Palaeoenvironmental and tectonic significance of Miocene lacustrine and palustrine carbonates (Ait Kandoula Formation) in the Ouarzazate Foreland Basin, Morocco

The Ouarzazate Basin is the southern foreland basin to the High Atlas Mountains in Morocco. The sedimentary fill records a sequence dating from the Eocene to Pleistocene that records the interplay between tectonics and climate. This study presents the first stable isotope and facies analyses of the Middle to Late Miocene Aït Ibrirn lacustrine Member (Aït Kandoula Formation). These data test whether the basin was internally draining and enable the development of palaeoenvironmental models for the Middle to Late Miocene. Five sedimentary facies of lacustrine and Carbonates of the Ouarzazate Basin. 2 palustrine limestones are interbeddded with extensive sequences of palaeosols and fluvial sandstones and conglomerates, often associated with evaporite (gypsum) development. These facies can be divided into two facies associations related to water depth and sub-aerial exposure within the basin. In the Serravalian and early Tortonian shallow water successions dominate the stratigraphy, typical of underfilled foreland basin settings. Furthermore, carbonate δ18O and δ13C isotopes from the sections show covariance confirming that these carbonates were deposited within a hydrologically closed basin. However, late Tortonian to Messinian carbonates do not demonstrate the covariance typical of endorheic basins. Additionally, the facies association indicates the presence of deeper water lake systems demonstrating that the basin was externally draining at this time. These results question the established view of tectonic stagnation in the Late Miocene and suggest that the Cenozoic sediments of the Ouarzazate Basin contain a rich and untapped record of climate change and tectonic evolution on the edge of the Sahara desert.


Introduction
Terrestrial carbonates have long been recognised as being excellent archives of climatic, environmental and tectonic information. Such carbonates can be found in extensional, compressional and cratonic settings and form in a wide variety of conditions from deep and shallow permanent lakes, palustrine conditions, to fluvially dominated plains. As a result the boundaries between different terrestrial sub-environments are not always clear (Alonso-Zarza, 2003) especially when there is no clear link between lake size, salinity, and climatic humidity (Herdendorf, 1984). This is especially true in semi-arid and arid environments where sub-aerial exposure and evaporation are common, which can result in pedogenic overprinting of previously deposited lacustrine carbonates forming the palustrine facies characteristic of seasonal wetlands (Platt and Wright, 1992;Wright and Platt, 1995). However, detailed sedimentology and petrography (i.e., Freytet and Verrecchia, 2002), combined with a robust stratigraphic framework (Bohacs et al., 2000) can allow the reconstruction of the morphology and type of palaeolakes. In addition, the geochemistry of primary carbonates can record the interplay between autogenic factors such as basin hydrology and biogenic productivity and allogenic effects of climate change, tectonics and drainage network evolution (i.e., Talbot, 1990;Talbot and Kelts, 1990;Valero-Garcés et al., 1995). Thus, providing records of terrestrial palaeoenvironmental changes that occurred through the evolution of lake systems.
In foreland basin settings, available accommodation space is controlled by the competition between subsidence, driven by loading, and uplift, resulting from thickening and rebound (DeCelles and Giles, 1996). While sediment supply reflects climate, uplift rate and river catchment size (Allen et al., 2013). Thus the balance between subsidence and sediment flux results in underfilling, filling or overfilling of the available accommodation space in the basin and is preserved in the sedimentary record through facies patterns and grainsize trends (i.e., Duller et al., 2010;Parsons et al., 2012). Therefore, lacustrine-palustrine wetlands in foreland basins can be sensitive recorders not only of palaeoenvironments within the basin but also reflect the uplift and erosion of the adjacent mountain front and the evolution of foreland basin drainage configurations. Here, Miocene limestones from the Ouarzazate Foreland Basin of Morocco ( Fig. 1) are described using standard facies descriptions for the first time. In addition, the first stable isotope data from these sediments are presented challenging the long held hypothesis that the palustrine-lacustrine sediments were deposited entirely within a closed basin environment (Görler et al., 1988;El Harfi et al., 2001). These data not only provide new insights into the palaeoenvironments of the Ouarzazate Basin in the Middle to Late Miocene but also have implications for our understanding of the evolution of the adjacent High Atlas Mountains.

