll-back to the west of an east-dipping subducting slab, with the Alboran block extending in a back arc setting [Royden, 1993]. Royden suggested that arc migration may have proceeded past Gibraltar. No evidence of such subduction zone migration through the Gulf of Cadiz was found by recent marine surveys [Hayward, 1997]. More recent models propose that subduction stopped retreating in the Gibraltar region [Sengör, 1993; Lonergan and White, 1997].
30 4.1.2. Slab break-off. An alternative subduction related model was initially proposed by Blanco and Spakman [1993] based tomography results and further modified by Zeck [1996]. Zeck suggested that a NW dipping slab was subducted beneath SE Iberia during the eastward movement of Africa and Iberia relative to Eurasia during the Cretaceous and Paleogene. This subducted slab was proposed to have broken off during the earliest Miocene initiating uplift, extension, and peripheral thrusting, and now is oriented vertically beneath southern Spain. This hypothesis is attractive as it provides a mechanism for considerable lithospheric thickening and/or subduction of significant quantities of lithosphere beneath Spain, as the significant length of the imaged high velocity anomaly (~450-500 km) is hard to explain given the limited convergence between Africa and Iberia. However, it should be noted that most of the eastward motion of Africa and Iberia probably occurred before the Mid Cretaceous [Srivastava et al., 1990], requiring any subducted lithosphere to remain hanging in the mantle for 60 million years before finally detaching and descending into the mantle.
4.1.3. Convective removal. P-T paths determined for rocks in the Betics and Alboran as they were exhumed to the surface lead Platt and Vissers [1989] to propose convective removal of mantle lithosphere as a possible mechanism. They proposed that lithospheric thickening as a result of convergence had generated a collisional ridge with a substantial upper mantle root by Oligocene time. In their model, this gravitationally unstable lithospheric root was convectively removed in the Late Oligocene-Early Miocene and replaced by hot low-density asthenospheric material, resulting in uplift and extension of the overlying crust. Thermal modeling of additional P-T paths derived from
31 data from ODP drill holes in the Alboran Sea have recently lead Platt et al. [1998] to propose that lithospheric material was removed at approximately 60 km depth, very close to the base of the over-thickened crust. Platt et al. [1998] comment that removal of lithosphere at such a shallow depth is more in keeping with a delamination (s.s.) model, but may still be modeled by a convective removal process given specific mantle rheology and temperature boundary conditions.
4.1.4. Delamination. The apparent inability of a convective removal model to explain the predominant westward migration of deformation lead García-Dueñas et al. [1992] to hypothesize that a delamination model [Bird, 1979] might be applicable to the Alboran. Docherty and Banda [1995] and Tandon et al. [1998] proposed a SE migrating delamination model based on marine seismic reflection profiles and wells. Observations of the 3-D pattern of earthquake hypocenters, gravity modeling and regional wave propagation characteristics also lead Seber et al. [1996a] to propose a delamination model.
The N-S line of intermediate-depth seismicity apparently overlain by an
aseismic, attenuating, low velocity mantle and underlain by a region of high velocity mantle [Seber et al., 1996b] was considered evidence that Iberian and African lithosphere peeled back from the east to west [Seber et al., 1996a] perhaps initiated by detachment of a gravitationally unstable thickened lithospheric root beneath the Internal zones. Mezcua and Rueda [1997] and Morales et al. [1997] supported this hypothesis based on their earthquake relocations. Seber et al. [1996a] further suggested that the delamination process is still active today in the western Alboran.
32 4.2. Implications of Tomography Results The most striking feature of the velocity model we obtained is the significant positive anomaly that appears to extend from lithospheric depths beneath the Strait of Gibraltar and southern Spain to depths of about 350-400 km beneath the Alboran Sea. We interpret this anomaly as a lithospheric body that has descended into the upper mantle.
Although a low velocity anomaly is present above this body beneath the
southeastern coast of Spain, the body appears to be attached to Iberia to the west and possibly north (Figures 11 and 12). The continuation of intermediate depth seismicity from the top of this body into the Iberian crust may be a further indication of this attachment (Figure 4). Slab break off or convective removal models proposing complete detachment of a lone lithospheric body in the Early Miocene and no further mantle processes are probably not sufficient to explain such a geometry. The tomography results alone do not allow discrimination between retreating subduction and delamination processes, as both would have very similar velocity signatures.
However, these
processes would have produced a different thermal structure in the Alboran lithosphere during extension in the Late Oligocene and Early Miocene [Platt et al., 1998]. A retreating subduction model would predict stretching of the overriding plate in a back arc setting whereas delamination (or detachment) models would predict a sudden heating at the base of a thinned lithosphere associated with upwelling of asthenosphere to replace the sinking body.
Platt et al. [1998] present strong evidence that lithosphere was
removed from near the base of an over thickened crust and explicitly state that a retreating subduction mechanism cannot explain the P-T path followed by rocks
33 presently beneath the western Alboran Sea. The above process of elimination leaves a delamination based model as the simplest way to explain these observations. The above argument is probably an over simplification, as it does not allow more than one active process during the Neogene. The peridotites in the western Betics and Rif (Figure 2) show evidence of cooling in the mantle at 70 km depth before emplacement at the base of the crust [Van der Wal and Vissers, 1993]. This cooling signature is interpreted by Vissers et al. [1995] to result from the location of these peridotites in the hanging wall of a subduction zone some time before the Neogene, suggesting that subduction probably played a role in the formation of the collisional belt. The existence of significant high velocity anomalies at depths of 550-650 km, coincident with the few deep earthquakes, and perhaps detached from the shallower body raises the possibility that some lithosphere may have become detached completely during the Neogene. It is also possible that the pull of this hanging lithosphere may be allowing some subduction of Iberian continental crust as Africa and Iberia continue to converge [Morales et al., 1999].
4.3. Synthesized model Figure 15 shows our modified version of the working hypothesis model proposed by Platt and Vissers [1989] (compare with Figure 14). Lithospheric thickening occurred during Paleogene convergence and probably involved limited subduction. At the end of Oligocene and Early Miocene the gravitationally unstable upper mantle root partially detached beneath the thickened Alboran crust and peeled back to the west with a portion of the lithosphere perhaps detaching completely and descending into the upper mantle.
34 Crust above the inflowing asthenospheric material was heated, uplifted, and extended. Today, although this peeling back has significantly reduced in speed or perhaps terminated, a large lithospheric body remains beneath the Alboran Sea attached to Iberia beneath the western Betics and possibly the Rif. The intermediate depth earthquakes appear to be occurring near the top surface of this hanging body. Whereas the deep earthquakes may be located in detached pieces of lithosphere perhaps as they undergo phase transitions at the base of the upper mantle.
