Magnetostratigraphy of the Ouarzazate Basin: Implications for the timing of deformation and mountain building in the High Atlas Mountains of Morocco.

Two magnetostratigraphic profiles (450 samples) have been carried out to constrain the age of synorogenic formations in the southern foreland of the High Atlas of Morocco. The Amekchoud profile covers the Aït Ouglif and Aït Kandoula alluvial formations that form the bulk of the Ouarzazate basin fill, indicating an age between the upper Langhian and the Messinian (Miocene). Data obtained in the previously unexplored Hadida formation profile covers the oldest terms of the foreland basin succession, but the low quality of the magnetic record only allows proposing a tentative age between the middle Lutetian and an undetermined middle to late Eocene. The correlation of the Amekchoud profile is based on the recognition of the long C5n chron (Tortonian) in the middle part of the section studied and a new vertebrate locality of upper Tortonian age found in the upper part. These results indicate a discontinuous record of foreland basin development in the southern Atlas domain from mid Eocene to late Miocene times, punctuated by an intermediate large hiatus of 20-25 ma (late Eocene to mid Miocene). Thrusting in the Sub-Atlas frontal thrust belt began before the Aït Ouglif and Kandoula formations, probably during the Oligocene, and extends up to recent times. The alternation of periods of deposition with others of no sedimentary record, which does not coincide with specific tectonic events, results probably from the interference of orogenic deformation and the mantle-related thermal uplift events that have been described for the Moroccan Atlas.


Introduction
The High Atlas of Morocco is a high intracontinental mountain chain that has been uplifted by two mechanisms: the first is thrusting with subsequent crustal thickening, and the second is lithospheric buoyancy induced by a thermal anomaly in the mantle. The present-day expression of both processes has been extensively described in the Atlas literature ( [1,2,3,4,5] and references therein), but chronologic constraints are scarce and often contradictory. To unravel the relative chronology of both processes it is crucial to understand the history of mountain building and the evolution of relief in this mountain chain.
The timing of the long-wavelength component of uplift caused by the lithospheric thinning has been deduced either indirectly, i.e., based on the age of related mantle-sourced alkaline magmatic pulses as a first approximation (40-45 ma and past 15 ma), [2,5] or directly from geomorphic/paleoelevation markers (post-Miocene for the current episode [6]).
Much of the dispersion in the published ages of deformation results from poor dating of synorogenic sedimentary formations, which were deposited in continental environments with very scarce fossil remains. The only precise dating available is that by Benammi et al. [17] and Benammi and Jaeger [18] for Neogene sediments in the Aït Kandoula basin (Fig.1), a small thrust-top basin north of the study area. We have undertaken a study of the larger Ouarzazate basin, which contains over 1000 m of fluvial-alluvial sediments, and have complemented it with a study of the frontal imbricates of the Sub-Atlas Zone that display the oldest terms of the molasse succession. The new ages for the syntectonic sediments obtained in this work constitute the first chronological data for the Ouarzazate basin and allow us to accurately define the timing of thrusting in the vicinity and to extract some general inferences on the chronology of the Atlas orogeny during the past 45 Ma.

Geological setting
The High Atlas is a 100 km wide, nearly E-W trending thrust and fold belt, localized in the foreland of the Rift-Tell Orogen in the NW African plate (Fig. 1). It was generated by the Cenozoic inversion of a Mesozoic transtensional rift ( [1,12,19,20,21], and references therein).
The High Atlas is built by Paleozoic and Mesozoic rocks, essentially Jurassic in age. Cenozoic foreland basin deposits around Atlas Mountains are usually scarce, probably due to the surface uplift generated by the thermal anomaly beneath [2]. Among these, the most complete and best exposed successions are in the Ouarzazate basin, a small elongated basin (145 km long, maximum 35 km across and 800 m deep) at the southern side of the High Atlas orogen (Fig. 1).