Geological Background and stratigraphic framework
The High Atlas Mountains are a ~ W -E trending intracontinental mountain belt formed through the inversion of a Mesozoic rift system owing to N-S directed compression in the Cenozoic (e.g., Jacobshagen et al., 1988;Frizon de Lamotte, 2000). The South Atlas Fault (SAF) and the North Atlas Fault (NAF) form the southern and northern margins of the High Atlas, respectively (Fig. 1). The Anti-Atlas Mountains to the south form the forebulge to the southern foreland basins.
Mountain building is thought to have commenced in the Eocene, although the exact timing of deformation is still a matter of debate owing to the range of evidence used to investigate the development of the orogeny (Fig. 2). Some authors advocate multiple phases of uplift primarily based upon sedimentological observations (i.e., Görler et al., 1988;Frizon de Lamotte et al., 2000;El Harfi et al., 2001). While others propose continuous deformation from the Oligo-Miocene onward (i.e., Babault et al., 2008;Teson and Teixell, 2008;Balestrieri et al., 2009) based upon structural relationships and apatite fission track data.
The Hadida Formation, was interpreted by Görler et al. (1988) as being deposited in proximal braided rivers and alluvial fans during the Early Oligocene to Early Miocene. While, the overlying Aït Kandoula Formation was subdivided into three lithostratigraphic units; an 'alluvial base member', the 'lacustrine member' and the 'alluvial top member' loosely dated as being Early Miocene to Pliocene in age (Görler et al., 1988). Görler et al. (1988) described the lacustrine member as being a sequence of mudstones with various interbeds of conglomerates, limestones and gypsum, which they interpreted as representing environmental changes between freshwater lakes and perennial saline lakes in a hydrologically closed basin. Teson and Teixell (2008) and Teson et al. (2010)], showing the location of the sections described herein. B) Inferred stratigraphic age and extent of the measured sections (01, 02 etc.) and lithostratigraphic units of the Ouarzazate basin (El Harfi et al., 2001;Teson et al., 2010) against the geological timescale and magnetic polarity of Gradstein et al., (2012 Teson and Teixell (2008) established a new stratigraphic framework (Fig. 3B) they did not substantially advance the sedimentology and palaeoenvironmental interpretation of the Aït Ibrirn member from that previously described by Görler et al. (1988).
At the present, the lacustrine deposits in the Aït Kandoula and Aït Seddrat basins are located at higher topographic elevations than the locations sampled in the main Ouarzazate Basin (Fig. 3), yet structural analyses of the fold and thrust belt indicate that range front thrusting activated during the deposition of the Aït Ibrirn Member. (Teson & Teixell, 2008). While tectonic deformation may have resulted in some compartmentalization of lacustrine depo-centres, it is not clear if there were substantial changes in elevation between lakes in different parts of the foreland basin system. The main phase of thrusting seems to have occurred during the deposition of the subsequent Aït Seddrat Member (Teson & Teixell, 2008), leading to the suggestion that the lakes could have been connected prior to that time (Görler et al., 1998).

Methods
Field stratigraphic and petrographic observations of the Aït Ibrirn Member were accomplished by sedimentary logging of five key sections, which form a vertical transect through the member. Petrographic analysis of 28 thin sections from carbonate beds were used to identify carbonate microfacies based upon sedimentary, petrographic and textural features (Dunham, 1962;Flugel, 2004

Stratigraphic correlation
For this study, five representative sections of the Aït Ibrirn Member (Fig. 3) were selected owing to the existence of previous age constraints (sections 1, 2, 6) or by being located nearby to previously described localities allowing an approximate stratigraphic correlation (Benammi et al., 1996;Benammi and Jaeger, 2001;Teson et al., 2010). In addition, the sections provide a vertical sequence through the Middle to Late Miocene and a lateral sequence west to east through the Ouarzazate, Aït Kandoula and Aït Seddrat Basins facilitating the evaluation of palaeoenvironmental trends in time and space.
Sections 2 and 4 are from the northern and southern limb of the Amekchoud anticline, respectively (Fig. 3A). Existing dating (Teson et al., 2010) of Section 2 allows us to assign an age to the section measured here as ~12.5 -10 Ma (Fig. 3B).
Section 4 has not been previously described and along strike correlations show that section 4 is stratigraphically higher than section 2, suggesting a possible middle to late Tortonian age for this section. Sections 1 and 6 are located within the Aït Kandoula Basin (Fig. 3A) and form part of a continuous sequence of lacustrine sediments that span > 5 Ma (Benammi et al., 1996). Section 6, located in the centre of the basin, correlates to part of Benammi et al.'s (1996) Afoud section that has been dated to the Tortonian and Messinian (~ 10 -5 Ma). Whereas, section 1 is equivalent to Benammi et al.'s (1996) Oued Tabia section; therefore, this section is probably Tortonian in age (~ 10 -7 Ma). Section 5 is located in the adjacent Aït Seddrat Basin; although no age constraints are available for this section, stratigraphic similarities to the Oued Tabia section suggests that this section could have also been deposited during the same time interval.