5. Conclusions A tomographic inversion of local, regional, and teleseismic phase arrivals recorded by Spanish and Moroccan regional networks images significant P velocity anomalies beneath the Betics, Rif, and Alboran Sea. Evaluation of the resulting velocity model using bootstrap resampling, block shifting, and synthetic tests suggest that a robust high velocity anomaly is located beneath southern Spain steeply dipping to the SE from depths of about 60 km to 400 km. Below this depth, the high velocity anomaly becomes diffuse.
A concentrated high velocity anomaly in the 570-650 km layer, although
probably poorly constrained, correlates well with the location of the few but significant deep earthquakes beneath southern Iberia. We interpret these high velocity bodies to represent cold lithospheric material recycled into the upper mantle. Previous tomographic studies have imaged a single high velocity body beneath the Betics and Alboran from about 200 – 650 km depth, but detailed synthetic and error tests that we performed suggest that the apparent continuity of this high velocity anomaly may be partially caused by smearing from shallower and deeper depths owing to poor station coverage within the Alboran Sea. The high velocity upper mantle body that we
35 mapped is overlain by a region of low velocity uppermost mantle beneath the Betics. This velocity model together with recent results from the latest Alboran ODP leg provide strong evidence that lithospheric delamination (s.s.) played an important role in the Neogene evolution of the Alboran region.
36
Acknowledgments. This work was supported by National Science Foundation Grant EAR-9627855. We would like to thank the members of Instituto Andaluz de Geofisica (Spain), Instituto Geografico Nacional (Spain), Centre National de Coordination et de Planification de la Recherche Scientifique et Technique (Morocco); and the Physique du Globe at Mohammed V University (Morocco) for their hospitality and for providing the waveform data used in this study; Claudia Piromallo for enlightening discussions about the different tomographic studies in the Alboran region; Harmen Bijwaard and Wim Spakman for providing the Alboran portion of their global model; Tony Watts for insights into the tectonics of the region and for providing preprints of upcoming papers; and Rob Van der Hilst for valuable discussion and review. The authors would also like to thank Andrea Morelli, Wim Spakman, and Rinus Wortel for their careful reviews and suggestions, and Graham Brew, Francisco Gomez, and José Morales for their reviews of early drafts. Institute for the Study of the Continents contribution number 255.
37
References Abers, G.A., Three-dimensional inversion of regional P and S arrival times in the east Aleutians and sources of subduction zone gravity highs, Journal of Geophysical Research, 99, 4395-4412, 1994. Abers, G.A., and S.W. Roecker, Deep structure of an arc-continent collision: earthquake relocation and inversion for upper mantle P and S wave velocities beneath Papua New Guinea, Journal of Geophysical Research, 96, 6379-6401, 1991. Aki, K., A. Christoffersson, and E.S. Husebye, Three-dimensional seismic structure of the lithosphere under Montana LASA, Bulletin of the Seismological Society of America, 66, 501-524, 1976. Alvarez, W., T. Cocozza, and F. Wezel, Fragmentation of the Alpine orogenic belt by microplate dispersal, Nature, 248, 309-314, 1974. Andriessen, P.A.M., and H.P. Zeck, Fission-track constraints on timing of Alpine nappe emplacement and rates of cooling and exhumation, Torrox area, Betic Cordilleras, S. Spain, Chemical Geology, 131, 199-206, 1996. Andrieux, J., J.-M. Fontbote, and M. Mattauer, Sur un modele explicatif de l'arc de Gibraltar, Earth and Planetary Sciences Letters, 12, 191-198, 1971. Balanyá, J.C., V. García-Dueñas, and J.M. Azañón, Alternating contractional and extensional events in the Alpujarride nappes of the Alboran Domain (Betics, Gibraltar Arc), Tectonics, 16, 226-238, 1997.
38 Banda, E., J. Gallart, V. García-Dueñas, J.J. Dañobeitia, and J. Makris, Lateral variation of the crust in the Iberian Peninsula: new evidence from the Betic Cordillera, Tectonophysics, 221, 53-66, 1993. Bellon, H., and R. Brousse, Le magmatisme périméditerranéan occidental. Essai de synthèse, Bulletin de la societe géologique de France, 19, 469-480, 1977. Ben Sari, D., Connaissance Geophysique du Maroc, 207 pp., Editions Marocaines et Internationales, Rabat, Morocco, 1987. Bijwaard, H., W. Spakman, and E.R. Engdahl, Closing the gap between regional and global travel time tomography, Journal of Geophysical Research, 103, 30,05530,078, 1999. Bird, P., Continental delamination and the Colorado Plateau, Journal of Geophysical Research, 84, 7591-7571, 1979. Blanco, M.J., and W. Spakman, The P-Wave velocity structure of the mantle below the Iberian Peninsula: evidence for a subducted lithosphere below southern Spain, Tectonophysics, 221, 13-34, 1993. Bouillin, J.P., M. Durand-Delga, and P. Olivier, Betic-Rifian and Tyrrhenian arcs : Distinctive features, genesis, and development stages, in The Origin of Arcs, edited by F.C. Wezel, pp. 281-304, Elsevier, New York, 1986. Buforn, E., P. Coca, A. Udías, and C. Lasa, Source mechanism of intermediate and deep earthquakes in southern Spain, Journal of Seismology, 1, 113-130, 1997. Buforn, E., C. Sanz de Galdeano, and A. Udías, Seismotectonics of the Ibero-Maghrebian region, Tectonophysics, 248, 247-261, 1995.
39 Canas, J.A., L. Pujades, J. Badal, G. Payo, F. de Miguel, G. Alguacil, J. Ibañez, and J. Morales, Lateral variation and frequency dependence of coda-Q in the southern part of Iberia, Geophysical Journal International, 107, 57-66, 1991. Carbonell, R., V. Sallarés, J. Pous, J.J. Dañobeitia, P. Queralt, J.J. Ledo, and V. GarcíaDuenas, A multidisciplinary study in the Betic chain (southern Iberia Peninsula), Tectonophysics, 288, 137-152, 1998. Carminati, E., M.J.R. Wortel, P.T. Meijer, and R. Sabadini, The two-stage opening of the western-central Mediterranean basins: a forward modeling test to a new evolutionary model, Earth and Planetary Science Letters, 160, 667-679, 1998a. Carminati, E., M.J.R. Wortel, W. Spakman, and R. Sabadini, The role of slab detachment processes in the opening of the western-central Mediterranean basins: some geological and geophysical evidence, Earth and Planetary Science Letters, 160, 651-665, 1998b. Chalouan, A., R. Saji, A. Michard, and A.W. Bally, Neogene tectonic evolution of the southwestern Alboran basin as inferred from seismic data off Morocco, American Association of Petroleum Geologists Bulletin, 81, 1161-1184, 1997. Choubert, G., and A. Faure-Muret, Moroccan Rif, in Mesozoic-Cenozoic Orogenic Belts, edited by A. Spencer, pp. 37-46, Scottish Academic Press, Edinburgh, 1974. Chung, W.-Y., and H. Kanamori, Source process and tectonic implications of the Spanish deep-focus earthquake of March 29, 1954, Physics of the Earth and Planetary Interiors, 13, 85-96, 1976. Comas, M.C., V. Garcia-Duenas, and M.J. Jurado, Neogene Tectonic Evolution of the Alboran Sea from MCS Data, Geo-Marine Letters, 12, 157-164, 1992.