To the north, the Ouarzazate basin is flanked by a marginal thrust belt called the Sub-Atlas Zone, an imbricate thrust fan constituted by Triassic to Neogene sedimentary rocks, [7,8,13,22]. The sedimentary infill of the basin, dominated in outcrop by Neogene alluvial fan to lacustrine sediments and Quaternary pediments and terraces, onlaps to the south the Precambrian rocks of the Anti-Atlas massif.
In the Sub-Atlas thrust belt pre-and syndeformational sediments are well exposed. Above the Mesozoic, the predeformational succession ends with 150 to 400 m-thick bioclastic limestones and marls of late Maastrichtian-Paleocene to mid Lutetian age [23,24], and an undated conformable unit of red shales with gypsum (Hadida formation, up to 700 m thick) passing laterally to microconglomerates and sandstones (Aït Arbi formation) [9,22]. These Cenozoic units are completely affected by the thrust structures, although local unconformities have been described in the base of the limestone formation [7,25]. However, the regional distribution of this marine formation across the High and Middle Atlas indicates that sedimentation was not in an orogenic setting but rather in a large epicontinental basin related to the Atlantic margin [26], where only fold ridges of local extent were developed. On the other hand, the Hadida and Aït Arbi formations, which are never found within the High Atlas hinterland, contain the first significant amounts of detrital material deriving from the orogen, and display thickness variations that altogether allow to interpret them as the first foreland basin deposits of the Atlas system [10,12,13]. Good exposures of the Hadida succession can be found in the frontal imbricates of the Sub-Atlas Zone.
The leading edges of the Sub-Atlas thrust sheets emerged in the proximal Ouarzazate basin margin in post-Eocene times [13]. Post-Eocene sedimentary units are unconformable on fold and thrust structures in the Sub-Atlas belt, and contain evidences of contemporaneous deformation [13]. They include the 30-40m thick Aït Ouglif conglomerate formation [8], and the up to 1000 thick Aït Kandoula formation [9], which dominates the outcrop of the Ouarzazate basin and is constituted by red shales, sandstones, conglomerates and minor lacustrine limestone beds. The age of the Aït Ouglif formation is poorly constrained, having been attributed to the upper Eocene [22], to the lower Miocene [9] or to the Oligocene [8,11]. The Aït Kandoula formation has been traditionally attributed to the Mio-Pliocene; a condensed succession of this formation in the Aït Kandoula intramontane basin, ~450 m thick, has been dated by Benammi et al. [17] and Benammi and Jaeger [18] as middle Miocene to lowest Pliocene. The Amekchoud anticline is a km-scale fold within the Ouarzazate basin ( Fig. 1) that exposes the Hadida formation in its core and allows the continuous outcrop of a thick (up to 1000 m) succession of Aït Ouglif and Aït Kandoula formations, making it suitable for a magnetostratigraphic study.

Stratigraphy and sedimentology
A profile of the Hadida formation has been sampled for magnetostratigraphy in the Oued Mgoun valley, west of the Hadida village (Sub-Atlas Zone) (Fig. 2a). The profile starts in the upper beds of the Eocene marine succession (31º19'50.59"N; 6º11'19.19"W), which grade transitionally to the red beds of the Hadida formation, ending at the base of the paraconformable Aït Ouglif conglomerates (31º20'25.35"N; 6º11'04.59"W). The total measured section is of 465 m, and represents the maximum thickness of the Hadida formation preserved in the entire Sub-Atlas Zone. The lower part of the formation, with thickness of 325 m, corresponds to a red argillaceous succession with interlayered beds of siltstone, sandstone, nodular gypsum and laminated carbonate (Fig. 3a, b). Two main facies types can be recognized in these intercalations; the first one is characterized by basal channelized cross-bedded sandstone followed by rippled fine-grained sandstone and siltstone, and capped by laminated carbonates with gypsum crusts on top. The other facies type consists of amalgamated bodies of cross-bedded fine-grained and well-sorted sandstone and gypsarenite interpreted as aeolian deposits. This lower part of the Hadida formation corresponds to the distal part of an aggradational arid alluvial fan system were episodic floods dug ephemeral channels that where partially filled by fluvial sandstones. Laminated carbonates and gypsum crusts related to a playa-lake/coastal sabkha environment developed on the abandoned channels. The sheet-like beds of amalgamated aeolian sandstones could be interpreted as the result of episodic development of sand dune fields.