Sedimentary Facies Analysis
Fifteen sedimentary facies have been identified in the studied exposures. Summary sedimentary descriptions of the facies are given in Table 1, with facies abbreviations following convention with G for conglomerates, S for sandstones and M for mudstones and siltstones, and L for limestones.

Carbonate facies field description
In the field, the carbonate beds are lime mudstones or wackestones, rich in fragmentary bioclastic material and whole gastropods. Bed thickness is variable from 1.0 ± 0.2 m to < 0.1 m in thickness, the thinner beds are often laterally discontinuous, while bed boundaries are sharp and conformable. Sedimentary structures are generally rare but some horizons do exhibit wavy and undulating lamination and many beds have a rubbly texture.

Carbonate microfacies analysis
Five carbonate microfacies have been identified from detailed petrographic analysis of 18 samples taken from the logged sections, to give greater insight into the carbonate depositional environments. This analysis shows that mudstones are the most common microfacies (and are volumetrically underrepresented in the thin-section analysis as the coarser-grained samples were preferentially selected for further study), followed by wackestones with a variable bioclastic component.
Microfacies have been characterised using Flugel's (2004) lacustrine microfacies (LMF) criteria. The majority of the samples show evidence for post-depositional pedogenic alteration consistent with palustrine environments.

Lime Mudstones/fossiliferous micrite with pedogenic development (LMF1)
This microfacies is composed of a dense micrite matrix exhibiting glaebule (i.e., The original lime mudstone facies is indicative of deposition through settling from suspension, where the micrite likely originated from either cyanobacterial or algal blooms or from abrasion of limestones (Flugel, 2004). This facies is found in deeper protected parts of lacustrine systems as well as in shallower water areas.
The samples studied here are typical of lacustrine carbonates that were affected by later pedogenesis and calichefication typical of palustrine environments.
Circumgranular cracking occurs when the sediments are subjected to seasonal wetting and drying cycles, while the presence of Microcodium and irregular pore space indicates root activity within the sediment (Wright et al., 1995). By contrast, lenticular (moldic) porosity could indicate where evaporitic crystals have been removed, supported by the presence of rare rhombic zoned calcite crystals that are possibly pseudomorphs after dolomite or gypsum (Fig. 4A).

Densely packed peloidal wackestone (LMF 5)
This microfacies is composed of a nodular, peloidal micrite matrix with some recrystallisation to microspar. There is a detrital quartz component and fragmentary bioclasts of ostracods, gastropods, charophytes, and algal crusts. Sparite variably infills the original irregular porosity, rarely some infills have a lens of opaque minerals at the base forming a gravitational infill. As many of these infilled pore spaces have flat floors these can be described as fenestrae.
The presence of micrite peloids and fragmentary fossil material indicates that this facies was deposited as the result of current reworking in shallow water, probably near the lake shoreline. This is supported by the presence of fragmentary algal mat material that forms on the sediment surface within the photic zone (Freytet & Verrecchia, 2002). Peloids are potentially also the result of reworking of primary precipitates or faecal pellets (Burne and Ferguson, 1983). The open-space structures maybe the result of subsequent bioturbation, sub-aerial dissolution (Flugel, 2004) or be part of the original algal mat structure.

Charophycean wackestones (LMF 7) with pedogenic development
This microfacies consists of a dense micritic matrix with variable nodule (glaebule) development with occasional circumgranular cracks beginning to form where nodule development is the most advanced (Fig. 5). There is minor recrystallisation to microspar but generally little sparite formation. Charophytes ( Charophytes are a type of green algal that inhabit freshwater and brackish environments and are highly characteristic of low energy, shallow water, lacustrine environments (i.e., Platt and Wright, 1991). Combined with the presence of ostracods and encrusting algal material this facies is characteristic of the shoreline region of many lakes (Flugel, 2004).
The primary sedimentary fabric is overprinted by secondary features characteristic of caliche development, including dense micrite/glaebules, intergranular cracking, and Microcodium aggregates, indicating subaerial exposure and pedogenesis after deposition. Microcodium are calcified root cells and indicate that root activity had an important role to play in the secondary alteration of these sediments (Wright et al., 1995).

