The Tell-Rif orogenic system (Morocco, Algeria, Tunisia) and the structural heritage of the southern Tethys margin

The Tell-Rif (Tell in Algeria and Tunisia; Rif in Morocco) is the orogenic system fringing to the south the West Mediterranean basins. This system comprises three major tectonic-palaeogeographic zones from north to south: (1) the internal zones (AlKaPeCa for Alboran, Kabylies, Peloritan, Calabria) originating from the former northern European margin of the Maghrebian Tethys, (2) the “Flyschs zone” regarded as the former cover of the oceanic domain and (3) the external zones, forming the former southern Maghrebian Tethys margin more or less inverted. The Tell-Rif is interpreted as the direct result of the progressive closure of the Maghrebian Tethys until the collision between AlKaPeCa and Africa and, subsequently, the propagation of the deformation within Africa. This gives a consistent explanation for the offshore Neogene geodynamics and most authors share this simple scenario. Nevertheless, the current geodynamic models do not completely integrate the Tell-Rif geology. Based on the analysis of surface and sub-surface data, we propose a reappraisal of its present-day geometry in terms of geodynamic evolution. We highlight its non-cylindrical nature resulting from both the Mesozoic inheritance and the conditions of the tectonic inversion. During the Early Jurassic, we emphasize the development of NE-SW basins preceding the establishment of an E-W transform corridor connecting the Central Atlantic Ocean with the Ligurian Tethys. The Maghrebian Tethys developed just after, as the result of the Late Jurassic-Early Cretaceous left-lateral spreading between Africa and Iberia. By the Late Cretaceous, the occurrence of several tectonic events is related to the progressive convergence convergence between the two continents. A major pre-Oligocene (pre-35 Ma) compressional event is recorded in the Tell-Rif system. The existence of HP-LT metamorphic rocks associated with fragments of mantle in the External Metamorphic Massifs of the Eastern Rif and Western Tell shows that, at that time, the western part of the North-African margin was involved in a subduction below a deep basin belonging to the Maghrebian Tethys. At the same time, the closure of the West Ligurian Tethys through east-verging subduction led to a shift of the subduction, which jumped to the other side of AlKaPeCa involving both East Ligurian and Maghrebian Tethys. Slab rollback led to the development of the Oligo-Miocene back-arc basins of the West-Mediterranean, reworking the previous West Ligurian Tethys suture. The docking of AlKaPeCa against Africa occurred during the Late Burdigalian (17 Ma). Subsequently, the slab tearing triggered westward and eastward lateral movements that are responsible for the formation of the Gibraltar and Tyrrhenian Arcs respectively. The exhumation of the External Metamorphic Massifs occurred through tectonic underplating during the westward translation of the Alboran Domain. It resulted in the formation of both foredeep and wedge-top basins younger and younger westward. The lack of these elements in the eastern part of the systems signs a different evolution dominated by frontal accretion. In the discussion, we precisely address the origin of the non-cylindrical behavior of the orogenic system and question the mechanisms explaining at large scale the phases of coupling/uncoupling between the major plates.

The Tell-Rif (Tell in Algeria and Tunisia; Rif in Morocco) or Maghrebides is the orogenic system fringing the West Mediterranean basins to the south (Durand- Delga, 1969;Wildi, 1983). This system comprises 3 major tectonic-palaeogeographic zones, i.e., from north to south: (1) the internal zones (AlKaPeCa) originated from the former northern European margin of the LMT; (2) the "Flyschs zone" regarded as the former sedimentary cover of the LMT and (3) the external zones, the former North African passive margin of the LMT (Fig. 2).
In the geodynamic frame of the West Mediterranean basins formation, the Tell-Rif is interpreted as the direct result of the progressive closure of the LMT until the collision of AlKaPeCa with Africa at 17 Ma. Such a scenario gives a consistent explanation for the offshore geodynamics and is now shared by almost all the authors. Some discussions remain about the importance of the westward movements (i.e. the initial position of Alboran terrane [the "Al" of the AlKaPeCa system]) and their precise timing. The difference between reconstructions by Van Hinsbergen et al. (2014) and Crespo-Blanc et al. (2016) are considerable in terms of westward migration of the Alboran domain but are end-members of a single model. This consensus has been recently broken by Vergés and Fernàndez (2012) followed by Casciello et al. (2015), who propose a completely different model. Their model suggests that Alboran was initially (i.e. during the Mesozoic) located along the African margin and consequently separated from the Kabylie, which, together with the Calabrian and Peloritan terranes, were on the European side of the Maghrebian Tethys. This provocative proposition is interesting but disregards several geological evidences: À the strong palaeogeographic affinities between Alboran and the Kabylie during the whole Mesozoic, especially during Triassic time, these two domains being characterized by "Alpine" facies contrasting with the "Germanic" facies which characterize both African and Iberian margins (Bouillin, 1986;Durand-Delga, 1969); À offshore data (Booth-Rea et al., 2007;Medaouri et al., 2014), suggesting rather WSW than NW movements of the Alboran Domain; À seismic data showing that Kabylie-like terranes can be currently followed offshore up to the Oran meridian , i.e. in the assigned root zone of Alboran in the Vergés and Fernàndez (2012) model.
For these reasons, we do not retain this scenario and the present paper follows the frame of classical models consider-    ing the European origin for all the AlKaPeCa domain. Nevertheless, these latter geodynamic models fail to integrate at least two major issues: they do not incorporate all recent geological developments in the Betics geology and forget partially or totally the Tell-Rif geology. We will not develop the Betics geology here. As a matter of fact, dealing with the Rif history for Mesozoic and Cenozoic times makes more sense in comparison with the evolution of the Tell, as they both belong to the South Tethys margin. The comparison between Rif and Betics is an important topic but mainly for the recent post-Early Miocene stages when Rif and Betics share several common events. For the sake of geological coherence since we consider the evolution of the South Tethys margin, we propose a renewed view of both the Tell and the Rif evolution, with a particular emphasis on the following points: À the importance of the Paleozoic and early Mesozoic structural inheritance, which strongly influenced the southern Tethys margin geometry; À the significance of pre-Late Oligocene  compressional events not only in the Atlas System, where they are well established (see a review in Frizon de Lamotte et al., 2000, 2009, 2011, but also in the Tell-Rif system, where their effects were often ignored or minimized; À the existence of high pressure-low temperature (HP-LT) metamorphic suite, typical of subduction zones,  associated with fragments of serpentinites in the Eastern Rif (Michard et al., 1992(Michard et al., , 2007 and likely in the Western Tell (Fenet, 1975;Guardia, 1975;Michard et al., 2006Michard et al., , 2007. This suggests that the southern Maghrebian Tethys margin was involved in subduction. The metamorphic rocks are preserved in the so-called "Massifs Externes à schistosité" of the French authors (hereafter called External Metamorphic Massifs [EMM]), strung out over Eastern Rif and Western Tell (Fig. 2); À the presence of wide Miocene basins associated with the EMM and developed on top of the orogenic system in the Rif and western Tell (Fig. 2). Among these basins, the Chelif Basin occupies a large part of the western Tell in Algeria (Neurdin-Trescartes, 1992;Perrodon, 1957).
To address these topics, we will first examine the structure and geometry of the Tell-Rif system, emphasizing its noncylindrical character. Then, we will establish new and/or revised facts explaining the changes in the lateral organization of the orogenic system. On this basis, we will propose an updated set of palaeogeographical maps for the Cenozoic evolution of the West Mediterranean. Key geodynamics consequences of our review will be discussed at the end of this contribution.
2 Geological setting À overall geometry of the Tell-Rif The Tell is a linear mountain belt striking E-W. By contrast, an important curvature exists in the Rif, which swings from E-W in the eastern Rif to N-S in the western Rif (Fig. 1). This feature is the southern branch of the "Gibraltar Arc" or "Betic-Rif Arc" (Durand- Delga and Fontboté, 1980), which extends northward in the Betics where another 90°apparent clockwise swing is observed. These are late-acquired Neogene features related to the Alboran motion and associated vertical axis rotations (e.g. Cifelli et al., 2016;Platt et al., 2013). The Betic-Rif belts thus defines a very arcuate orogenic system (Lonergan and White, 1997). The Tell also extends in Sicily towards the Apennines, with a "symmetrical" 90°anticlockwise swing within the Calabrian arc. In the literature, the Tell-Rif belt is presented as a cylindrical orogenic system with continuous palaeogeographic-palaeotectonic zones, namely the internal zones (AlKaPeCa), the Flyschs zone and the external zones (see the masterly synthesis of Wildi, 1983). However, these zones are not continuous and very important along strike variations occur in the external zones ( Figs. 2 and 3). Relying on the available literature and new cross-sections along the belt, we will describe the first-order geometry of the system and its lateral variations from the internal zones toward the external ones in the following sections.

