Evolution of the Newfoundland–Iberia conjugate rifted margins

Earth and Planetary Science Letters 273 (2008) 214–226

Contents lists available at ScienceDirect

Earth and Planetary Science Letters j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / e p s l

Evolution of the Newfoundland–Iberia conjugate rifted margins Alistair Crosby a,⁎, Nicky White a, Glyn Edwards b, Donna J. Shillington c a b c

Bullard Laboratories, Department of Earth Sciences, University of Cambridge, CB3 0EZ, UK BP Exploration Operating Co. Ltd., Sunbury-on-Thames, TW18 7LN, UK Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY 10964, USA

A R T I C L E

I N F O

Article history: Received 7 November 2007 Received in revised form 19 June 2008 Accepted 20 June 2008 Available online 4 July 2008 Editor: C.P. Jaupart Keywords: continental margins stretching strain rate inversion paleobathymetry rheology

A B S T R A C T It is accepted that mildly extended sedimentary basins form by largely uniform thinning of continental lithosphere. No such consensus exists for the formation of highly extended conjugate rifted continental margins. Instead, a wide range of models which invoke differing degrees of depth-dependent thinning have been proposed. Much of this debate has focussed on the well-studied Newfoundland–Iberia conjugate margins. We have tackled the problem of depth dependency at this pair of margins in three steps. First, we have reconstructed water-loaded subsidence histories by making simple assumptions about changes in water depth through time. Secondly, we have used these reconstructed subsidence histories to determine the spatial and temporal variation of lithospheric strain rate. An inversion algorithm minimizes the misfit between observed and predicted subsidence histories and crustal thicknesses by varying strain rate as a smooth function of distance across the margin, depth through the lithosphere, and geologic time. Depthdependent thinning is permitted but, crucially, our algorithm does not prescribe its existence or form. Given the absence of significant volumes of syn-rift magmatism, we have also applied a minimal melting constraint. Inverse modeling has yielded excellent fits to both reconstructed subsidence and crustal observations, which suggest that rifting occurred from ∼ 150–135 Ma and at rates of up to 0.3 Ma− 1. Strain rate distributions are depth-dependent, suggesting that lithospheric mantle thins over a wider region than the crust. Beneath highly extended parts of the margin, crustal strain rates greatly exceed lithospheric mantle strain rates. Thirdly, we have tested our strain rate histories by comparing the total horizontal extension with the amount of extension inferred from normal faulting patterns. Both values agree within error. We freely acknowledge that there are important uncertainties in reconstructing the subsidence histories of deep-water margins. Nevertheless, stratigraphic records remain the only, albeit imperfect means of determining how crust and lithospheric mantle thin through time and space. © 2008 Elsevier B.V. All rights reserved.

1. Introduction Over the past 40 years, there has been much interest in the development of passive continental margins, which are a primary manifestation of plate tectonics. It is generally accepted that passive margins form by extension and cooling of continental lithosphere. Simple kinematic models have been developed which describe the growth of margins in terms of either uniform or depth-dependent stretching (e.g. McKenzie, 1978; Le Pichon and Sibuet, 1981; Hellinger and Sclater, 1983; Keen and Dehler, 1993). The temporal and spatial variation of crustal and lithospheric thinning across a margin control variations in temperature, heatflow, subsidence, extensional faulting and decompression melting (White et al., 2003). Despite acceptance of the basic model, there is much disagreement about the spatial distribution of lithospheric extension. This controversy has particularly focussed on the Newfoundland–Iberia conjugate margins which have been studied in detail using deep-seismic reflection ⁎ Corresponding author. Tel.: +44 1223 337102. E-mail address: [email protected] (A. Crosby). 0012-821X/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.epsl.2008.06.039

and wide-angle surveys, calibrated by deep-water boreholes. There are two substantive and unresolved issues. First, a significant discrepancy between the amount of extension measured from upper crustal faulting and the amount of extension measured from crustal thickness is often reported (e.g. Davis and Kusznir, 2004). Secondly, broad tracts of exhumed and mostly unmelted lithospheric mantle occur at deep-water margins (e.g. Whitmarsh et al., 2001; Minshull et al., 2001). These two observations are mutually incompatible but if either or both are correct, they imply that various forms of depth dependency occur, perhaps at different stages of margin evolution. A closely related puzzle concerns the existence and importance of lithospheric detachment faults (e.g. Wernicke, 1985). These gently dipping detachment faults may offset stretching of the crust and lithospheric mantle, generating asymmetric patterns of syn- and postrift subsidence. It is generally accepted that detachment faults do not control lithospheric extension in mildly extended basins (White, 1989). However, spectacular detachment faults are sometimes imaged on rifted margins and mapped in collisional mountain belts where the original rift architecture can sometimes be inferred (e.g. Hoffmann and Reston, 1992; Pérez-Gussinyé and Reston, 2001; Hölker et al., 2003). Symmetric

A. Crosby et al. / Earth and Planetary Science Letters 273 (2008) 214–226

shear zones have also been invoked as a mechanism for extending, but not significantly melting, lithospheric mantle (Froitzheim et al., 2006). The existence and form of depth dependency have been explored in two different ways. The first approach is observation-based and relies upon making detailed measurements of upper crustal faults and comparing them with crustal thickness measurements. The extension discrepancy is reported for highly extended margins and there is much debate about its cause. One school of thought argues that it is real, which implies that the brittle upper crust barely extends at deepwater margins compared with the rest of the crust and lithospheric mantle (Davis and Kusznir, 2004). The unlikely inference is that there is a major change in the depth distribution of extension at the brittle– plastic transition. A second school of thought argues that the extension discrepancy is not real but arises from the limitations of seismic imaging of highly deformed rocks. Once stretching factors exceed about 1.7, a second generation of faulting forms which dissects the older generation. Reston (2007) has shown convincing examples similar to those observed in the field but, in general, multiple generations of faulting are exceedingly difficult to image. Moreover, 40–70% of brittle deformation occurs on scales which cannot be resolved by seismic imaging (Walsh et al., 1991; Marrett and Allmendinger, 1992). A second approach uses numerical experiments to investigate the dynamic evolution of conjugate margins. The starting point of these thermo-mechanical models is an assumed rheological structure of the continental lithosphere. Typically, the upper crust is assumed to deform by frictional failure according to Byerlee's Law while the lower crust and lithospheric mantle deform by power-law creep (e.g. Braun and Beaumont, 1987; Chéry et al., 1992; Bassi et al., 1993; Tett and Sawyer, 1996; Huismans et al., 2001; Huismans and Beaumont, 2003; Rosenbaum et al., 2005; Lavier and Manatschal 2006; Burov, 2007; Harry and Grandell, 2007). These models are intricate and it is important to bear in mind that rheological descriptions of the lithosphere are at best educated guesses. The inevitable uncertainties have led to a major debate about lithospheric strength (e.g. Jackson, 2002). Fig. 1 summarizes two end members from a recent set of numerical simulations which were designed to investigate the effects of frictional-plastic and viscous strain softening (Huismans and Beaumont, 2003). Strain softening promotes asymmetric extension of that portion of the lithosphere controlled by the dominant rheology. By increasing or decreasing strain rate, Huismans and Beaumont (2003) showed that the locus of the dominant rheology moves, promoting varying degrees of asymmetry. The first simulation was generated using an unrealistically slow extensional velocity of 0.06 cm/yr over 400 km (i.e. ε̇xx ∼ 0.002 Ma− 1). Extension is highly asymmetric and non-uniform with depth, closely resembling the lithospheric simple shear model of Wernicke (1985). In contrast, the second simulation was generated using a much faster velocity of 10 cm/yr over 400 km (i.e. ε̇xx ∼ 0.3 Ma− 1) and deformation is dominated by pure shear (McKenzie, 1978). These simulations may, or may not, be directly applicable to the Newfoundland–Iberia rift but it is clear that radically different lithospheric geometries can be obtained simply by varying the rate of deformation. Asymmetric extension is evidently an attractive way of producing tracts of exposed and largely unextended lithospheric mantle. Fig. 1a also makes a crucial point: if lithospheric mantle is largely unextended on one margin, then the opposite must be true on the conjugate margin. Thus, if the lithospheric mantle under the Iberian margin is relatively unextended, the Newfoundland margin should show evidence consistent with a highly extended lithospheric mantle (e.g. a large ratio of post-rift to syn-rift subsidence and evidence for significant volumes of decompression melting). We all seek a dynamic understanding of conjugate margin formation but existing models share three important drawbacks. First, the rheological descriptions