Gastropod Wackestone and Packstone (LMF8)
This microfacies is characterised by whole gastropods (Fig. 4F) and well-preserved charophytes (mostly oogonia but some stem material is also present), with fragmentary bioclastic material mostly derived from ostracods and bivalves/gastropods. There is some primary porosity in shell cavities but otherwise lacks other pore space in the micrite matrix, which is fairly texturally uniform but does contain rare silt-sized quartz grains. There is some recrystallisation of shelly material to sparite and some micrite to microspar. The micrite matrix also exhibits minor glaebule development with associated cracking in some areas.
The bioclastic component of this microfacies is consistent with shallow water lacustrine deposition and is similar to the Charophycean wackestone but with a much higher gastropod fauna suggesting more permanent water conditions (i.e., Burne and Ferguson, 1983).

5.2.5.Carbonate cemented siltstone.
This facies is a siltstone with ~ 70 % clasts and 30 % cement ( This facies is interpreted as the calcrete K-horizon. The sparite cement is consistent with alpha microfabrics dominated by non-biogenic features (Wright, 1990) characteristic of palaeosols in arid or semi-arid environments.

Summary of carbonate facies
The carbonate beds were deposited across a range of very shallow/intermittently inundated environments to slightly deeper water lake environments. The presence of abundant charophytic material is indicative of low-energy shallow water conditions (Platt and Wright, 1991;Flugel, 2004) and along with the presence of ostracods, is characteristic of freshwater or brackish environments. Slightly deeper water facies show evidence of the reworking of bioclastic material and the addition of allocthonous lithic and fine-grained organic matter. While shallow water facies are coarser-grained and contain material sourced from algal mats; seen in the field as irregular laminations. All carbonate facies have evidence of secondary nodule development, dissolution textures or root traces due to subaerial exposure and vegetation growth after deposition leading to the formation of typical palustrine limestone features.

Siliciclastic and Sulphate Facies
Grey to black mudstones and marls ( This facies is characteristic of soil development in semi-arid and arid environments. Mottling is likely the result of the remobilisation of Fe/Mn oxides/hydroxides due to oscillations in the water table (Freytet, 1973) resulting in wetting and desiccation (Alonoso-Zarza et al., 2012). While caliché formation is the result of carbonate leaching and precipitation within the soil profile (i.e., Wright, 1990;Goudie, 1996).
Planar laminated sandstones (facies Sp) are coarse-grained to very coarse-grained, grey litharenites. Beds are < 0.5 m thick with sharp bedding contacts and parallel laminations. This facies likely represents deposition within the upper flow regime owing to the coarse-grain size of the deposits coupled with laminations. Evidence for current flow suggests that these sandstones were deposited from a fluvial system.

Granular to pebbly sandstones (facies Sch) facies is composed of sub-rounded to
well-rounded, poorly-sorted lithic clasts with a clast-supported texture. The bases of the beds are sharp and often erosional into lower strata, while the beds are laterally discontinuous over 10s of metres. The morphology of these beds is highly suggestive of fluvial channels with a later sediment fill, possibly of small bar forms but no clear sedimentary structures were observed in the field.
Clast-supported conglomerates (facies Gc) are composed of sub-angular to subrounded, poorly-sorted lithic clasts with a clast-supported texture. Beds are massive with sharp bases and variable erosion into underlying sediments. This facies is characteristic of sheet floods, and could represent distal alluvial fan deposition.
Fibrous gypsum (Gf), beef or satin spar is composed of needle shaped crystals forming bedding parallel sheets up to a few centimetres in thickness, with the crystals orientated perpendicular to the bedding planes. This gypsum form is common in mudstone facies. This type of gypsum is a secondary precipitate from meteoric water. Growth of such bedding parallel veins is thought to relate to vein opening probably relating to dehydration of gypsum elsewhere in the basin or the effect of tectonic stress (Cobbold et al., 2013). A gypsarenite (Sg) facies has been identified in a single location and is composed of sub-rounded, very fine to fine-grains of detrital gypsum. Bedding is massive. A lack of sedimentary structures makes it difficult to firmly establish depositional process beyond current reworking.

Section 1 -Oued Tabia
Section 1 (Fig. 3) is located 5.7 km NE of Toundout, exposed to the north of the road

Section 2 -Amekchoud
Section 2 lies ~1.5 km west of Amekchoud (Figs. 3 and 8), exposed in a series of wadis and associated badland topography, and correlates to part 1 of the Amekchoud profile sampled by Teson et al. (2010).

