Characterizing and identifying structural domains at rifted continental margins: application to the Bay of Biscay margins and its Western Pyrenean fossil remnants

Abstract We use the Bay of Biscay and Western Pyrenees as a natural laboratory to develop and apply an approach to characterize and identify distinctive rifted margin domains in offshore and onshore settings. The Bay of Biscay and Western Pyrenees offer access to seismically imaged, drilled and exposed parts of one and the same hyperextended rift system. Offshore, we use gravity inversion and flexural backstripping techniques combined with seismic interpretation to provide estimates of accommodation space, crustal thickness and lithosphere thinning. Onshore, we focus on key outcrops of the former rift domain to describe the nature of sediment and basement rocks, and of their interface. This qualitative and quantitative characterization provides diagnostic elements for the identification of five distinct structural domains at magma-poor rifted margins and their fossil analogues (proximal, necking, hyperthinned, exhumed mantle and oceanic domains). This new approach can be used to reconcile offshore and onshore observations, and to aid interpretation when only local observations are available. Onshore remnants can be placed in an offshore rifted-margin context, enabling the prediction of first-order crustal architecture. For the interpretation of offshore seismic reflection sections, geological insights into rift structures and basement nature can be suggested based on onshore analogies. Supplementary material: Sensitivity of backstripping results to flexural rigidity is available at http://www.geolsoc.org.uk/SUP18778.

Knowledge of magma-poor rifted-margin structure has significantly evolved over the past decades. Present-day models of continental rifted margin include the existence of hyperextended rifted domains made of extremely thinned continental crust and/or exhumed mantle, as described at present-day rifted margins (e.g. Lundin & Doré 2011;Reston & Manatschal 2011 and references therein) and in onshore fossil analogues preserved in collisional orogens (e.g. the Alps: Manatschal 2004; Mohn et al. 2010Mohn et al. , 2012Masini et al. 2011Masini et al. , 2012; the Pyrenees: Jammes et al. 2009;Lagabrielle et al. 2010; the Caledonides: Andersen et al. 2012). Both marine and onshore observations offer complementary datasets; however, the different scale and resolution of observations restrain straightforward correlations, and most studies are either focused on offshore or onshore examples.
As very few drill holes allow the direct exploration of distal parts of offshore continental rifted margins, their investigation mainly relies on indirect approaches such as the interpretation of reflection and refraction seismic sections (e.g. Osmundsen & Ebbing 2008;Franke 2013;Péron-Pinvidic et al. 2013;Sutra et al. 2013), potential field methods or geophysical quantitative techniques (e.g. Pérez-Gussinyé et al. 2001Reston 2009;Roberts et al. 2013). In contrast, fossil analogues of rifted margins preserved in collisional orogens provide direct observation of the nature of sediments and basement, and of their interface. Remnants of ancient rifted margins preserving primary relationships provide valuable insights not only into the interaction between tectonic and sedimentary processes but also into the subsidence history during and after rifting (e.g. Mohn et al. 2010;Masini et al. 2013).
In this study, we aim to characterize distinctive structural domains at continental rifted margins in order to propose qualitative and quantitative criteria to identify and interpret hyperextended domains onshore and offshore. We use the Bay of Biscay and Western Pyrenees to develop and apply this geological/geophysical approach. This natural laboratory preserves different evolutionary stages of a hyperextended rifted-margin system that are seismically imaged, drilled and exposed because of its subsequent inversion and partial integration into the Pyrenees.
This combined approach can be used to reconcile offshore and onshore observations on continental rifted margins. First-order relationships can be established between accommodation space creation, lithosphere thinning, extensional settings and field observations on the nature of sediment and basement rocks, and of their interface for the different domains. As a consequence, this approach can be used to aid interpretations of local observations both onshore and offshore. Onshore fossil remnants can be placed back in the context of the whole rifted margin predicting the first-order crustal architecture. Onshore analogies can be used to suggest geological insights into rift structure and basement nature for offshore seismic interpretations.
Finally, we believe that this approach may also be used, if limitations are respected, in a more general way to characterize, identify and interpret hyperextended rift systems, either buried under thick sedimentary sequences at present-day margins or disrupted in collisional orogens.

Rifted margins: terminologies, approaches and observations
In spite of the wide spectrum of rift architectures observed at continental rifted margins, first-order observations can be used to describe, characterize and subdivide them into different domains (Reston 2009;Péron-Pinvidic et al. 2013). The identification of different domains is either based on seismic studies (e.g. Dean et al. 2000;Osmundsen & Ebbing 2008;Minshull 2009;Péron-Pinvidic et al. 2013, Sutra et al. 2013, quantitative approaches (e.g. Reston 2009) or rheological considerations on how the crust and lithosphere deform (Pérez-Gussinyé et al. 2001Sutra & Manatschal 2012). As a consequence, terminologies and definitions of domains depend on the approach, as well as on the resolution and scale of observations ( Fig. 1). Most studies of present-day rifted margins rely on geophysical techniques (e.g. seismic reflection or refraction and potential field methods) that give large-scale images of the crustal structure but no direct insights into the nature of the rocks and structures that form the margins. In contrast, the investigation of remnants of fossils margins exposed in internal parts of collisional orogens offers direct access to outcrops corresponding to 'high-resolution' observations of the former margin section. However, since these outcropping remnants represent disconnected samples from the former rifted margin, the relationship between them is often difficult to determine and they can be interpreted only if they can be restored back to their initial position (e.g. Jammes et al. 2009;Mohn et al. 2010) (Fig. 1). Drill-hole observations are the only dataset that offers a local georeferenced in situ access to detailed geological information within a present-day rift section, enabling direct outcrop observations and indirect geophysical observations to be bridged. The Bay of Biscay and West Pyrenean rift system offers the unique possibility to have access to seismically imaged, drilled and exposed parts of one and the same hyperextended rift system. Using this example, we aim to demonstrate how onshore (high-resolution but discontinuous) and offshore (low-resolution but continuous) observations can be combined. The challenge is to define both geophysical and geological diagnostic elements that may be used to define the genetic rift domains proposed in this work (proximal, necking, hyperthinned, exhumed mantle and oceanic domains: Fig. 1).

Geophysical approaches
Geophysical studies such as seismic reflection and refraction surveys or potential field data provide first-order continuous images of rift systems based on the mapping of physical properties of rocks forming continental rifted margins. Since different rocks can have similar densities and/or velocities, the interpretation of geophysical data is non-unique. Nevertheless, first-order interfaces, juxtaposing materials with different densities and/or velocities, can be imaged in seismic sections ( Fig. 2a): (1) sea level; (2) seafloor (water -top of sedimentary infill); (3) top basement (base of sediments -top of acoustic basement); and (4) seismic Moho (crust -mantle velocity interface). These first-order observations can be used to estimate accommodation space (as defined in Fig. 2a) and crustal thickness evolution. Since both are related to extension and subsidence, they provide input parameters for analytical methods linking lithosphere thinning and subsidence history (e.g. Kusznir & Karner 2007;Roberts et al. 2013). In spite of limitations, owing to the restricted knowledge of lithospheric processes and of the thermal and density structure of the lithosphere, these methods can quantify the evolution of accommodation space, crustal thickness and lithosphere thinning, providing a first-order description of the spatial and temporal evolution of rifting.
Accommodation space. Considering a section through a present-day thermally equilibrated rifted margin, the most straightforward observation is to consider the evolution of potential accommodation space (between sea level and seafloor) and total accommodation space (between sea level and top basement), as shown in Figure 2. At a thermally equilibrated magma-poor rifted margin, accommodation space commonly increases oceanwards as continental crust thins and is controlled, to first order, by crustal thickness, post-breakup thermal subsidence and sediment thickness (contributing to compaction and flexural loading). Second-order controls, such as eustatic sea-level variations and mantle processes (dynamic topography), become important when attempting to restore the palaeobathymetry at time of break-up.
Lithosphere thinning. Lithosphere thinning variations at rifted continental margins may be determined from the lithosphere stretching factor, b, of McKenzie (1978). This model assumes a pure shear deformation of the lithosphere (i.e. lithosphere thinning is equal to whole crustal thinning). For continental margin lithosphere, b is defined as the ratio of the initial continental crustal thickness to the present-day crustal thickness. For oceanic lithosphere (i.e. in oceanic domains), b factor values tend to infinity. In order to take into account the possible occurrence of magmatic addition resulting from decompression melting during rifting and continental break-up, a correction for volcanic addition thickness may be applied to the presentday crustal thickness to calculate the lithosphere stretching factor. Hereafter, lithosphere thinning is represented by the thinning factor (g ¼ 1-1/b), which is more convenient to represent graphically since it only varies between 0 (no crustal thinning) and 1 (complete thinning of the continental crust).