The upper part of the Hadida formation, 140 m thick, corresponds to a succession of red microconglomerates, sandstones and bioturbated siltstones and shales that can be assimilated to a fluvial environment. It starts with a thick large-scale cross-bedded sandstone level followed by thinner beds of massive sandstone, usually with cross-bedding in their lower part and occasionally showing hydroplastic deformation, interlayered with bioturbated siltstones and shales. The microconglomerate and sandstone bodies show erosional basal contacts and fining-upward trends from conglomerates to cross-bedded sandstones and, finally, to massive or bioturbated fine-grained sandstones at the top. This upper part of the Hadida Formation displays a general thinning and fining upwards trend in the base, which turns to a thickening and coarsening sequence in the last 40 m.
The Aït Ouglif and the Aït Kandoula formations have been sampled for magnetostratigraphy and paleontology in the northern part of the Ouarzazate basin, west of the Amekchoud village. A profile of the succession has been studied in two separate parts to include the whole series (Fig. 2b). The lower part of the profile (783 m) starts in the unconformity between the Hadida Fm and the Aït Ouglif Formation, in the core of the Amekchoud anticline (31º12'39.42"N; 6º36'23.05"W) (Fig. 3c), and ends within a shaly succession with thin lacustrine limestone beds below a Quaternary terrace (31º13'26.27"N; 6º36'44.01"W). The upper part of the profile (256 m) has been studied in a small creek in the southeast flank of the Tifarouine hill (Fig. 2b). This part of the profile starts in shales and lacustrine beds that are laterally equivalent to the top of the previous log (31º14'59.77"N; 6º38'26.19"W), and ends 50 m beneath the preserved top of the conglomerates forming the Tifarouine hill (31º15'13.15"N; 6º39'43.89"W).
Just below the Aït Ouglif conglomerate, a subaerial erosive surface on top of the Hadida Formation is marked by pedogenetic carbonates. The Aït Ouglif conglomerates, slightly erosive over this pedogenic horizon at the outcrop scale, are represented by a thin succession (9 m) of amalgamated channels of poorly-sorted conglomerates including pebbles up to 30 cm in diameter, and of coarse-grained sandstones. The finergrained Aït Kandoula fm overlies conformable the Aït Ouglif conglomerate. It consists of a thick sequence of distal alluvial and lacustrine deposits that culminates with the progradation of coarse alluvial fan deposits (Tifarouine hill). Accordingly, two members can be distinguished in this formation: the basal Aït Ibrirn member composed mainly by red shales with sandstone, conglomerate and limestone intercalations, and the Aït Seddrat member, composed predominantly by conglomerates [13].
In the analyzed profile, the Aït Ibrirn member (~800 m in thickness), is organized in three main sequences of alluvial sandstones grading to distal alluvial shales with lacustrine carbonates. The lower one is represented by 185 m of alluvial channels filled by cross-bedded sandstones with conglomerate lags and interbedded with red shales, followed by 210 m of shales including sporadic alluvial sandstone channels and thin levels of limestone with nodular gypsum. The second sequence is represented by 50 m of alluvial channels filled by conglomerates and coarse sandstones in a thickening and coarsening upwards macrosequence, followed by 75 m of shales with sporadic channelized sandstones and conglomerates in the lower part, and lacustrine carbonates, nodular gypsum and gypsarenites in the upper part. The third sequence is represented by 250 m of shales with conglomerates and sandstones in their lower part and lacustrine carbonates in their upper part.