Section 4 -Oued Madri tributary
This section lies 3.5 km north of the small village of Aït Said O Mansour (Fig. 3), along a tributary of the River Madri (UTM Zone 29N 3449413N 725043W). The section is exposed in cliffs formed by the Quaternary incision of the river (Fig. 6C).
Bedding is horizontal, lying on the southern limb of the Amekchoud anticline. The base of the member is not exposed in this location.   11A). This facies association is characteristic of the 'Evaporitive' lacustrine association of Carroll and Bohacs (1999) and Bohacs et al. (2000) and represents shallow water lake margin sedimentation.

Deeper water carbonate lake facies association
The sedimentology of the Oued Madri (Section 4) and upper Afoud (Section 6) sections still show abundant evidence of pedogenic modification, yet the limestone facies in these sections are thicker and more developed than observed in the marginal lacustrine facies associations. Charophyte material is abundant and suggests that the photic zone was relatively free of sediment (Dunagan and Turner, 2004), supported by the low amount of detrital quartz observed in thin section.
Abundant bioclastic material and burrowing suggests also that the water bodies were normally oxygenated. Although the presence of evaporites in lower part of the Oued Madri section indicates that at times evaporation would have been high likely leading to salinity variations, also previously suggested by Görler et al. (1988). In addition, the presence of glaebule and intergranular cracking observed in thin section indicates that limestones were still affected by later pedogenesis and subaerial exposure (palustrine conditions).
The Oued Madri section in particular is interesting as there are defined sequences of organic rich mudstone, limestone, followed by mudstone plus or minus sandstones.
These are interpreted as transgressive systems tracts (TST) and highstand sequences as described by Bohacs et al. (2000). Where the thin clastic beds are interpreted as sheet flows representing the rejunvenation of fluvial systems during the TST. These deposits are overlain by typical lake sediments, commonly enriched in organic material, recording the rapid inundation of a low relief surface. Peak organic enrichment has been recognised in many evaporitive lake systems just above the TST probably the result of an increase in primary productivity (i.e., Wilkins Peak Mb., Green River Formation., Bohacs, 1998), represented in this section by the thin black mudstone horizons. Overlying carbonate mudstones and wackestones represent deposition from suspension or reworking of material from the littoral zone during the maximum lake extent.
Therefore, these sections represent areas in the basin where more persistent lakes developed in the Tortonian (Fig. 11B), as indicated by facies associations and bed thicknesses. The marls likely represent the periods of more persistent and deeper water lacustrine conditions, while limestone beds reflect shallower water deposition.
The presence of pedogenic features indicates that even in these areas desiccation and pedogenic alteration was common during lowstands.