Onshore field geology
Fossil remnants

Present-day margins
Depth (km)

Geological observations
Hyperextended domain Fig. 1. Simplified architecture of a hyperextended magma-poor rifted margin (based on the IAM 5 refraction line from the West Iberia margin: Afilhado et al. 2008) and terminology adopted in this study (central part). A comparison with terminologies from other recent studies relying either on geophysical approaches (upper part of the figure) or geological field observations (lower part of the figure) is shown (Dean et al. 2000;Pérez-Gussinyé et al. 2003;Jammes et al. 2009;Minshull 2009;Osmundsen & Ebbing 2008;Mohn et al. 2010;Sutra & Manatschal 2012;Péron-Pinvidic et al. 2013;Sutra et al. 2013).
high-b systems, can be distinguished (Fig. 2) based on basin architecture and associated extensional structures (Wilson et al. 2001). Low-b extensional systems correspond to classical rift structures. They are characterized by high-angle normal faults and tilted blocks delimiting graben-and half-grabentype basins filled by wedge-shaped sedimentary sequences. Low-b systems can only accommodate small amounts of extension and therefore are frequently found over weakly thinned crust (i.e. in proximal margins: Fig. 2). In contrast, high-b extensional systems are associated with large-offset normal faults (e.g. Wernicke 1985;Lister et al. 1986;Brady et al. 2000), also referred to as top-basement detachment faults (Hölker et al. 2003). In these systems, faults and their underlying footwall can be exhumed to the seafloor, leading to the creation of new surfaces tectonically exhumed to the seafloor. This type of fault is difficult to recognize on seismic sections because, in contrast to high-angle faults, they form relatively smooth fault-related topography. As these faults can accommodate large displacements and thin the crust, they are interpreted to floor large sedimentary basins within the hyperextended distal parts of magma-poor rifted margin (Fig. 2). These 'hyperextended sag' basins are characterized by overlying subparallel sedimentary sequences and a low-angle onlap of the strata on to basement (Fig. 2). This type of basin is described from several deep-water rifted margins: for example, the Mid Norwegian margin (high-b basins: Osmundsen & Ebbing 2008), and the Angola and Iberia margin (sag basins: Péron-Pinvidic et al. 2007, 2013Unternehr et al. 2010).

Geological approaches
The discovery of dislocated fragments of ancient rifted margins in collisional orogens gives access to detailed geological observations and unlimited sampling, enabling the characterization of the nature and age of sediments, the type of basement rocks, as well as the nature of the contact between sediments and basement rocks. The combination of sedimentological, petrological and structural observations provides information on the depositional environment, aggradation history, kinematics and timing of faulting, and the thermal and isostatic history of a rift system. However, well-exposed and   Afilhado et al. 2008) showing first-order interfaces (seafloor, top basement and Moho), and the potential and total accommodation space (AS). (b) & (c) Architecture and characteristics of half-graben and hyperextended sag basins, also referred to as low-and high-b extensional systems. Note the difference in potential and total accommodation space between the two types of basins and the nature of the sediment-basement contact.
preserved outcrops of former rift systems are only locally preserved due to the reactivation of former rift structures during their emplacement in the orogenic system. Therefore, the restoration of fossil remnants of rift systems back into their original position within the margin is often difficult and related to large uncertainties. Studies of remnants derived from the former Jurassic Tethys margins, today exposed in the Alps, provide detailed descriptions of rift structures, and their link to sediments and underlying basement rocks. Because alpine fossil remnants show many similarities with structures that were drilled and seismically imaged along the Western Iberia margin, they are considered as 'analogues' of low-and high-b extensional rift systems (Fig. 2) (e.g. Wilson et al. 2001;Manatschal 2004). While low-b systems can be well defined from seismic data, high-b systems and in particular the exhumation faults (i.e. topbasement detachment fault) related to these systems are only unambiguously recognized at an outcrop scale. Therefore, being able to resolve the nature of top basement (fault or stratigraphic contract) and the type of hanging-wall and footwall rocks is a key to interpreting seismic data. In offshore systems, this information can be established only locally by drill-hole data.
Nature of sediments and the creation of accommodation space. In order to quantify the total accommodation space created during extension, the palaeobathymetry and decompacted stratigraphic thickness need to be determined. In particular for deep-water environments, palaeobathymetry estimations are associated with large uncertainties. Therefore, field data can determine the formation of accommodation space qualitatively only by estimating the thickness of sedimentary sequences and the deepening of the depositional environment.
In proximal margins characterized by low-b extensional systems, the crust is weakly thinned. Accommodation space is therefore usually relatively small, and eustatic sea-level variations may result in local subaerial exposure and erosion of pre-rift sediments and basement. As a consequence, the occurrence of thin shallow-marine post-rift sediments onlapping directly on to pre-rift sediments or basement is characteristic of this domain. Indeed, thick sequences of deep-marine sediments may indicate a more important total accommodation space but restricted to graben-and half-graben-type basins (Fig. 2b).
High-b extensional systems are associated with distal margins (Wilson et al. 2001) or highly extended rift settings (Friedmann & Burbank 1995). Well-preserved remnants of high-b extensional systems are described from the Alps and Pyrenees (i.e. supra-detachment basins in Masini et al. 2011). The complex architectures of these systems, including large-offset normal faults, break-away blocks and extensional allochthons (see Masini et al. 2013, fig. 3), are reflected in the sedimentary infill history that strongly depends on the sedimentary supply (Masini 2011). In sediment-starved systems, these basins are filled by deep-water turbidites interleaved with pelagic or hemi-pelagic material, whereas, in sediment-rich systems, these basins show thick aggradational sequences.

Nature of the top basement.
A key observation to distinguish between low-and high-b extensional settings is the nature of the basement rock and of its contact with the overlying sediments. This contact can either be a stratigraphic contact, a highangle fault or a tectonically exhumed basement, in which case the top basement is either an exhumed fault surface or an erosion surface (Masini et al. 2011) (Fig. 2b, c).
The sharp topography observed along low-b extensional systems is due to the small offset of the top basement along high-angle normal faults. Since the faults can accommodate only little extension, they cannot exhume deep pre-rift lower crustal levels or mantle rocks to the surface. Hence, in proximal domains, low-b extensional systems dominantly preserve pre-rift upper crust in rotated fault blocks ( Fig. 2).
High-b extensional systems can be best recognized in outcrops as top-basement detachment fault surfaces. These large-offset normal faults can exhume deep crustal levels and mantle rocks while thinning the crust. The presence of exhumed subcontinental mantle at the seafloor suggests an entire removal of the continental crust corresponding to a complete crustal thinning.
Geometry and kinematics of extensional systems. While on seismic sections the recognition of the large-scale architecture of extensional systems mainly relies on the study of the infill stratigraphic geometry; in orogens, the overall architecture of rift basins is usually not preserved. Insights may come from the study of the relationship between fault geometry and the related syntectonic sediments. Low-b settings are classically recognized by wedge-shaped syntectonic sediments (Wilson et al. 2001) that can be interbedded with footwall-derived breccia along fault scarps. The syntectonic sediments typically onlap on to the tilted hanging wall, and the contact with the normal fault is at a high angle (Fig. 2b). In high-b settings, the relationships between the syntectonic sediments and the fault surface are very different (Fig. 2c) (Wilson et al. 2001). Top-basement detachment fault surfaces are associated with a characteristic sequence of rocks that include cataclastic basement and gouges, overlain by tectonosedimentary breccias; grading upwards into sedimentary breccias and turbidites reworking hanging-wall-and footwall-derived material . The occurrence of clasts derived from pre-rift lower crust and mantle in synrift breccias provide indirect evidence for the existence of exhumation faults in hyperextended domains. Another characteristic of top-basement detachment faults is the occurrence of hangingwall-derived extensional allochthons overlying the fault surface. These blocks can range in size from tens of kilometres to tens of metres (e.g. see ).

Drill-hole data: the link between direct and indirect approaches
Drill-hole observation is a unique dataset providing a direct link between rock-derived data (physical properties, ages and kinematics) and seismic data at present-day rifted margins. In contrast to outcropsampling remnants of former rifted margins that are disrupted and usually out of context, drill-hole data are geolocalized and sample well-defined parts of the margins. At present, most observations based on drill-hole data come from proximal margins, while data coming from distal margins are limited to the conjugate Iberian and Newfoundland margins, which at present is the only margin system where drill holes have penetrated the hyperextended basement. The Ocean Drilling Program (ODP) drill-hole data (legs 103, 149, 173 and 210) from the Iberian -Newfoundland margins have already enabled deep structures seismically imaged along these margins to be linked with outcrops exposed in the Alps (e.g. Wilson et al. 2001;Manatschal 2004). However, since these margins formed independently from each other and their post-rift evolution is very different, the integration of the two datasets, beyond using them as analogues, is not possible.

The Bay of Biscay and Western Pyrenees natural laboratory
In this paper we use the Bay of Biscay -Western Pyrenees as a natural laboratory to show how geophysical and geological field observations can be integrated into a coherent and consistent approach to characterize and interpret hyperextended rifts systems both offshore and onshore. The Bay of Biscay-Western Pyrenees preserves, along one and the same system, different evolutionary stages of a rift and compressional system (Fig. 3). Thanks to decades of exploration, the access to critical drillhole observations is possible, making this a unique natural laboratory to link onshore and offshore observations.