40 Comas, M.C., J.P. Platt, J.I. Soto, and A.B. Watts, The origin and tectonic history of the Alboran Basin: Insights from ODP Leg 161 results, in Proceedings of the Ocean Drilling Program, Scientific Results, edited by R. Zahn, M.C. Comas, and A. Klaus, pp. 555-580, Ocean Drilling Program, College Station, 1999. Dewey, J., M. Helman, E. Turco, D. Hutton, and S. Knott, Kinematics of the western Mediterranean, in Alpine Tectonics, edited by M. Coward, D. Dietrich, and R. Park, pp. 265-283, Geological Society of London, London, 1989. Docherty, C., and E. Banda, Evidence for the eastward migration of the Alboran Sea based on regional subsidence analysis: A case for basin formation by delamination of the subcrustal lithosphere?, Tectonics, 14, 804-818, 1995. Evans, J.R., and U. Achauer, Teleseismic velocity tomography using the ACH method: theory and application to continental-scale studies, in Seismic Tomography: Theory and Practice, edited by H.M. Iyer, and K. Hirahara, pp. 319-360, Chapman and Hall, London, 1993. Frizon de Lamotte, D., J. Andrieux, and J.-C. Guézou, Cinématique des chevauchements néogènes dans l'Arc bético-rifain: discussion sur les modèles géodynamiques, Bulletin de la Société géologique deFrance, 162, 611-626, 1991. García-Dueñas, V., J.C. Balanyá, and J.M. Martínez-Martínez, Miocene extensional detachments in the outcropping basement of the northern Alboran basin (Betics) and their tectonic implications, Geo-Marine Letters, 12, 88-95, 1992. Grimson, N.L., and W. Chen, The Azores-Gibraltar plate boundary: focal mechanism, depths of earthquakes and their tectonic implications, Journal of Geophysical Research, 92, 2029-2047, 1986.
41 Guerrera, F., A. Martin-Algarra, and V. Perrone, Late Oligocene-Miocene syn-/-lateorogenic successions in the western and central Mediterranean chains from the Betic Cordillera to the southern Appennines, Terra Nova, 5, 525-544, 1993. Gurria, E., J. Mezcua, and M.J. Blanco, Crustal and upper mantle velocity structure of Southern Iberia, the sea of Alboran, and the Gibraltar arc determined by local earthquake tomography, Annali di Geofisica, 40, 123-132, 1997. Hayward, N., Marine geophysical study of the Eurasian-African plate boundary in the vicinity of the Gorringe Bank, Ph.D. thesis, 225 pp., University of Oxford, Oxford, 1997. Hearn, T.M., and J.F. Ni, Pn velocities beneath continental collisional zones: the TurkishIranian Plateau, Geophysical Journal International, 117, 273-283, 1994. Hernandez, J., and H. Bellon, Chronologie K-Ar du volcanisme Miocéne du Rif oriental (Maroc): implications tectoniques et magmatologiques, Revue de géologie dynamique et géographie physique, 26, 85-94, 1985. Houseman, G.A., D.P. McKenzie, and P. Molnar, Convective instability of thickened boundary layer and its relevance for the thermal evolution of continental convergent belts, Journal of Geophysical Research, 86, 6115-6132, 1981. Jackson, D.D., and M. Matsu'ura, A Bayesian approach to nonlinear inversion, Journal of Geophysical Research, 90, 581-591, 1985. Johnson, C., N. Harbury, and A. Hurford, The role of extension in the Miocene denudation of the Nevado-Filábride complex, Betic Cordillera (SE Spain), Tectonics, 16, 189-204, 1997.
42 Jones, C.H., H. Kanamori, and S.W. Roecker, Missing roots and mantle "drips": Regional Pn and teleseismic arrival times in the southern Sierra Nevada and vicinity, California, Journal of Geophysical Research, 99, 4567-4601, 1994. Kennett, B.L.N., and E.R. Engdahl, Traveltimes for global earthquake location and phase identification, Geophysical Journal International, 105, 429-465, 1991. Kirker, A.I., and J.P. Platt, Unidirectional slip vectors in the western Betic Cordillera: implications for the formation of the Gibraltar arc, Journal of the Geological Society, London, 155, 193-207, 1998. Lévêque, J.-J., L. Rivera, and G. Wittlinger, On the use of the checker-board test to assess the resolution of tomographic inversions, Geophysical Journal International, 115, 313-318, 1993. Lonergan, L., and N. White, Origin of the Betic-Rif mountain belt, Tectonics, 16, 504522, 1997. Martínez-Martínez, J.M., and J.M. Azañón, Mode of extensional tectonics in the southeastern Betics (SE Spain): Implications for the tectonic evolution of the periAlborán orogenic system, Tectonics, 16, 205-225, 1997. Masson, F., and J. Trampert, On ACH, or how reliable is regional teleseismic delay time tomography?, Physics of the Earth and Planetary Interiors, 102, 21-32, 1997. Mazzoli, S., and M. Helman, Neogene patterns of relative plate motion for AfricaEurope: some implications for recent Mediterranean tectonics, Geologische Rundschau, 83, 464-468, 1994.