The Aït Seddrat member reaches a maximum preserved thickness of 264 m in the study area, of which 214 were suitable for magnetostratigraphic sampling. It overlies the last lacustrine carbonate of the Aït Ibrirn Member and appears as a sudden input of coarse clastic sediments derived from the High Atlas, representing prograding proximal alluvial fan environments with more humidity indicated by hydromorphic soil horizons. The lower part of the Aït Seddrat Member is represented by conglomerate and sandstone channels with pebbles up to 10 or 30 cm in diameter. They are interlayered with multicolored shales associated to the development of the hydromorphic soils. Channel thickness increase progressively to the upper part of the section, where the interlayered shaly levels become thinner, redder and bioturbated. The conglomeratic channels are multiepisodic, with sharp basal contacts, often with deep erosive scours but also with large flat basal contacts, occasionally showing internal gravel bars. The conglomerates are clast-supported but poorly sorted, with occasional imbricated pebbles. Pebbles derive dominantly of Jurassic carbonates, and their diameter increases from the base to the top of the section where they can attain up to 120 cm.
Myocricetodon cf. afoudensis is represented by first upper molars with well subdivided anterocone (Fig. 4 d, e, and f). The anterolophule is medially connected to the anterocone and sometimes the longitudinal ridge is interrupted between the paracone and the hypocone. A well developed enterostyle is present in part of the sample. Myocricetodon cf. afoudensis from PAM 207 coincides in size with the sample from the type-locality Afoud 6, from the Aït Kandoula basin [29].
The murid Paraethomys cf. miocaenicus from PAM 207 (Fig. 4 h, i, j) corresponds to a very archaic representative of the genus, best fitting in size and morphology the sample identified with the same name from Amama 2 (Algeria; Coiffait-Martin [30]). From the five upper molars recovered from this level, no one has developed a T6-T9 connection, while the T9 is well developed in all the teeth. Zramys cf. hammamai is represented by two teeth which exhibit large size and hypsodonty (Fig. 4k).
With respect to Oligosoricini indet. (Fig. 4 l, m, n), the first lower incisor is very small. Only Paenelimnoecus and Suncus are known to reach such reduced dimensions, however, none of both include any species displaying a bended root-crown disposition in their first lower incisors. Such condition is considered to be plesiomorphic within the shrews, and better fits the profile of the archaic Oligosoricini crocidosoricines (sensu Furió et al., 2007 [31]). Considering also that the first lower molar is also peculiar in its talonid by being relatively long compared with the recent white-toothed African shrews, the two soricid teeth recovered are ascribed to an archaic member of the family Crocidosoricinae.
According to its murid content, sample PAM 207 can be placed in an biochronological intermediate position between the Turolian sites of Amama 2 (7.7 ma according to radiometric dating by Coiffait-Martin [30]) and Khendek el Ouaich [27] in Morocco, on the one hand, and those of Argoub Kemellal (Algeria; Coiffait-Martin et al. [32]) and Afoud 5 (> 6.3 ma, after magnetostratigraphy [17] in the Aït Kandoula basin, on the other hand. Radiometric dating of a volcanic tuff layer above the fossiliferous level of Kendech el Ouaich has been correlated with the dated volcano at Guilliz (north Morocco) that has an age of 7.58 +/-1.2 Ma according to Choubert et al. [33] radiometric data after Harland et al. [34] correction. In addition, Paraethomys cf. miocaenicus from PAM 207 is certainly less derived than the first Paraethomys entering Spain within chron C3An.2n, at about 6.2 Ma [35]. Therefore, the age of PAM 207 should be bracketed between 6.2 and 7.58+/-1.2 ma according to Paraethomys contents, although it may be perhaps placed near the lower boundary, taking into account the age of Afoud 6 (10.1 to 10.5 ma after magnetostratigraphy of Benammi et al. [17]) with similar Myocricetodon content.