General setting
The Bay of Biscay, located between France and Spain, is characterized offshore by an embryonic oceanic basin, bounded to the north by the Western Approach and Armorican margins, and to the south by the North Iberian margin. At the eastern termination, several Mesozoic rift basins preserving geophysical and geological evidence of extreme crustal thinning are identified both offshore (e.g. the Parentis Basin: Pinet et al. 1987;Bois & Gariel 1994;Tomassino & Marillier 1997;Jammes et al. 2010a, b, c) and onshore (e.g. the Arzacq -Mauléon Basin: Grandjean 1992; Daignières et al. 1994;Jammes et al. 2009;Lagabrielle et al. 2010; and the Basque-Cantabrian Basin; Pedreira et al. 2007).
The present-day structuring of the region results from the succession of several extensive and compressive tectonic events that initiated with the Variscan Orogen and early Permian orogenic collapse (Burg et al. 1994a, b), followed by Triassic -Lower Cretaceous rift phases. The Late Cretaceous northwards movement of the African Plate resulted in the partial closure of the Bay of Biscay -Pyrenean domain.
As this work will be mainly focused on the Early Cretaceous hyperextensional rift event that led to oceanic crust formation at Aptian -Albian time in the Bay of Biscay (Montadert et al. 1979b;Boillot 1984), we will subdivide the sedimentary record in the study area into pre-, syn-and post-hyperextension sedimentary sequences.

Plate kinematic context, Cretaceous hyperextension and Pyrenean compression
The palaeogeographical and plate kinematic evolution related to the opening of the Bay of Biscay remains strongly debated at present. Large uncertainties about this evolution stem from a lack of reliable magnetic anomalies (including the M-series and the following magnetic quiet zone) indicative of the break-up time in the Iberia-Newfoundland margins (Bronner et al. 2011) and the Bay of Biscay (Norton et al. 2007). Furthermore, the discovery of hyperextended crust and exhumed mantle at the Iberia -Newfoundland margins over hundreds of kilometres continentwards of the first unambiguous magnetic anomaly, related to break-up (for a review see Tucholke et al. 2007or Péron-Pinvidic & Manatschal 2009) is at present not taken into account by most restorations (e.g. Olivet 1996). As a result, several geodynamic models are suggested for the Late Jurassic -Early Cretaceous evolution of the Bay of Biscay, including a back-arc setting (e.g. Sibuet et al. 2004;Vissers & Meijer 2012) or a transtensional setting forming pull-apart basins within the Pyrenean domain (e.g. Choukroune & Mattauer 1978). Based on the mapping of the Iberia -Newfoundland margins (e.g. Péron-Pinvidic & Manatschal 2009;Sutra et al. 2013) and combining it with field observations from the Mauléon and Parentis basins, Jammes et al. (2009Jammes et al. ( , 2010a proposed an alternative plate kinematic scenario. Their model, similar to the one proposed by Williams (1975), suggested that the main rifting episode is linked with a counter-clockwise rotation of Iberia during Aptian time (Gong et al. 2008), overprinting a Late Jurassic -Early Aptian sinistral strike-slip or transtensional event.
The kinematics of the area is reasonably well constrained after the magnetic quiet period of the Cretaceous with the identification of the magnetic anomaly 34 (about 84 Ma, Campanian -Santonian) in the Bay of Biscay (e.g. Rosenbaum et al. 2002).
First evidence of compressional deformation was recorded in Santonian-Campanian time in the Pyrenees (Garrido-Megías & Ríos 1972;McClay et al. 2004), and is associated with heterogeneous and weak reactivations in the northern and eastern part of the Bay of Biscay (Pinet et al. 1987;Thinon et al. 2001). After Eocene time, the study area acquired a structure close to that of the present day. The southern Cantabrian margin, and the Basque -Cantabrian and Mauléon basins, were reactivated and integrated into the Pyrenean Orogen.
A natural laboratory to observe, characterize and interpret hyperextensional systems The Bay of Biscay and Western Pyrenees do not only preserve different evolutionary stages of a rift system, including hyperthinned crust, exhumed mantle and oceanic domains (terminology defined in Fig. 1), but also show different steps of reactivation along one and the same system. In the following, we will consider three sections preserving different genetic and reactivation steps. A 'margin' stage is investigated using the Norgasis 11 -12 seismic section across the Western Approach margin (Avedik et al. 1996;Thinon 1999) (for the location see Fig. 3). The 'hyperextended' stage is imaged on the ECORS Bay of Biscay profile across the Parentis Basin (Pinet et al. 1987) (for the location see Fig. 3). Finally, the integration of the Mauléon rift basin into the Western Pyrenees enables the reactivated stage of a hyperextended rift system to be explored (Jammes et al. 2009) (for the location, see Fig. 3). The three sections result from the interpretation of different datasets with different resolutions and observational scales providing geological and geophysical diagnostic elements to identify structural domains at rifted continental margins.

The margin stage: the Western Approach margin
The Western Approach margin corresponds to the NW segment of the Bay of Biscay (Fig. 3). It shows typical 'building blocks' of a hyperextended magma-poor rifted margin (e.g. de Charpal et al. 1978;Montadert et al. 1979b;Avedik et al. 1982;Le Pichon & Barbier 1987), including indirect hints supporting the existence of exhumed mantle domains (Thinon 1999;Thinon et al. 2003;Bullock 2004;Bullock & Minshull 2005). Only few drill holes penetrated the sedimentary cover (leg 48: Montadert et al. 1979a), enabling some seismic stratigraphic correlations across the margin to be proposed and allowing them to be extended towards the adjacent Armorican margin (Thinon et al. 2002). We focus on the Norgasis 11-12 sections across the Western Approach margin, corresponding to consecutive seismic sections acquired in 1994 by Ifremer during the Norgasis campaign (Avedik et al. 1996;Thinon 1999). In the following, we will mainly rely on the mapping of first-order interfaces (seafloor, top basement and Moho). The mapping of these interfaces allows us to define low-b and high-b extensional systems (Wilson et al. 2001), and also to quantify crustal thickness and accommodation space, and to describe and characterize genetic domains across the margin.

Seismic interpretation: first-order interfaces
First-order interfaces in seismic sections may be defined between materials with different physical properties: the seafloor (water -sediment interface); the top basement (between sediments and seismic basement); and the seismic Moho (crust-mantle boundary). The first limit can be well observed in the time-migrated Norgasis 11-12 sections ( Fig. 4). Top basement is also rather well imaged along most of the section, and corresponds to the interface between the base of syn-hyperextensional sediments (Berriasian -Aptian: Thinon et al. 2002) and the acoustic basement, characterized by strong low-frequency reflectors either corresponding to older well-cemented sediments, metasediments or crystalline basement. However, in some places, top basement is uncertain either because of multiples (to the north, where the section cuts through the Western Approach basin, SP 5160-4400) or because of the lack of well-defined basement reflections (southward of Trevelyan, SP 2000-1400). Deep low-frequency subparallel reflections are also observed in the seismic basement between 9 and 10 s (two-way travel time (twtt)). These reflections are highly discontinuous north and south of the Meriadsek terrace (SP 4400-3800 and SP 3000-2600) but more continuous oceanwards, between SP 16 and SP 1400 at the southernmost termination of the section. In the Trevelyan area (SP c. 2400(SP c. -2000, where compressional deformation is recognized in the overlying sediments (Thinon et al. 2001), similar strong reflectors can be described but at shallower levels of around 7.5 s (twtt), probably due to the tectonic inversion. These deep reflections may be interpreted as the velocity contrasts between continental or oceanic crust and the underlying mantle rocks, corresponding to seismic Moho.

Stratigraphic and basin architecture
A detailed description of the morphology and stratigraphic architecture of the Western Approach margin was provided by Thinon (1999). In this study we focus on observations enabling the distinction between high-or low-b extensional systems (Fig.  2). Hence, only the base of post-hyperextensional sediments is discussed (Late Aptian: Thinon et al. 2002). The Western Approach margin is classically only interpreted as a succession of titled blocks (e.g. Avedik et al. 1979;Montadert et al. 1979b) characteristic of low-b extensional settings. However, based on the descriptions of Wilson et al. (2001), two high-b extensional domains can be recognized on the Norgasis sections. Over the Meriadzek terrace, almost subparallel syn-hyperextensional sediment sequences are either onlapping the underlying basement at a low angle (SP c. 3800) or terminate towards a basement high (SP c. 3600) delimiting a large sedimentary basin (Fig. 4b). The second one, which is more distal, is located southwards of Trevelyan. In this domain, the youngest syn-hyperextension sequence (Early Aptian: Thinon et al. 2002) directly overlies the almost flat top basement and shows a stratigraphic architecture similar to that of classical post-rift sediments. However, it is important to note that these sediments are older than the first sediments that overlie oceanic crust (Late Aptian -Albian in age: Thinon et al. 2002) (Fig. 4c). This 'hyperextended sag'-type architecture is indicative of high-b extensional settings that may be floored by top-basement detachment faults (Fig. 2c) (Wilson et al. 2001;Masini et al. 2012;Péron-Pinvidic et al. 2013), and related to thinning and exhumation processes.