43 Medialdea, T., E. Suriñach, R. Vegas, E. Banda, and J. Ansorge, Crustal structure under the western end of the Betic cordillera (Spain), Annales Geophysicae, 4, 457-464, 1986. Mezcua, J., and J. Rueda, Seismological evidence for a delamination process in the lithosphere under the Alboran sea, Geophysical Journal International, 129, F1F8, 1997. Michard, A., H. Feinberg, D. El-Azzab, M. Bouybaouene, and O. Saddiqi, A serpentinite ridge in a collisional paleomargin seting: the Beni Malek massif, External Rif, Morocco, Earth and Planetary Science Letters, 113, 435-442, 1992. Morales, J., I. Serrano, A. Jabaloy, J. Galindo-Zaldivar, D. Zhao, F. Torcal, F. Vidal, and F. Gonzalez-Lodeiro, Active continental subduction beneath the Betic Cordillera and the Alboran Sea, Geology, 27, 735-738, 1999. Morales, J., I. Serrano, F. Vidal, and F. Torcal, The depth of the earthquake activity in the central Betics (Southern Spain), Geophysical Research Letters, 24, 32893292, 1997. Pavlis, G.L., and J.R. Booker, The mixed discrete-continuous inverse problem: application to the simultaneous determination of earthquake hypocenters and velocity structure, Journal of Geophysical Research, 85, 4801-4810, 1980. Piromallo, C., and A. Morelli, Imaging the Mediterranean upper mantle by P-wave travel time tomography, Annali di Geofisica, 40, 963-979, 1997. Platt, J.P., J.I. Soto, M.J. Whitehouse, A.J. Hurford, and S.P. Kelley, Thermal evolution, rate of exhumation, and tectonic significance of metamorphic rocks from the floor
44 of the Alboran extensional basin, western Mediterranean, Tectonics, 17, 671-689, 1998. Platt, J.P., and R.L.M. Vissers, Extensional collapse of thickened continental lithosphere: A working hypothesis for the Alboran Sea and Gibraltar arc, Geology, 17, 540543, 1989. Plomerová, J., G. Payo, and V. Babuska, Teleseismic P-residual study in the Iberian Peninsula, Tectonophysics, 221, 1-12, 1993. Polyak, B.G., M. Fernàndez, M.D. Khutorskoy, J.I. Soto, I.A. Basov, M.C. Comas, V.Y. Khain, B. Alonso, G.V. Agapova, I.S. Mazurova, A. Negredo, V.O. Tochitsky, J. de la Linde, N.A. Bogdanov, and E. Banda, Heat flow in the Alboran sea, western Mediterranean, Tectonophysics, 263, 191-218, 1996. Rehault, J.-P., G. Boillot, and A. Mauffret, The western Mediterranean basin, in Geological Evolution of the Mediterranean Basin, edited by D.J. Stanley, and F.C. Wezel, pp. 101-129, Springer-Verlag, New York, 1985. Roecker, S.W., Velocity structure of the Pamir-Hindu Kush region: Possible evidence of subducted crust, Journal of Geophysical Research, 87, 945-959, 1982. Roecker, S.W., T.M. Sabitova, L.P. Vinnik, Y.A. Burmakov, M.I. Golvanov, R. Mamatkanova, and L. Munirova, Three-dimensional elastic wave velocity structure of the western and central Tien Shan, Journal of Geophysical Research, 98, 15779-15795, 1993. Roecker, S.W., Y.H. Yeh, and Y.B. Tsai, Three-Dimensional P and S wave velocity structures beneath Taiwan: Deep structure beneath an arc-continent collision, Journal of Geophysical Research, 92, 10547-10570, 1987.
45 Royden, L.H., Evolution of retreating subduction boundaries formed during continental collision, Tectonics, 12, 629-638, 1993. Seber, D., M. Barazangi, A. Ibenbrahim, and A. Demnati, Geophysical evidence for lithospheric delamination beneath the Alboran Sea and Rif-Betic mountains, Nature, 379, 785-790, 1996a. Seber, D., M. Barazangi, B.A. Tadili, M. Ramdani, A. Ibenbrahim, and D. Ben Sari, Three-dimensional upper mantle structure beneath the intraplate Atlas and interplate Rif mountains of Morocco, Journal of Geophysical Research, 101, 3125-3138, 1996b. Sengör, A.M.C., Some current problems on the tectonic evolution of the Mediterranean during the Cainozoic, in Recent evolution and seismicity of the Mediterranean region, edited by E. Boschi, E. Mantovani, and A. Morelli, pp. 1-51, Kluwer, Amsterdam, 1993. Spakman, W., S. van der Lee, and R. van der Hilst, Travel-time tomography of the European-Mediterranean mantle down to 1400 km, Physics of the Earth and Planetary Interiors, 79, 3-74, 1993. Srivastava, S.P., W.R. Roest, L.C. Kovacs, G. Oakley, S. Lévesque, and J. Verhoef, Motion of Iberia since the Late Jurassic: Results from detailed aeromagnetic measurements in the Newfoundland Basin, Tectonophysics, 184, 229-260, 1990. Tandon, K., J.M. Lorenzo, and J. de La Linde Rubio, Timing of rifting in the Alboran Sea basin - correlation of borehole (ODP Leg 161 and Andalucia A-1) to seismic reflection data: implications for basin formation, Marine Geology, 144, 275-294, 1998.
46 Tarantola, A., and B. Vallette, Generalized nonlinear inverse problems solved using the least squares criterion, Reviews of Geophysics and Space Physics, 20, 219-232, 1982. Thurber, C.H., and W.L. Ellsworth, Rapid solution of ray tracing problems in heterogeneous media, Bulletin of the Seismological Society of America, 70, 11371148, 1980. Tichelaar, B.W., and L.J. Ruff, How good are our best models? Jackknifing, bootstrapping, and earthquake depth, Eos, Transactions American Geophysical Union, 70, 605-606, 1989. Torres-Roldan, R.L., G. Poli, and A. Peccerillo, An Early Miocene arc-tholeiitic magmatic dike event from the Alboran Sea- Evidence for precollision subduction and back-arc crustal extension in the westernmost Mediterranean, Geologische Rundschau, 75, 219-234, 1986. Van der Wal, D., and R.L.M. Vissers, Uplift and emplacement of upper mantle rocks in the western Mediterranean, Geology, 21, 1119-1122, 1993. Vissers, R.L.M., J.P. Platt, and D. van der Wal, Late orogenic extension of the Betic Cordillera and the Alboran Domain: A lithospheric view, Tectonics, 14, 786-803, 1995. Watts, A.B., J.P. Platt, and P. Buhl, Tectonic evolution of the Alboran Sea, Basin Research, 5, 153-177, 1993. Working group for deep seismic sounding in the Alboran sea, Crustal seismic profiles in the Alboran sea- preliminary results, Pure and Applied Geophysics, 116, 167-180, 1978.
47 Yilmaz, P.O., I.O. Norton, D. Leary, and R.J. Chuchla, Tectonic evolution and paleogeography of Europe, in Peri-Tethys Memoir 2: Structure and prospects of alpine basins and forelands, edited by P.A. Ziegler, and F. Horvàth, pp. 47-60, Mém. Mus. natn. Hist. nat., Paris, 1996. Zeck, H.P., Betic-Rif orogeny: subduction of Mesozoic Tethys lithosphere under eastward drifting Iberia, slab detachment shortly before 22 Ma, and subsequent uplift and extensional tectonics, Tectonophysics, 254, 1-16, 1996. Zeck, H.P., Mantle peridotites outlining the Gibraltar Arc - centrifugal extensional allochthons derived from the earlier Alpine, westward subducted nappe pile, Tectonophysics, 281, 195-207, 1997.