The scarce literature about Neogene insectivores from Morocco does not permit a shorter bracketing of this age. The gap of determined species comprised between the Middle Miocene record of Lartetium dehmi from Beni-Mellal [36] and the Upper Pliocene records of Asoriculus, Crocidura and Suncus [37,38], only permits to ascertain that the presence of an Oligosoricini crocidosoricinae is not in disagreement with the biochronological estimation provided by rodents. Fig. 6: Examples of thermal demagnetization in samples from the Amekchoud profile. For each sample it is displayed the orthogonal plot, the stereographic (equal area projection) and the thermal decay curve (°C) of the total intensity of the magnetization (in 10 -6 A/m). Five samples display normal polarity and four reverse polarity. Three distinctive samples come from the important N10 local chron. All samples are projected in a geographic coordinate system, although in most of them this orientation coincides with the paleogeographic reference frame (horizontal bedding).

Sampling and laboratory procedures
Two magnetostratigraphic profiles have been analyzed to achieve a better chronology of the Ouarzazate basin infill. The Hadida profile consists in 465 meters along the Hadida formation. On the other hand, the Amekchoud profile is made of two sub sections, with 783 and 306 meters respectively, and spans along the Aït Ouglif and Aït Kandoula formations [8,9,11]. A total of 450 in situ-oriented cores (using a magnetic compass), were collected with water cooling drilling machine (348 in Amekchoud and 102 in Hadida). The lithology of the samples is as follows: 202 sandstones, 24 microconglomerates, 169 reddish mudstones, 42 limestones and 13 taken from laminated gypsum with shales, giving an average of one site (1 core) every three meters.
Standard specimens were measured in the Paleomagnetic Laboratory of the Earth Sciences Institute "Jaume Almera" (CSIC-University of Barcelona). Thermal stepwise demagnetization, was carried out to unravel the magnetic paleodirections, was performed with a 2G-DC squid-magnetometer and a MMTD oven. After pilot demagnetizations, magnetic cleaning steps were set in 30°-50° C intervals from room temperature up to 690°C; around 14-15 steps were measured in each sample.
Demagnetization orthogonal diagrams were interpreted using the PALDIR software package (Utrecht University). Characteristic remanent directions (ChRM) were fitted using principal component analysis (PCA, Kirschvink, [39]) and demagnetization circles analysis [40] at the sample scale. Fisher [41] statistics was used to calculate mean values at the section scale.
The visual inspection of orthogonal demagnetization diagrams (Fig. 6) allows the definition of two main magnetic components: 1) between 20° and 250°C: a recent magnetic recording of the present geomagnetic field often influenced by the drilling direction, 2) an intermediate temperature component (up to 400-500°C) usually detected in limestone's and represents about 10% of the total dataset. The absence of further reliable information and the unblocking temperatures A typical scattering from terrestrial clastic series [42,43,44,45] is found in the paleomagnetic signal at the sample scale. This is responsible for the apparent equatorial VGP latitudes recorded in some samples and can be caused by, 1) the differential delaying in the final blocking of the magnetization for different grain-sizes in the same rock. 2) Inclination error caused by the sedimentary load. 3) Overlapping of temperature spectrum of primary and presentfield directions at low-intermediate temperatures, which unlikely affects the primary polarity.
The Hadida profile is strongly influenced by these problems (≈50% of ambiguous paleolatitudes; VGP Lat <30° or >-30°) and has been taken with caution to constrain the age of the base of the Ouarzazate Basin, nevertheless the presence of the Aznag fossil locality [24] in the base of the sequence is a good anchor point for this profile. In contrast the Amekchoud profile only displays 25% of ambiguous data (usually belonged to classes II or III) and can be used to build a feasible local polarity sequence.
The paleomagnetic directions fitted with the PCA method [39] have been classified in three distinct groups (Fig. 6): class I samples, where the direction points to the origin, the polarity is clear and the maximum angular deviation is below 10°, class II samples, which display less quality but unambiguously define the magnetic polarity of the magnetic field, and class III samples with doubtful polarities that were not used.