The AlKaPeCa and Flyschs domains
The internal zones of the Tell-Rif belt, shared with the Betics, belong to the AlKaPeCa domain, as defined by Bouillin (1986). AlKaPeCa is a continental domain with European affinities (Bouillin, 1986;Michard et al., 2002Michard et al., , 2006. Except from a dismembered Mesozoic to Eocene cover (the so-called "Dorsale Calcaire"), unconformably capped by upper Oligocene to lower Miocene clastics, it mainly shows Paleozoic rocks distributed into two major nappes complexes. They crop out in Lesser Kabylie (Bouillin, 1979;Bouillin and Perret, 1982;Caby and Hammor, 1992;Mahdjoub et al., 1997), Greater Kabylie (Saadallah and Caby, 1996) and Internal Rif (review in Michard et al., 2006). The upper nappe complex is poorly or non-metamorphic, only recording Variscan events. However, close to the basal contact with the lower nappe complex an Early Miocene metamorphic event widely overprinting also the lower nappe can be recognized (see Chalouan and Michard, 1990;Negro et al., 2006 for the Rif;Saadallah and Caby, 1996 for the Lesser Kabylie; among others Caby et al., 2001;Fernandez et al., 2016 for the Edough Massif). In contrast, the lower nappe (internal Rif: Sebtides; Greater Kabylie: Lower Unit; Lower Kabylie: Beni Ferguen nappe) shows a general higher grade and a clear polyphased history. It is separated from the upper nappe through an extensional shear zone. This lower nappe still records remnants of the Variscan events (e.g. Fernandez et al., 2016;Mahdjoub et al., 1997;Montel et al., 2000;Peucat and Bossière, 1981;Saadallah and Caby, 1996;Sánchez-Gómez et al., 1999;Zeck, 1999). However, it is mainly characterized by the signature of the Alpine events, ranging from HP-LT for the earliest events (Bruguier et al., 2017;Caby et al., 2014;Mahdjoub et al., 1997) to HT-LP for the most recent ones during Early to Middle Miocene (Fernandez et al., 2016;Michard et al., 2006;Platt et al., 2013;Rossetti et al., 2010 with references therein). The Edough massif presents a peculiar case. It has either been interpreted as an external or an internal massif (Bruguier et al., 2017;Caby et al., 2001;Caby et al., 2014;Hilly, 1962;Michard et al., 2006;Vila, 1970). This difficulty stems from the poorly defined contact between the external zones and the Edough massif. For Vila (1970) Fig. 3. Cross-sections of the Tell-Rif system. Sections were built using surface geology, wells and property reflection seismic data. From west to east: (a) eastern Rif belt section cutting through the Flyschs unit, the Ketama unit, intrachain Neogene basins (Taounate) and associated Jurassic units, and Tsoul nappe on top of the Late Miocene foredeep. Main map reference is Suter (1980aSuter ( , 1980b completed by our team fieldwork; (b) Western Algerian Tell section, from the offshore, the Dahra Massif, Miocene Chelif Basin, and Tellian nappes resting on the early Miocene foredeep. Major map references are within the works by Delteil (1974), Fenet (1975), Polvêche (1960) and Thomas, (1985); (c) Eastern Algerian Tell section, from the offshore through the Lesser Kabylie, Flyschs unit, the Tellian nappes, and the Atlas domain. Main map references are within the work by Bouillin (1979) and Vila (1978Vila ( , 1980; (d) Tunisian Tell section, from the offshore through the Numidian nappe, the imbricated frontal units, the Medjerda salt area, the Atlas domain. Principal map reference is the map by Rouvier (1977) for the Tunisian Tell. metamorphic cover belonging to the South Tethys margin. Caby et al. (2001) demonstrated that this contact is not a stratigraphical one but tectonic, still approving the external nature of the Edough Massif. They considered it to be an African metamorphic core complex, developed after the emplacement of the Tellian nappes (and following them, among others, Caby et al., 2014;Fernandez et al., 2016;Bruguier et al., 2017). Following Michard et al. (2006) on the basis of metamorphic data and Paleozoic remnants within the massif, it is considered as belonging to the internal zones rather than the external ones. The recent geochemical and Oligocene-Miocene dating results also converge more with an internal origin (Bruguier et al., 2017;Fernandez et al., 2016). Sandwiched between AlKaPeCa and the external domain (Figs. 2 and 3), the Flyschs domain is consensually viewed as the former cover of the Maghrebian Tethys and partially of the East Ligurian Tethys (i.e. the "Bassin des Flyschs" of the French authors), separating Europe and Iberia from the African plate up to the Miocene (Bouillin et al., 1970;Durand-Delga and Fontboté, 1980;Guerrera and Martín-Martín, 2014 for a review). The flyschs are divided into two units, namely an internal "Mauretanian" unit, whose stratigraphic age spans from Jurassic to Late Miocene, and an external "Massylian" unit including the so-called "Numidian Formation", often detached from its substratum and dated as Late Oligocene to Early Miocene (Guerrera and Martín-Martín, 2014;Thomas et al., 2010). The Massylian unit shows a progressive lateral transition towards the external domains of the South Tethys margin.
Contrary to the older parts of the flyschs, deposited in a passive margin environment, the post-Eocene sedimentation directly recorded the orogenic activity. The basin is fed by immature turbiditic sandstones in the internal parts (Guerrera et al., 2005) and ultramature sandstones from the African craton in its external part (Belayouni et al., 2013;Guerrera et al., 2012;Guerrera and Martín-Martín, 2014;Thomas et al., 2010). The question of whether or not the Maghrebian Tethys was floored by an oceanic substratum is still open. Mafic and mantle rock remnants can be found along with the Jurassic strata of the Mauretanian flyschs (Durand- Delga et al., 2000;Gélard, 1979;Guerrera and Martín-Martín, 2014) but they are very scarce. Durand- Delga et al. (2000) proposed that a probable rifted and hyper-extended continental crust only partially oceanized could have floored the Maghrebian Tethys without having given rise to a true oceanic lithosphere. More data about the mafic and mantle rock remnants in the Kabylies is still required to bring crucial arguments about the nature of the Maghrebian Tethys.
In map view, the Flyschs and AlKaPeCa domains are discontinuous. To the west, they disappear onshore between Al Hoceima (Central Rif) and Mostaganem (western Tell) (Figs. 2 and 3). In this area, a problem arises due to the apparent disruption of these domains corresponding to the offshoreonshore limit (Fig. 3). Offshore studies by Domzig et al. (2006) show that the internal zones and the flyschs extend up to some distance offshore ( Fig. 2) but the major part of this disruption is of tectonic origin, due to the post-Early Miocene motions of the westernmost part of the AlKaPeCa domain (mostly the Alboran microdomain). Interpretation of this disruption is a major issue. To the east, on the contrary, the Kabylie-like terranes are continuous from Eastern Tunisia to Sicily and southern Apennines (Auzende et al., 1974;Bouillin, 1986;Bouillin et al., 1999;Catalano, 1987;Catalano et al., 1985;Compagnoni et al., 1989;Mascle et al., 2004;Tricart et al., 1994;Zitellini et al., 1989). Also, the Flyschs domain, notably the so-called Numidian flysch, develops widely in Eastern Algeria and Tunisia.

The external domains
The external domains, representing the former African paleomargin, are described as a pile of thrust-sheets resting on top of a Miocene foredeep in which allochthonous blocks (olistostromes) are included (e.g. Caire, 1957;Vidal, 1971). This foreland basin, well developed at the front of the Rif (Gharb Basin) and in western and central Algeria, becomes narrower in between (in the Gareb-Kebdana promontory [Hervouët and De Luca, 1980]) and even disappears in eastern Algeria and Tunisia, in areas where the foreland itself has been impacted by major inversion tectonics (Fig. 2). The foreland basin, when it exists, rests unconformably on a more or less folded and deeply eroded Mesozoic substratum, namely either the Atlas belt or the "Mesetas" (Fig. 1). It is locally overlain by thin thrust-sheets (few hundred meters to no more than 1.5 km), which constitute erosional outliers of the uppermost Tell-Rif units by contrast to lower units interpreted as tectonic windows.
The uppermost units, called here "higher Rif-Tell nappes", are rather homogeneous all along the Tell-Rif. In general, they are detached along an Upper Cretaceous décollement and comprise a typical lithostratigraphic column of deep-marine sediments from Upper Cretaceous up to Lower Miocene without significant unconformity. In the Rif, they comprise the Aknoul, Tsoul, Ouezzane and Prerif nappes or units and correspond to the former cover of the Ketama-Tanger unit belonging to the Intrarif zone (Suter, 1980a(Suter, , 1980b. The Aknoul nappe crops out in the eastern Rif and presents the most complete sedimentary pile from Lower Cretaceous to Miocene. The Tsoul nappe (Leblanc, 1979) and Ouezzane nappes (Durand Delga et al., 1962;Hottinger and Suter, 1962) are twin-units detached along a décollement in the uppermost Cretaceous. In western Algeria, they find an equivalent in the so-called "Senonian", "Albo-cenomanian", "Chouala" and "Oligo-Miocene" nappes of the Oran area (Delteil, 1974;Fenet, 1975;Guardia, 1975;Polvêche, 1960). In central Algeria, these units correspond to the complexes "B" and "C" (Caire, 1957;Mattauer, 1958), complex "B" being more or less equivalent to the Aknoul nappe and complex "C" of Ouezzane-Tsoul nappe. In Eastern Algeria and Tunisia, a Miocene frontal basin only scarcely underlines the limit between the Tell and the Atlas. The higher nappes (Kasseb units in Tunisia, Rouvier, [1977]) correspond mostly to tectonic imbrications resting directly over the Lower Cretaceous rocks of the foreland (Vila, 1980). In this case, the sole-thrust of the nappes is a major Upper Cretaceous décollement level allowing the complete decoupling of the upper part of the sedimentary pile.
Below and/or laterally, the lower Tell-Rif units are exposed. They occur in three different structural settings, i. e., from outer to inner zones: À in the most external zones, they constitute tectonic windows corresponding to late culminations of the African foreland piercing the core of late anticlines refolding the over-lying nappes. This is typically the case from west to east for the "internal Prerif", the "external Mesorif" (Suter, 1980a(Suter, , 1980b and for the tectonic windows of the Ouarsenis (Mattauer, 1958) and Biban (Caire, 1957) Massifs. A similar configuration is observed more to the east in the Constantinois (Vila, 1980), in Tunisia (Rouvier, 1977), as well as in productive structures for oil and gas in Sicily and Southern Apennines (Casero, 2004 and references therein), with the following particularities: in the Constantinois, there are no Miocene deposits between the autochthonous and allochthonous domains; by contrast, in Italy Pliocene deposits have been preserved attached to the lower plate (Casero et al., 1991). These differences are due to the diachronism of the main overthrusting episode across the Maghrebian-Apenninic orogen. As indicated above, the sole-thrust of the "higher Rif-Tell nappes" remains a décollement (and not strictly speaking a thrustfault) in this case, with only a lateral displacement of younger units over older ones, without any real tectonic duplication (older over younger rocks), except near the tectonic fronts where the nappes can thrust over or be intermingled within young Miocene sediments; À more to the north, units with significantly stronger deformation than the "higher Rif-Tell nappes" and the above-mentioned tectonic windows are exposed in the eastern Rif and western Tell (Figs. 2 and 3). These so-called External Metamorphic Massifs (EMM) comprise the Temsamane Massif (Andrieux, 1971;Frizon de Lamotte andLeikine, 1985, 1987) in the Eastern Rif and the Oran (Ciszak, 1993;Fenet, 1975;Guardia, 1975), Chelif (Lepvrier, 1978) and Blida (Bles, 1971) massifs in the Western and Central Tell. The best exposures of the EMM are located in the Temsamane Massif (Booth-Rea et al., 2012; Frizon de Lamotte and Leikine, 1985Leikine, , 1987Frizon de Lamotte et al., 1991;Jabaloy-Sánchez et al., 2015;Negro et al., 2007Negro et al., , 2008. They are made up of a stack of tectonic slices with Paleozoic rocks exhibiting HP-LT metamorphic type (7-9 kbar and 330-430°C; Negro et al., 2007) at the top of the system. The organization of the Oran Massif roughly mimics the Temsamane Massif with increasing metamorphic conditions toward the NW showing the highest grade in the highest Paleozoic slices (Fenet, 1975;Guardia, 1975). In the Temsamane Massif, the HP-LT metamorphism is older than 23 Ma ( 40 Ar/ 39 Ar on white micas: Negro et al., 2008) in agreement with previous ages published by Monié et al. (1984) ( 40 Ar/ 39 Ar on white micas). Jabaloy-Sánchez et al. (2015) ( 40 Ar/ 39 Ar on white micas and amphiboles) retain a Late Oligocene-Early Miocene age for the peak metamorphic event, but an older age (Late Eocene) remains possible. The EMM disappear west of the Nekor fault in the Rif and east of the Greater Kabylie in the Tell (Fig. 2). It is not yet established whether these units are buried at depth below more internal units or if this absence is due to major transition zones that would separate significantly different tectonic domains; À north of the EMM, the "northernmost Rif-Tell units" are cropping out. They correspond roughly to the mid-Tellian and partly the high-Tellian units of Wildi (1983). From west to east, they comprise: the Ketama-Tanger unit, the Dahra unit between Oran and Alger and, finally, the Babor unit between the Greater and Lesser Kabylie (Leikine, 1971;Obert, 1981). In general, these units comprise Triassic to Lower Cretaceous sedimentary rocks involved in large recumbent folds and separated into sub-units by important thrust-faults. Importantly, these massifs are tectonically unroofed and their former Upper Cretaceous to Cenozoic cover makes up the current "higher Rif-Tell nappes".