215

Fig. 1. Sketches of two dynamical models redrawn from numerical experiments 8 and 9 of Huismans and Beaumont (2003). In both cases, a complicated rheological stratification of the lithosphere was used which includes frictional-plastic strain softening. Model (a) was generated using a very slow strain rate and closely resembles highly depth-dependent stretching of the lithosphere (Wernicke, 1985). Note that the form of depth dependency at each margin must be of opposite sense in order to conserve mass. Model (b) was generated using a moderate strain rate and closely resembles uniform stretching of the lithosphere (McKenzie, 1978). Stippled shading indicates crust, and grey shading lithospheric mantle.

upon which they are based rely upon extrapolation of laboratory experiments by up to ten orders of magnitude. Secondly, they are very slow to run and so the more interesting inverse problem is not yet tractable. Thirdly, dynamic models are seldom concerned with fitting geologic observations in a formal sense. Instead, they are essentially ‘thought experiments’ which permit the consequences of particular rheological frameworks to be explored. In order to determine the spatial and temporal evolution of deepwater margins, we argue that a kinematic approach is more fruitful. Our strategy is threefold. An obvious starting point is that conjugate margins must be modeled simultaneously since their structural evolution should be compatible. We then suggest that the best way to identify the existence and form of depth dependency is to model subsidence histories. Although present-day crustal structure is an important constraint, it cannot reveal how strain has been partitioned between the crust and lithospheric mantle. Subsidence histories, on the other hand, contain valuable, albeit indirect, information about the distribution of thermal anomalies and thus deformation throughout the lithosphere in both space and time. Finally, we seek smooth models which minimize the misfit between observed and predicted subsidence and crustal histories. An important corollary is that no prior assumptions should be made about either the existence or the form of depth dependency. These aims are best achieved using an inverse model. We focus on the well-studied Newfoundland–Iberian conjugate margin pair, but our inverse strategy is generally applicable. 2. A general strategy Heterogeneous sedimentary records are used to calculate the water-loaded (i.e. tectonic) subsidence histories of basins and margins. For many years, these tectonic subsidence histories have been investigated using forward models which assume either instantaneous or finite-duration stretching.

216

A. Crosby et al. / Earth and Planetary Science Letters 273 (2008) 214–226

White (1993) showed that one-dimensional subsidence data can be inverted by varying lithospheric strain rate as a function of time. Strain rate inversion is easily applied to well-log information and mapped stratigraphic sections but inevitably many simplifying assumptions must be made (e.g. Airy isostasy, one-dimensional heatflow) and it provides only a partial view of lithospheric deformation. This approach was extended into two dimensions by White and Bellingham (2002) and Bellingham and White (2002) who showed that the spatial and temporal variation of lithospheric strain rate can be recovered by inverting water-loaded stratigraphic profiles. First, a starting distribution of strain rate is calculated by assuming a thermal expansion coefficient, α, of zero and by stacking a set of onedimensional solutions. The misfit between observed and calculated water-loaded subsidence is then minimized by varying the strain rate in the forward model as a function of distance along the profile and geologic time. The smoothest solution which yields the minimum misfit is found using a conjugate gradient search algorithm (Press et al., 1986). The resultant variation of strain rate through time and space shows how sedimentary basins grow and provide important dynamic insights (e.g. Jones et al., 2004). The two-dimensional algorithm makes several important assumptions. First, inherently smooth solutions, which do not model the small-scale details of the faulting process, are sought. Secondly, Bellingham and White (2002) assume that lithospheric stretching is uniform with depth (i.e. a thin-sheet approximation with ∂u/∂z = 0). This approximation is probably reasonable for mildly extended sedimentary basins where β ≤ 2. However, at highly extended deepwater margins, we cannot assume that uniform stretching applies. It is equally crucial that no particular form of depth dependency is imposed. Instead, we seek the smoothest distribution of lithospheric deformation which optimizes the fit between observed and predicted water-loaded subsidence. In this respect, our work appears to build upon the one-dimensional, depth-dependent analysis of Keen and Dehler (1993) and Dehler and Keen (1993). However, they state that their analysis does not conserve volume, which is a significant flaw. Instead we use algorithms developed and applied by Edwards (2006), Edwards et al. (in preparation) and Shillington et al. (2008) who use a spectral approach to determine the variation of lithospheric strain rate along a profile as a function of distance, depth and geologic time such that volume is automatically conserved. The cornerstone of their algorithm is a forward model which calculates subsidence and crustal thinning from the variation of strain rate. This calculation is broken down into three steps. First, the temperature history of the deforming lithosphere is calculated for a given distribution of strain rate. Using the following equation @T ¼ κ∇2 T−v  ∇ T @t

ð1Þ

which has advective and diffusive terms in x and z. Relevant parameters are defined in Table 1. Strain rate G is related to velocity by G¼

@υx @υz ¼− @x @z

ð2Þ

Eq. (1) is solved using a combined Eulerian and Lagrangian technique where the solution is decoupled from the grid movement. The appropriate boundary conditions are @v ¼ 0 at x ¼ 0; wðt Þ @z v ¼ 0 at x ¼ 0 υz ¼ 0 at z ¼ a Secondly, the temperature history is used to calculate loads as a function of time and space. Tracking the movement of major density boundaries (e.g. the base of the crust) is not trivial when ∂u/∂z ≠ 0. Edwards (2006) and Edwards et al. (in preparation) have used an

Table 1 Parameters used for inverse modeling Quantity Meaning T t v G x z w tc a κ k ρm ρc ρw

Temperature Time Velocity Strain rate Offset Depth Box width Initial crustal thickness Initial lithospheric thickness Thermal diffusivity Thermal conductivity Mantle density Crustal density Water density Model surface temperature Model base temperature Thermal expansivity Number of Fourier terms Number of spline heights Number of x-nodes where G specified Number of z-nodes where G specified Number of timesteps where G specified Misfit penalty if average stratigraphic error is 1 km Misfit penalty if average crustal thickness error is 1 km Misfit penalty if maximum dry solidus overstep is 1 °C Stratigraphic smoothing filter width

Value

See Fig. 4 See Fig. 7 8 × 10−7 m2 s− 1 3.1 W m− 1 K− 1 3350 kg m− 3 at 0 °C 2850 kg m− 3 1030 kg m− 3 0 °C 1333 °C 3.28 × 10− 5 K− 1 20 5 40 10 16 20 1 0.05 20 km

Lithospheric thickness is chosen so that the reference column is in isostatic balance with a mid-ocean ridge. Penalty components in the misfit function are cumulative. Values of the other components (which relate to strain rate smoothness and positivity) are given in Edwards (2006).

‘immersed boundary’ method which enables boundaries to be tracked at a higher resolution than is required for the thermal calculation itself. Thirdly, the load history is used to calculate the subsidence history given an initially flat Moho by assuming either Airy isostasy or by assuming that the lithosphere has a constant flexural rigidity. Since we are primarily interested in the smoothest (i.e. long-wavelength) deformation history, an Airy approximation suffices. Subsidence is calculated with respect to a reference lithospheric column which is prescribed using the parameters of Table 1. For simplicity, the crust is assumed to have a uniform composition with no internal heat generation. The kinematic forward model is simple and fast. It was designed so that subsidence profiles can be calculated for all possible forms of depth dependency. It has also been carefully bench-marked against instantaneous and finite-duration analytical solutions. Edwards (2006) has used this forward model as the basis for solving the far more challenging inverse problem. Strain rate inversion is relatively straightforward when uniform stretching of the lithosphere is assumed. In general, however, strain rate, G(x, z, t), varies as a function of x, the distance along the profile, z, the depth through the lithosphere, and t, geologic time. The principal difficulty is that generalized strain rate distributions produce velocity fields which do not necessarily satisfy the boundary conditions. Thus G(x, z, t) must be parameterized in such a way which ensures that mass conservation is obeyed and that boundary conditions are satisfied. Edwards (2006) has shown that these conditions are automatically fulfilled provided G(x, z, t) is recast in spectral form. Thus the starting distribution of G(x, z, t) can be rewritten as a set of Fourier series. The coefficients of this set of periodic functions are then systematically varied until the misfit between observed and calculated subsidence is minimized. Linear splines are used for interpolation and inverse modeling is subject to the usual smoothness and positivity constraints. We are conscious that more sophisticated kinematic and dynamic descriptions of lithospheric stretching exist. However, such descriptions are slow to run and cannot yet be used as the basis of inversion. We are concerned with solving the inverse problem which requires the forward model to be run many thousands of times. Our forward model contains