Stable Isotopes
Alteration of primary depositional and pedogenic phases can be determined through petrographic and cathodluminescence microscopy. Non-luminescent cements and an abundance of vadose and subaerial exposure features are suggestive of carbonates that have undergone early diagenesis (Valero-Garcés and Aguilar, 1992). Analysis of samples shows mainly dull to moderate luminescence of the micrite with most variability related to glaebule formation (Fig. 5). Combined with a range of features related to subaerial exposure, suggests that these sediments underwent early diagenetic stabilisation and variable post-depositional alteration related to palustrine evolution.
Therefore, 120 bulk samples were drilled for stable isotope analysis from areas of micrite away from obvious areas of palustrine alteration. Lacustrine carbonates typically have depleted ratios of δ 18 O and δ 13 C when compared to marine carbonates (Keith and Weber, 1964), and the limestones of the Aït Ibrirn Member are no exception (Fig. 12).
Both sections 1 and 4 ( Fig. 12) exhibit relatively constant isotope ratios, with no stratigraphically coherent shifts in either δ 13 C or δ 18 O (Fig. 12, Table 2). Carbon isotope ratios are similar in both sections (-6.0 ± 1.0‰ VPDB), while the average δ 18 O ratio is slightly lower in section 1 (-8.1 ± 0.5‰ VPDB) than section 4 (-6.3 ± 1.1‰). Section 6 is somewhat more variable; a ~3‰ negative shift in both δ 13 C and The isotopic composition of samples has also been considered by carbonate facies type at each location (Fig. 12), as not all samples were thin-sectioned this has been simplified into either lime mudstone or wackestone facies recognisable in hand specimen. In addition, the amount of palustrine alteration has been qualitatively assessed as either being low (no or negligible visible alteration) or present (obvious visible alteration in the hand specimen). Most specimens show some evidence for palustrine alteration apart from in section 6, where little obvious alteration is present in some samples.
Section 2 is the oldest section dating to ~12.5 -10 Ma (Teson et al., 2010) with facies dominated by lime mudstone. These isotopic data plot in a single domain and have a clear covariant trend with an r 2 of 0.75, and the regression line has a gradient of 0.79 (Fig. 12).
Sections 4 and 6 date to ~ 10 -7 and 9 -6 Ma, respectively (Benammi et al., 1995;1996;Benammi and Jaeger, 2001). These sections have the most extensive carbonate records and have more variable isotope ratios than the other sections sampled. Section 4, dominated by lime mudstone, exhibits variable δ 18 O, while δ 13 C is somewhat less variable. Overall, the δ 13 C and δ 18 O show a weak co-variant trend with an r 2 value of 0.15 (Fig. 12).
Section 6 has the greatest sample density across wackestones, and low to clearly However, the r 2 values of the two data sets is different with the clearly altered sediments having an r 2 of 0.79 and the low alteration set a r 2 of 0.92. This observation is an indication that palustrine alteration increases the scatter on the δ 13 C values in particular; similar observations have been made in the Miocene lacustrine systems of Spain that were also subjected to post-depositional pedogenesis (Arenas et al., 1997;Alonso-Zarza et al., 2012). By contrast, fewer carbonate beds are present above 120 m but these plot in a similar domain to the wackestones and the regression line has an r 2 value of 0.36. Of note is that the slope of theis regression line (0.26) is lower than for all the regressions on data lower in the sequence.
Sections 1 and 5 are the youngest studied intervals, and are thought to date to ~ 9 -5 Ma (Benammi et al., 1996), although the age of section 5 is poorly constrained.
Section 1 is dominated by gastropod-bearing wackestone with fewer lime mudstones than in the other sections; however, both facies plot in the same isotopic domain.
When the two variables are cross-plotted, there is a weak covariant trend with an r 2 of 0.13 when all isotope data for this location are taken into account (Fig. 12). By contrast, the isotopic analysis from section 5 shows little variance in either δ 18 O or δ 13 C with an r 2 of 0.07 (Fig. 13). In addition, the data plot in a different isotope domain with lower δ 13 C (< 4.5 ‰) than the other samples tested, although the slope of the regression line is 0.17 and therefore similar to sections 1, 4 and the upper part of section 6.

Figure 12 (next page). Stable isotope compositions of the micrite from sections 1 to 4 studied plotted by section height showing the vertical change in isotopic composition through the sections combined with isotope cross-plots for each section, facies association is indicted by colour where orange represents shallow water and blue represents deep water facies. Key to logs is shown on figure 7. Also included is a location map showing location of the Ouarzazate and Ait Kandoula Basins and approximate geological timescale.
Isotope results are reported in standard delta-per-mil (‰) relative to PDB. Figure 13. δ 18 O / δ 13 C cross-plot of section 5.

Implications of isotopic variations laterally and vertically
Section 2 is the oldest section studied (~12.5 -10 Ma; Teson et al., 2010), and is located in the main Ouarzazate Basin. The covariant trend with an r 2 of 0.75 ( Fig.   12) is strong evidence that these carbonates were deposited in a hydrologically closed lake, as it has previously been demonstrated that strong positive correlations (r 2 > 0.7) are characteristic of carbonate precipitation in closed lake environments (Talbot, 1990;Talbot and Kelts, 1990). This interpretation is supported by the presence of evaporites typical of such environments. Stratigraphically higher, section 4, represents the evolution of the Ouarzazate Basin lacustrine system.
Although there is still a weak covariant trend the r 2 (= 0.15) indicates a hydrologically open lake (Talbot, 1990). Yet a positive δ 18 O value indicates that there was either intense evaporation at the base of the section or that residence times were relatively long (Fig. 14), as hydrologically open lakes typically have only small variations in δ 18 O (Talbot, 1990). It is also possible that these basal sediments may still be  (Dunagan and Driese, 1999;Tanner, 2000), yet Alonso-Zarza et al. (2012) demonstrated the pedogenesis does not completely obscure the original lacustrine isotopic signature. This conclusion is also supported by the data from section 6 studied here, where increased pedogenic alteration increases the data variability but does not obscure original trends. Therefore, we interpret that the Ouarzazate Basin lacustrine system evolved from a closed basin in the Serravalian to an open system in the Tortonian. Talbot, 1990;Arenas et al., 1997 Interestingly, there is also a decrease in steepness of the regression line from s = 0.79 to s = 0.17 from section 2 to section 4, suggesting that the area/depth ratio of the lacustrine system had shifted (Fig. 14). It is unlikely that the lake became shallower over time given the sedimentary evidence of deeper water facies, instead we favour the interpretation that the area of the lakes became much greater at this time (Talbot, 1990). The origin of the covariant trends from the two sections is similar suggesting that the original isotopic compositions of the water masses feeding the lacustrine system was unaltered.