Oceanization steps in the Tethys realm
In all the recent kinematic and palaeogeographic restorations (Dercourt et al., 2000;Handy et al., 2010;Schettino and Turco, 2009;Sibuet et al., 2012, Nirrengarten et al., 2018Biari et al., 2017), Iberia was located far east of its present position relatively to Africa at the beginning of the Jurassic (Fig. 5A). So, at first order, the Maghrebian Tethys is interpreted by the authors as a transform oceanic sea way connecting the Central Atlantic and the Ligurian (Alpine) Tethys. This corridor allowed the left lateral movement of Africa relatively to Iberia during the Jurassic and Lower Cretaceous (Fig. 5).
In detail, the evolution is more complex. Geological data from Iberia and Africa show a set of NE to ENE trending "en echelon" Mesozoic basins, which were more or less subject to tectonic inversion during the Cenozoic. These basins are: the Algarve (Ramos et al., 2016;2017a, 2017b and Subbetic (García-Hernández et al., 1989;Vera, 2001) basins on the Iberian side and the Khemisset-Rif (Favre, 1995;Favre et al., 1991), Middle-Atlas, High-Atlas, Sahara-Atlas and Tunisia-Atlas basins on the African side (Frizon de Lamotte et al., 2000, 2009 and references therein; Fig. 5A). All of them are slightly oblique to the Atlantic and Alpine Tethys ENE trend but also to the E-W connection (i.e. the so-called Maghrebian Tethys) between these two oceanic domains. It is interesting to try to understand the geometrical and chronological relationships between the intracontinental rift basins and the oceanic domains. An interesting comparison comes to mind with the Aden Gulf opening reworking obliquely former Jurassic-Cretaceous basins (d'Acremont et al., 2005). Another likely similar geological setting might be found in the Gulf of California where the ongoing right-lateral displacement between the Pacific and the North America plates gave birth to pull-apart basins. There, the crust was stretched to various extents. The Salton Sea Basin, located on continental crust, is located along the San Adreas fault and is showing extreme thinning, though being bound by continental crust of normal thicknesses (Han et al., 2016 and references therein). This setting where only portions of the transform zone shows extreme crustal thinning might prefigure the Maghrebian Tethys before the left-lateral motion of Africa after the Bajocian (van Wijk et al., 2017).
In between the Central Atlantic and the Ligurian Tethys, the Iberian and North-African basins present a similar Triassic to Middle Jurassic evolution. They developed firstly during the Upper Triassic at the same time as the rifting in Central Atlantic (Leleu et al., 2016;Olsen, 1997;Withjack et al., 1998 and references therein) and in the Alpine domain (Bertotti et al., 1993). Then, whereas slow oceanic accretion or mantle exhumation began in the Central Atlantic (Labails, 2007;Sibuet et al., 2012), they recorded a new rifting phase in the Lower Jurassic. This second rifting event lasted until the late Bajocian in the Rif (Favre et al., 1991) and in the Middle Atlas (Sani et al., 2000;Zizi, 1996aZizi, , 1996b. In the Tell, seismic lines published by Bracène and Frizon de Lamotte (2002) suggest that the rifting was still active during the lower part of the Dogger. In southern Iberia, the timing is less clear due to a rifting revival characterizing the Iberian Plate during the Late Jurassic-Lower Cretaceous (e.g. in the Algarve Basin: Rey, 2009 and references therein; in the Betics: [Vera, 1988[Vera, , 2001). Anyhow, we can postulate that at the end of the Jurassic rifting the North African basins (Rif, Middle Atlas, Sahara Atlas) and their Iberian counterparts (Algarve, Cadix, Subbetic) were already face-to-face (e.g. for a recent reference, Nirrengarten et al., 2018). The development of an E-W transform zone connecting the Alpine Tethys to the Central Atlantic occurred just after.
Given the kinematics of the Central Atlantic (Labails, 2007;Sibuet et al., 2012), it is likely that a "pure" transform movement lasted until 155 Ma (Tithonian). Then, a net decrease of the spreading rates (down to 2.6 cm.yr À1 ) occurred, coeval to a change in the plate kinematics. The net result was an increase of the distance between Iberia and Africa. From a chronological point of view, this scenario is consistent with the formation of the Ketama-Tanger oceanic domain at the end of Jurassic times. Thus, this basin filled with flyschs outcropping in the Ketama unit simply appears as an appendix belonging to the Flyschs Basin itself. It is worth noting that a similar timing is also recognized for mantle exhumation along the south Iberian margin in the Gorringe Bank area (Girardeau et al., 1998;Jiménez-Munt et al., 2010).
Such a kinematic scenario could explain why mantle exhumation and onset of oceanic crust accretion occurred later in the Maghrebian Tethys than in the main Atlantic domain and possibly later than in the Ligurian domain. The left-lateral movement of Africa associated with little divergence lasted during the whole Early Cretaceous and the beginning of the Upper Cretaceous, until the onset of the Africa-Eurasia convergence at 84 Ma (Santonian). 175-195 Ma (Sahabi et al., 2004;Labails et al., 2010)

Spatial distribution and meaning of the External Metamorphic Massifs (EMM)
From the general configuration described above (Sect. 2.2), we think that the existence of the EMM cannot be understood without underlining their link with the "northernmost Rif-Tell" units on the one hand and with their wedge-top Miocene basins on the other hand. The relationships between the Ketama-Tanger unit and Temsamane massif, in the eastern Rif, well depict the first point.
The Ketama-Tanger unit (Rif, Morocco) is particularly interesting because it shows serpentinite bodies along its eastern border (the Beni Malek massif), interpreted as exhumed subcontinental mantle (El Azzab et al., 1997;Michard et al., 1992Michard et al., , 2007. Therefore, it is likely that the northeastern part of the Ketama-Tanger unit might be floored by an oceanic basement. In contrast, its southernmost part exhibits Upper Triassic salt and Jurassic breccias consistent with a deep margin setting (Favre, 1992). The precise nature of the basement is out of scope of our study but might be heterogenous, which could be important in terms of mechanical behavior during the Cenozoic inversion (e.g. salt vs. no salt). In any case, the whole Ketama-Tanger unit represents a former deep basin (Andrieux, 1971;Favre, 1995;Frizon de Lamotte and Leikine, 1985), rather than a continental block (micro-continent) as it has been recently proposed (Michard et al., 2007 in their palaeogeographic proposals to account for the existence of the peridotites Although slightly metamorphic in its eastern part (no more than early greenschist facies, see Andrieux, 1971; Frizon de Lamotte and Leikine, 1985), the Ketama-Tanger unit does not exhibit HP metamorphism. Therefore, we must consider that the adjacent Temsamane units were buried and then exhumed, still maintaining the Ketama-Tanger unit as a roof structure. This means that the subduction explaining the HP-LT metamorphism of the Temsamane units was an intra-oceanic NW-dipping subduction allowing the burial of the thinned African margin (Temsamane) at intermediate pressures but keeping close to the surface the "oceanic" Ketama-Tanger basin (Fig. 6). Such a configuration is unusual because, in general, the subducted domain is the oceanic one. However, a comparison comes to mind with the Oman belt where tectonic windows exhibiting HP-LT metamorphism affecting continental crust support the non-metamorphic Semail ophiolite (see Duretz et al., 2016;Goffé et al., 1988;Rioux et al., 2016).
The "oceanic" nature of the northeastern part of the Ketama-Tanger Basin raises two questions: (1) how this NEtrending basin was ending to the SW, toward the African Plate, and (2) how this deep basin was connected to the "Flyschs Basin", interpreted as the main boundary between Africa and Iberia?
Regarding the first question, a "V" shape is well identified for the Ketama-Tanger Basin (Favre, 1992;Favre et al., 1991). The tip of the "V" is located south of the Prerif Ridges (formerly, "Rides Prérifaines" of the French authors) in the Khemisset Basin (Faugères, 1978;Favre, 1995). Between the Prerif Ridges and Ketama, evidences of units with "oceanic affinities" have been described in the Sofs (Benzaggagh, 2011;Bulundwe Kitongo et al., 1987) and Senhadja units (Favre, 1992;Papillon, 1989), belonging to the internal Prerif and external Mesorif (Benzaggagh et al., 2014). These units, possibly showing remnants of the SW extension of the former Ketama Basin (the Mesorif Suture Zone of Michard et al., 2014), present a chaotic aspect with a complex association of magmatic rocks (gabbro, basalt), Lower-Middle Jurassic carbonates (more or less metamorphosed; from our own  observations they can locally be marbles, but are still poorly described in the literature [Benzaggagh et al., 2014;Michard et al., 2014]) and Upper Jurassic turbidites. They have no equivalent in the Tell, except possibly in the Chouala Nappe (western Tell; Delteil, 1974;Fenet, 1975;Guardia, 1975;Polvêche, 1960) in which "blocks" of Jurassic carbonate are described.
Considering the second question, elements that may correspond to the Beni Malek serpentinites and the Temsamane units, but more scattered, are known in the western Tell (Oran region) (Ciszak, 1993;Fenet, 1975;Guardia, 1975;Fig. 2). They may also extend up to the westernmost boundary of the Greater Kabylie in the Blida Massif, across the Chelif Basin (Bles, 1971;Lepvrier, 1971;Lepvrier, 1978). These structural units display more intense deformation than the upper Tell units and the external tectonic windows, but still less than the Oran and Temsamane EMMs and are not associated to peridotites. Nonetheless, from Oran to the Chelif Basin, numerous Triassic salt plugs are bringing up to the surface xenoliths of deep metamorphic crustal or mafic rocks and peridotites (Kornprobst and Delteil, 1976;Midoun, 1989;Midoun and Seddiki, 2016). Although the salt may have undergone significant horizontal displacements during tectonic transports, it testifies for a former widespread extensional context, likely linked to the Triassic-Jurassic rifting phase.
All these observations suggest that the Ketama-Tanger Basin developed and probably enlarged to the NE. The Ketama-Tanger Basin shares with the Flyschs domain the turbiditic sedimentation that invaded the whole Ligurian-Maghrebian Tethys during the Early Cretaceous (Bouillin, 1986;Durand-Delga et al., 1999;Guerrera et al., 1993Guerrera et al., , 2005Guerrera and Martín-Martín, 2014). In both cases, the Cenozoic clastics sourced in Africa (Wildi, 1983). So, it appears realistic to consider that the Ketama-Tanger Basin and the Flysch basin were in fact a unique basin connected through a complexly rifted and segmented zone along the southern Tethys (at the present-day location of Morocco and West/ Central Algeria; Fig. 5). Their differences concern mainly the state of deformation, higher in Ketama-Tanger than in the flyschs, and their initial position, Ketama being located more to the west but not their initial sedimentary infill.