A. Crosby et al. / Earth and Planetary Science Letters 273 (2008) 214–226

217

Fig. 2. Shaded relief bathymetric map of the North Atlantic Ocean between Newfoundland and Iberia overlain by free-air gravity anomalies which were low-pass filtered to exclude wavelengths shorter than ∼700 km (GRACE satellite-only gravity model GGM02). These anomalies show that there is a difference of ∼40 mGal between each side. If the admittance, Z, is ∼ 30 mGal/km, then the difference in dynamic support is ∼750 m which reflects the spatial pattern of convective circulation beneath the lithospheric plate (e.g. (Crosby et al., 2006) and references therein). Thick annotated black lines show the wide-angle seismic lines discussed in text; medium black lines indicate plate boundaries; thin annotated black lines delineate traces of the M0 (i.e. 121 Ma) magnetic anomaly. The inset shows line locations on a reconstruction of the Iberia–Newfoundland conjugate margins at M0 time (after Srivastava et al. (2000)).

the essential features needed to describe general depth-dependent deformation along a two-dimensional profile. It was carefully designed so that it runs in less than about a second. Optimization is carried out using Powell's algorithm (Press et al., 1986); typically, 2–4 × 105 iterations of the forward model are required for convergence, which takes approximately 24–48 h of CPU time on a 2.8 GHz Intel P4 Dual Core PC running a Linux operating system with an open source compiler. As before, the resulting strain rate distributions are smooth on timescales of 1–5 Ma and on length scales of 10–20 km. Stratigraphic profiles are usually smoothed beforehand. Edwards (2006) applied the depth-dependent strain rate algorithm to a number of sedimentary basins (including the North Sea and Gulf of Suez) and well-sedimented passive margins (the northern margin of the South China Sea and the Black Sea.). His results show that these basins and margins exhibit modest depth dependency. However, before applying this algorithm to the Newfound-Iberian conjugate margins, we must first reconstruct water-loaded stratigraphic profiles. 2.1. Paleobathymetry at deep-water margins Sedimentary basins and passive margins are loaded by a combination of water and sedimentary rocks of variable density. In onedimensional analyses, this heterogeneous load is converted into waterloaded subsidence of underlying basement rocks, Sw. At a given time,  Sw ¼ Ss

ρm −ρs ρm −ρw

 þ dw

ð3Þ

where Ss is the thickness of a heterogeneous sedimentary column after decompaction, P ρs is the mean density of this column which is calculated using the method of Sclater and Christie (1980), and dw is the water depth at the time of deposition (i.e. paleobathymetry). In two dimensions, it is more convenient to model water-loaded stratigraphy rather than basement subsidence. Otherwise, two-dimensional subsidence analysis is similar (i.e. stratigraphic horizons are decompacted

and water-loaded by the standard methods). Bellingham and White (2002) showed that uncertainties in compaction parameters have little effect on the water-loaded stratigraphy unless clastic and carbonate rocks are interchanged. The greatest source of uncertainty stems from our sketchy understanding of paleobathymetry. In mildly extended sedimentary basins, there is considerable debate about paleobathymetric constraints. In the North Sea basin, the depositional water depth of chalk remains uncertain and estimates range from 50–800 m. In contrast, platform carbonate deposits form in shallow water, although a conservative range of 0–50 m is often used. Similar problems afflict the paleobathymetry of clastic rocks. In coastal shelf deposits, there are excellent paleontological and sedimentological indicators which can be used to demonstrate gradual increases in water depth from nearshore to the shelf edge. However, once water depths exceed several hundred meters, fine-grained sedimentary rocks generally do not contain sensitive paleobathymetric indicators and it is difficult to determine water depth to within one order of magnitude. An improved understanding of the depth ranges of benthic foraminifera would be of considerable help. These problems are exacerbated at continental margins where the highly extended distal portions are often sediment-starved, poorly drilled, and well below the range of depths where sedimentological and micropaleontological information are diagnostic of paleobathymetry. We suspect that this apparently insurmountable problem has led many to neglect subsidence records of highly extended margins. We acknowledge these concerns, but given that the only temporal information about margin evolution is contained in this sedimentary record it is important to devise a strategy. We have reconstructed the water-loaded subsidence history of margins by applying three simple and transparent rules. The first rule is obvious: we know the presentday water depth across a margin. It is also reasonable to assume that the water depth was negligible at the start of continental rifting. In the absence of any other information, we could assume that paleobathymetry simply grades linearly through time. Fortunately, we know how water-loaded oceanic lithosphere subsides from global depth-age studies (e.g. Parsons and Sclater, 1977; Crosby et al., 2006). This second rule gives

218

A. Crosby et al. / Earth and Planetary Science Letters 273 (2008) 214–226

us a vital ‘pin’ at the distal end of a continental margin where paleobathymetric estimates are poorest. The third rule concerns the geometry of sedimentary infill. If paleobathymetry varied smoothly with time, sedimentary packages will consist of parallel-bedded strata as seen on the Newfoundland–Iberian margins. In contrast, rapid changes in water depth due to sediment influx are expected to be manifest by development of easily identifiable clinoformal packages. These three rules provide important (and often the only) constraints on paleobathymetric evolution at margins, which subsequently evolve into oceanic basins. We have adapted this general strategy to reconstruct the paleobathymetry of the Newfoundland–Iberian conjugate margins (Figs. 2–4). First, an estimate of the present-day dynamic topography was subtracted from the stratigraphic data. Paleobathymetry of the ocean–continent boundary since initiation of seafloor spreading was calculated using a combination of Eq. (3) and a thermal plate model. Finally, we assumed that the scaling function (i.e. water depth as a fraction of present-day bathymetry) is constant across the margin. This assumption is reasonable since there is no evidence for periods of rapid infill. In this study, mapped stratigraphic horizons above the basement pick are younger than both the oldest ocean crust and the most significant period of stretching, so it is straightforward to configure the paleobathymetry of each horizon (Table 2). If older stratigraphic horizons existed, it would be necessary to specify paleobathymetry in some other way and to correct for layer strain. The paleobathymetric scheme shown in Fig. 5 is simple and transparent. It is also corroborated by water depth estimates which were inferred from nearby borehole information (e.g. Shipboard Scientific Party, 2004; Péron-Pinvidic et al., 2007). Since we model the water-loaded stratigraphy using an inverse strategy, it is straightforward to investigate how paleobathymetric uncertainties affect our conclusions. The crucial point is that paleobathymetry is estimated independently rather than determined as part of the modeling process.

3. Application The Edwards (2006) algorithm was initially applied to intracontinental sedimentary basins. In order to model highly extended margins, a number of modifications are needed. The forward calculation, which is the cornerstone of the inverse algorithm, is solved spectrally and to avoid high frequency ringing, it is important to model conjugate profiles or profiles which have been reflected about the ocean–continent boundary. At the landward margin, we have added gradual tapers to incomplete stratigraphic profiles. Normally, the spatial and temporal variation of strain rate is calculated from water-loaded stratigraphy alone. We have formally included the misfit between observed and predicted crustal thickness as an additional constraint since the variation of crustal thickness across Newfoundland–Iberian margins is accurately known. The algorithm does not include decompression melting and we do not model the growth of oceanic crust. Since we are modeling ‘cold’ margins, which are characterized by the absence of significant volumes of basaltic magmatism, it is important that we ensure that the temperature of the lithosphere–asthenosphere boundary does not significantly overstep the solidus for dry peridotite (McKenzie and. Bickle, 1988). If overstep occurs during inversion, an appropriate penalty is added to the misfit function (see Table 1 and Edwards et al. (in preparation) for details). For comparative purposes, we have also run each inversion without a melting constraint. 3.1. Crustal structure and rift duration The Newfoundland–Iberia conjugate margins have been carefully studied by acquisition programs over the last 10–15 years, which include deep-reflection profiles, wide-angle seismic surveys and deep-water drilling (Péron-Pinvidic et al., 2007). Here we model five

Fig. 3. Summary of crustal structure along conjugate SCREECH and Iberian lines, redrawn from wide-angle models (SCREECH 1: Funck et al. (2003); SCREECH 2: Van Avendonk et al. (2006); SCREECH 3: Lau et al. (2006a); IAM-9: Dean et al. (2000); and GP-101: Whitmarsh et al. (1996), Zelt et al. (2003)). Red shading indicates continental crust; yellow shading indicates transitional crust; blue shading indicates bona fide oceanic crust; stippled shading indicates syn-rift and post-rift sedimentation; vertical lines delineate regions where stratigraphic horizons could be digitized from published seismic reflection interpretations (the landward line marks the limit of the data, the seaward line is the estimated continent– ocean transition, see Table 2 and text for details); finally, thin dashed lines show the crustal thickness variation predicted by inverse modeling of water-loaded subsidence record.