Estimations of crustal thickness and lithosphere thinning
We depth-converted the seafloor, base post-hyperextension and top seismic basement defined on the Norgasis 11 -12 section. The depth conversion uses a constant velocity for water (1500 m s 21 ) and includes an increasing velocity model with depth for sediment thickness (detailed methodology in Japsen 1993). Moho depth, crustal thickness and continental lithosphere thinning factor (Fig. 5) have been determined using a gravity inversion technique incorporating a thermal gravity anomaly correction and a parameterization of decompression melting to predict volcanic addition (detailed scheme described by Greenhalgh & Kusznir 2007;Alvey et al. 2008;Chappell & Kusznir 2008b;Cowie & Kusznir 2013) (parameters for this study are given in Table 1). The data used in the gravity inversion are public domain free-air gravity (Sandwell & Smith 2009) and bathymetry (Smith & Sandwell 1997), the depth-converted sediment thickness from the seismic profiles, and oceanic isochrons (Müller et al. 1997). A compactioncontrolled density -depth relationship is assumed for sediment (Chappell & Kusznir 2008a).
Two end-member results are presented in Figure  5a, b. The first one corresponds to a magma-starved model and the second one to a model with 'normal' volcanic additions. The magma-starved solution (Fig. 5a) implies that no melt is generated during rifting and break-up; hence, the crustal basement thickness derived from gravity inversion only represents continental crust. The solution computed with a 'normal' volcanic addition (i.e. maximum oceanic crustal thickness of 7 km) (Fig. 5b) assumes that decompression melting occurs during rifting, producing magmatic additions to the extended continental crust once the lithosphere is thinned below a critical value (Chappell & Kusznir 2008b, fig. 3). Therefore, the crustal basement thickness derived from gravity inversion represents both the residual continental crust and volcanic additions. We are aware that neither of these two models completely represents the Norgasis 11 -12 section but they provide end members. Considering gravity inversion results for the Norgasis sections, the Moho inflection at the termination of the residual continental crust (Fig. 5b) is compatible with a basement composed of serpentinized mantle.
Comparing the two thinning factor profiles ( Fig.  5d) and thickness variations of the residual continental crust (Fig. 5c) of both models, three main trends can be observed delimiting different domains (Fig. 5). In the most proximal domain, the thinning factor is relatively low and constant (g ¼ c. 0.2) in both cases, coincident with a limited thinning of the continental crust. Contrastingly, the initiation of extreme thinning of the continental crust coincides with the high-b domain identified in Figure  4 in the Meriadsek terrace. In this domain a strong increase in thinning values (from g ¼ c. 0.2 to g ¼ c. 0.8 for the magma-starved model) is observed that extends oceanwards until complete thinning of the crust is achieved (resulting in g ¼ c. 1 for the 'normal' volcanic addition model). This domain corresponds to the beginning of the ocean-continent transition, where mantle exhumation is proposed on the Western Approach margin (Thinon 1999;Thinon et al. 2003) and where another more distal high-b setting is identified (Fig. 4).

Quantification of accommodation space evolution
The estimation of potential and total accommodation space for the Norgasis 11 -12 section relies only on the identification of seafloor and top basement from seismic reflection data (Fig. 6a, b). In order to determine the sediment-corrected post-rift bathymetry at a thermally equilibrated continental rifted margin, two-dimensional (2D) flexural isostatic backstripping and decompaction has been performed on the Norgasis 11-12 section. This post-rift sediment-corrected bathymetry enables the evolution of accommodation space without the loading and compaction effects of post-rift sediments to be determined (Fig. 6c).
Flexural backstripping and reverse post-rift thermal subsidence modelling has also been applied to the section in order to predict bathymetry at the start of post-break-up subsidence. The distribution  of reverse thermal subsidence is determined using the model of McKenzie (1978) and the thinning factor profile resulting from gravity inversion (using normal volcanic addition model, Fig. 5c). The methodology for flexural backstripping, decompaction and reverse thermal subsidence modelling is described in Kusznir et al. (1995) and Roberts et al. (1998). A break-up age of 110 Ma for the Bay of Biscay, corresponding to the Late Aptian (Montadert et al. 1979b;Boillot 1984), is used to define the reverse thermal subsidence time. This age is assumed to represent the transition from tectonicdominated subsidence to thermal subsidence (the parameters used are presented in Table 2). The modelled bathymetry at the break-up time prior to thermal equilibration (Fig. 6c) enables an evaluation to be made of the available syn-to post-rift accommodation space.
Palaeobathymetries determined from flexural backstripping (Fig. 6c, d) for our case study provide isostatically balanced estimates of how much accommodation space can be created during continental margin rifting and break-up. The general trend of bathymetries predicted can then be related to the thinning domains previously described (Figs 5c & 6). In the most proximal domain, where the crust is only weakly thinned, potential accommodation space created during rifting is minor (Fig.  6) and local erosion is predicted before post-rift thermal subsidence (Fig. 6d). Contrastingly, total accommodation space can record greater pre-to synrift sediment aggradation. An increase in the potential accommodation space is related to the extreme thinning of the continental crust; the total accommodation slightly varies, depending on basement topography. The transition from continental to oceanic domains (interpreted as exhumed mantle: Thinon et al. 2003) preserves locally the largest total accommodation space. Oceanwards, the total accommodation space slightly diminishes, recording the isostatic response due to the increasing thickness of magmatic additions in the oceanic domain (Fig. 6).

The hyperextended stage: the Parentis Basin
Located at the eastern termination of the Bay of Biscay, the offshore Parentis Basin is a roughly 100 km-wide rift basin bounded by the Landes High to the south and the Armorican shelf to the north (Fig. 3). The extreme crustal thinning of the basin to less than 10 km is relatively well captured by numerous geophysical surveys, including the ECORS Bay of Biscay profile (Pinet et al. 1987;Bois & Gariel 1994;Tomassino & Marillier 1997).
In the following, we will use the ECORS Bay of Biscay seismic profile (Pinet et al. 1987) to map first-order interfaces, such as seafloor, top basement and Moho (relying on seismic reflection and refraction) similarly as for the previous example. Additional constraints on the stratigraphic architecture and sedimentary evolution are provided by numerous industry drill holes. These data enable us to define several key stratigraphic horizons that can be mapped across the basin (Fig. 7).

Seismic interpretation of top basement and Moho
Even though no drill hole is located directly on the ECORS Bay of Biscay profile, some drilling results can be projected (Pinet et al. 1987;Bois et al. 1997;Jammes et al. 2010a, b, c) to facilitate stratigraphic correlations (Fig. 7a). For the purpose of this study, we defined stratigraphic units that are pre-hyperextension (Triassic-Late Jurassic), syn-hyperextension (c. Valanginian/Barremian-Aptian), post-hyperextension (Albian-Upper Cretaceous) and syncompressional (Tertiary) sedimentary sequences. However, except for the Danu well over the Landes High, no drill hole reached the Palaeozoic basement and only three of them (the Cormoran, Ibis and Pelican wells) penetrated Upper Jurassic sediments. Furthermore, the identification of the top basement is complicated by the presence of Triassic salt, associated with diapiric and gravitational structures (Biteau & Canérot 2007), strongly affecting the imaging of the southern border of the basin.
Insights on the seismic Moho depth are provided by the study of the expanding spread profiles (ESP), acquired at the same time as the ECORS Bay of Biscay section (location shown on Fig. 7) (Marillier et al. 1988;Tomassino & Marillier 1997). The interface between the crustal basement and the seismic Moho is marked by a major increase in velocities  (from c. 6 km s 21 in the crust to c. 8 km s 21 in the mantle).