Relationships between the EMM and wedge-top Neogene basins
The occurrence of intra-mountain Miocene to Pliocene basins on top of the nappes stack characterizes the central and eastern Rif as well as the western Tell area (Figs. 7 and 8). In the Rif, these basins (namely the Taounate, Rhafsaï, Tafrant, Dahar Souk, Arbaa de Taourirt, Kert and Boudinar basins, see Fig. 7) constitute the north-Rif corridor. During the Upper Miocene but before the Messinian crisis, this corridor was a marine connection between the Mediterranean and the Atlantic (Capella et al., 2017;Wernli, 1988). In the western Tell, the wider Chelif Basin occupies a similar structural position on top of the Tell nappes (Roure et al., 2012;Thomas, 1985;Figs. 2, 3 and 8). The Chelif is older than the Rif basins: below the Late Miocene, the sedimentary pile is known down to the Aquitanian(?)-Burdigalian (Neurdin-Trescartes, 1992).
Classically, the central and estern Rif and Chelif basins were interpreted as "post-nappes basins", that means basins post-dating the main tangential phases and the nappes emplacement, subsequently reworked during a late compres-   sional event (Capella et al., 2017;Morel, 1987;Tejera de Leon et al., 1995). However, such a designation is not suitable because these basins have the same age as the respective adjacent flexural basins supporting the nappes. As illustrated by the cross-sections A and B (Fig. 3), they are better described as wedge-top basins developed during the activity of the orogenic prism.
Focusing on the mechanisms of formation of the central Rif wedge-top basins, Samaka et al. (1997) evidenced, using industrial seismic data, early extensional faults delineating sedimentary wedges (Fig. 7D and E). These faults are bounding the southern or southwestern flanks the basins (Fig. 7A). The basins are consequently not only synclines as promoted by Suter (1980aSuter ( , 1980b, Tejera de Leon et al. (1995) or Capella et al. (2017) but initially developed as extensional basins (Flinch, 1994;Samaka et al., 1997), subsequently involved in compressional deformation. In the central Rif, the basins rest over the Mesorif tectonic windows. In the eastern Rif, the basins are fringing the Temsamane EMM to the NW following the Nekor Fault. At large scale, the basins consequently underline the "external or Mesorif suture zone" defined by Michard et al. (2014). From a chronological point of view, they developed during the late stages of the Temsamane Massif exhumation, which occurred between 15 and 10 Ma . Given that this exhumation developed in a context characterized by intense ENE stretching and top to the west shearing (Frizon de Lamotte, 1987;Frizon de Lamotte et al., 1991;Jabaloy-Sánchez et al., 2015;Negro et al., 2007Negro et al., , 2008, the formation of the Rif wedge-top basins corresponds to the final lateral collapse of this system. By the Late Tortonian, they are involved in a compressional setting explaining their present syncline-like shape. The tectonic history of the Chelif Basin appears longer and more complicated. A three-step scenario can be proposed. During the first step (Aquitanian(?)-Burdigalian to Langhian), the basin developed as a piggy-back basin over the moving higher Tell nappes, which in turn emplaced southward over a Lower Miocene foredeep. During a second step (Late Langhian-Serravallian up to Tortonian), a set of low-angle normal faults with a possible left-lateral movement developed (Meghraoui et al., 1986;Roure et al., 2012;Thomas, 1985). These faults controlled the Serravalian to Late Miocene filling of the basin. As in the Rif, these two first steps can be related to the progressive exhumation of the EMM (Oran and Chelif massifs) through major north-dipping fault connecting the EMM deformation to the Chelif infill during the whole Miocene (Figs. 3 and 8). Due to the lack of recent structural and petrographic work, the precise kinematics of these tectonic processes is unknown. However, given the ubiquitous existence of an E-W to ENE trending and low dipping lineation in the Algerian EMM (Guardia, 1975;Kireche, 1977), it is likely that, as in the Temsamane, the exhumation occurred in a transcurrent context. Finally, as in the Rif, during a third step, the Chelif region suffered a late contractional event by the Late Tortonian(?)-Pliocene (Meghraoui, 1982;Thomas, 1985). This compressional tectonic activity has remained operating up to now.

The first tectonic inversions of the Southern
Tethys margin, importance of the Middle-Late Eocene event  The convergence between Africa and Eurasia follows a counter-clockwise rotation for fixed Eurasia around a pole located in the Central Atlantic, at small distance from the Moroccan margin (Handy et al., 2010;Rosenbaum et al., 2002;Fig. 9). Consequently, between 67 and 35 Ma, the amount of convergence is two times greater in Tunisia than in Morocco. During this period, the trajectory is almost and globally N-S with an inflection toward the NW by 45 Ma. The convergence is accommodated by the subduction of the different oceanic domains: the Maghrebian Tethys and the two branches of the Ligurian Tethys (Fig. 10). The subduction process is not continuous but disturbed by compressional adjustments indicative of periods of strong coupling between the two plates and jumps from an oceanic domain to another one with the AlKaPeCa domain in between.
As everywhere along the South Tethys margin, a first compressional tectonic event is recorded at the onset of Africa-Eurasia convergence (84 Ma) (Frizon de Lamotte et al., 2011 and references therein). This event is known everywhere in the northern part of Africa as the "Santonian Event"  We st L ig u r ia n T e t h y s   , 2005). It is also known in Europe (among others, Kley and Voigt, 2008;Jolivet et al., 2016; with references therein). Nonetheless, its effects remain weak or overprinted by subsequent events in the Tell-Rif. A major compressional tectonic event occurred during the Middle-Late Eocene (45 to 35 Ma), i.e. at the end of the period of oblique convergence indicated above (Fig. 11). Laffitte (1939), who called it "Atlas Event", firstly identified this event in the Aurès Mountains (Algerian Atlas).  Khomsi et al. (2004Khomsi et al. ( , 2016 and Leprêtre et al. (2018).

Ma gh re bia n Te th ys
In the external Tell-Rif, the unconformity of the Oligocene (?)-Miocene sediments of the foredeep over the Atlas structures is recognized since a long time (Caire, 1957;Mattauer, 1958;Suter, 1980aSuter, , 1980b; and references therein). More precisely, Fenet et al. (1969), Delteil (1974), Fenet (1975) and Guardia (1975) defined an "intra-Lutetian" event responsible for tangential movements, strong deformation and development of regional metamorphism in the western Algerian Tell. An analog configuration exists in the eastern Algerian Tell (Vila, 1980), even if the Priabonian age proposed by this author is a little bit younger. In the eastern Rif, in the Temsamane EMM (Figs. 2 and 6), unambiguous trace of HP-LT metamorphic type have been described  and is associated with a subduction or deep underthrusting of this continental domain below the Ketama-Tanger oceanic unit (see above). The age of burial or exhumation related to this stage is not very well defined. It should be at least Oligocene and possibly older (Monié et al., 1984;Negro et al., 2007). Laterally, in the Mesorif zone, Oligocene (?) and Lower Miocene conglomerates rest unconformably over the older Mesozoic rocks (Favre, 1995;Favre et al., 1991;Leblanc, 1979;Marçais, 1936;Vidal, 1971). The subsequent exhumation of the Temsamane, occurring in a transcurrent context, is dated as Langhian to Late Serravalian/ Early Tortonian (Frizon de Lamotte, 1987;Negro et al., 2008;Jabaloy-Sánchez et al., 2015). The deformation of the Mesorif is better correlated to a Late Eocene-earliest Oligocene deformation with a stratigraphic gap from the Early to the Late Oligocene. As a consequence of this interpretation, we relate the Oran up to the Chelif EMMs to an eastward prolongation of the area affected by compression during this period. This explains the similar structural setting comprising peridotites and the metamorphic units of the Oran EMM (Michard et al., 2006). Further works of dating for these metamorphic units are required to confirm this hypothesis. deformation is less intense, the EMM would result from the northward underthrusting of the African plate, but it may not have reached such HP conditions like in the western parts in spite of experiencing the same type of gradient. Therefore, during the Middle-Late Eocene, North Africa was affected by more or less intense shortening, not only in the Atlas (the concept of "Atlas" event) but also in the external Tell-Rif where it is locally associated with HP-LT metamorphic type (Fig. 11). This event is coeval with the so-called "Pyrenean" event, evidenced not only in the Pyrenees but also in the whole Iberian Plate and even over the entire Western European domain (Bergerat, 1985;Casas Sainz and Faccenna, 2001). It correlates with the age of the HP-LT metamorphic type in the Alps ( , we consider that the collision between AlKaPeCa and the eastern Iberian margin, that is to say the closure of the West Ligurian Tethys, occurred at the same time (see also Schmid et al., 2017).
However, this major tectonic reorganization did not correspond to a direct collision between Iberia and Africa. At that time, the Maghrebian Tethys was still an oceanic seaway connecting the Atlantic and the East Ligurian Tethys (Fig. 11). Moreover, in the Flyschs domain and in the "northernmost Rif-Tell units", the former sedimentary cover of the Maghrebian Tethys, the Eocene-Oligocene transition corresponds to a change from carbonate to clastic facies but without any angular unconformity (Wildi, 1983). The stillexisting Maghrebian Tethys after 35 Ma is also consistent with the fact that the subduction recorded in the Temsamane Massif (under-thrust segment of the African margin) and likely in the Oran area had begun as a subduction within the Ketama-Oran Basin. We consequently propose that from 67 Ma to 35 Ma, the oblique convergence between Africa and Iberia was accommodated (1) by the progressive closure of the West Ligurian Tethys (see also Michard et al., 2002) and (2) by the subduction of the western part of the African margin (Temsamane and Oran units) below the Ketama-Flyschs oceanic basin.