A. Crosby et al. / Earth and Planetary Science Letters 273 (2008) 214–226

219

Fig. 4. Total water-loaded subsidence plotted against stretching factor, β, for each of lines shown in Fig. 1. In each case, estimates were made from wide- angle models and dynamic topography was subtracted (see text). Red points are thinned continental crust; yellow points are transitional crust; and blue points are bona fide oceanic crust, which was included for comparison purposes only. The solid black line is the Airy isostatic prediction of total water-loaded subsidence as a function of β using a reference crust with average density of 2850 kg m− 3 and thickness as indicated (see expressions of McKenzie (1978) and parameter values of Table 1). The dashed black line is the Airy isostatic prediction, including the effect of a 5 km thick layer of serpentinization which reduces the average density of sub-Moho mantle by 150 kg m− 3. In each case, the inclusion of serpentinization improves the fit. Small deviations remain, which are probably consequence of pre-existing Moho topography, small but finite flexural rigidities and uncertainties in wide-angle models, particularly where crustal thickness gradients are steep. These deviations can impede our ability to simultaneously fit stratigraphy and crustal thicknesses. Finally, the dotted line in (e) is the Airy isostatic prediction when a more realistic reference crustal thickness of 36 km is used. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

transects whose evolution is much debated (Figs. 2 and 3). It is generally accepted that at least two phases of rifting occurred (Tucholke et al., 1989). The older phase of rifting occurred in PermoTriassic times and is largely confined to onshore sedimentary basins which fringe both continental margins. It is commonly observed on many passive margins that an early phase of moderate rifting ends, giving way to a later phase which steps seaward; the Newfoundland– Iberian conjugate margins are no exception. This later phase began in Late Jurassic to Early Cretaceous times (150–135 Ma) and it is of far greater significance since it gave rise to break-up and initially very slow seafloor spreading by ∼ 127 Ma (i.e. magnetic anomalies M0–M4 (Russell and Whitmarsh, 2003; Lau et al., 2006a). It is this rifting phase which has caused so much controversy and which we model in the following sections. The structure of the crust has been constrained by at least five modern wide-angle seismic experiments and we summarize the results in Figs. 3 and 4. On the Newfoundland margin, the SCREECH experiments carried out in 2000 yielded three high-quality profiles (Funck et al., 2003; Van Avendonk et al., 2006; Lau et al., 2006a). On the Iberian margin, profile IAM-9, which is conjugate to SCREECH 2, was acquired in 1995 (Dean et al., 2000). A wide-angle seismic profile, coincident with deep-

reflection profile GP101 (Mauffret and Montadert, 1988) and conjugate to SCREECH 1, was acquired in 1988 (Whitmarsh et al., 1996) although a more recent coincident wide-angle survey also exists Zelt et al. (2003). The maximum crustal thickness is about 31 km landward of SCREECH 1 and GP101, 33 km landward of SCREECH 2 and IAM-9, and about 35 km landward of SCREECH 3. Unstretched crustal thicknesses beneath Newfoundland and Iberia are about 35–40 km (Marillier et al., 1994) and 28–33 km (Paulssen and Visser, 1993; Pedreira et al., 2003; Fernández-Viejo et al. 1998), respectively. Continental crust on both margins has been dramatically and asymmetrically thinned over Table 2 Regional seismic stratigraphic horizons fitted by inverse modeling (after Péron-Pinvidic et al. (2007) and Shipboard Scientific Party (2004)) Horizon

Boundary

Assigned age (Ma)

Basement Top-A Top-B Top-C Seabed

Latest Jurassic Aptian–Albian Turonian–Coniacian Eocene–Oligocene Present day

150 112 90 34 0

Common name

U Au

220

A. Crosby et al. / Earth and Planetary Science Letters 273 (2008) 214–226

Fig. 5. The paleobathymetry scaling function for each horizon in Table 2, calibrated using the known water-loaded subsidence of the oldest ocean floor. Grey lines are illustrative since water-loaded stratigraphy is only specified at the ages marked by black circles. The palaeo-water depth almost certainly increased more rapidly given the low syn-rift sediment thickness and high initial strain rates.

distances of 50–100 km. There is excellent evidence for pervasive normal faulting, especially on the Newfoundland side where the ocean– continent transition is consequently indistinct. Sometimes, notably at ODP sites 897, 899 (20 km north of IAM-9) and 1277 (at the ocean– continent transition of SCREECH 2), crustal thinning is complete and serpentinized peridotites are exposed at the seafloor. The P-wave velocity of the top 5 km of lithospheric mantle directly beneath highly thinned crust is often 0.5–1 km s− 1 slower than normal, which is consistent with hydrothermal alteration promoted by pervasive faulting. Minor amounts of alkaline magmatism have been drilled on both margins but, crucially, there is scant evidence for voluminous basaltic volcanism typical of margins in the North Atlantic Ocean (e.g. seaward-dipping reflections, magmatic underplating), even though stretching factors exceed 3–4 over wide tracts (Jagoutz et al., 2007). Possible exceptions include portions of the Iberian margin where Russell and Whitmarsh (2003) have interpreted isolated magnetic anomalies as syn-rift igneous intrusions. In contrast, Sibuet et al. (2007) argue that linear magnetic anomalies M3– M17 (i.e. as old as 140 Ma) were acquired during amagmatic seafloor spreading and serpentinization. 3.2. Dynamic topography Several unexpected problems arose when we began to analyze the conjugate margins. The first of these problems concerns the difference in dynamic topographic support on either side of the Atlantic Ocean. Dynamic topography is the long-wavelength component of topography which cannot be accounted for by crustal and/or lithospheric thickness variations but is generated by convective circulation beneath plates. It is difficult to measure accurately, although intermediate (i.e. 500–4000 km) wavelength free-air gravity anomalies are a useful guide. In the North Atlantic Ocean, there is a significant longwavelength positive gravity anomaly over the Newfoundland margin and small negative anomalies over the relevant part of the Iberian margin (Fig. 2). In the oceans, the ratio of the free-air gravity anomaly to the dynamic seafloor topography is 25–40 mGal km–1 (Crosby et al., 2006). Thus we estimate 0.6–1.2 km of dynamic uplift on the Newfoundland side and 0.25–0.5 km of dynamic subsidence on the Iberian side. This differential support is consistent with the shift required to align the two pairs of conjugate water-loaded stratigraphic profiles at the ocean–continent boundary (−0.6 km for SCREECH 2, +0.55 km for IAM-9; −0.6 km for SCREECH 1, +0.35 km for GP101). These values are broadly consistent with the subsidence of oceanic lithosphere which abuts both margins. The mean water-loaded depth of unperturbed ∼ 125 Ma oceanic lithosphere is 5.6 ± 0.2 km when the

mean crustal thickness is 7.1 km (Crosby et al., 2006; White et al., 1992). After applying an isostatic correction for unusually thin oceanic crust, we calculated the water-loaded subsidence of the oldest oceanic lithosphere and found that Newfoundland oceanic lithosphere is 0.6– 1.0 km shallower than expected, while Iberian oceanic lithosphere at the end of Line IAM-9 is 0.5–1.0 km deeper. Oceanic lithosphere at the end of GP101 is 0.2–0.6 km deeper but crustal thickness is poorly constrained. Our three different estimates of dynamic support are similar but not identical. For simplicity, we have chosen the correction required to match the two pairs of conjugate water-loaded stratigraphic profiles. SCREECH 3 has no conjugate profile and was shifted by the same amount as SCREECH 2. We note that our estimates disagree in both sign and magnitude with those of Conrad et al. (2004), who calculated 100–300 m of subsidence for the Newfoundland side and 200–300 m of uplift for the Iberian side. The calculation of dynamic topography is fraught with uncertainties and for the purposes of this paper we prefer to use direct observations. 4. Serpentinization Differential elevation across a margin can also be generated by intrinsic changes within the mantle lithospheric column. The most obvious source of elevation change is serpentinization of lithospheric mantle rocks which has occurred on both margins. Reduction in the average density of a thickness of mantle, ts, by an amount Δρ causes an isostatic reduction in water-loaded subsidence, Δsw,where Δsw ¼