G. Alguacil, Instituto Andaluz de Geofisica, Campus Universitario de Cartuja, Universidad de Granada, 18071-Granada, Spain. (
[email protected]) M. Barazangi, A.S. Calvert, E. Sandvol, and D. Seber, Institute for the Study of the Continents, Snee Hall, Cornell University, Ithaca, NY 14853. (
[email protected]) N. Jabour, CNCPRST, 52 Avenue Omar Ibn Khattab, B.P. 1346, Agdal-Rabat, Morocco. T. Mourabit, Tetouan University, Tetouan, Morocco. (
[email protected]) S. Roecker, Department of Earth and Environmental Sciences, Rensselaer Polytechnic Institute, Troy, NY 12180. (
[email protected]) F. Vidal, Instituto Geografico Nacional, General Ibañez de Ibero, 3 28003-Madrid, Spain. (
[email protected])
48
Table 1. P and S reading uncertainties assessed by comparing independent picks of the same arrival Data Type
Impulsive (P phase /S phase) sec a b Local, IGN and IAG 0.2/1.0* Local, CNCPRSTc 0.5/0.8 Local, Mohammed Vd 0.4/0.8 Teleseism, IGN and IAG 0.3 ‡ 0.3 Teleseism, CNCPRST + Phase not reported as either impulsive or emergent
Uncertainties Unreported+ (P phase /S phase) sec 0.4*/0.8*/0.7 0.4/0.8 0.3 0.6
Emergent (P phase /S phase) sec 0.4/0.8 0.8/0.7 0.4/0.8 -
*Limited data a
Instituto Geografico Nacional, Madrid, Spain
b
Instituto Andaluz de Geofisica, Granada, Spain
c
Centre National de Coordination et de Planification de la Recherche Scientifique et Technique,
Rabat, Morocco d
Physique du Globe, Mohammed V University, Rabat, Morocco
‡ Teleseismic residuals for IFR, TAF, & RBA were found to be particularly scattered hence arrivals reported to these stations were assigned an increased reading uncertainty (1 second).
Table 1
Table 2. Initial and final 1-D models Interface Depth Initial P vel Final 1-D P vel Initial S vel Final 1-D S vel (km) (km/s) (km/s) (km/sec) (km/s) -5 - 15 5.80 5.75 3.40 3.32 15 - 30 6.50 6.53 3.75 3.80 30 - 45 7.90 7.89 4.50 4.54 45 - 60 8.00 8.09 4.50 4.64 60 - 100 8.10 8.12 4.50 4.67 100 - 150 8.10 8.12 4.50 4.67 150 - 200 8.15 8.15 4.52 4.67 200 - 260 8.20 8.20 4.55 4.67 260 - 330 8.40 8.40 4.65 4.67 330 - 410 8.80 8.80 4.80 4.80 410 - 490 9.45 9.45 5.15 5.15 490 - 570 9.80 9.80 5.30 5.30 570 - 650 10.15 10.15 5.50 5.50 The final best-fit 1-D model was obtained from inversion of local data. The average velocity in each layer of final 3-D model was insignificantly different from 1-D inversion results.
Table 2
Table 3. Criteria used to remove poorly constrained events Criteria No. Criteria Value 1 Condition Number < 100 2 Weighted RMS residual < 2.0 sec 3 Closest Station Distance < 150 km 4 Azimuth Gap < 200 degrees 5 Average Residual < 1.0 sec 6 Theoretical Location Error < 10 km 7 Final Iteration step < 3 km 8 Depth > 0.1 km 9 Horizontal Shift < 10 km 10 Vertical Shift < 10 km The criteria were applied after relocation. Criteria 1,3, and 4 check that a good station geometry is used to determine the location. Criteria 2 and 5 check that the location provides a reasonable fit to the data. Criterion 7 checks that the solution converges. Criterion 8 was required to remove a significant number of events that attempted to converge to a location above ground. Criteria 9 and 10 were used during the 1-D and 3-D inversion iterations to remove any events that were shifted a significant amount when relocated in the new velocity models.
Table 3
49
Figure Captions
Figure 1. Maps showing regional setting of the Alboran Sea and schematic paleogeographic reconstructions of western Mediterranean [Modified from Alvarez et al., 1974; Rehault et al., 1985; Dewey et al., 1989; Yilmaz et al., 1996; Lonergan and White, 1997]. (a) The Alboran Sea is one of a number of Neogene extensional basins that formed along the Africa-Eurasia plate boundary zone despite continued convergence between the two plates since the Cretaceous. MA = Middle Atlas and SA = Saharan Atlas. Box shows Alboran region and location of Figure 2. (b) Extension behind a subduction zone retreating to the E and S in the LateOligocene/Early Miocene resulted in the collapse of a thickened mountain belt that used to extend along the coast of Spain and France. The Paleogene location of the thickened crust that later formed the Betic and Rif Internal zones and the westerly extent of the subduction zone are not well constrained. Arrows show the transport direction. (c) Extension persisted until the Early/Middle Miocene with continued westward transport of the Internal zones in Alboran region.
Figure 2. Simplified tectonic map of Alboran region. AH = Al-Hoceima, AR = Alboran Ridge, EAB = East Alboran Basin, NF = Nekor fault, SAB = South Alboran Basin, SG = Strait of Gibraltar, WAB = West Alboran Basin, p = location of Rhonda (Spain) and
50 Beni-Bousera (Morocco) peridotites, b = Early Miocene basaltic dykes, c = Mid Miocene calc-alkaline volcanism, s = Late Miocene potassic volcanism, and b = Pliocene and Quaternary alkali basalt flows [Bellon and Brousse, 1977]. Note absence of volcanism west of Malaga and concentration of calc-alkaline volcanism in the center of the Alboran Sea. [Map modified after Lonergan and White, 1997; Platt et al., 1998; Comas et al., 1999].
Figure 3. Seismic networks in the region used for this study. CNCPRST = Centre National de Coordination et de Planification de la Recherche Scientifique et Technique (Morocco); MOH V = Mohammed V University (Morocco); IAG = Instituto Andaluz de Geofisica (Spain), IGN = Instituto Geografico Nacional (Spain). Bracketed term in legend describes the type of data used in this study. D = Digital, A = Analog, and B = Bulletin.
Figure 4. Block diagrams showing seismicity in the Alboran region (same area as Figure 2). (a) Data compiled from local (CNCPRST) and global (ISC & NEIC) catalogs and relocations made by authors. NS and EW projections of seismicity located between the dashed lines are shown on block sides with x6 vertical exaggeration. Note the N-S line of intermediate depth seismicity that dips beneath the Alboran Sea from beneath city of Malaga (see Figure 2 for location) and the three deep earthquakes located beneath the central Betics. Scattered intermediate depth earthquakes are also located beneath the Gulf of Cadiz. The horizontal lines of seismicity at depths of 10 and 33 km are result of ISC and NEIC operators fixing the depth to obtain an epicentral location with limited
51 data. Data shown formed the initial pool of local/regional data used for the velocity inversions. (b) Events relocated in derived 3D velocity model after removal of poorly constrained events (failed criteria 1-8 shown in Table 3).
Figure 5. Map showing the 1600 teleseismic earthquakes used for the velocity inversion. The size of the symbol marking each event indicates the number of stations in the Alboran region that recorded an impulsive arrival from this event. Note good E-W azimuth coverage. Illumination is limited from N and S especially from the south. Circles mark 30 and 90 degrees from the center of the Alboran Sea.
Figure 6. 1-D Velocity model used for earthquake relocation and other representative velocity models. (a) Velocities for the uppermost layers were chosen based on local refraction experiments. Refraction velocities are from a compilation presented by [Carbonell et al., 1998] and references therein. (b) Velocities at greater depths were chosen to lie close to PM2 [Spakman et al., 1993]. Jefferys-Bullen and IASPEI91 models are shown for reference.