The Oligocene-Miocene (r)evolution (Figs. 12-14)
The Eocene-Oligocene transition (35 Ma) is marked by a new change in the kinematics of the convergence between Africa and Eurasia, which got back to its previous N-S trend. The onset of a new subduction on the eastern and southern sides of AlKaPeCa involving both East Ligurian and Maghrebian Tethys (i.e. the "Flyschs Basin") accommodated the convergence (Fig. 12). In parallel, the two previously active subduction zones (i.

Sedimentary record of the Oligocene geodynamic change
In the Maghrebian and East Ligurian Tethys, the change of the geodynamic context is recorded by the transition from a passive-margin-bounded basin to a compressional foreland basin (see review by Guerrera and Martín-Martín, 2014). The Oligocene to Early Miocene sedimentary infill of this basin comprises immature turbiditic sandstones in the north ("Flysch gréso-micacé" of the French authors; Mauretanian Domain) and the mature to ultra-mature sandstones of the famous Numidian Formation to the south (Massylian Domain; see about the "Numidian question", the review by Thomas et al., 2010). It is established that the Mauretanian Domain was sourced from the north, i.e. AlKaPeCa, and the Massylian from the south, i.e. Africa. Mixed sedimentation exists in-between these two domains (Belayouni et al., 2013 and references therein).
The links between the former oceanic units (Mauretanian/ Massylian) and the coeval more external deposits are not straightforward. In the Tunisian part of the African margin (Tunisian Atlas), the wide Fortuna delta system development was contemporaneous with the Numidian flysch deposition; and it represents one of its main sources (Yaich et al., 2000). The Numidian Formation extends in Eastern Algeria with similar characteristics (Vila, 1980). From the Greater Kabylie westwards, due to the structural complexity, the presence of intermediate facies between the Lower Miocene rocks cropping out in the foredeep and tectonic windows and the Numidian is not so clear. In the Western Tell, Oligo-Miocene deposits are found in two distinct structural positions. On the one hand, they rest unconformably over the allochthonous external units as it is the case for the wedge-top Chelif Basin (Neurdin-Trescartes, 1992;Roure et al., 2012; on the southern nappes of Fig. 3B) or for small lower Miocene basins (Courme-Rault, 1984). On the other hand, they are involved in thrust-sheets whose palaeogeographic origins are difficult to precisely track. In the latter case, the Oligo-Miocene formations should represent transitional or lateral facies of the Numidian. In the Western Tell, Fenet (1975) and Guardia (1975) suggest that the so-called "Oligo-Miocene" unit could correspond to the former cover of the "Senonian" units (forming the main body of the higher Rif-Tell nappes) and could consequently represent such lateral and transitional facies. In the Eastern Rif, it is worth noting that conformable Numidian-like facies are known at the top of the higher Tell-Rif nappes, e.g. in the Aknoul (Leblanc and Feinberg, 1982), Tsoul and Ouezzane units.

Extensional deformation in AlKaPeCa, far effects in the foreland
The onset of the northward to northwestward subduction of the ELT and Maghrebian Tethys below Corsica-Sardinia and AlKaPeCa was associated with a considerable slab retreat, leading to the formation of back-arc basin (Réhault et al., 1984). The net result was the translation/rotation of these two micro-continents and the coeval opening of the West Mediterranean basins either made of stretched continental or oceanic crust (Figs. 12-14). After the counter-clockwise rotation of the Corsica-Sardinia that opened the Liguro-Provençal Basin, the Algerian Basin resulted from the dislocation of the Alboran-Kabylie following N-S to NW-SE trends (Arab et al., 2016;Booth-Rea et al., 2007;Leprêtre et al., 2013;Mauffret et al., 2004 and references therein).
Sedimentary record and geochronological data evidence the onset of a generalized Upper Oligocene/Lower Miocene rifting. The widespread extension is witnessed on the upper plate involved in the subduction process. It was firstly recognized on Kabylie basement through the deposition of the syn-rift "Oligo-Miocene Kabyle" dated from Late Oligocene to Burdigalian (Aïte and Gélard, 1997;Géry et al., 1981). These ages of rifting agree with seismic interpretations by Leprêtre et al. (2013) and Arab et al. (2016) for areas along the Algerian margin where tilted blocks are preserved. Recent seismic studies in the West Alboran Basin have also evidenced this large-scale extensional event. Based on seismic analysis coupled with onshore data, Suades (2016) documented this extensional event in the northern Alboran Basin and onshore southern Spain (Malaga area). Do Couto et al. (2016) proposed new seismic analysis with revised calibration for the stratigraphy of the West Alboran Basin and found a similar timing with major N-S extension during the Late Oligocene-Early Miocene period.
During the Oligocene and Early Miocene, in the foreland of the Tell-Rif, the Atlas system is also the site of an extensional regime or, at least, tectonic quiescence (Bracène and Frizon de Lamotte, 2002;Frizon de Lamotte et al., 2000). We discuss the meaning of this unexpected deformation on the lower plate in the discussion section (see Sect. 5.2).

"Collision" and lateral slab tearing along the African margin
The onset of Oligocene subduction along the southeastern side of AlKaPeCa led to the development of an accretionary prism involving the sedimentary cover of the Maghrebian and East Ligurian Tethys. The docking of the Kabylie against the African margin occurred at 17 Ma (Late Burdigalian) for the oldest as witnessed by the Collo granite emplacement crosscutting both the Lesser Kabylie and the African basement (Abbassene et al., 2016;Aïssa et al., 1998Aïssa et al., , 1999Chazot et al., 2017). This result is in line with the stratigraphic ages of the youngest Miocene sediments sealing the frontal deformation of the Tell nappes in central and eastern Algeria (Caire, 1957;Courme-Rault, 1984;Mattauer, 1958;Polvêche, 1960;Vila, 1980;Wildi, 1983). This event marked the closure of the central part of the Maghrebian Tethys and the onset of the delamination of the sub-continental African mantle (Roure et al., 2012). In fact, the initial orogen propagated outward with early thrusting of the Kabylie-Flysch domains on top of the external domain during the Late Burdigalian, whereas the frontal parts of the nappes glided in the Miocene foredeep up to Langhian times in central and eastern Algeria (Courme-Rault, 1984). The consequence of this closure was also a change in the structural style. From that time onward, the shortening has been accommodated through both frontal accretion and tectonic underplating (used hereafter in the sense of Platt, 1986Platt, , 1988, for accretion prisms, in coincidence with deep duplexing at the base of accretion prism), through the activation of different décollement levels, from the thinned African crust to the sedimentary levels upward.
Under North Africa margin, the southward mantle delamination (Roure et al., 2012) allowed propagation of the deformation toward the foreland. However, very rapidly slab rupture occurred below the African margin and it facilitated its tearing in two different directions: towards the present-day Calabrian Arc, creating the Tyrrhenian Basin, and towards the present-day Gibraltar Arc, creating the Algerian Basin (Jolivet et al., 2008;Lonergan and White, 1997;Mauffret et al., 2004;Spakman and Wortel, 2004;Van Hinsbergen et al., 2014).
The eastward tearing of the African slab toward Sicily, Calabria and southern Apennines is well described by different authors (see a review in Carminati et al., 2012). After a swing of the tectonic transport direction, it roughly corresponds to the continuation of the progressive subduction of the eastern part of the Maghebian-Ligurian Tethys and its African continental margin. It is characterized by the continuing eastward incorporation of more and more external units in the orogenic wedge and the coeval migration of the foredeep (Casero, 2004;Casero and Roure, 1994;Di Paolo et al., 2014). On the Mediterranean side of the system (Tyrrhenian Sea), a parallel eastward migration of the back-arc basin is observed at the rear of the Calabrian Arc (among others Faccenna et al., 2004).
The westward propagation of the slab tearing allowed the opening of the West Algerian Basin from Burdigalian to Tortonian times, in relationship with the westward motion of the Alboran block, bearing the already formed West Alboran Basin (Booth-Rea et al., 2007;Do Couto et al., 2016;Medaouri et al., 2014;Spakman et al., 2018). The precise kinematics of the West Algerian Basin opening remains unclear because of the complexity of the associated processes. The migration of the Alboran block is nicely recorded onshore within the Miocene foredeep, fossilizing the displacement of the higher Tell-Rif nappes (Fig. 4). The age of the Miocene sediments evidence a migration that began in the Middle-Late Serravallian in the Oran area, in the Late Tortonian in the central-eastern onshore foreland Rif Basin and even to the Pliocene in the Western Rif (Feinberg, 1986). An important point must be underlined here regarding the position of the Alboran block before the westward migration. The area showing the oldest sediments post-dating the Kabylie docking can be identified eastward up to Mostaganem (Fig. 2). This gives an approximate westernmost location for the western boundary of the Alboran block before its migration. This proposal is quite close to the proposal by Domzig et al. (2006) based on seismic and bathymetry data.
The coeval exhumation of the metamorphic rocks of the EMM during the Alboran block migration is a very important feature of the Rif-western Tell segment of the system. We propose a two-steps scenario for the exhumation, which globally occurred in a context of left-lateral movement (Booth-Rea et al., 2012;Frizon de Lamotte andLeikine, 1985, 1987;Frizon de Lamotte et al., 1991;Jabaloy-Sánchez et al., 2015;Negro et al., 2007Negro et al., , 2008. In our view, after the subduction stage (Fig. 6b), an imbrication stage with underplating developed within the accretion prism (Fig. 6c). It resulted in a doming that created the slope responsible for the detachment and the southward gliding of the higher Tell-Rif nappes. The age of the youngest sediments of the foreland basin just below the nappes gets younger to the west-southwest: from Early-Middle Miocene in the Tell to Upper Miocene in the Central Rif and Pliocene in the Western Rif. In the western Tell, this early doming stage is responsible for the Lower Miocene infill of the Chelif Basin in a piggyback configuration on top of the nappes. The second stage, in continuity or partially synchronous with the former one, allowed the development of Upper Miocene wedge-top basins like the second Miocene sequence of the Chelif Basin and the Neogene Basins Central-Eastern Rif (Figs. 7 and 8). Deposition within these basins was controlled by extenseional deformation (Roure et al., 2012;Samaka et al., 1997;Figs. 3, 7 and 8) associated to strike-slip faulting. Here again, the Rif basins (Tortonian-Messinian) are younger than the West Algerian ones (Late Langhian/Serravallian-Tortonian; Fig. 4).  In summary, the Oligo-Miocene period corresponds to the closure of the Maghrebian Tethys and the coeval opening of the West Mediterranean Basin (Figs. 10-15). A very important point, which is not taken into account in most of the kinematic models, is that along the North-African margin, the deformation mechanism is not the push linked to AlKaPeCa motion but the pull linked to slab retreat in a first time, and the delamination of the African sub-continental mantle in a second time (Roure et al., 2012). It is the reason why the deformation can develop at distance from the AlKaPeCa front. As a matter of fact, the present West Mediterranean slab is disconnected from the suture between the Alboran domain and the African margin (de Lis Mancilla et al., 2015;Palomeras et al., 2014;Thurner et al., 2014, and references therein). The westward retreat of the slab explains why the External Rif is moving more rapidly in the west than in the east (Pérouse et al., 2010;Figs. 14-15).