Although no significant fold test can be performed in the data set due to the very small dipping attitudes along the Amekchoud profile, the occurrence of two polarities with pseudo-antiparallel means (Fig. 7) 42). On the other hand, and despite of the local scattering, the detected mean directions do not display any significant rotation when compared with the expected declination in northern Africa during the studied time span [46]. Future data coming from more suitable lithologies should confirm this observation.

Magnetic polarity and correlation with GPTS
Several polarity changes were found along the two studied sections (Fig. . 8 and 9). The following criteria have been considered to define the local sequence of magnetozones: 1) only class I and II directions were considered, 2) three or more sites are needed per zone, and 3) if only three sites are to define the magnetozone, at least one site should belong to class I.
Following these criteria, the Hadida section has to be taken with caution for a detailed correlation with the GPTS, since only 25% of the samples belong to class I and II, and  Kirschvink, 1980) are shown together with the stratigraphic profile. Black, grey, and whitish dots correspond to the sample quality type (classes I, II and III respectively). most local magnetozones are defined by few samples (Fig. 8); this difficulty should be considered in future studies of the Hadida formation geochronology.
On the other hand, the Amekchoud section of the Aït Ouglif and Aït Kandoula formations, which actually form the bulk of the Ouarzazate basin, yielded better results, where more than two thirds of the samples are reliable to define a polarity. This, together with the larger density and length of the profile, allows a better definition of the local magnetozone pattern (Fig. 9). When attempting to correlate with the GPTS, we have reduced noise by removing the equatorial VGPs (<|30°|) from the local sequence ( Fig. 9b and 10). It is worth mentioning that most equatorial data had not been considered for the local zone definition in figures 9b and 10 since they generally (>70%) belonged to class II or III. This filtered local sequence of polarity fits almost perfectly when it is compared to the polarity column exclusively built with the more stable and reliable data from the reddish mudstones. Except for N7, R8 and R11 zones, where less than three sites define the polarity, the remaining magnetozones are characterized by a significant number of sites (between 6 and 34). N10 plus N11 represent the longest and best-defined magnetozone in the entire profile and is a significant anchor point for correlation.
Chron C5n is undoubtedly present and most probably can be correlated to the longest normal local zone N10-N11, characterized by more than 40 reliable samples (classes I and II) along two hundred meters of sequence (Fig. 10); accordingly, R10 are correlated to C5r. Within this assumption, the pattern of N8, R9 and N9 would correlate to C5An. The R8 magnetozone in our section would correspond to chron C5Ar, while N7 and R7 are correlated to C5AAn and C5Aar respectively. Therefore the lower part of Amekchoud profile could be placed in chron C5ABr. Consequently, the age of the base of Aït Ouglif formation in the Amekchoud section is upper Langhian. In the upper part of this profile the correlation has the anchor point of the fossil level PAM 207 (bracketed between 6.2 and 7.58+/-1.2).
As is discussed in the biostratigraphy chapter, the lower limit in age for PAM 207 corresponds to an indirect dating [33], so it must be taken with caution. PAM 207 is in the R13 magnetozone, so we correlate it to C4r taking into account the biostratigraphic constraints. This correlation gives to our PAM 207 level an age of 8.1 to 8.75ma, near the lower limit of the proposed bracket. This implies that the noisy R12 and N12 local zones correspond to the entire chron C4A. In the uppermost part of the Amekchoud section, N13 may be correlated to C4n; R14 and R15 to C3Br and C3Ar (Fig. 10). The resulting average accumulation rate for whole Amekchoud section is of about 15cm/ka.