Figure 7. Map showing station delays obtained from 1-D inversion. Note significant delays for stations in the western Betics and Rif, and early arrivals near the Spanish Alboran coast.
52 Figure 8.
Block diagram showing model parameterization and locations of cross-
sections (Figure 12). 50 x 50 x 15 km blocks were used in upper 60 km. 100 x 100 km blocks increasing from 40 to 70 km in thickness with depth were used from 60 to 650 km. Dashed line marks the area shown in Figures 10-13.
Figure 9. Maps showing hit count for 50 km width blocks throughout the model volume. Note that for the inversion 100 km blocks were used below 60 km depth.
Figure 10. Cross sections showing an example section through the final 3-D P velocity model and the different attributes used to assess it. The cross sections are all located in the same position parallel and 25 km north of BB' along the southern coast of Spain (See Figure 8 for location of BB’). A topographic profile along the line of the cross section is shown for reference. (a) Final velocity model calculated using block shift-and-average approach (average of 100 separate inversions conducted with 10 km block shifts, see text for details). Anomalies shown are relative to the 1-D model shown in Table 2. Earthquakes (circles) are projected in from a maximum distance of 50 km perpendicular to the section. The most striking feature is the high velocity anomaly in the upper mantle. This anomaly reaches its shallowest point in the west where it is separated from the low velocity region by the intermediate depth seismicity. (b) Result of a single inversion using the model parameterization shown in Figure 8 (no block shifting). The first order features shown in Figure l0a, although more difficult to interpret, are visible.
53 (c) Standard error in block velocities calculated using a posteriori covariance matrix. (d) Uncertainty in block velocities calculated using non-linear bootstrap resampling. The high velocity mantle anomalies are found to be of the order or greater than two times the bootstrap error (See Figure 10b) suggesting that they are robust features for this model parameterization. (e) Diagonal elements of the resolution matrix. High diagonal elements indicate strong control of observations on inversion result. Diagonal elements are higher in the west owing to the larger number of illuminating rays. (f) Spike test input model. 100 km blocks with +/- 0.5 km/s perturbations relative to reference model were used, separated by a block with zero perturbation. A depth slice would also show a similar pattern of perturbations. (g) Result of inversion using synthetic arrival times calculated using spike model shown in Figure 10f. No noise was added to data and block boundaries used for inversion were coincident with the boundaries of the perturbations. Recovery is almost perfect in areas of illumination. (h) Identical to 10g except representative Gaussian distributed noise was added to the data. The addition of noise degrades the recovery but not substantially. (i) Result of shift-and-average inversion with noise added to arrival times. Recovery is very poor throughout the upper mantle layers showing that degree of recovery of anomalies the size of a model block is strongly dependent on the model parameterization. Note the surprising recovery of velocity structure at the base of model (See text for discussion).
54 (j) Modified synthetic model with spikes the same width as in Figure 10f but now extending over two layers with a two layer vertical gap. Horizontal size and spacing of perturbations was not changed. (k) Shift-and-average recovery of spikes spanning two layers. Limited structure is now imaged in the mid-upper mantle but smearing is significant, indicating that depth resolution may be limited.
Figure 11. Horizontal slices through the center of each layer of the final P velocity model. Bracketed term in the title is the reference velocity. Darker and lighter gray shades indicate regions of higher and lower velocities relative to reference model, respectively.
Earthquakes were relocated from the starting data set using the 3-D
velocity model (same locations as Figure 4b) with poorly constrained events removed. Main features (see text for detail discussion): (a) Low velocities in western Betics and Rif; (b) High velocity anomaly off southern coast of Spain east of Malaga; (c) ‘Horseshoe’ of low velocities surrounding a high velocity region in the central Alboran; (d) Low velocities beneath Betics and Rif; (e) High velocity beneath western Betics and Rif separated from a low velocity along southern coast of Spain by intermediate depth earthquakes; (f-i) Progression of high velocity anomaly to south and east with depth beneath the Alboran Sea; (j and k) High velocity anomaly becomes more diffuse; (l and m) more concentrated anomalies imaged in the vicinity of deep earthquakes.
Figure 12. Cross sections: (a) AA', (b) BB', (c) CC', and (d) DD' of P velocity model (left) determined using block shift-and-average and ray coverage (right). See Figure 8
55 for locations. Darker and lighter gray shades indicate regions of higher and lower velocities relative to reference model (Table 2), respectively. Gray rays are associated with local events and black rays with teleseismic events. Earthquakes shown with ray plots are events actually used for the inversion. Earthquakes shown on the velocity model plots were relocated from the starting data set in the 3-D velocity model with poorly constrained events removed. Triangles are stations used for the inversion. Only earthquakes and ray segments located within 50 km of the cross sections are shown. A significant high velocity anomaly is imaged in the upper mantle that appears to dip to the SE from lithospheric depths to approximately 350 km before becoming more diffuse. The intermediate depth earthquakes mark a west to east transition from high to low velocity in the uppermost mantle. Significant anomalies are also imaged at the base of the model near the location of the deep earthquakes (Marked by arrows on AA’ and DD'). Only one of the three deep earthquake locations in the data set was sufficiently well constrained by arrivals in the region to be plotted.
Figure 13. Horizontal and vertical sections through the synthetic (SYN) and recovered (REC) models used to investigate smearing of upper mantle anomalies. Synthetic travel times were calculated using model SYN designed to include the principle features of the derived velocity model (Compare with Figures 11 and 12). However, Layers 5, 10, and 11 (e, j, and k) were left with no perturbation in the synthetic model to determine whether anomalies would be smeared into these layers from other layers. Perturbations are +/5% in the synthetic model. Inversion of the synthetic travel times after the addition of suitable noise produced model REC. Limited smearing occurs into layer 5. However,
56 significant smearing occurs into layers 10 and 11 indicating that determining slab continuity at these depths may be beyond the resolution of the data set used for this study.
Figure 14. Models previously proposed to explain the geodynamic evolution of the Alboran region. These models may be categorized in terms of roll-back or detachment of lithosphere and whether or not crust is involved in recycling of the lithosphere into the mantle. (a) Retreat from east to west of an east dipping subducting slab during the Early Miocene [Lonergan and White, 1997]. (b) Break-off of a NW dipping slab at the Oligocene-Miocene boundary perhaps subducted during earlier (mainly Cretaceous) eastward movement of Iberia relative to Eurasia [Blanco and Spakman, 1993; Zeck, 1996; Zeck, 1997]. (c) Delamination (s.s.) of lithosphere thickened during Paleogene convergence initiated during Early Miocene but still active today [Seber et al., 1996a]. (d) Convective removal during the Early Miocene of mantle lithosphere thickened during Paleogene convergence [Platt and Vissers, 1989].