The Late Tortonian-Pliocene-to-Present (r) evolution
Currently, the direction of convergence between Africa and Eurasia is NW-SE (Nocquet and Calais [2004] and references therein). This oblique convergence is accommodated in the Tellian onshore by reverse motion on NE-SW trending faults and by right-lateral strike-slip motion on E-W trending faults as indicated by earthquakes focal mechanisms and by tectonic studies Rabaute and Chamot-Rooke, 2014). All these faults are inherited from the previous geodynamic history.
In the Tell offshore, high resolution seismic data reveal a set of young or still active fault-related folds at the toe of the Central and Eastern Algerian margin interpreted as the trace of a possible subduction initiation of the young Algerian Basin beneath Africa (Déverchère et al., 2005;Domzig et al., 2006;Kherroubi et al., 2009). If correct, it should represent the inception of a new slab reversal after the one occurring at 35 Ma after the closure of the western Ligurian Tethys. The age of this tectonic inversion of the Algerian margin is suggested by Arab et al. (2016) to have initiated as soon as the Late Tortonian. We show below that this is consistent with the generalized resumption of compression at the scale of the northern Maghreb and even in South Spain at this age.
In Eastern Algeria and Tunisia, the tectonic regime remains consistent with the eastward escape controlled by the Tyrrhenian slab (Rabaute and Chamot-Rooke, 2014). In Western Algeria, on the contrary, the new tectonic regime strongly contrasts with the previous one dominated by leftlateral movements and the westward translation of the Alboran Late Tortonian (c.a. 9-8 Ma)  domain (Fig. 16). As soon as the Alboran Domain had made its WSW drifting, the Africa-Eurasia convergence imposed NNW-SSE compressional deformation. Following Thomas (1985), the transition between the two regimes should have occurred around 8 Ma, during the Tortonian. It is also evidenced in the Betics (e.g. Do Couto et al., 2014;Janowski et al., 2017). Paleomagnetic measurements performed on volcanic rocks younger than 8 Ma show that they suffered systematic clockwise vertical axis rotations in agreement with right-lateral movement along E-W faults (Derder et al., 2011(Derder et al., , 2013. Following Morel and Meghraoui (1996), the deformation observed in the Tell is an element of a longer right-lateral deformation zone that they called GALTEL (Gorringe-Alboran-Tell) (Fig. 16). They interpret this 50 to 100 km wide zone as the present diffuse plate boundary between Africa and Europe. In fact, the right lateral movement is well documented along the Tunisian and Algerian margin up to the Yussuf Fault Zone (Fig. 16). South and west of this limit, the recent and active deformation remains under the control of the westward retreat of the delaminated sub-continental mantle (Pérouse et al., 2010). This westward translation is accommodated in the Central Rif by vertical axis counterclockwise rotation measured in the Upper Miocene wedge-top basins (Cifelli et al., 2008) and by development of N-S extensional faults (Lafosse et al., 2017). The NE-SW left-lateral Nekor Fault, inherited from the Lower Jurassic rifting (Figs. 2 and 7A), is a major element of this system and marks the southern limit of the domain affected by the westward movement.

Discussion
The review of the evolution of the Tell-Rif system in the frame of the West Mediterranean realm allows us highlighting some general questions. Since the initial Pangea configuration in the early Mesozoic, the history can be described as a succession of extensional phases with high coupling (e.g. during Jurassic described in Sect. 3.1), leading subsequently to major decoupling between the different plates ( Fig. 5a and b showing the extensional coupling and the independent motions from Late Jurassic onwards, respectively). The individualization between Iberia, Eurasia, AlKaPeCa and Africa plates reached its climax during the Early Cretaceous. Since the Late Cretaceous, the convergence began and progressively led to the current configuration. This convergence was not steady state but occurred through successive phases of coupling/ decoupling between the plates.
The first aim of this discussion is to better define how the Tell-Rif system can be integrated in the regional kinematic scenario. This clarification helps understanding the important mechanical couplings that existed during the Cenozoic stages. In a second time, we will attempt bringing an explanation for the strong non-cylindrical behavior of the Tell-Rif belt. We will emphasize the respective role of the structural inheritance on one hand and specific Cenozoic kinematics on the other hand.

Major tectonic events and kinematic reorganizations in the Tethys realm
After the first short compressional event ("Santonian Event"; e.g. Guiraud et al., 1987 for the Africa plate), at the onset of Eurasia-Africa convergence, the movement of Africa relative to Eurasia continued up to the Middle Eocene without major visible compressional structures. During that period, the poor mechanical coupling between the plates enabled subduction to completely accommodate the convergence. The location and the vergence of this (these) subduction(s) remain(s) currently the matter of debates. Some authors (Faccenna et al., 2014;Jolivet and Faccenna, 2000;Lacombe and Jolivet, 2005;Stampfli and Marchant, 1997) place a single and permanent west/northwest-verging subduction east and south of AlKaPeca. This hypothesis minimizes or ignores the existence of a western Ligurian Tethys. Besides, this hypothesis copes with two difficulties: (1) the absence of tectonic activity along the AlKaPeCa east and south margin (the "Dorsale Calcaire") up to the end of the Eocene and (2) the presence of HP-LT metamorphic rocks within AlKaPeCa, which is difficult to explain if this domain always remained in the same upper plate position. For these reasons, we favor another hypothesis suggesting that the Africa-Eurasia convergence was firstly accommodated along the eastward subduction of the WLT, which would simply represent the western continuation of the Alpine subduction (Figs. 10 and 11). Thus, by the end of this first collision, AlKaPeCa may have incorporated pieces of the former lower plate, even HP units that will be exhumed during the West Mediterranean opening. The remnants of the WLT subduction are scarce but well described (located as blue stars in our palaeogeographic proposal in Fig. 11). They are mainly present within the Betics (Li and Massonne, 2018;Platt et al., 2005;Puga et al., 1999), in Alpine Corsica (Vitale-Brovarone et al., 2013) and possibly also in Calabria (Heymes et al., 2010;Rossetti et al., 2004) and East Algeria (Bruguier et al., 2017). Regarding the metamorphic record, the WLT hypothesis nonetheless suffers from the same limits as the "no-WLT" hypothesis, namely (1) the poor representation of HP-LT and other remnants from this oceanic domain which makes difficult the vergence interpretation of the general structures and (2) the lack of consensus on the age of the HP-LT event(s).