In summary, our proposed correlation suggests that the Aït Ouglif formation starts in the upper Langhian and the Aït Kandoula formation ends in the lower Messinian, with the limit between Aït Ibrirn and Aït Seddrat members in the upper Tortonian in the Ouarzazate basin. In the Amekchoud  Kirschvink, 1980) are shown together with the stratigraphic profile. Black, grey, and whitish dots correspond to the sample quality type (classes I, II and III respectively). B) Paleolatitude filtered of low quality data: selecting only class I and II, and removing equatorial VGP's. C) Local magnetic sequence from "fine-material" (reddish mudstones and siltstones). area, the Aït Ouglif formation is represented by only 9 m of conglomerate, but its thickness increases near the Sub-Atlas thrust belt, reaching locally up to 60 m [8,11]. This formation has onlapping geometry; consequently the age lower limit of Aït Ouglif formation remains imprecise. Nevertheless, extending the sedimentation rates of the correlation down to the maximum observed thickness, this lower limit remains in Langhian times.
In spite of the low quality of the paleomagnetic record of the Hadida section, the lower 200 m (with a simple pattern of magnetozones; N1 and R1), coupled with the biostratigraphic constraints at the base, allow to make some inferences about the age of the Hadida formation. The Aznag fossil locality ( Fig. 2A), located in the marine beds immediately under the Hadida formation, was dated as belonging to the P11 foraminiferal zone (upper part of Chron C20n and lowest part of C20n) [24]. Hence, the long N1 and R1 magnetozones in the profile may correspond to the mid-late Lutetian C20n -C18r chrons (Fig. 8) giving an accumulation rate for this lower 2/5 of the formation of 7 to 10 cm/ka. From there upwards we cannot define an unambiguous correlation for the rest of the profile, but if we extrapolate the observed accumulation rates as "best-fit accumulation rate-dependent" of Talling and Burbank [47], we may conclude that the rest of the Hadida formation still belongs to the Eocene.

Tectonic implications
Combination of published paleontological data [24] and our magnetostratigraphic correlation indicate that the first alluvial influx from the High Atlas took place in mid Lutetian times (~ 42-44 Ma, chron C20n), marked by the transition of Eocene marine limestones to Hadida/Aït Arbi continental shales, sandstones and conglomerates (Fig. 11). According to the published age of the limestone formation, platform marine sedimentation occurred at an average accumulation rate of 0.7-2 cm/ka, while with our tentative magnetostratigraphic correlation we obtain increased rate of 7-10 cm/ka for the Hadida terrestrial formation. We do not favor an interpretation of the marine deposits as representatives of an early foreland basin stage, because of their reduced sedimentation rate and their present-day distribution in the interior of the High and Middle Atlas [48]. In agreement with El Harfi et al. [11], we interpret the Hadida/ Aït Arbi formations as the first foreland basin in front of the southern High Atlas, thus installed some 42-44 Ma ago. It is worth noting that foreland basin development coincided with marine to continental transition: in addition to an increased terrigenous influx and progradation, and to the effect of an eustatic sea-level fall during the late Lutetian [49], we may speculate that the regression in the area is largely due to mantle-related regional uplift in the whole Atlas domain, which is suggested by the coincidence with the end of the first pulse of alkaline magmatism occurring 40-45 ma ago. A nearly equilibrated competition between buoyant and thrust loads prevented the maintenance of marine conditions in the foreland basin.
According to the relative chronology of deformation proposed by Tesón and Teixell [13], thrusting began in the Sub-Atlas Zone some time between the end of sedimentation of the Hadida formation (Eocene-Oligocene boundary) and the Aït Ouglif conglomerate (late Langhian-early Serravallian). The large hiatus produced during this interval (20)(21)(22)(23)(24)(25) prevents precising the onset of thrusting, but the amount of shortening recorded (70% of the total measured from balanced sections in Tesón and Teixell [13]) suggests it might have started in the Oligocene.