Figure 15. Evolutionary model showing map view and two cross sections modified from working model proposed by Platt and Vissers [1989] (compare with convective removal model shown in Figure 14) to explain results from this tomography study and Platt et al.'s [1998] more recent results.
Thick dashed lines show suggested position for
connection of delaminated lithosphere to surface lithosphere. G.B. = Guadalquivir Basin. Circles show locations of present subcrustal seismicity. Note the body shown in the upper mantle in the "Present" panel of BB' is attached to Iberia to the west of the cross
57 section. Patterns are the same as those shown in Figures 2 and 14. The crooked arrows shown in the Strait of Gibraltar marks the relative motion of Africa to Iberia remaining before present.
10°W
5°W
0°
5°E
10°E
15°E
Ap
Bay of Biscay
en
gh
Pyrenees c ia
Tr ou
Tyrrenian Sea rc nA
Va ea an S
r
r Alge Tell Atlas
CretaceousPaleogene Thrust Belts Neogene Thrust Belts
Rif
tic an n l t A ea Oc
a
sin
Ba çal
ven
B Albo
Calabria
Pro ian-
s etic
35°N
es
len
Iberia
40°N
nin
Neogene Basin
A
M
SA
0
s High Atla
Late Oligocene Early Miocene
Atlas Mountains
Africa
km
500
Thrust
Neogene Oceanic Crust Neogene Extended Continental Crust Tethyan Oceanic Crust Stable Block
Subduction
du c
tio n
Middle Miocene
b Mesozoic Su Oceanic Crust
?
? Miocene Oceanic Crust
?
? ?
? ? Flysch
b
?
km
c
km
Figure 1
Figure 2
Gulf of Cadiz
G
id
M
le
d
Rif
p
WAB
p
5°W
c
s
la At
NF
AH
c
AR
as a
s
c
SAB
cc c c c c
bb b Malaga
b b u S
High Atlas
arb Gh in s Ba
SG
a uad
iv lqu
etic
sin
a ir B
5°W
Iberia
Moroccan Meseta
Atlantic Ocean
35°N
10°W
c
EAB
c
sa
c
a
s
b Prse
a
c c
a s
etic
e Bal
nds
Isla c i r a
0°
0
km
200
a = alkali basalt flows b = basaltic dykes c = calc-alkaline s = shoshinite/lamprophyre (potassium enriched) p = peridotite
Internal Zone Flysch Nappes External Zone Basin Atlas Mountains Stable Block
Tell Atlas
Mediterranean Sea
0°
10°W
5°W
0°
40°N
Mediterranean
35°N
Atlantic Network Administrator (Data type)
CNCPRST (DAB) MOH V (AB) IAG & IGN (DB) Other (B)
0
km
200
Figure 3
N or th
EQ Depth < 50 km 50-500 km
A Depth (km)
0 30 60 90 120 150
> 500 km
0
100
km
B
Figure 4
eg re
30 d
es
# of Stations 6-8 9-12 13-20 21-30 31-50
Figure 5
(a) 0
S
Depth (km)
20
P
40 60 80 100
3
4
5
6 Velocity (km/s)
(b)
7
8
0
Jefferys-Bullen (ISC) IASPEI91 PM2 Refraction 1D Relocation
100 Depth (km)
9
200 300 400 500 600 700
3
4
5
6
7 Velocity (km/s)
8
9
10
11
Figure 6
10°W
5°W
0°
40°N
Mediterranean
35°N
P delay (sec) Atlantic
0
km
200
>1 0.5 -1.0 0.2 - 0.5 -0.2 - 0.2 -0.5 - -0.2 -1.0 - -0.5 < -1.0
Figure 7
Figure 8
Depth (km)
S
N-
600
600
500
400
300
200
100
0
0
400
B
)
m
(k
200
e
at
din
or
Co
200
el
M od
400
A
600
C'D'
800
600
400
W E-
200
400
)
600
km ( 200 ate n i 0 ord o C el d o M
CD
A' B'
800
Figure 9
j) Layer 10: 330-410 km
i) Layer 9: 260-330 km
1
f) Layer 6: 100-150 km
e) Layer 5: 60-100 km
m) Layer 13: 570-650 km
b) Layer 2: 15-30 km
a) Layer 1: -5-15 km
10
100
50 km block hit count
k) Layer 11: 410-490 km
g) Layer 7: 150-200 km
c) Layer 3: 30-45 km
1000
l) Layer 12: 490-570 km
h) Layer 8: 200-260 km
d) Layer 4: 45-60 km
a) Shift P velocity perturbation (%) Iberia/Alboran
W Atlantic
4
E
Med.
3 0
2
100
1
Depth (km)
200
0
300
-1
400
-2
500
-3
600 500
400
300
200
100
0
100
0
0
100
100
200
200
300
300
400
400
500
500
400
500
-4
600
500
400
-3
300
-2
200
100
-1
0
100
0
200
1
300
400
2
500
3
500
4
d) Bootstrap Error (%)
0
400
300
0.5
200
1
100
1.5
0
100
2
200
2.5
300
400
3
500
3.5
4
e) P resolution diagonals
0
0
100
100
200
200
300
300
400
400
500
500
600 500
300
c) Standard Error (%)
600
0
200
W-E Model Coordinate (km)
b) No shift P velocity perturbation (%)
-4
Percent velocity perturbation relative to 1-D model
600 400
0.5
300
1
200
100
1.5
0
2
100
200
2.5
300
3
400
500
3.5
500
4
0.5
400
300
0.6
200
100
0.7
0
100
0.8
200
300
400
0.9
500
1
Figures 10a-e
f) Single layer input spike model
g) Single layer spike: No shift and no noise
0
0
100
100
200
200
300
300
400
400
500
500
600
600
500
400
300
200
100
0
100
200
300
400
500
500
0
0
100
100
200
200
300
300
400
400
500
500
600
200
100
0
100
200
300
400
500
300
400
500
400
500
600 400
300
200
100
0
100
200
300
400
500
500
j) Double layer input spike model
0
100
100
200
200
300
300
400
400
500
500
600
600 400
300
200
100
0
100
-4
200
400
300
200
100
0
100
200
k) Double Layer spike: Noise & Shift
0
500
300
i) Single layer spike: Noise & Shift
h) Single layer spike: No shift & with noise
500
400
300
-3
400
-2
500
500
-1
0
1
2
400
300
3
200
100
0
100
200
300
4
% P velocity Perturbation
Figure 10f-k
Figure 11a-d
400
300
100
200
300
400
-1
500
-4
-3
300
400
500
-2
1
400
400
2
100
0
100
200
300
300
3
200
4
100
0
100
200
300
d) Layer 4, 45 -60 km < 8.09 km/s>
200
b) Layer 2, 15-30 km < 6.53 km/s>
300