ts Δρ ρm −ρw

ð4Þ

The age of serpentinization is not known although it is likely to have occurred within several million years of rifting (e.g. Sibuet et al., 2007; Skelton and Valley, 2000). It is easiest to correct for serpentinization before inverse modeling. A practical method for estimating this correction is to plot the total water-loaded subsidence as a function of crustal stretching (Fig. 4). On thermally and isostatically equilibrated margins, we expect a simple relationship between total water-loaded subsidence and crustal thickness. Visual inspection demonstrates that highly extended continental and transitional crust have generally subsided less than expected (Fig. 4). This difference is consistently 0.3 km of water-loaded subsidence and cannot be caused by differential stretching of the lithosphere which changes the ratio of syn-rift to postrift subsidence but not the total subsidence. It is easily explained by decreasing the density of the lithospheric mantle. If Δρ is −150 kg m− 3 (Escartín et al., 2001) and if the average P-wave velocity of the anomalous layer is 7.5 km s− 1, we obtain ts ∼ 5 km for crustal thinning factors of greater than 4–6, which is consistent with wide-angle models. Fig. 4 shows that serpentinization is not required when crustal thinning factors are smaller than 3 or 4. Deviations from an isostatic relationship at lower stretching factors are harder to explain (e.g. SCREECH 3; Fig. 4e). We suspect that they partly arise from flexural support of short-wavelength, fault-bounded blocks in the upper crust and pre-existing Moho topography, from short-wavelength variations in crustal density, and from the inability of ray-based seismic algorithms to model steep velocity gradients. Either way, these deviations present a problem when we simultaneously fit waterloaded stratigraphic profiles and crustal thicknesses. The biggest problem occurred with SCREECH 3; if we use a realistic unextended crustal thickness (36 km), then the residual misfit for moderately stretched crust is large. One explanation for this discrepancy is a change in lithosphere thickness across the line (Priestley and McKenzie, 2006). As a result, we are forced to choose a reference crustal thickness of 30 km, which is not strictly correct even though it reconciles observed crustal thinning and water-loaded stratigraphy when β N 3. On most profiles, water-loaded subsidence of oceanic

A. Crosby et al. / Earth and Planetary Science Letters 273 (2008) 214–226

lithosphere is consistent with measured crustal thickness. The obvious exception is SCREECH 1 where water-loaded subsidence is about 0.6 km smaller than expected because serpentinization has penetrated beneath oceanic crust where anomalously slow velocities are observed (Funck et al., 2003). 4.1. Stratigraphic records The temporal and spatial evolution of a stretched margin is partially preserved by the stratigraphic record. The goal of our analysis is to use this incomplete stratigraphic record to reconstruct the waterloaded subsidence of a deep-water continental margin, which is then

221

used to recover the temporal and spatial history of deformation. The difficulties of reconstruction appear formidable but are tractable if broken down into a series of transparent steps. The regional seismic stratigraphy is summarized by Péron-Pinvidic et al. (2007). We have exploited their internally consistent stratigraphic records which are based upon a set of mappable seismic reflections whose ages are known from nearby boreholes (Table 2). The age of the base of this stratigraphic record is uncertain and we have chosen the oldest likely age (150 Ma). On both margins, the stratigraphic record forms sub-parallel packages and there is no evidence for the development of clinoformal geometries which suggests that paleobathymetry evolved smoothly with age. Digitized stratigraphic horizons were taken

Fig. 6. Observed and predicted water-loaded subsidence for the five lines located on Fig. 2. (a). Conjugate line pair SCREECH 1 and GP 101 which join at the continent–ocean boundary indicated by a vertical grey bar. Thin black lines show water-loaded subsidence calculated from observed stratigraphy and corrected for changing paleobathymetry, for serpenitinization, and for dynamic topography (see text for details); open circles and dashed lines show water-loaded subsidence calculated using the strain rate distribution shown in Fig. 7a and b and obtained by inverse modeling. (b) Conjugate line pair SCREECH 2 and IAM-9. (c) SCREECH 3 does not have a conjugate and was modeled by reflecting data about the continent–ocean boundary. Note the quality of fit between theory and reconstructed subsidence in all three cases.

222

A. Crosby et al. / Earth and Planetary Science Letters 273 (2008) 214–226

from the following publications: SCREECH 1: Hopper et al. (2004); SCREECH 2: Shillington et al. (2006); SCREECH 3: Lau et al. (2006b); IAM-9: Dean et al., 2000; GP-101: Whitmarsh et al. (1996) and Mauffret and Montadert (1988). Where appropriate, we converted two-way travel time into depth using interval velocities from the relevant wideangle model. Internally consistent stratigraphic records were available for the central portions of the profiles 3. In each case, we calculated the water-loaded stratigraphy by decompacting and backstripping the observed stratigraphic record in the standard way (Eq. (2); Sclater and Christie (1980)). For simplicity, we assumed that sediment composition varied linearly from 100% sand at the proximal end of a profile to 100% shale at the distal end. Previous analysis has demonstrated that large changes in sediment composition have minor effects. Our chosen paleobathymetry is

summarized in Fig. 5. The starting water-loaded depth of the oldest oceanic lithosphere is 3 km, which is deeper than average because oceanic crust here is unusually thin (Lecroart et al., 1997). The bestfitting thermal plate thickness is about 100 km and our resultant paleobathymetry exceeds 80% of its present-day value by ∼110 Ma. Our paleobathymetric estimates are not easy to corroborate but we note that deep-water benthic foraminifera recovered at ODP Site 1276 imply that water depths at the SCREECH 2 ocean–continent transition exceeded 2 km since Late Aptian–Early Albian times (115–110 Ma) (Shipboard Scientific Party, 2004). Where they have been drilled, postrift carbonate rocks have evidently been deposited below the calcite compensation depth (Péron-Pinvidic et al., 2007). Furthermore, sediments at Galicia Bank ODP sites 901, 1065 and 1069 of Tithonian age appear to have been deposited at moderate water depths, which

Fig. 7. Strain rate functions which correspond to the best-fitting water-loaded subsidence shown in Fig. 6. (a) Contoured strain rate distribution at the surface of the lithosphere plotted as a function of range, x, and age, t, for SCREECH 1 and GP-101 conjugate pair. Black crosses show positions in x–t where the strain rate function is calculated during inversion and their displacement is a measure of strain; white areas indicate masking of artificial tapering of subsidence profile at either end of conjugate pair; the solid line is the continent– ocean boundary. (b) Contoured strain rate distribution plotted as a function of range, x, and depth, z, for first time step when strain rates are greatest. (c) and (d) SCREECH 2 and IAM-9 conjugate pair. (e) and (f) SCREECH 3. Note the form of depth dependency which concentrates highest strain rates within the crust toward the continent–ocean boundary. The width of the corresponding regions of stretched mantle under less-stretched crust is a function of the width of the box in the model.

A. Crosby et al. / Earth and Planetary Science Letters 273 (2008) 214–226

223

implies rapid deepening immediately after the start of rifting. The water at these sites then continued to deepen in Berriasian and Barremian time (145–125 Ma) (Tucholke and Sibuet, 2007). 5. Inverse modeling Fig. 6 shows the reconstructed water-loaded stratigraphies for both pairs of conjugate margins and for the SCREECH 3 profile. In each case, total subsidence is consistent with the water-loaded estimates summarized in Fig. 4. Each stratigraphic profile was inverted using the Edwards (2006) algorithm. We sought the smoothest distribution of lithospheric strain rate as a function of distance, depth and time, G(x, z, t), which minimized the misfit between reconstructed and calculated stratigraphies and between observed and calculated crustal thicknesses. The calculated stratigraphies and crustal thickness values are shown in Figs. 3 and 6, respectively. Fits are generally excellent, yielding χ2 values of less than 2. The corresponding strain rate histories are shown in Fig. 7 where we show two slices through each G(x, z, t) cube. In each case, the x–t slices show that a period of rapid strain rate (∼ 0.3 Ma− 1) occurred between 150 and 140 Ma. Strain rate values increase towards the deeper parts of each margin. Since syn-rift sediments are thin, we expect the water to have deepened rapidly after the start of rifting. After ∼ 140 Ma, strain rates are much diminished and consistent with the growth of unfaulted thermal subsidence. The x–z slices are more interesting since they are directly pertinent to the debate about depthdependent stretching. Depth dependency takes two forms. Along the mildly extended portions of the margins, lithospheric mantle has been extended over a wider region than the crust. We have obtained similar results elsewhere. Higher strain rates within the lithospheric mantle are a direct consequence of a lower than expected ratio of syn-rift to postrift subsidence. As the lithosphere becomes increasingly stretched, we pass into a narrow zone, usually about 25–30 km wide, where strain rate is approximately uniform with depth. At the distal end of each margin, there is a significant switch in the form of depth dependency and crustal strain rates are much higher than those in the lithospheric mantle (contra Dehler and Keen (1993)). This distribution is supported by three arguments. First, the reconstructed ratio of syn-rift to postrift subsidence is now significantly greater than expected for uniform stretching. This change is obviously contingent upon the chosen paleobathymetric configuration. Secondly, the combination of a highly extended crust and a lack of syn-rift magmatism must force strain to be concentrated within the crust. At depth, the lithospheric mantle strains at less than 0.1 Ma− 1, which suppresses adiabatic decompression and melting of the asthenosphere. Note that small amounts of enriched melting are easily generated by decompressing fusible material trapped against the wet solidus in the lower half of the lithospheric mantle. Thirdly, isostasy dictates that the crust cannot have extended as slowly as lithospheric mantle since we know that water depths increased very rapidly during syn-rift times; in any case there is a short window of time available for rifting prior to sea-floor spreading. The cumulative extension for each line is shown in Fig. 8. In each case, although extension in the lithospheric mantle is clearly distributed over a wider region, the total horizontal extension accumulated within the crust and within the lithospheric mantle are in good agreement. A supplementary figure shows the resultant strain rate functions when the misfit function is not penalized for overstepping the solidus of dry peridotite. As before, subsidence and crustal thinning observations are well matched but peak strain rates are higher and there is substantially less depth-dependent stretching (i.e. crustal and lithospheric mantle extension factors are more similar). Under highly stretched crust, however, the dry solidus has been over-stepped by 30– 160 °C for 10–20 km thickness of lithosphere. This overstep will generate substantial volumes of basaltic magmatism which have not