The Middle-Late Eocene tectonic crisis
The Middle-Late Eocene period is known as a period of major kinematic reorganization with a strong coupling between Africa and Eurasia plates as evidenced by far-field compressional stress widely distributed in the two major plates as well as in intermediate plates like Iberia (Bergerat, 1985;Guiraud et al., 2005). More precisely, it corresponds to an orogenic climax in the Pyrenees (Vergés and García-Senz, 2001) (the concept of Pyrenean phase), the Iberian Chain (Guimerá et al., 2004), the Catalan Coastal range (López-Blanco et al., 2000) and in Southern Portugal (Algarve region: Ramos et al., 2016, Ramos et al., 2017a. Mirroring the Iberian intra-plate deformations, North Africa also recorded a strong middle-upper Eocene intraplate compressional event, present at the scale of the whole Atlas belt (the concept of "Atlas event" [Frizon de Lamotte et al., 2000, 2009, 2011).
In the Tell-Rif, the Middle-Late Eocene tectonic crisis was often minored, suggesting that the convergence was mainly accommodated in intra-plate settings (Frizon de Lamotte et al., 2000). However, we have presented some evidence (see Sects.
3.2 and 4.1), showing that the outer part of the external Tell-Rif underwent significant deformation during Paleogene times. More importantly and according to Negro et al. (2007Negro et al. ( , 2008, a fossil "subduction" acted along the African margin, allowing the Temsamane Massif (Eastern Rif) to reach MP-LT metamorphic conditions during at least the Early Oligocene and probably earlier. This fossil "subduction" of the African margin can be followed eastward up to the western Tell (Michard et al., 2006. Therefore, the Paleogene deformation was important along the former African margin, although not homogeneously distributed. The reasons of this peculiar distribution are discussed just below. From the Rosenbaum et al. (2002) reference, the rotation pole of Africa relative to Europe stayed within the [28.9-36.7]°N /[19.2-15.5]°W area from 75 to 35 Ma. That means a rotation pole located close to north Morocco present-day coastline, in the Central Atlantic Ocean. From this pole, the absorbed shortening along the Africa-Europe boundary was thus progressively increasing toward the east during this period. On fixed African points with respect to stable Europe, Rosenbaum et al. (2002) estimated convergence rates ranging from 5 ± 1 km.Myr À1 in Morocco to 10 ± 1 km.Myr À1 in Sicily for the Eocene period. During this convergence period and in the eastern regions of the study area, we agree that the convergence was absorbed by intra-plate deformation and along the Ligurian subduction front, here considered to follow the western/southwestern boundary of AlKaPeCa. To the west, along the Iberia/Africa plate boundary we have to integrate a complex kinematic configuration in which the Ligurian subduction front disappears westwards and the shortening becomes distributed along minor to intermediate compressional structures including: incipient subduction in the Rif and Oran basins and inverted former extensional structures (Gorringe, Algarve, Middle Atlas, High Atlas). The complexity of this area is on one hand, inherited from the Mesozoic rifting stages (Fig. 5), and on the other hand related to the Africa plate motion from the Late Cretaceous to the Middle-Late Eocene (Fig. 9). An interesting point is the fact that in spite of the overall shortening processes during this pre-Oligocene phase, the sedimentary cover of the Maghrebian Tethys itself (Flyschs domain) remained almost unaffected, which implies to behave as a rigid domain, the convergence being accomodated elsewhere.
The Eocene plate configuration of Northwest Africa (Fig. 11) can be compared to extensional or compressional zones in a triangle configuration (either opening or closing, respectively). The example of the North Atlantic northern branch, though being an extensional example, provides such a configuration. Its rotation pole is localized within north Siberia (Gaina et al., 2002). The oceanic ridge ends northward with the Gakkel Ridge (Rowley and Lottes, 1988;Fig. 17), before being relayed onshore by the numerous normal faults of the Ust-Lena Rift. In this extensional case, one can see a single "fault" zone concentrating most of the extension at the ridge location, whereas approaching the rotation pole with decreasing linear speed, the deformation becomes a lot more diffused, distributed over an array of minor normal faults over a wide area. Furthermore, the extension turns into compression right after the pole (Fig. 17). In the case studied here, the estimates of the Tunisian convergence rate is quite low (10 ± 1 km. Myr À1 ) but could be enough to sustain a very slow subduction. Compared to the Tunisian rate, the North Morocco convergence rate is extremely low (5 ± 1 km.Myr À1 ) and could hardly sustain a continuous subduction (e.g. Hall et al., 2003 for minimum convergence rate avoiding thermal re-equilibration). Fig. 16. Neotectonic map of Northwest Africa. This map was built using the following studies: Arab et al. (2016), Cadet et al. (2004), Domzig (2006), Meghraoui et al. (1996), Rabaute and Chamot-Rooke (2014), Lafosse et al. (2017) and Yelles et al. (2009). GALTEL abbreviation is for Gorringe-Alboran-Tell after Morel and Meghraoui (1996). In spite of considering an extreme case in terms of convergence rates, the western tip of the Tethys shows an example where accommodation is progressively changing from a typical subduction along the AlKaPeCa northern boundary to a diffuse and more distributed accommodation zone westward (Fig. 11c). The Gorringe Bank, Algarve Basin, Mesorif windows and subductions in the Ketama-Oran Basin are individual pieces of this triangle "accommodation zone".

The Oligocene-Miocene reorganization
After the consumption of the WLT, a second generation of "sutures" developed from the post-Eocene stages. This part of the history is better documented and corresponds to the subduction of the ELT and the Maghrebian Tethys below the drifted AlKaPeCa and Corsica-Sardinia blocks (Doglioni et al., 1999;Handy et al., 2010;Lacombe and Jolivet, 2005;Michard et al., 2002;see Figs. 13-15). An "oceanic" suture is recognized by the authors in the Flyschs domain, considered to represent the former cover of the Maghrebian and East Ligurian Tethys (Bouillin, 1979;Bouillin et al., 1977;Durand-Delga et al., 2000;Gélard, 1979;Guerrera and Martín-Martín, 2014). The oceanic nature of this "suture" that preserves remnants of the former Maghrebian Tethys and ELT crusts is poorly evidenced through few basic rocks remnants within Sicily, the Rif and the Tell (Durand- Delga et al., 2000 and references therein). A distinctive aspect of this suture is the scarcity of HP/LT remnants associated to it.
The major event regarding the post-Eocene stages is the Oligocene to Early Miocene extensional phase. By its geographical extent and the amplitude of extension, it represents one of the two periods of strong mechanical coupling between the Africa and Europe/Iberia plates, together with the Middle-Late Eocene tectonic crisis. The end of the Middle-Late Eocene showed a compressive deformation at the scale of the future West Mediterranean at least from the Pyrenees to the Atlas belt (Fig. 11). The Oligocene-Early Miocene event either fossilized or reworked in extension many of the former compressional structures (among others, see for the Pyrenees: Fillon and van der Beek, 2012;Vergés and García-Senz, 2001; for the Catalan Range: Roca, 1994; for the Baleares: Etheve et al., 2016;Roca, 1994; for AlKaPeCa: Aïte and Gélard, 1997;Arab et al., 2016;Caby et al., 2001;Comas et al., 1999;Heymes et al., 2008Heymes et al., , 2010Rossetti et al., 2004Rossetti et al., , 2010Saadallah and Caby, 1996; for the Atlas belt: Bracène and Frizon de Lamotte, 2002;Ghandriche, 1991;Khomsi et al., 2016;Muratet, 1991;Torbi and Gélard, 1994).
An interesting issue regarding the Oligocene-Early Miocene extension concerns the fact that both upper and lower plates underwent extension. As it is expected mainly for the upper plate, the question of their mechanical coupling remains crucial. Studies compiling field data Lallemand et al., 2005;Sdrolias and Müller, 2006) and laboratory experiments (Clark et al., 2008;Heuret et al., 2007) regarding back-arc extension pointed the importance of the slab dip to explain the occurrence of back-arc extension as a crucial parameter. The convergence rate seems poorly significant to influence the extensional regime at the rear of the subduction zone. Also, as revealed by kinematic reconstructions (Handy et al., 2010;Rosenbaum et al., 2002), this rate remained very low during the Late Oligocene-Early Miocene (0.4-0.6 to 1.2-1.4 cm.yr À1 ). Chamot-Rooke et al. (1999) considered very low convergence rates around few mm.yr À1 for the African plate relative to Europe, at odds with the southward trench migration, reaching up to 6 cm.yr À1 . We must notice that the usual explanations for this quick retreat lies within the slab dynamics in the upper mantle . In this view, at the Late Eocene-Oligocene transition, the slab fell down by gravity in the upper mantle and given the poor Africa convergence rate, the only way to accommodate this was to produce intense extension with back-arc opening at the rear of the system. However, we have proposed in this work to consider two successive subducting domains during the Late Cretaceous-Eocene period and then during the Oligocene-Miocene period, making this explanation unlikely. Although the driving force (s) required for triggering extension in our scenario remain(s) unknown, we suggest that mantle evolution of the AlKaPeCa terrane after the Eocene collision might have played a crucial role for the onset of slab retreat. One can imagine progressive delamination of the Eocene belt ( Fig. 11) with associated crustal extension and enhanced trench migration toward the south.

The Tell-Rif: a fundamentally non-cylindrical belt
One of the most striking features, when looking at the structural organization of the Tell-Rif (Figs. 2 and 3), is the difference between the western and eastern segments. This difference, contrasting with the almost linear shape of the belt (Fig. 2), involves the occurrence or the absence of at least three parameters: (1) Fig. 17. Structural map of the Arctic Ocean, focused on the Gakkel ridge that ends the North Atlantic ridge (after Dobretsov et al., 2013). Pale grey areas represent the oceanic domains. The ridge is relayed through the Ust-Lena Rift and the extension become diffuse before becoming compressive toward the SE after crossing the rotation pole.