On the other hand, the reason for this long lasting and widespread sedimentary gap remains difficult to explain. El Harfi et al. [10] suggested that it is indicative of a period of tectonic quiescence. However, the unconformable nature of the Aït Ouglif formation on well-characterized folds and thrusts [13] indicates that thrust loads should have been significant at least during part of this period of no sedimentary record. In addition, the hiatus largely coincides with a gap in the Atlas alkaline magmatism, which was not resumed until 15 ma ago. We suggest that during the late Eocene, at the end of the Hadida formation, the area reached a critical high elevation as a delayed response to the lithospheric thinning process recorded by the magmatism few millions years older. We cannot objectively tell whether the Atlas entered in a period of tectonic quiescence, because the Eocene deformation structures themselves have never been sorted out. From this time, the products of erosion from the growing Atlas Mountains and proximal Ouarzazate basin bypassed the foreland, which was probably externally drained since the marine stages of the early Paleogene. An accentuation effect of the climatic changes at the Eocene-Oligocene transition (with global cooling and sea level fall [50,51]) cannot be discarded.
According to our best fit correlation, the south Atlas foreland trapped sediments again, allowing the accumulation of the Aït Ouglif and Aït Kandoula formations from Langhian to Messinian times. During this interval the Ouarzazate basin became internally drained [9] and received the most important influx of sediments. El Harfi et al. [11] suggest but do not demonstrate that this indicates a period of renewed tectonic activity, whereas Missenard et al. [5] follow Frizon de Lamotte et al. [12] in considering it within a period of tectonic quiescence attributing the terrigenous input to the mantle-related thermal doming. However, on the basis of geometrical features displayed in detailed geological cross-sections, Tesón and Teixell [13] argue that the Aït Kandoula sedimentation does mark a stage of generalized thrusting in the Sub-Atlas belt, and that thrust loading was effective in creating a net subsiding basin at Ouarzazate.
In the Pliocene, the sedimentary record is again interrupted in the Ouarzazate basin; no sediments of this age are preserved in front of the thrust sheets. The only sediments of probable Pliocene age in the southern Atlas region are the lacustrine deposits of the hamadas, name that designates limestone tablelands that occur several hundreds kilometers to the ESE of the Ouarzazate basin, bordering the Anti-Atlas chain in southern Morocco and Algeria [52]. The propagation wave of the mantle-related uplift reached the Hamada area in late Pliocene times as discussed by Babault et al. [6].

Conclusions
New magnetostratigraphic and paleontological data from the Ouarzazate basin and flanking thrust belt provide constraints about the age of the terrestrial synorogenic successions of the southern foreland of the High Atlas Mountains. Magnetic stratigraphy, anchored by dated marine sediments just underneath, suggests a mid Lutetian age for the onset of foreland basin deposits; the end of the Hadida sedimentation is less constrained, being probably still in the Eocene. The Aït Ouglif and Aït Kandoula syntectonic formations that form the bulk of the infill of the Ouarzazate foredeep are dated on the basis of magnetic stratigraphy and a new vertebrate fossil locality of late Miocene age. According to our correlation, these formations are comprised between the Langhian and the Messinian in the Ouarzazate basin, although in the Aït Kandoula intramontane basin, north of Ouarzazate, the Aït Kandoula formation may reach the base of the Pliocene if we follow Benammi et al. [17].
Within this chronological frame, a long sedimentary hiatus around 20-25 Ma is defined between the Hadida and Aït Ouglif formations. On top of the Aït Kandoula formation another interruption of the sedimentary record is observed below quaternary pediments and terraces. While the precise causes for the first interruption remain speculative, the mantlerelated uplift documented by Babault et al. [6] after Miocene times may account for the second hiatus and the end of the Tertiary sedimentation in the Ouarzazate basin, which became externally drained some time during this period.
The main thrusting episode leading the generation of the Sub-Atlas external fold-and-thrust belt begins during the Oligocene and extends up to recent times. The initiation of orogeny in the High Atlas hinterland is placed in mid Eocene times (Lutetian), as recorded by the terrigenous influx and subsidence increase of the Hadida and the equivalent Aït Arbi formations.