% P velocity Perturbation
0
400 500 200
400 500
100
300
300
0
200
200
100
100
100
200
0
0
300
100
100
400
200
200
400
300
c) Layer 3, 30-45 km < 7.89 km/s>
300
400
400 500
KM
400 500 0
300
300
100
200
200
200
100
0
100
0
100
100
KM
200
200
400
300
a) Layer 1, -5-15 km < 5.75 km/s>
300
400
400
400
500
500
Figure 11e-h
300
200
100
0
100
200
300
400
-1
500
0
100
200
-3
300
-2
400
500
-4
1
400
400
2
100
0
100
200
300
300
3
200
4
100
0
100
200
300
h) Layer 8, 200-260 km < 8.20 km/s>
200
400
400
f) Layer 6, 100 -150 km < 8.12 km/s>
300
% P velocity Perturbation
0
400 500
100
400 500 200
300
300
300
200
0
0
200
100
100
100
200
200
100
300
300
400
400 500
400 500
400
300
300
g) Layer 7, 150-200 km < 8.15 km/s>
200
200
400
100
100
0
0
200
300
400
100
400
e) Layer 5, 60-100 km < 8.12 km/s>
100
200
300
400
500
500
Figure 11i -l
400
-1
500
100
200
-3
300
-2
400
500
1
400
400
2
300
300
100
0
100
200
300
400
3
200
4
100
0
100
200
300
400
l) Layer 12, 490-570 km < 9.80 km/s>
200
j) Layer 10, 330-410 km < 8.80 km/s>
% P velocity Perturbation
0
400 500
400 500
-4
300
200
200
300
100
100
0
0
0
100
100
100
200
200
200
300
300
300
400
400
400
k) Layer 11, 410-490 km < 9.45 km/s>
400 500
400 500 300
300
300
200
200
200
100
100
100
0
0
0
100
100
100
200
200
200
300
300
300
400
400
i) Layer 9, 260-330 km < 8.40 km/s>
400
500
500
m) Layer 13, 570 -650 km
400
300
200
100
0
100
200
300
400 500
400
300
200
100
0
100
200
300
400
-4
-3
-2
-1
0
1
2
3
4
500
% P velocity Perturbation
Figure 11m
Figure 12a
Depth (km)
500
300
-3
400
0
100
200
300
400
500
-4
-1
0
1
2
% P velocity Perturbation
-2
W-E Model Coordinate (km) 3
4
500
600
400
300
200
100
0
600 100
Mediterranean
500
200
Iberia
500
400
300
200
100
0
Atlantic
a) A-A', W-E Cross section
Depth (km)
400
300
100
0
100
200
W-E Model Coordinate (km)
200
300
a) Rays, A-A', W-E Cross section
400
500
Figure 12b
Depth (km)
100
200
300
400
500
400
300
200
100
0
500
300
-3
400
-4
-1
0
1
2
% P velocity Perturbation
-2
W-E Model Coordinate (km) 3
4
500
600 0
Mediterranean
600 100
Alboran
500
200
Gibraltar
500
400
300
200
100
0
Atlantic
b) B-B', W-E Cross section
Depth (km)
400
300
100
0
100
200
W-E Model Coordinate (km)
200
300
b) Rays, B-B', W-E Cross section
400
500
Figure 12c
Depth (km)
300
-3
400
0
100
200
300
400
-4
-1
0
1
2
% P velocity Perturbation
-2
S-N Model Coordinate (km) 3
4
400
600
400
300
200
100
0
600 100
Iberia
500
200
Alboran Sea
500
400
300
200
100
0
Morocco
c) C-C', S-N Cross section
Depth (km)
300
100
0
100
200
S-N Model Coordinate (km)
200
300
c) Rays, C-C', S-N Cross section
400
Figure 12d
Depth (km)
300
200
100
0
100
200
300
400
-3
-4
-1
0
1
2
% P velocity Perturbation
-2
S-N Model Coordinate (km) 3
4
400
600
600
400
500
400
300
200
100
0
500
400
300
200
100
0
d) D-D', S-N Cross section Morocco Alboran Iberia Sea
Depth (km)
300
100
0
100
200
S-N Model Coordinate (km)
200
300
d) Rays, D-D', S-N Cross section
400
Figure 13a-h
-1
0
1
2
3
4
g)REC: Layer 7: 150-200 km
% P velocity Perturbation
-2
f)REC: Layer 6: 100-150 km
e)REC: Layer 5: 60-100 km
-3
f)SYN: Layer 6: 100-150 km
e)SYN: Layer 5: 60-100 km
-4
c)REC: Layer 3: 30-45 km
b)REC: Layer 2: 15-30 km
a)REC: Layer 1: -5-15 km
g)SYN: Layer 7: 150-200 km
c)SYN: Layer 3: 30-45 km
b)SYN: Layer 2: 15-30 km
a)SYN: Layer 1: -5-15 km
h)REC: Layer 8: 200-260 km
h)SYN: Layer 8: 200-260 km
d)REC: Layer 4: 45-60 km
d)SYN: Layer 4: 45-60 km
Figure13i-s
m)REC: Layer 13: 570-650 km
-4
-3
-1
0
1
2
3
4
r)REC: West-East: BB’
r)SYN: West-East: BB’
% P velocity Perturbation
-2
q)REC: West-East: AA’
q)SYN: West-East: AA’
Cross sections
k)REC: Layer 11: 410-490 km
j)REC: Layer 10: 330-410 km
i)REC: Layer 9: 260-330 km
m)SYN: Layer 13: 570-650 km
k)SYN: Layer 11: 410-490 km
j)SYN: Layer 10: 330-410 km
i)SYN: Layer 9: 260-330 km
s)REC: South-North: CC’
s)SYN: South-North: CC’
l)REC: Layer 12: 490-570 km
l)SYN: Layer 12: 490-570 km
Figure 14
Crust & mantle recycling
Mantle recycling
ns
otatio
N
Asthenosphere
Lithosphere
Moho
Africa Iberia
S
W
Paleocene
Lonergan & White (1997)
Extension
East
Seber et al. (1996a)
Eocene-Oligocene E. Miocene-Present
Delamination
Rapid Roll Back Early-Mid Miocene
g&R tenin Shor
West
Retreating Subduction
Progressive Rolling or Peeling Back
SE Betic Lithos.
200 km
SSE
Oligocene
100 km
NNW
?
ge l rid iona collis
Extension
oma
Present
D ran Albo
in
Platt & Vissers (1989)
Early Miocene
teau g pla ults ndin nt fa extetachme de
Zeck (1997)
Present
200 km
Convective Removal
Breakoff of subducted slab L. Oligocene/E.Miocene
200 km
Iberian Lithosphere
NW
Slab Break off
Detachment
Sinking Slab
Paleogene
L. Oligocene/ E. Miocene
Present
B A dge al ri
n llisio
co
?
A'
eau plat ults g n i a tf nd exte chmen a t e d
G.B.
n
mai
Do oran
Alb
?
B' A
A'
Crust
G.B.
Mantle Lid Asthenosphere
?
?
?
?
Inflow
100 km
B
B' ?
100 km
Figure 15