Fig. 8. Total model extension factors (i.e. 1 − 1 / β and 1 − 1 / δ) for (a) SCREECH 1/GP-101, (b) SCREECH 2/IAM-9, and (c) SCREECH 3. Red lines refer to the crust and blue lines refer to the original lithospheric mantle. The areas under the two curves are constrained to be equal. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

been observed. Thus the depth-dependent stretching observed in Fig. 7 is a direct consequence of restricting decompression melting. Are strain rate patterns well resolved? In order to investigate the extent to which strain rate variation is sensitive to small uncertainties in the water-loaded stratigraphy, we have inverted fifty perturbed versions of the SCREECH 2/IAM-9 conjugate lines and analyzed the variance of the resultant strain rate distributions. Reassuringly, Fig. 9 shows that the variation of strain rate with depth and distance is insensitive to these perturbations (the standard deviation of the strain rate population is everywhere less that 0.05 Ma− 1). As expected, sensitivity increases with depth. By inference, Fig. 9 also shows that the choice of initial smoothing is unimportant. We acknowledge that systematic errors and/or asymmetries in the reconstructed water-loaded subsidence, which are not ruled out by the sparse geologic data, could affect our inferred strain rate histories in important ways. We simply point out that our reconstructions have the virtue of simplicity. 6. Discussion Do these strain rate distributions yield ‘balanced’ models? The Edwards (2006) algorithm automatically conserves mass at all times and so the total amount of strain taken up within the crust is identical to that taken up by the lithospheric mantle (Fig. 8). Thus

∫

þ∞ 

−∞

1−

 1 dx β

ð5Þ

is conserved and equals the total amount of extension. It has often been argued that the total extension across basins and margins

224

A. Crosby et al. / Earth and Planetary Science Letters 273 (2008) 214–226

Fig. 9. Bootstrap (i.e. Monte Carlo) modeling of water-loaded subsidence for SCREECH 2 and IAM-9 conjugate pair, with pseudo-topographic perturbations generated using a fractal method similar to that described by Turcotte (1999). (a) 50 perturbed versions of water-loaded subsidence with maximum perturbation of ±0.5 km. Perturbations decrease smoothly to zero at distal end of each margin where paleobathymetry can be accurately determined from global age-depth models (Crosby et al., 2006). The grey vertical bar is the continent– ocean boundary. The crustal thickness was also perturbed in a similar way. (b) Strain rate map in x–z space. Solid blue lines are the 0.15 Ma− 1 strain rate contours obtained from the 50 inversions; red shading defines the region within which the mean strain rate exceeds 0.15 Ma− 1. The gross form of strain rate function is robust and relatively insensitive to stratigraphic perturbations. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

determined from subsidence and crustal observations greatly exceeds the horizontal extension determined from normal faulting (e.g. Davis and Kusznir (2004)). If correct, this argument has dramatic consequences for the form of depth dependency particularly at highly extended margins where the extension discrepancy is thought to be particularly large. Fig. 10 shows fault interpretations of the five profiles. Simplified versions of published interpretations are shown apart from SCREECH 2 and 3 which we interpreted ourselves. The minimum amount of horizontal extension along each profile can be

gauged by summing heaves (i.e. horizontal components of displacement) across all visible faults. This simple method ignores unresolved internal deformation and multiple generations of faulting and is probably a gross underestimate. A better estimate can be made if we assume that 50% of the extensional deformation occurs on small-scale faulting which cannot be resolved (40–70% is thought to be a reasonable range (Walsh et al., 1991; Marrett and Allmendinger, 1992)). Our results are summarized in Table 3. Apart from SCREECH 1, we find good agreement between the predicted and observed

Fig. 10. Sketched interpretations of the five profiles (see Fig. 2 for location). Red, yellow and blue shading indicate stretched continental crust, transitional crust, and oceanic crust respectively; grey shading is syn-rift subsidence with demonstrable fault-bounded stratigraphic growth; stippled shading is unfaulted post-rift subsidence; solid and dashed vertical lines are on-line and projected Ocean Drilling Program (ODP) drill holes; thin sub-vertical lines are mappable normal faults (dashed lines delineate regions where faults could not be confidently identified). The total amounts of horizontal extension across each conjugate pair are given in Table 3. (a) SCREECH 1 and GP-101 conjugate pair. Fault interpretations redrawn from (Hopper et al., 2004) and (Whitmarsh et al., 1996). (b) SCREECH 2 and IAM-9 conjugate pair. Fault interpretation for IAM-9 redrawn from (Dean et al., 2000). Note difficulty in interpreting faults along portions of SCREECH 2. (c) SCREECH 3 has clearly identifiable faulted syn-rift packages and is straightforward to interpret.

A. Crosby et al. / Earth and Planetary Science Letters 273 (2008) 214–226 Table 3 Estimates of upper crustal extension determined from summation of fault heaves Line

Length, km

Fault extension, km

Model extension, km

SCREECH 1 SCREECH 2 SCREECH 3 GP-101 IAM-9

50 150 289 105 208

13–26 37–74 126–251 56–112 109–218

45 65 160 75 200

Lower bounds are calculated by assuming that all faulting is visible; upper bounds are calculated by assuming that only 50% is visible (e.g. Walsh et al. (1991) and Marrett and Allmendinger (1992)).

horizontal extension. Although our line of argument is different, we share Reston's (2007) view that extension discrepancies at highly extended margins are a consequence of poor acoustic imaging. Some zones of dramatically extended upper crust where closely spaced listric faults detach into serpentinized mantle probably owe more to large-scale landsliding than to crustal extension (e.g. Clark et al. (2007)). We also suggest that multiple generations of normal faulting are sufficient to locally exhume mantle in regions where the crust has been most thinned. Within the lithospheric mantle, it is difficult to determine whether stretching occurs by plastic creep or whether it is accommodated on sets of discrete shear zones (Tucholke et al., 1989). Do our results shed any light on the debate about lithospheric strength? Does a strong crust or a strong lithospheric mantle govern margin evolution? Temperature is widely regarded as the key parameter. The rheologic behavior of a material is principally controlled by τ, the ratio of its temperature to its melting temperature, both measured in Kelvin. We suggest that the evolution of the highly extended Newfoundland–Iberia margins must be primarily controlled by the strength of lithospheric mantle. A key observation is the existence of wide tracts of extremely thin and pervasively faulted crust which overly lithospheric mantle which stretched very slowly because it did not produce substantial volumes of decompression melt prior to oceanic basin formation. If rapid stretching was mostly focussed within the crust, the temperature of the sub-Moho mantle must have decreased rapidly. Serpentinization of ∼ 5 km of lithospheric mantle is unlikely to have played a significant limiting role since 25 km of lithospheric mantle cools by ∼ 250 °C in our best-fit models. Finally, the particular form of depth dependency recovered by inverse modeling does depend upon the geological interpretation of transition zone crust. If the transition zone consists of highly extended continental crust and if syn-rift magmatism is largely absent then the recovered form of depth dependency is a logical consequence, given the short interval of time between initiation of continental stretching and seafloor spreading: the crust must have been rapidly extended and the lithospheric mantle must have either not extended or extended isothermally (Minshull et al., 2001). This conclusion is insensitive to the choice of starting lithospheric template. If the transition zone consists of serpentinized continental mantle, a similar form of depth dependency is required since the crust must have been excised. Finally, if the transition zone consists of exhumed mantle with variable amounts of intruded gabbro and basalt, which is formed by ultra-slow seafloor spreading, then the Newfoundland–Iberian margins could have been formed by largely uniform stretching. 7. Conclusions The temporal and spatial distribution of lithospheric extension beneath the Newfoundland–Iberian conjugate margins is much debated. We have tackled this problem in two stages. First, we have estimated the water-loaded stratigraphy along each margin. The problem of paleobathymetric variation through time has been tackled by adopting a simple and smooth model which is tied to the subsidence history of adjacent oceanic lithosphere. Secondly, we have fitted