Stratigraphic architecture and insights into the sedimentary infill history
Drill-hole observations constitute a critical dataset to investigate the infill history of the basin (Bois et al. 1997) and also to model the related subsidence evolution of the Parentis Basin (as proposed by Brunet 1984Brunet , 1997. In this work, the sedimentary sequences are simplified with reference to the Early Cretaceous hyperextensional rift event preceding the opening of the Bay of Biscay. The pre-hyperextensional sediments (from Triassic to Upper Jurassic) record a diffuse and widely distributed Triassic -Liassic rifting episode evidenced by the local deposition of more than 2 km-thick siliciclastic and evaporitic sequences, followed by the accumulation of thick Jurassic shallow-marine limestones, dolomite and shales (Brunet 1984;Desegaulx & Brunet 1990). In the northern part of the basin this pre-hyperextensional sequence is continuous, as shown in the seismic sections (Fig. 7, SP c. 2500-4500); whereas over the Landes High and towards the Aquitaine shelf, this sequence is either absent or very thin (e.g. the Danu borehole: Bois et al. 1997).
The syn-hyperextensional succession records a Valanginian/Barremian -Late Aptian subsidence (Desegaulx & Brunet 1990) related to rifting in the Bay of Biscay. During this phase of hyperextension, deltaic sequences were deposited in the basin (Curnelle & Dubois 1986). The synhyperextension sequence is also absent or highly reduced over the Landes High (Danu borehole: Bois et al. 1997), contrasting with the thick (c. 2-3 km) time-equivalent succession deposited in the basin. Furthermore, the syn-hyperextension infill is asymmetrical in the Parentis Basin. In the northern part of the basin, the sequence thickens towards the centre of the basin (Fig. 7, SP c. 3000), while, in the southern part (Fig. 7, SP c. 5000 -4500), the sequence is thinner (Ibis borehole: Jammes et al. 2010a, b, c).
The deposition of the post-hyperextension sequence (Albian -Upper Cretaceous) is related to the thermal subsidence following extreme lithosphere thinning in the Bay of Biscay, and can be subdivided into a rapid Albian subsidence and reduced rates in Late Cretaceous time (Curnelle 1983;Desegaulx & Brunet 1990). During this postrift phase, the basin was differentiated into a platform domain (to the south, over the Landes High: Fig. 7, SP c. 7900-5000) preserving shallow-water carbonates. In contrast, the northern basin (Fig. 7, SP c. 5000-2500) is characterized by the deposition of more than 2.5 km of turbidites and shales (Curnelle 1983) at a deep-water depth (c. 2 km, estimations of Curnelle 1983).
The syn-compressional sequences of the Parentis Basin record the progressive formation of the Pyrenean Orogen starting from Late Cretaceous and lasting until Oligocene-Miocene time. In the central part of the Parentis Basin only a little deformation is observed, related to a complex network of thrusts, sealed by Eocene -Oligocene sequences. The southernmost sediments of the ECORS Bay of Biscay section are deformed by thrust sheets and, over the Cantabrian shelf (Fig. 7, SP c. 6500-6000), present the classical architecture of a foreland basin (Roca et al. 2011). Crustal architecture and thinning structures: insights from depth sections and drill-hole data Different interpretations have been proposed to explain the geometry of the Parentis Basin, ranging from a synclinal basin (Bois & Gariel 1994;Bois et al. 1997) to a half-graben-shaped basin bounded to the south by one major crustal-scale listric fault (interpretation of the MARCONI 3 profile: Ferrer et al. 2008). However, as already highlighted by Pinet et al. (1987), the amount of extension inferred from the few normal faults observed across the Parentis Basin cannot explain the observed extreme crustal thinning. This crucial observation led Jammes et al. (2010a, b, c) to propose an alternative interpretation, suggesting that the architecture of the basin and its extreme crustal thinning are controlled by a top-basement detachment fault, flooring the southern part of the basin. This interpretation is based on analogies with the Mauléon Basin (Jammes et al. 2009) and supported by drill-hole data eastwards of the ECORS Bay of Biscay profile (Saint Girons en Marensin and Contis boreholes: Jammes et al. 2010b). Hence, the authors suggested that the architecture of the Parentis Basin is strongly asymmetrical, related to a different evolution between an upper plate to the north and lower plate setting to the south. The upper and lower plate positions are defined with respect to a major, north-dipping detachment system (Jammes et al. 2010a, b, c).
The interpreted depth section presented in this work relies on published drill-hole data (Bois et al. 1997) and includes some of the observations previously described by Jammes et al. (2010a, b, c). Based on the identification of the main stratigraphic horizons and on the interpretation of the top basement (Fig. 7a), a depth section is constructed (Fig. 7b), using the same velocity structure as described above for the Norgasis 11-12 sections. Insights into the Moho depth come from the ESP studied by Tomassino & Marillier (1997).
The evolution of the crustal architecture of the section is well recorded in the syn-to posthyperextension sediments (a shallow carbonate platform to the south and a deep basinal domain to the north). As previously highlighted by Jammes et al. (2010a, b, c), sedimentary sequences from the northern part of the Parentis Basin show a relatively simple sag geometry, and no major fault structures can be identified. Crustal thinning is progressive from approximately 30 km (on the Armorican shelf ) to less than about 7 km in the central part of the basin (towards the Ibis Fault), coinciding with a thickening of the syn-hyperextension sediments and a deepening of syn-to post-hyperextension sequences (Curnelle 1983), indicating a progressive transition from a proximal to hyperthinned domain (Fig. 7). The architecture of the southern border of the basin is more complex. Over the Cantabrian shelf and Landes High, the post-hyperextension shallowwater carbonate platform (Curnelle 1983) overlies an approximately 30 km-thick crust and the Moho is at a depth of 35 km, indicating only a weak crustal thinning during the rift event (i.e. proximal domain). Major crustal thinning occurs only northwards of the Landes High. The Saint Girons en Marensin and Contis boreholes (Jammes et al. 2010b) that penetrated sequences of rocks characteristic of top-basement detachment faults are located in a similar structural position, at the southern border of the Parentis Basin. Hence, the initiation of thinning, northwards of the Landes High, is coincident with the existence of a high-b extensional system. To the north, the extreme crustal thinning continues towards a major crustal discontinuity, the Ibis Fault (Jammes et al. 2010a, b, c), delimiting a southwards, approximately 12 km-thick, crust from an approximately 5 km-thick crust to the north, and may indirectly indicate the occurrence of another major high-b extensional system (Fig. 7).

The reactivated stage: the Arzacq -Mauléon Basin
The Arzacq -Mauléon Basin is located in the NW part of the Pyrenean Orogen. It is one of the best-documented examples of a reactivated hyperextended rift basin (Fig. 3). Evidence of extreme crustal thinning and local mantle exhumation has been reported and discussed by several authors based on geophysical and geological approaches (e.g. Grandjean 1992;Daignières et al. 1994;Canérot 2008;Jammes et al. 2009Jammes et al. , 2010aDebroas et al. 2010;Lagabrielle et al. 2010). Jammes et al. (2010a, b, c) suggested that prior to its Pyrenean overprint the Arzacq -Mauléon Basin showed a similar basin architecture to the presentday Parentis Basin, relying on comparable geological and geophysical (mainly gravity) observations in the two basins. The Mauléon Basin was inverted as a large tectonic pop-up structure along north-and south-vergent thrusts (Fig. 8). In the north, it was thrust over the Grand Rieu High and southern Arzacq Basin along the north-vergent Saint Palais and Sainte Suzanne thrust systems (Daignières et al. 1994;Teixell 1996Teixell , 1998, and in the south over the Axial Zone and southern foreland basin along the south-vergent Lakoura thrust system (Fig. 8) (Muñoz 1992;Teixell 1998). The inversion and erosion of the Mauléon Basin enable us to have local access to the basal sedimentary sequences and basement flooring the deepest parts of the basin (e.g. Jammes et al. 2009;Debroas et al. 2010;Lagabrielle et al. 2010). In the following, we will focus on several outcrops preserving primary relationships between the sedimentary cover and underlying basement. We will characterize the nature of sediments and basement rocks, define the nature of the top basement, and identify the position of these units within the present-day nappe stack (see the  Daignières et al. 1994) that is extended into the Axial Zone and southern foreland basin, as proposed by Teixell (1996). Ga, Gavarnie Thrust; Lak, Lakoura Thrust; GR, Grand Rieu High; Gu, Guarga Thrust; SPFT, South Pyrenean Frontal Thrust; St P, Saint Palais Thrust; Ste S, Sainte Suzanne Thrust, forming the NPFT, North Pyrenean Frontal Thrust system. The geological section C-C ′ illustrates the pop-up inversion of the Mauléon Basin and the interaction between the two phases of deformation. Outcrops and field photographs are presented in Figure 10. Ab, Abarratia quarry; Ka, Kalkuetta Canyon; Me, Mendibelza Massif; Ur, Urdach Quarry. Localities: Bo, Bonloc; M, Mauléon; O, Oloron-Sainte-Marie; SP, Saint Palais. This map is modified after the BRGM (1/50 000) geological map of: Arthez de Béarn (Dubreuilh & Karnay 1997), Argelès-Gazost (Ternet 1980), Gavarnie (Majesté-Menjoulàs & Debon 1999), Hasparren (Burger et al. 1972), Iholdy (Boissonnas et al. 1974), Laruns-Somport (Ternet et al. 2004), Larrau (Casteras et al. 1970c), Lourdes (Casteras et al. 1970b), Oloron-Sainte-Marie (Casteras et al. 1970a), Orthez (Henry et al. 1989), Mauléon-Licharre (Boltenhagen et al. 1976), Morlaàs (Delfaud et al. 1982), Pau (Bureau de Recherches Géologiques et Minières 1963), Saint-Jean-Pied-de-Port (Le Pochat et al. 1978) and Tardets (Casteras et al. 1971).