L o m o n o s o v R id g e Po dv on iko v Ba sin U s t -L e n a R if t A m u n d s e n B a s in N a n s e n B a s in
the Miocene foredeep in front of the belt. The coexistence of these three parameters coincides with the western segment (Rif and western Tell) whereas their absence is typical of the eastern segment from the Babor Massif to Tunisia (Eastern Tell) (Fig. 2).
Three main parameters allow understanding this contrasted evolution. The first one is the inherited Mesozoic palaeogeography leading to the direct overlap (i.e. without rigid blocks in between) of the Tell over the Eastern Algerian and Tunisian Atlas (Fig. 1). In western Algeria and Morocco, these rigid blocks ("Mesetas") represent structural highs with thin Meso-Cenozoic deposits, escaping deformation during the orogeny. In eastern Algeria and Tunisia, a permanent high, with long-lasting development of a carbonate platform ("Constantine Platform" [Vila, 1980], Fig. 5), existed during the Mesozoic but is currently integrated within the Tell and partly within the Atlas. The second point is the tectonic history: the accommodation of the convergence was different with a subduction involving the African margin in the western regions. The last point is the kinematic history. Contrary to the eastern segment, the western segment structure results not only from frontal displacements, AlKaPeCa colliding with Africa, but also from post-Burdigalian stages showing a westward lateral escape of the Alboran block (Fig. 4). These considerable movements explain why the Rif displays a particularly complex structure with a frontal wedge in its western part whereas a combination of compressive and strike-slip structures gives its structural complexity to the eastern Rif (Frizon de Lamotte and Leikine, 1985Leikine, , 1987Jabaloy-Sánchez et al., 2015).
The presence of the EMM, exhibiting HP-LT metamorphic type, is linked to the former development of a subduction involving the segment of the margin located between the Temsamane and Oran-Chelif areas (Figs. 2, 5B and 6). They are consequently restricted to these western zones. The exhumation of the EMM occurred during the Middle Miocene in the Rif when the orogenic prism incorporated them by underplating. The exhumation is not precisely dated yet in the Tell, but following the first appearance of EMM pebbles (Fenet, 1975), it would have occurred as early as the Serravallian up to the middle Tortonian (Thomas, 1985). As indicated above, the exhumation occurred in a context of leftlateral movements. It was related to the westward translation of the Alboran domain, dragged by the progressive westward detachment of the African sub-continental mantle (Pérouse et al., 2010;Roure et al., 2012). From a geographic point of view, the wedge-top basins grew close to the EMM, flanking them laterally (Chelif, Eastern Rif) or frontally (Central Rif) (Fig. 2). They result from lateral and frontal collapse during the late stages of the EMM exhumation (e.g. for the Temsamane Massif: Booth-Rea et al., 2012). In the eastern regions, the absence of subduction of the margin explains the absence of both EMM and wedge-top-basins.
The western Tell and Rif Miocene foredeeps present the typical characteristics of flexural basins, with outward progressing onlaps well-observed in the Hodna Basin, for instance (Bracène and Frizon de Lamotte, 2002). The absence of any foredeep in the eastern Tell (Fig. 2) is consequently a challenging issue. This classical configuration for a fold-thrust belt is not "lost" in the eastern segment, it is likely to have never existed (e.g. Vila, 1980). This segment is the most distinctive in the entire West Mediterranean, because in Sicily and in the Appenines also, one can see another well-developed foredeep basin, largely over-thrusted by former basinal units (see Roure et al., 2012 for a review).
In fold-and-thrust belts, the development of a foredeep results from the interactions between few parameters. For a given tectonic load, the lithosphere rigidity (Watts, 2001) and the surface processes (i.e. mass transfer by erosion and sedimentation) control an important part of the deformation style (i.e. Bonnet et al., 2007;Fillon et al., 2013;Malavieille and Konstantinovskaya, 2010;Mary et al., 2013;Simpson, 2006). The existence and efficiency of several décollement levels add complexity for the structural style of the prism (Malavieille, 2010;Sherkati et al., 2006). Given these parameters, the reduction of the available space to generate a foredeep can be achieved through different ways. On the one hand, the more important the lithosphere rigidity, the more the flexure will tend to be weak, resulting in a large foredeep but thin available space for deposition. On the other hand, two competing effects have to be considered for the surface processes. Erosion role is very effective within the orogenic hinterland but exerts a less significant role for the external foldand-thrust belt (Beaumont et al., 1992;Willett, 1999). On the contrary, syn-tectonic sedimentation plays a first-order role in the deformation style of the fold-and-thrust belts (Fillon et al., 2013;Stockmal et al., 2007;Storti and McClay, 1995). It is worth noting that the mass transfer from hinterland to foreland can be achieved through tectonic unroofing of the hinterland, leading to the accumulation of gravity-driven nappes in the foreland. This mechanism was particularly efficient in the Rif and Western Tell: the duality between the northernmost Tell-Rif units and the Tell-Rif higher nappes (Figs. 2 and 3).
From the map in Figure 2, the Tell-Rif foredeep extents from the Hodna Basin to the Gharb Basin at the western end of the system (Fig. 2). In between the Tell and the Rif, the foredeep is poorly developed north of the Gareb-Kebdana embayment (Fig. 2). Nonetheless, as a whole, the Tell-Rif foredeep is narrow and deep. This is indicative of a small effective elastic thickness of the lithosphere combined with a significant subsidence. The youth of the basement, which is Variscan in age, could explain the low value of the effective elastic thickness. The tectonic loading, responsible for the flexure, results from the duplexing observed in the EMM and subsequently from stacking, which led to the formation of the Tell-Rif external windows (Figs. 2 and 3 [Sects. a and b]). Therefore, the mechanism responsible for the flexure is the tectonic underplating of the African margin. The emplacement of the Tell-Rif higher nappes in the foredeep contributed to accentuate the subsidence. Our Rif and western Tell sections ( Fig. 3A and B) illustrate the combination of these processes.
In the western Tell, the foredeep is Lower Miocene to Langhian (Fig. 4;Delteil, 1974;Roure et al., 2012), synchronous with the Lower Miocene infill of the Chelif Basin (Perrodon, 1957;Roure et al., 2012). At this stage, the Chelif developed as a piggy-back basin, in front of underplated basement slices of the EMM. In the Gareb-Kebdana Recess, the foredeep is limited to the so-called "Gareb-Kebdana Chaotic Unit" (Hervouët and De Luca, 1980;Hervouët, 1985) of Lower to Middle Tortonian age. The absence of Tell-Rif higher nappes in this segment of the belt is noteworthy. This is possibly one of the factors explaining the low subsidence of the foredeep. To the west, the foredeep is Upper Miocene (Upper Tortonian-Messinian) (Suter, 1980a(Suter, , 1980b and coeval to the development of the wedge-top collapse-basins of the central and eastern Rif (Fig. 4). This period partly corresponds to the second step of the Chelif Basin, combining lateral collapse and left lateral displacement. It appears consequently that the Tell-Rif foredeep developed mostly on a slope perpendicular to the main movement. Its development occurred during two main steps (Late Burdigalian-Langhian and Upper Tortonian-Messinian, respectively), corresponding to two steps of Alboran lateral escape (Figs. 14 and 15). In between these two steps, the emplacement of the Gareb-Kebdana Chaotic Unit likely marks a period of rapid translation, coeval to the E-W opening of the Algerian oceanic basin (Mauffret et al., 2004;Medaouri et al., 2014).
In summary, the development of a foredeep in front of the western Tell and Rif depends on the tectonic underplating, which led to the formation of the EMM and external tectonic windows, combined with a thin effective elastic thickness of the lithosphere. These two elements are lacking in the eastern Tell. On the one hand, there is no EMM nor external tectonic windows in this part of the belt ( Fig. 2; Fig. 3, Sects. c and d).
On the other hand, this region is out of the Variscan Front (see the postulated position of this front in Frizon de Lamotte et al. [2011], Fig. 10). The crust is consequently older (at least Panafrican) and we can expect a thicker effective elastic thickness of the lithosphere. If we look at the Late Oligocene configuration (Fig. 13), it is interesting to note the development of a wide basin (p.p. the Numidian Basin) at the junction between the Ligurian and Maghrebian Tethys. The basin, which covers not only the oceanic domain but also the African margin and a large part of the Atlas domain (Fig. 13), develops at the front of the southward moving AlKaPeCa. Such a wide basin implied an important rigidity of the lithosphere at this place. The basin is subsequently involved in a progressive deformation leading to its dismembering. During the formation of the orogenic prism, frontal imbrication dominated underplating as evidenced by the absence of tectonic culminations ( Fig. 2; Fig. 3, Sects. c and d).
Finally, during the Upper Miocene the eastern Tell directly overlapped the Atlas belt, without barriers such as the "Mesetas" present in front of the western segments (Fig. 15). We know that the Atlas belt presents a very efficient Triassic salt décollement, which had likely reactivated the system. Therefore, the Atlas system absorbed the shortening by distributing and translating it southward without development of a foredeep in between. In addition, the reactivation of some basement highs thanks to efficient décollement layers in the Atlas (within the Triassic salt but also within the basement) could alter or prevent the deposition of a classical foredeep, as described by Catuneanu (2004) in the Karoo province or by Aschoff and Steel (2011) in the Sevier-Laramide foreland.

Conclusion
Classically, the Tell-Rif belt has been presented as an orogenic system developed since the end of the Eocene as the result of the slab roll-back of the Maghrebian Tethys lithosphere. The analysis of the rich geological (stratigraphical, tectonic, metamorphic) records improves this picture and allows us to propose a more comprehensive integration in the West Mediterranean geodynamics through a revised structural map for the belt and new cross-sections (Figs. 2 and 3). Conversely, a better knowledge of the Tell-Rif may lead to an enhanced understanding of the geodynamic mechanisms active in the West Mediterranean.
Our main goal was to explain the non-cylindrical behavior of the belt. At least two first-order parameters appeared significant. The structural inheritance is a first key point. The southern Tethys paleo-margin, currently included in the Tell-Rif external zones, was a Middle-Late Jurassic E-W highly oblique margin reworking Early-Middle Jurassic rifted basins oblique (i.e. NE-SW) to the main eastward motions of Africa (Fig. 5), themselves inherited from Variscan trends. The Atlas orogenic system preserves some of them in the form of slightly inverted intracontinental belts (Fig. 1). The other basins, more distal and partly oceanic, were strongly inverted and currently are integrated in the western Tell and Rif external zones. Their closure gave rise to the development of secondary subductions involving only the western part of the margin (Figs. 10 and 11).
The subsequent kinematic evolution led to rework this inherited structural frame. Indeed, the EMM occurrence together with the western Tell-Rif foredeep development resulted from the Jurassic frame and the Eocene kinematics allowing the development of subduction of the African paleomargin in the part of the system located west of the West Ligurian Tethys subduction front (Fig. 11). The Miocene kinematics of the Alboran block is also essential in understanding the time lag in the foredeep age (here the oldest age for the emplacement of the nappes) between the western Tell and the Rif, from the Late Burdigalian to the Late Tortonian, respectively (Figs. 12-14).
The Africa-Europe convergence led to the closure of the two branches of the Ligurian Tethys. During the Paleogene, the West Ligurian Tethys has been consumed through subduction (Fig. 10). Its subduction ended with the collision of the AlKaPeCa block with Iberia. In the Rif and the western Tell, the subduction of the African paleo-margin was still active. The peak of HP-LT metamorphism and deformation occurred during the Middle-Late Eocene (45 to 35 Ma). It corresponds to a major event at the scale of the whole belt and is correlated to a period of strong coupling between the Eurasian, Iberian and African plates with development of intra-continental belts (Fig. 11) and far-field compressional deformation.
The Eocene-Oligocene transition (35 Ma) marks a complete kinematic and geodynamic reorganization, the onset of a new cycle. This part of the history corresponds to the closure of the Maghrebian Tethys (Figs. 12-15). The slab roll-back mechanism, associated with back-arc extension, explains the asymmetry of the belt, characterized by a single vergence for each transect (Fig. 3). The progressive accretion of sedimentary slices led to the building of a classical tectonic wedge exposed in the eastern regions. In the western regions, the underplating of metamorphic rocks (inherited from the Middle-Late Eocene event), the coeval development of wedge-top basins and the preeminence of lateral movements led to a more complex configuration. A new kinematic change occurred at 8 Ma. It is particularly spectacular in Western Tell with a switch from left-lateral to right-lateral displacements along E-W faults (Fig. 16).