225

stratigraphic and crustal measurements by varying lithospheric strain rate as a function of distance, depth and time with the condition that the dry peridotite solidus is not substantially exceeded. Our inverse algorithm allows for depth-dependent stretching but does not prescribe its existence or form. Excellent fits have been obtained and the resultant distributions of strain rate suggest that the style of depth dependency changes as lithospheric extension increases. At the proximal end of each margin, the highest strain rates occur in the lithospheric mantle. Toward the distal, more highly stretched, end of each margin, the style of deformation switches and the highest strain rates are found in the crust. Our results conserve mass and preclude gross lithospheric asymmetry. The total horizontal extension across each margin is in good agreement with estimates of upper crustal extension obtained from summation of fault heaves on seismic reflection profiles. Acknowledgements This research is funded by BP Exploration and forms part of the BPCambridge Margins Research Programme. We are grateful to P. Carragher, R. Corfield, R. Gibson, S. Jones, H. Lau, D. Lyness, L. Mackay, S. Matthews, M. Thompson and R. White for their help. C. Jaupart, J. Hopper and two anonymous reviewers provided thorough and helpful reviews. Department of Earth Sciences Contribution Number 9278. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.epsl.2008.06.039.

References Bassi, G., Keen, C.E., Potter, P., 1993. Contrasting styles of rifting: models and examples from the eastern Canadian margin. Tectonics 12, 639–655. Bellingham, P., White, N., 2002. A two-dimensional inverse model for extensional sedimentary basins 2. Application. J. Geophys. Res.107, 2260. doi:10.1029/2001JB000174. Braun, J., Beaumont, C., 1987. Styles of continental rifting: results from dynamic models of lithosphere extension. In: Beaumont, C., Tankard, A.J. (Eds.), Sedimentary Basins and Basin-Forming Mechanisms. Mem. Can. Soc. Pet. Geol., vol. 12, pp. 241–258. Burov, E., 2007. The role of gravitational instabilities, density structure and extension rate in the evolution of continental margins. In: Karner, G.D., Manatschal, G., Pinheiro, L.M. (Eds.), Imaging, Mapping and Modelling Continental Lithosphere Extension and Breakup. Geol. Soc. Lond. Spec. Pub., vol. 282, pp. 139–156. Chéry, J.F., Lucazeau, M., Daigniéres, M., Vilotte, J.P., 1992. Large uplift of rift flanks: a genetic link with lithospheric rigidity? Earth Planet. Sci. Lett. 112, 195–211. Clark, S.A., Sawyer, D.S., Austin Jr., J.A., Christeson, G.L., Nakamura, Y., 2007. Characterizing the Galicia Bank-Southern Iberia Abyssal Plain rifted margin segment boundary using multichannel seismic and ocean bottom seismometer data. J. Geophys. Res. 112, B03408. doi:10.1029/2006JB004581. Conrad, C.P., Lithgow-Bertelloni, C., Louden, K.E., 2004. Iceland, the Farallon Slab, and dynamic topography of the North Atlantic. Geology 32, 177–180. Crosby, A.G., McKenzie, D., Sclater, J.G., 2006. The relationship between depth, age and gravity in the oceans. Geophys. J. Int. 166, 553–573. Davis, M., Kusznir, N.J., 2004. Depth-dependent lithospheric stretching at rifted continental margins. In: Karner, G.D. (Ed.), Proceedings of NSF rifted margins theoretical institute. Columbia University Press, pp. 92–136. Dean, S.M., Minshull, T.A., Whitmarsh, R.B., Louden, K.E., 2000. Deep structure of the ocean–continent transition in the southern Iberia Abyssal Plain from seismic refraction profiles: The IAM-9 transect at 40 20'N. Dehler, S.A., Keen, C.E., 1993. Effects of rifting and subsidence on thermal evolution of sediments in Canada's east coast basins. Can. J. Earth Sci. 30, 1782–1798. Edwards, G.R.H., 2006. Inverse modelling of extentional sedimentary basins and margins, Ph.D. thesis, University of Cambridge. Edwards, G., White, N., Haines, J., in preparation. Inverse modelling of extentional sedimentary basins and margins, J. Geophys. Res. Escartín, J., Hirth, G., Evans, B., 2001. Strength of slightly serpentinized peridotites: implications for the tectonics of oceanic lithosphere. Geology 29, 1023–1026. Fernández-Viejo, G., Gallart, J., Pulgar, J.A., Gallastegui, J., Dañiobietia, J.J., Córdoba, D., 1998. Crustal transition between continental and oceanic domains along the North Iberian margin from wide-angle and gravity data. Geophys. Res. Lett. 25, 4249–4252. Froitzheim, N., Pleuger, J., Nagel, T.J., 2006. Extraction faults. J. Struct. Geol. 1388–1395. Funck, T., Hopper, J.R., Larsen, H.C., Louden, K.E., Tucholke, B.E., Holbrook, W.S., 2003. Crustal structure of the ocean–continent transition at Flemish Cap: seismic refraction results. J. Geophys. Res. 108, 2531. doi:10.1029/2003JB002434.

226

A. Crosby et al. / Earth and Planetary Science Letters 273 (2008) 214–226

GRACE satellite-only gravity model GGM02, see http://www.csr.utexas.edu/grace/ gravity/ for further documentation. Harry, D.L., Grandell, S., 2007. A dynamic model of rifting between Galicia Bank and Flemish Cap during the opening of the North Atlantic Ocean. In: Karner, G.D., Manatschal, G., Pinheiro, L.M. (Eds.), Imaging, Mapping and Modelling Continental Lithosphere Extension and Breakup. Geol. Soc. Lond. Spec. Pub., vol. 282, pp. 157–172. Hellinger, S.J., Sclater, J.G., 1983. Some comments on two-layer extensional models for the evolution of sedimentary basins. J. Geophys. Res. 88, 8251–8269. Hoffmann, H.J., Reston, T.J., 1992. Nature of the S reflector beneath the Galicia Banks rifted margin: Preliminary results from prestack depth migration. Geology 20, 1091–1094. Hölker, A.B., Manatschal, G., Holliger, K., Bernoulli, D., 2003. Tectonic nature and seismic response of top-basement detachment faults in magma-poor rifted margins. Tectonics 22. doi:10.1029/2001TC001347. Hopper, J.R., Funck, T., Tucholke, B.E., Larsen, H.C., Holbrook, W.S., Louden, K.E., Shillington, D., Lau, H., 2004. Continental breakup and the onset of ultraslow seafloor spreading off Flemish Cap on the Newfoundland rifted margin. Geology 32, 93–96. Huismans, R., Beaumont, C., 2003. Symmetric and asymmetric lithospheric extension: relative effects of frictional-plastic and viscous strain softening. J. Geophys. Res. 108. doi:10.1029/2002JB002026. Huismans, R.S., Podladchikov, Y.Y., Cloetingh, S., 2001. Transition from passive to active rifting: relative importance of asthenospheric doming and passive extension of the lithosphere. J. Geophys. Res. 106, 11271–11291. Jackson, J., 2002. Strength of continental lithosphere: time to abandon the jelly sandwich? GSA Today 4–9 (September). Jagoutz, O., Müntener, O., Manatschal, G., Rubatto, D., Péron-Pinvidic, G., Turrin, B.D., Villa, I.M., 2007. The rift-to-drift transition in the North Atlantic: a stuttering start of the MORB machine? Geology 35, 1087–1090. Jones, S.M., White, N.J., Faulkner, P., Bellingham, P., 2004. Animated models of extensional basins and passive margins. Geochem., Geophys., Geosyst. 5. doi:10.1029/2003GC000658. Keen, C.E., Dehler, S.A., 1993. Stretching and Subsidence: rifting of conjugate margins in the North Atlantic Region. Tectonics 12, 1209–1229. Lau, K.W.H., Louden, K.E., Funck, T., Tucholke, B.E., Holbrook, W.S., Hopper, J.R., Larsen, H.C., 2006a. Crustal structure across the Grand Banks–Newfoundland Basin Continental Margin — I. Results from a seismic refraction profile. Geophys. J. Int. 167, 127–156. Lau, K.W.H., Louden, K.E., Deemer, S., Hall, J., Hopper, J.R., Tucholke, B.E., Holbrook, W.S., Larsen, H.C., 2006b. Crustal structure across the Grand Banks–Newfoundland Basin Continental Margin — II. Results from a seismic reflection profile. Geophys. J. Int. 167, 157–170. Lavier, L.L., Manatschal, G., 2006. A mechanism to thin the continental lithosphere at magma-poor margins. Nature 440, 324–328. Lecroart, P., Cazenave, A., Ricard, Y., Thoraval, C., Pyle, D.G., 1997. Along-axis dynamic topography constrained by major-element chemistry. Earth Planet. Sci. Lett. 149, 49–56. Le Pichon, X., Sibuet, J.C., 1981. Passive margins: a model of formation. J. Geophys. Res. 86, 3708–3720. Marillier, F., Hall, J., Hughes, S., Louden, K., Reid, I., Roberts, B., Clowes, R., Coté, T., Fowler, J., Guest, S., Lu, H., Luetgart, J., Quinlan, G., Spencer, C., Wright, J., 1994. Lithoprobe East onshore-offshore seismic refraction survey — constraints on interpretation of reflection data in the Newfoundland Appalachians. Tectonophysics 232, 43–58. Marrett, R., Allmendinger, R.W., 1992. Amount of extension on “small” faults: an example from the Viking graben. Geology 20, 47–50. Mauffret, A., Montadert, L., 1988. Seismic stratigraphy off Galicia. Proc. Ocean Drill. Program Sci. Results 103, 13–30. McKenzie, D., 1978. Some remarks on the development of sedimentary basins. Earth Planet. Sci. Lett. 40, 25–32. McKenzie, D., Bickle, M.J., 1988. The volume and composition of melt generated by extension of the lithosphere. J. Petrol. 29, 625–679. Minshull, T.A., Dean, S.M., White, R.S., Whitmarsh, R.B., 2001. Anomalous melt production after continental break-up in the southern Iberia Abyssal Plain. In: Wilson, R.C.L., Whitmarsh, R.B., Taylor, B., Froitzheim, N. (Eds.), Non- volcanic Rifting of Continental Margins: A Comparison of Evidence from Land and Sea. Geological Society, London, Special Publications, vol. 187, pp. 537–550. Parsons, B., Sclater, J.G., 1977. An analysis of the variation of ocean floor bathymetry and heat flow with age. J. Geophys. Res. 82, 803–827. Paulssen, H., Visser, J., 1993. The crustal structure of Iberia inferred from P-wave coda. Tectonophysics 221, 111–123. Pedreira, D., Pulgar, J.A., Gallart, J., Díaz, J., 2003. Seismic evidence of Alpine crustal thickening and wedging from the western Pyrenees to the Cantabrian Mountains (north Iberia). J. Geophys. Res. 108. doi:10.1029/2001JB001667. Pérez-Gussinyé, M., Reston, T.J., 2001. Rheological evolution during extension at nonvolcanic rifted margins: onset of serpentinization and development of detachments leading to continental breakup. J. Geophys. Res. 106, 3961–3975.