Figure 14. Summary of principal environmental controls on oxygen and carbon covariant trends in lacustrine sediments (modified after
Section 6 is the same age as section 4  Similarly section 1, along strike from section 6, has low and stable values of δ 18 O and δ 13 C suggesting that evaporative drawdown was relatively limited at these sites during intervals of carbonate deposition. This suggestion implies that water residency times were short and the poor correlation of the covariance trend indicates that during deposition the lacustrine system was hydrologically open (Talbot, 1990).
Furthermore, similar ranges in the δ 18 O and δ 13 C are seen across all four locations despite locations 1 and 6 being presently located in the Aït Kandoula thrust top and locations 2 and 4 in the Ouarzazate basin. This observation suggests that during the Miocene these different lakes formed part of a single hydrological system.
The age of section 5 is unclear as no age constraints have been published for the Aït Seddrat Basin. Samples for this area are from carbonates with significant pedogenic overprint, although data from elsewhere in the basin and Alonso-Zarza et al. (2012) demonstrate the original trends have been preserved. Therefore, our limited samples suggest deposition in an open lacustrine system and coeval deposition with the top of section 6, suggesting that section 5 is of late Tortonian to Messinian in age based upon existing dating (Benammi et al., 1995;1996). Also of note, is that the isotopes from section 5 sit in a different isotopic domain compared to the other sections. This observation can be explained as section 5 is located in the Aït Seddrat thrust-top basin, to the east of the Aït Kandoula and Ouarzazate basin sediments, indicating the Aït Seddrat Basin was not connected to the rest of the lacustrine basins at this time.
These new isotope data show for the first time that the Miocene Ouarzazate lake system evolved from an initial hydrologically closed condition in the Serravallian to early Tortonian into an open lake system in the late Tortonian. The facies associations identified in the Middle Miocene sediments of the Ouarzazate Basin fall into two groups, one indicating shallow, ephemeral lacustrine conditions typical of the 'evaporitive' facies of Carroll and Bohacs (1999) and the other more persistent and deeper water environments (Fig. 11). It is likely that the grey marls and overlying carbonate horizons indicate periods of highest water levels and freshest water in the basin. Combined with the stable isotope evidence, these data suggest that the Aït Ibrirn Member of the Aït Kandoula Formation was deposited in a lacustrine system that was initially hydrologically closed but over time became a hydrologically open system. Initial carbonate deposition started around 12.5 Ma, and is represented by section 2 and the base of sections 1, 4 and 6. The palaeolake system was possibly confined to smaller pools prone to evaporite deposition resulting in higher salinity water, with carbonate deposition taking place during highstands ( Fig. 11A and 15). These ponds were surrounded by a playa of mudflats and marshes, where soil development overprinted older sediments resulting in the widespread palustrine facies observed. Pedogenic alteration would have been especially pronounced during lake level lowstands (Fig. 11A). Similar palustrine systems with fluctuating water levels have been described elsewhere in the Mediterranean during the Miocene (i.e., in Turkey;

Miocene palaeoenvironments
Alciçek & Alciçek, 2014, and in Spain; Arenas and Pardo, 1999;Saez et al., 2007). Even though, lake levels were low there appears to have been limited transport of coarse alluvium into the basin in the areas studied. This lack of coarse sediment suggests that alluvial material was mainly being trapped in mountain front alluvial fans and that only sandy bedload and muddy suspended load was transported further out into the basin. Only occasional sheet floods and small channels transported coarse-grained material out into the foreland basin. As the Ouarzazate Basin was hydrologically closed at this time, it is reasonable to assume that high evaporation combined with low water input led to seasonal flow in the river systems and the termination of over ground flow in the remaining pools. These characteristics are typical of an underfilled lake basin (c.f., Caroll and Bohacs, 1999) where accommodation space exceeds sediment flux. As a result lake levels rarely reach sill levels and sediments are dominated by evaporite facies interbedded with alluvial-fluvial strata.
During the later Tortonian, it is likely that a lake covered a large portion of the Ouarzazate and Aït Kandoula Basins (indicated by El Harfi et al., 2001) (Fig 15).
δ 13 C and δ 18 O values indicate that at these times the lake waters were variable in terms of oxygen levels and salinity but overall the system was hydrologically open and had transitioned into a 'balanced-fill' lacustrine system (Carroll and Bohacs, 1999). In balanced-fill lakes, accommodation approximately equals sediment supply and periodically surface outflows are developed.
Similarities in isotope measurements also suggest that the Ouarzazate and Aït appears that the Aït Seddrat Basin was being fed by waters with a different isotopic signature and that a return to mainly fluvial deposition had taken place