Sedimentary evolution
Although the Parentis and Arzacq -Mauléon basins show many analogies from a structural point of view (Jammes et al. 2010a, b, c), the timing of extension and subsidence related to hyperextension is slightly younger in the Arzacq -Mauléon Basin (Late Aptian-Early Cenomanian) compared to the Parentis Basin (Early Cretaceous -Aptian).
The pre-hyperextension sequence, Permianlower Aptian in age, records ill-defined Triassic -Early Liassic rift events (Brunet 1984) leading to the deposition of shales, sandstones, conglomerates, limestones and evaporites (e.g. the Germanic facies : Curnelle 1983;Fréchengues 1993). The debated Jurassic -Early Cretaceous evolution is related to the deposition of carbonates and marls, locally eroded and associated with the formation of bauxites (Biteau & Canérot 2007;Canérot 2008). Thickness variations observed within these pre-hyperextension sequences range from a few hundred metres to a few kilometres. Upper Triassic evaporites (the Keuper Formation) represent an important rheological horizon that acted as a major decoupling level during Cretaceous hyperextension (Canérot et al. 2001;Fig. 9. Simplified structural map of the Mauléon Basin and of the southern part of the Arzacq Basin. The section A-A ′ is the same as in Figure 8 but only tectonic units are indicated. The location of the top-basement detachment faults identified by Jammes et al. (2009, in the Labourd Massif ) and Johnson & Hall (1989a, b, in the Mendibelza Massif ) are indicated. Outcrops and field photographs are presented in Figure 10. Ab, Abarratia quarry; Ka, Kalkuetta Canyon; Me, Mendibelza Massif; Ur, Urdach quarry. Localities: Bo, Bonloc; M, Mauléon; O, Oloron-Sainte-Marie; SP, Saint Palais. Jammes et al. 2010b) and Pyrenean compression (Casteras et al. 1970a(Casteras et al. , c, 1971Le Pochat et al. 1978). Subsalt formations are coupled and deformed together with the upper crustal basement rocks and mainly crop out south of the basin (e.g. south of the Labourd Massif, in the Jara -Arbailles area and south of the Layens: Fig. 8). In contrast, supra-salt formations are often decoupled from the underlying units and form almost east -west-trending ridges (e.g. Sarrance, Mail Arrouy: Fig. 8). The Arbailles Massif is an exception where precompressional stratigraphic sequences and their relationship with the underlying basement are locally preserved.
During the hyperextensional event, from Late Aptian to Cenomanian, sedimentation in the Arzacq Basin remained carbonate dominated, as indicated by the deposition of more than 3 km of carbonates and marls (Brunet 1984). In the Mauléon Basin, Upper Aptian -Lower Albian sequences record the progressive formation of half-graben basins, with the development of a carbonate platform over basement highs (Urgonian facies and Melobesia carbonates) and marls in the basins (e.g. Spicula marls). The progressive transition to a thick siliciclastic deltaic system records a rapid creation of accommodation space and strong subsidence during the onset of hyperextension. During this stage, the basin was filled in its southern part by a prograding deltaic system associated with conglomeratic facies (e.g. the Mendibelza conglomerates and lateral equivalent), while in the deeper parts of the basin further to the north turbidites are deposited (Fig. 8). These siliciclastic deposits, referred to as 'Flysch Noir deposits' (e.g. Souquet et al. 1985), progressively filled the Mauléon Basin until Cenomanian time. The main depocentre to the north of the basin (near the city of Mauléon) can reach a total compacted thickness of 4 km (Roux 1983;Fixari 1984;Souquet et al. 1985).
Alkaline magmatism occurs in the northern part of the Mauléon Basin at the transition from synto post-hyperextension. The post-hyperextension sequence (Cenomanian-Coniacian) records a major break in the sedimentation of the Mauléon Basin, with the transition from siliciclastic to carbonatedominated sedimentation. This source change is recorded by the development of a shallow-water carbonate platform (c. 400 m, e.g. the Calcaire des Cañons platform) to the south of the basin directly overlying the basement (in the Axial Zone). Concurrently, calciturbiditic and hemipelagic systems filled the northern part of the basin.
The deposition of Santonian breccia and calcarenites unconformably over a karstic substratum records the onset of Pyrenean convergence. Upper Cretaceous -Tertiary flysch-type sedimentation results from the uplift and erosion of the axial domain. Syn-orogenic deposits are mainly preserved in the Jaca foreland basin to the south and to the north of the Saint Palais and Sainte Suzanne frontal thrust system forming the presentday Aquitaine Basin. In the Mauléon Basin, the last evidence for compressional deformation are in the Eocene -Early Oligocene Teixell 1998), whereas it continued until the Late Oligocene -Early Miocene in the southern Pyrenees (Teixell 1996(Teixell , 1998.

Nature of basement rocks
Different types of basement rocks are cropping out in the Mauléon Basin. Palaeozoic metasediments locally intruded by carboniferous plutons (e.g. Eaux-Chaudes granite: Ternet et al. 2004) are representative of pre-hyperextension upper crustal rocks, cropping out in the Jara and Arbailles massifs, and further to the south and SW in the Axial Zone and the southern Basque Massif (e.g. the Aldudes Massif, Fig. 8). Metasedimentary and metabasic granulites (Boissonnas et al. 1974) described in the Labourd Massif by Vielzeuf (1984) indicate conditions of equilibration prior to hyperextension corresponding to a depth of approximately 20 km (mineral assemblages of the basic granulites 775 + 50 8C and 6 + 0.5 kbar) and, hence, representative of pre-hyperextension mid to lower crustal rocks. Numerous outcrops of mantle-derived rocks (of variable size) can also be observed in the basin, mainly derived from lherzolite and websterite ultramafic protolites, and showing a strong hydrothermal alteration with various degrees of serpentinization (Fabriès et al. 1998) (for the distribution of the mantle rocks, see Fig. 8). In the Mauléon Basin, mantle outcrops occur in the northern part of the basin within thrust sheets associated with strongly altered crustal rocks (,100 m thick) at the base of the pre-and syn-hyperextensional sediments (e.g. Sarrance, Mail Arrouy).

Deformation history
Any attempt to reconstruct the pre-Pyrenean riftrelated history depends on the ability to unravel the deformation history and to distinguish Pyrenean compressive structures from rift-related ones. In the following, we will focus mainly on a northsouth geological section located in the eastern part of the Mauléon Basin (Figs 8 & 9), where the different compressional events are well preserved, defining coherent compressive tectonic units within the large-scale Mauléon pop-up structure (Fig. 9). Two opposite directions of thrusting can be observed. An older set of south-directed thrusts are overprinted by north-directed ones (Figs 8 & 9, geological section and interpretation of the ECORS Arzacq profile, Fig. 8). Hence, the initial thrust sheet emplacement related to the initial shortening of the north -south Pyrenean compression is delimited by south-directed thrusts that can be mapped throughout the basin (e.g. the Sarrance, Mail Arrouy or Lakoura thrust systems that are all Late Cretaceous: Teixell 1998) (Figs 8 & 9). The recognition of this initial phase of deformation is crucial as it can be used to define coherent compressive tectonic units that sample remnants of the former rift system (Fig. 9). With increasing shortening, a second phase of deformation led to the formation of mainly north-directed thrusts in the Mauléon Basin that may either deform or overthrust former south-directed thrusts (e. g. the Ossau Thrust (Canérot et al. 2001) or north of the Labourd and Arbailles, or, at a larger scale, the Saint Palais and Sainte Suzannne thrusts) (Figs 8 & 9) and is also related to the formation of southdirected backthrusts in the Axial Zone, responsible for the formation of the Jaca foreland basin (Teixell 1996).