Péron-Pinvidic, G., Manatschal, G., Minshull, T.A., Sawyer, D.S., 2007. Tectonosedimentary evolution of the deep Iberia–Newfoundland margins: evidence for a complex breakup history. Tectonics 26, TC2011. doi:10.1029/2006TC001970. Press, W.H., Teukolsky, S.A., Vetterling, W.T., Flannery, B.P., 1986. Numerical recipes in FORTRAN. Cambridge University Press. Priestley, K., McKenzie, D., 2006. The thermal structure of the lithosphere from shear wave velocities. Earth Planet. Sci. Lett. 244, 285–301. Reston, T., 2007. Extension discrepancy at North Atlantic nonvolcanic rifted margins: depth-dependent stretching or unrecognized faulting? Geology 35, 367–370. Rosenbaum, G., Regenauer-Lieb, K., Weinberg, R., 2005. Continental extension: from core complexes to rigid block faulting. Geology 33, 609–612. Russell, S.M., Whitmarsh, R.B., 2003. Magmatism at the west Iberia non-volcanic rifted continental margin: evidence from analyses of magnetic anomalies. Geophys. J. Int. 154, 706–730. Sclater, J.G., Christie, P.A.F., 1980. Continental stretching: an explanation of the postmid-Cretaceous subsidence of the central North Sea basin. J. Geophys. Res. 85, 3711–3739. Shillington, D.J., Holbrook, W.S., Van Avendonk, H.J.A., Tucholke, B.E., Hopper, J.R., Louden, K.E., Larsen, H.C., Nunes, G.T., 2006. Evidence for asymmetric nonvolcanic rifting and slow incipient oceanic accretion from seismic reflection data on the Newfoundland margin. J. Geophys. Res. 111, B09402. doi:10.1029/2005JB003981. Shillington, D.J., White, N.J., Minshull, T.A., Edwards, R., Edwards, G., Jones, S.M., 2008. Cenozoic evolution of the eastern Black Sea: a test of depth-dependent stretching models. Earth Planet. Sci. Letts. 265, 360–378. Shipboard Scientific Party, 2004. Leg 210 summary. Proc. Ocean Drill. Program, Init. Rep. 210, 1–78. Sibuet, J.C., Srivastava, S., Manatschal, G., 2007. Exhumed mantle-forming transitional crust in the Newfoundland–Iberia rift and associated magnetic anomalies. J. Geophys. Res. 112. doi:10.1029/2005JB003856. Skelton, A.D.L., Valley, J.W., 2000. The relative timing of serpentinization and mantle exhumation at the ocean–continent transition, Iberia: constraints from oxygen isotopes. Earth Planet. Sci. Lett. 178, 327–338. Srivastava, S.P., Sibuet, J.C., Cande, S., Roest, W.R., Reid, I.D., 2000. Magnetic evidence for slow seafloor spreading during the formation of the Newfoundland and Iberian margins. Earth Planet. Sci. Lett. 182, 61–76. Tett, D.L., Sawyer, D.S., 1996. Dynamic models of multi-phase continental rifting and their implications for the Newfoundland and Iberia conjugate margins. Proc. Ocean Drill. Program Sci. Results 149, 635–648. Tucholke, B.E., Sibuet, J.C., 2007. Leg 210 synthesis: tectonic, magmatic and sedimentary evolution of the Newfoundland–Iberia Rift. Proc. Ocean Drill. Program 210. doi:10.2973/odp.proc.sr.210.101.2007 Sci. Res. Tucholke, B.E., Austen, J.A., Uchupi, E., 1989. Crustal structure and rift-drift evolution of the Newfoundland Basin. In: Tankard, A.J., Balkwill, H.R. (Eds.), Extensional tectonics and stratigraphy of the North Atlantic Margins, pp. 247–263. Turcotte, D.L., 1999. Fractals and Chaos in Geology and Geophysics, 2nd edition. Cambridge University Press. Van Avendonk, H.J.V., Holbrook, W.S., Nunes, G.T., Shillington, D.J., Tucholke, B.E., Louden, K.E., Larsen, H.C., Hopper, J.R., Seismic velocity structure of the rifted margin of the eastern Grand Banks of Newfoundland, Canada, J. Geophys. Res. 111 (2006) B11404. doi:10.1029/2005JB004156. J. Geophys. Res. 105 (2000) 5859–5885. Walsh, J., Watterson, J., Yielding, G., 1991. The importance of small-scale faulting in regional extension. Nature 351, 391–393. Wernicke, B., 1985. Uniform-sense normal simple shear of the continental lithosphere. Can. J. Earth Sci. 22, 108–125. White, N., 1989. Nature of lithospheric extension in the North Sea. Geology 17, 111–114. White, N., 1993. Recovery of strain rate variation from inversion of subsidence data. Nature 366, 449–452. White, N., Bellingham, P., 2002. A two-dimensional inverse model for extensional sedimentary basins 1. Theory. J. Geophys. Res. 107, 2259. doi:10.1029/2001JB000173. White, R.S., McKenzie, D., O'Nions, R.K., 1992. Oceanic crustal thickness from seismic measurements and rare Earth element inversions. J. Geophys. Res. 97, 19683–19715. White, N., Thompson, M., Barwise, T., 2003. Understanding the thermal evolution of deep-water continental margins. Nature 426, 334–343. Whitmarsh, R.B., White, R.S., Horsfield, S.J., Sibuet, J.C., Recq, M., Louvel, V., 1996. The ocean–continent boundary off the western continental margin of Iberia: crustal structure west off Galicia Bank. J. Geophys. Res. 101, 28291–28314. Whitmarsh, R.B., Manatschal, G., Minshull, T.A., 2001. Evolution of magma-poor continental margins from rifting to seafloor spreading. Nature 413, 150–154. Zelt, C.A., Sain, K., Naumenko, J.V., Sawyer, D.S., 2003. Assessment of crustal velocity models using seismic refraction and reflection tomography. Geophys. J. Int. 153, 609–626.