Remnants of the former rift system in the Pyrenean nappe stack
A structural analysis of the Mauléon Basin enables the mapping of compressive structures defining the present-day position of coherent tectonic units within the reactivated Mauléon Basin (Fig. 9). The recognition and restoration of these units back into their precompressional position is essential as it also defines their initial place in the previous rift system. The identification of precompressional relationships preserved in outcrops and observations on the nature of basement and sediments provide crucial information on the rift-related history that would not be obtained through the restoration of compressional structures only. Within each unit, key outcrops preserving the nature of the contact between basement and pre-, syn-and posthyperextensional sediments (e.g. Canérot 2008;Jammes et al. 2009;Debroas et al. 2010;Lagabrielle et al. 2010) can be found. In the following, we describe the main tectonic units and associated outcrops going from a lower to an upper structural position with reference to the first phase of compressional deformation (i.e. from more external to more internal positions in the orogen : Figs 9 & 10).
The most external position is preserved in the Axial Zone and is delimited to the north by the Lakoura thrust system (Teixell 1998) and lateral equivalents. The basement is made of pre-hyperextension upper crustal rocks directly onlapped by post-hyperextension shallow-water platform carbonates (for outcrop observations see Fig. 10a, b and for the location see Figs 8 & 9). The erosion of the top basement and the deposition of posthyperextension shallow-water sediments indicate that over this domain no accommodation space was formed during syn-hyperextension time. This observation suggest that the crust was not or was only weakly thinned, indicating that this domain was in a proximal position relative to the basin (Jammes et al. 2009).
The Bedous -Mendibelza and Layens-Labourd units correspond to higher (i.e. more internal) positions, and are separated from more internal ones by the south-directed Sarrance Thrust (Fig. 9). Western and eastern structural equivalents are delimited by north-directed thrusts (e.g. north of the Arbailles and the eastern part of the Labourd massifs: Fig. 9). On the geological section (Fig. 9), the two units are delimited from each other by a south-directed thrust preserved at the base of the Layens, overprinted westwards and eastwards by north-directed thrusts (e.g. north of the Mendibelza Massif: Fig. 9). The Bedous-Mendibelza unit is made of thick aggradational sedimentary sequences (c. 1 km) that are deposited on to pre-hyperextension upper crustal rocks and sediments corresponding to the Mendibelza and Igountze massifs. Syn-hyperextension sequences consist in their southern proximal part of massive conglomerates (the Mendibelza Formation: Fig. 10c), whereas their northwards time equivalents are turbidites and marls (Fig. 10d). The contact between the upper crustal rocks and the onlapping sediments is formed by a cataclastic fault zone (Fig. 10f )   Fig. 10. Field photographs and simplified cartoons of key observations from remnants of the former rift system preserved in the Pyrenean nappe stack (location of outcrops is shown in Fig. 8). associated with tectonosedimentary breccias (Fig. 10e). This contact was interpreted as a topbasement detachment fault by Johnson & Hall (1989a, b) and Miranda-Avilés et al. (2005). This interpretation was made in analogy with supradetachment basin structures described in the Basin and Range. Hence, the progressive thickening and deepening of sediments onlapping on to exhumed basement rocks indicate the progressive creation of accommodation space along large offset normal faults (Fig. 10g).
In contrast, in the Labourd Massif area, in the Layens-Labourd unit, granulites derived from the pre-hyperextension mid-to lower crust (Vielzeuf 1984) are onlapped by late syn-to post-hyperextension sediments. Jammes et al. (2009) described in detail the nature of the contact between basement and sediments, and identified a major brittle fault zone preserving syntectonic breccias and onlapping sediments (Abarratia quarry: Fig. 10h, i). The authors interpreted the mid-to lower crustal rocks as being exhumed along a top-basement detachment fault and later resedimented into syn-hyperextension sediments (e.g. Bonloc breccia : Claude 1990). This observation suggests that upper crustal rocks were removed and thinned along large-offset normal faults (Fig. 10j).
The Sarrance -Mail Arrouy unit (Fig. 9) and its structural equivalents to the west (e.g. Roquiague) correspond to the most internal units, delimited to the north by the Saint Palais Thrust (Fig. 9). Within this unit, crustal basement rocks are rare and only represent thin slices (usually less than 100 m thick) associated with mantle outcrops. Key observations come from the Urdach area ( Fig. 10) (Jammes et al. 2009;Debroas et al. 2010;Lagabrielle et al. 2010), where the occurrence of mantle clasts into a sedimentary breccia within the late syn-hyperextension sedimentary cover (Fig. 10k, l) suggests that, at least locally in the Mauléon Basin, mantle rocks had to be exhumed to the seafloor during Cretaceous hyperextension. The sediments of this unit are mainly represented by thick accumulations of syn-hyperextension deep-water sediments (in the western part of the basin) and prehyperextension sediments representing east-west ridges thrust southwards (e.g. Mail Arrouy and Sarrance) over upper Triassic evaporites, interpreted as extensional allochthons by Jammes et al. (2009). The large accumulation of syn-hyperextension sediments (e.g. near Mauléon city : Roux 1983;Fixari 1984;Souquet et al. 1985) indicates the creation of large accommodation space. This observation goes along with the occurrence of mantle rocks reworked in the sediments. All of these observations suggest that the central part of the Mauléon Basin may have been floored at least locally by exhumed mantle rocks (Fig. 10m).

Synthesis: observations and diagnostic elements for offshore and onshore systems
Interpretations of onshore data often use templates derived from present-day continental rifted margins, while seismic interpretations are often inspired by onshore observations. The use of analogues has a long tradition in Earth Sciences, and has been used successfully for the interpretation of proximal rifted margins and oceanic crust. However, in the case of hyperextended domains, the use of analogues is still hampered by the lack of well-studied examples and access to critical data, such as drill holes. We have focused on the Bay of Biscay and Western Pyrenean rift system to describe and characterize quantitatively and qualitatively onshore and offshore observations. Geophysical techniques (Western Approach margin: Figs 4, 5 & 6), geological observations (Mauléon Basin: Figs 8,9 & 10) or both types of approaches (Parentis Basin: Fig. 7) were applied to different stages of this rift system. Based on results from these examples, observationdriven diagnostic elements are pointed out and used to define different domains across rift sections, enabling geologically more coherent and physically better constrained interpretations to be proposed.

Offshore geophysical observations: defining limits of domains
Geophysical observations depend on the definition of first-order interfaces (sea level/seafloor/top basement/Moho), their morphology and structures affecting them to provide estimations of crustal thickness, lithosphere thinning, and accommodation space (Fig. 11). As exemplified by the Norgasis 11 -12 section, the quantification of crustal thickness and lithosphere thinning allows three domains to be distinguished: (1) the proximal domain that is weakly thinned; (2) the necking and hyperthinned domain where most of the thinning occurs; and (3) the exhumed mantle or oceanic domain where complete thinning of the continental crust is achieved (Figs 5 & 11). In magma-poor rifted margins, accommodation space is, to first order, controlled by crustal thickness, sedimentary load and postbreak-up thermal subsidence. The total accommodation space generally increases oceanwards, as continental crustal thinning increases (Fig. 6). The distinction between the exhumed mantle (oceancontinent transition) and the oceanic domains corresponds to a morphological boundary and step in basement topography (Figs 4 -6), with smaller values of accommodation space in the oceanic domain. Assumptions on extensional settings (highv. low-b extensional setting: Wilson et al. 2001) depend on basin architecture ('hyperextended sag' basin or half-graben-type basin: Fig. 2). Low-b extensional settings can be observed throughout the margin (e.g. Norgasis 11 -12, Parentis Basin), whereas high-b extensional settings are observed where major thinning occurs: in the necking domain, over hyperthinned crust or in the exhumed mantle domain (Figs 4 & 7).

Geological observations: definition of limits and domains
In remnants of hyperextended rift systems, the interpretation relies on geological observations of the nature of sediment and basement, and of their interface, as well as on their position within the compressional nappe stack (Fig. 11), as exemplified by the Mauléon Basin (Figs 8-10). Estimates of the total accommodation space created during rifting rely on combined observations from the postrift depositional environment and aggradation history (sediment thickness). Only considering the depositional environment could be misleading. Even if post-rift shallow-marine sediments are mainly observed in proximal domains, in the case of a large aggradation during rifting, shallowmarine sediments could mask the large creation of accommodation space. Similarly, evaporites are often used as an indicator of global sea level; however, where the base level of these basins is no longer controlled by the global sea level, even high-b basins overlying hyperthinned domains may be associated with the formation of evaporites (Reston 2010). Basement nature changes with crustal thinning but it cannot be used as a standalone criterion to define domains, as only the occurrence of mantle rocks can indicate a complete crustal thinning. Continental crust (upper and lower crust) can be encountered in all domains (except for the oceanic domain) either in a footwall position or as extensional allochthons in a hanging-wall position in the exhumed mantle domain. Finally, the nature of the contact between the basement and overlying sediments is critical as it enables the identification not only of fault structures (i.e. extensional settings) but also basement erosion and magmatic additions. The identification of top-basement detachment fault surfaces is critical as it occurs in the hyperextended domain within the necking or hyperthinned and exhumed mantle domains. Subareal erosion of pre-rift sediments or basement during rifting indicates that there was no creation of accommodation space and is characteristic of the proximal domain but may also occur in the necking domain during early stages of the rift evolution (Péron-Pinvidic & Manatschal 2009). Magmatic additions are dominant in the oceanic domain, may also occur in the exhumed mantle or hyperthinned domains, and possibly also in the necking and proximal domains in the case of magma-rich margins. None of the criteria presented above can be used as standalone criteria to unambiguously replace an outcrop back into its rift context. Only by combining the different observations can more precise assumptions about structural domains be made.

Additional constraints
Relying only on the direct and indirect approaches proposed in this study, the distinction between the necking and hyperthinned domain is not straightforward. The distinction between the necking and hyperthinned domain is based on the progressive oceanward embrittlement of the continental crust (Pérez-Gussinyé et al. 2001) controlling the deformation mode of the continental crust (Sutra et al. 2013). Deformation in the necking domain is decoupled at mid-crustal ductile levels, while deformation in the hyperthinned domain is coupled at the scale of the crust (faults can cut through the embrittled continental crust and penetrate into the mantle: Pérez-Gussinyé et al. 2001). Sutra et al. (2013) showed that the change from decoupled to coupled deformation is accompanied by polyphase deformation and a change in the age of the syntectonic sediments, which get younger in the hyperextended domain going oceanwards. This tectonic migration of the deformation is reported both offshore of the Iberian margin and also in onshore remnants preserved in the Alps (Mohn et al. 2010;Masini et al. 2013). In the Bay of Biscay, the deeper parts of the margin have not been drilled but it is likely that in this basin the syntectonic sediments observed on the proximal margin are older than the ones overlying the distal margin.
Implications of a combined geologicalgeophysical approach for the interpretation of offshore and onshore sections Indirect and direct approaches were applied to describe the different rift domains on three examples (Fig. 12). The Parentis Basin was the only case for which both geological and geophysical approaches could be used to interpret the rift-related structure and evolution of the basin. Interpretations for the Western Approach margin and the Mauléon palaeo-rift system are proposed hereafter based on the combination of geological insights and quantitative estimations of accommodation space and crustal thickness derived from the study of the three sections (Fig. 12).

The Parentis Basin: combined approaches
The combination of drill-hole observations with reflection and refraction seismic data and depth conversion enabled the identification of three different domains on the ECORS Bay of Biscay section. As previously proposed by Jammes et al. (2010a, b, c), the Parentis Basin appears strongly asymmetrical with a northern sag basin architecture, where domains are transitional and have a more complex southern geometry. to the south. Most of the crustal thinning occurs to the north and is interpreted as resulting from another top-basement detachment fault rooting towards the Ibis structure at the centre of the basin (Fig. 12). The poor imaging of the ECORS Bay of Biscay profile north of the Landes High can be related to important salt tectonics enabling the decoupling of pre-hyperextension sediments from the underlying basements and diapir formation (e.g. Mathieu 1986;Biteau et al. 2006).

The Western Approach margin: insights from outcrop observations
Based on indirect approaches (i.e. estimations of crustal thickness, lithosphere thinning and accommodation space) as well as on basin morphology, several domains were successfully defined. Nevertheless, as exemplified by the Parentis example, geological insights can greatly improve the usual low resolution of seismic interpretation (Fig. 12). On the Norgasis 11 -12 section, basement morphology is ill defined in the 'proximal domain' except for one major normal fault. Since no major thinning seems to be associated with it, it is interpreted to be decoupled at mid-crustal levels. Thinning is initiated over the Meriadsek terrace, where a high-b extensional setting can be identified (Figs 4b & 12) and interpreted as a top-basement detachment fault structuring the 'necking domain'. The transition to the 'hyperthinned domain' is not straightforward based on thinning arguments only, but further constraints may come from the description of the synrift stratigraphic architecture overlying the two domains. Over the Meriadzek terrace, in the necking domain, sediments within the basins show evidence of syn-to post-tectonic geometries, whereas southwards, in the hyperthinned domain, sediments of the same age are tilted in a pre-tectonic position. This change in synrift stratigraphic architecture from the necking to hyperthinned domain is also described from the West Iberia margin and related to the polyphased evolution of rifting (Sutra & Manatschal 2012). The transition to the 'exhumed mantle domain' corresponds to the complete thinning of the continental crust (Fig. 5d). Gravity inversion predicts an inflection in the gravity Moho on the depth section (Figs 5 & 12). The distal high-b extensional setting identified south of Trevelyan is suggested to represent a topbasement detachment fault responsible for the exhumation of mantle rock overlain by the youngest syn-hyperextensionsediments(Aptianinage: Thinon et al. 2002). Finally, the transition to the 'oceanic domain' is characterized by a ramp of the topbasement morphology and the direct onlapping of the post-rift sediments. This transition corresponds to the progressive magmatic overprint of the exhumed mantle domain (Fig. 12).
The Mauléon Basin: estimations of accommodation space and crustal thickness  Souquet et al. 1985) indicates the progressive creation of accommodation space that could be related to the thinning of the crust, as observed in the 'necking or hyperthinned domain'. Finally, remnants of exhumed mantle were also observed and described by Jammes et al. (2009). The occurrence of mantle outcrops preserving crust only a few hundred metres thick and their reworking into syn-hyperextension sediments indicate that, at least locally, the thinning of the continental crust was complete. Using quantifications of accommodation space at break-up time (Norgasis 11 -12 section, Fig. 6d), a total accommodation space of 3-4 km may be expected in the exhumed mantle domain, increasing to 6-7 km after post-rift thermal subsidence (Fig. 12).

Towards a reconciliation of observations from present-day and fossil analogues for rift domain definition
The investigation of different case studies using geophysical approaches, geological observations and a combination of both (Fig. 12) enabled us to Fig. 13. Conclusive diagram synthesizing the geophysical quantitative (upper part) and geological qualitative (lower part) characteristics of structural domains at continental rifted margins, as proposed in this study. In the central part this geological -geophysical approach is applied to aid the interpretation of an example of a continental rifted margin section (Norgasis 11-12 profile) and to suggest additional geological insights at an outcrop scale (based on geological observations). This combined approach can therefore be used as a first-order method to provide geological insights into rift structures, and the nature of basement and sedimentary rocks for seismic interpretations, and also to estimate crustal thickness and accommodation space in onshore fossil remnants.
propose a self-consistent approach to define rift domains that is both observation driven and semiquantitative (Fig. 13). Indirect quantification based on the Norgasis 11 -12 profiles are proposed in the upper part of the figure, whereas the lowermost part presents an overview of key geological observations relying on observations from onshore study (mainly the Mauléon Basin). Diagnostic elements (from direct and indirect datasets) enabling domain characterization have already been discussed; however, a critical point is that unambiguous characterization of domains relies on the combination of observations. We believe that this synthesis of geological and geophysical observations can be used to improve the interpretation of domains both in present-day rifted margins (e.g. the Western approach section (Fig. 12) and the central part of Fig. 13) and in collisional orogens that contain remnants of rift domains (e.g. the Mauléon Basin: Fig. 12).

Main limitations
The Bay of Biscay and Western Pyrenees case study described in this paper can be considered as representative of an equilibrated magma-poor hyperextended rift system: that is, magmatic additions are of subordinate importance during the hyperextensional episode (e.g. de Charpal et al. 1978;Montadert et al. 1979b;Avedik et al. 1982;Le Pichon & Barbier 1987;Thinon et al. 2003). In magmapoor thermally equilibrated margins, there is a first-order relationship between lithosphere thinning, continental crust thickness, 'rift-related accommodation space' and the nature of extensional setting (low-or high-b extensional settings of Wilson et al. 2001). Therefore, observations and interpretations resulting from direct and indirect approaches discussed in this study cannot be applied in a straightforward way to all rifted margin settings. Indeed, in the case of magma-rich systems (e.g. conjugate margins in the NE and southern South Atlantic: for a review of volcanic margins see Geoffroy 2005 and references therein) crustal thickness evolution during rifting depends on lithosphere thinning processes but also on magmatic additions. As a result, the relationship between accommodation space and lithosphere thinning is not straightforward and may be evaluated only if magmatic additions can be quantified. Furthermore, late magmatic additions to the system may mask and obliterate rift-related structures.
Similarly, in the case of thermally unequilibrated margins or in the case of margins that show a strong dynamic topography, the relationship between the creation of accommodation space and lithosphere thinning may be more complex. Indeed, the quantification of accommodation space during rifting (as in Fig. 6) depends on estimations of the dynamic topography (Table 2) or it may result in completely underestimated (dynamic uplift) or overestimated values (dynamic subsidence).
In-sequence rift systems are considered to imply that the deformation is polyphase and progressively focusing towards the location of final break-up (e.g. Péron-Pinvidic & Manatschal 2009). Therefore, the domains described here (Fig. 1) represent genetic domains: that is, recording different time steps that formed during the evolution of one rift system. Hence, in the example of early magmatic addition during rifting, some domains may not be observed (e.g. domain of exhumed mantle) or are masked by late magmatic additions.
As exemplified by the Mauléon case, the reconstruction of the rift-related history in deformed analogues depends on the ability to distinguish compressive structures from the rift-related ones but it is also necessary to unravel the different sequences of the deformation history. The restoration of the system prior to the first phase of deformation will define the initial sampling of the rift system. However, restorations of compressional structures only provide the original position of units relatively to each other as the horizontal movement along the thrust systems cannot be quantified precisely.

Conclusions
The aim of this study was to reconcile offshore-and onshore-type observations to further characterize structural domains at present-day continental rifted margins and in fossil examples exposed in collisional orogens. We used the Bay of Biscay and Western Pyrenees as a natural laboratory where access to seismically imaged, drilled and exposed fossil parts of one and the same hyperextended rift system is possible. We focused on offshore and onshore examples (the Western Approach margin, and the Parentis and Mauléon basins), relying on the interpretation of different type of datasets with different resolutions and observational scales (from outcrop to seismic scale). For the Western Approach margin and Parentis Basin examples, we used geophysical quantitative techniques (gravity inversion and flexural backstripping) combined with seismic interpretations. We estimated accommodation space, crustal thickness and lithosphere thinning, and identified two types of extensional settings (low-and high-b settings: Wilson et al. 2001). Based on the onshore Mauléon Basin and drilled parts of the Parentis Basin, we focused on key outcrops to describe the nature of sediments and basement rock, and of their interface (stratigraphic, tectonic or magmatic). These observations provide the necessary diagnostic elements to characterize qualitatively and quantitatively five domains: proximal, necking, hyperthinned, exhumed mantle and oceanic. Finally, we believe that this geological-geophysical approach (illustrated in Fig. 13) can be used in a more general way to: (1) place onshore fossil remnants into a rifted margin context, resulting in first-order predictions of the crustal architecture; and (2) suggest geological insights into rift structures, and the nature of basement and sediments for seismic interpretations of continental rifted margins where no drill-hole data are available.