Climatic and tectonic influences
Traditionally continental carbonates are thought to form when there is relative tectonic quiescence and when climate was neither too arid, nor too humid to prevent limestone formation (i.e., Cecil, 1990;Alonso-Zarza, 2003;Valero-Garcés et al., 2008;Pla-Pueyo et al., 2009;Valdeolmillos-Rodríguez et al., 2011;Cabaleri and Benavente, 2013;Ashley et al., 2014;de Wet et al., 2015). By contrast, the creation of tectonically induced relief has often been recognised in the rock record by the appearance of conglomerates (e.g., Heller et al., 1988;Flemings and Jordan, 1990;DeCelles et al., 1991;Carrapa and DeCelles, 2008). These ideas led El Harfi et al. However, other authors propose that tectonic activity in the High Atlas was continuous throughout the Miocene based upon structural relationships (i.e., Teson and Teixell, 2008) and fission track evidence of exhumation (i.e., Balestrieri et al., 2009) (Fig. 2). Indeed, the presence of underfilled (accommodation space > sediment supply) lacustrine deposits suggest high subsidence rates (Bohacs, 1999) incompatible with a model of total tectonic quiescence. High subsidence rates and tectonic activity have been shown to be essential for the formation of other lacustrine systems. For example, the deposition of the lacustrine Green River Formation (GRF) of North America is associated with periods of tectonism in the Eocene (Roehler, 1993;Pietras et al., 2003;Smith et al., 2015), not periods of inactivity.
Furthermore, facies models of non-marine foreland basins (e.g., Heller et al., 1988;Flemings and Jordan, 1990;Marr et al., 2000;Clevis et al., 2003;Densmore et al., 2007;Armitage et al., 2011;Allen et al., 2013) demonstrate facies retrogradation should occur when subsidence rates are high as a result of the increased lithospheric loading resulting from thrust emplacement. By contrast, phases of tectonic quiescence and erosional unloading cause the progradation of gravel facies across the basin and the displacement of longitudinal rivers distally away from the thrust front (Burbank, 1992;Burbank and Vergé, 1994).
If this is the case then high rates of subsidence during the Serravallian to Tortonian can account for the development of the lacustrine facies, while a Latest Miocene to Pliocene slowdown in thrusting, and reduction in load-induced subsidence, would explain the subsequent transition from the lacustrine sediments to the Plio-Quaternary conglomerates . A cessation of thrusting in this period has previously been suggested (i.e., Faissinet et al.,1988;Balestrieri et al., 2009) prior to another phase of uplift in the Quaternary (Görler et al., 1998;El Harfi et al., 2001;, associated with geomorphic evidence of landscape rejuvenation and gorge formation (i.e., Stokes et al., 2008, Boulton et al., 2014. Although, recent dating of fluvial terraces questions high rates of Quaternary uplift as incision rates are < 0.2 mmyr -1 over the last ~ 100 ka (Stokes et al., 2018).
The Climatic forcing could drive the change from an underfilled to balanced fill system if accommodation space was constant, or increasing, and where sediment/water supply was also increasing (Fig. 13A). The presence of more persistent lakes argues for more water into the basin, potentially due the change from a semi-arid to subhumid climate supporting this argument. Higher lake levels could then have breached a local basin sill. A trend towards more negative oxygen isotope values in the Late Miocene sediments lends support to an increase in precipitation at this time.
However, the Middle -Late Miocene period is notable for global climatic shifts towards colder and drier climates (Potter and Szatmari, 2009) and in North Africa the Late Miocene is considered at the period when aridification of the Sahara developed Sepulchre et al., 2006). Neither of these trends would account for greater lacustrine activity at this time that would suggest wetter not drier conditions ( Fig. 13 inset). Indeed Sepulchre et al. (2006) consider Early to Middle Miocene uplift key for developing the Atlas Mountains as a topographic barrier to moisture, and thus forming the Sahara Desert ~ 8 Ma.
A decrease in accommodation space could also result in the filling of the basin and drive the trend in the sedimentary facies to a balanced fill system, despite the sedimentary evidence of more persistent lake facies (Fig. 15). Leprêtre et al. (2015) recently presented evidence for Middle Miocene uplift until ~ 11 Ma, with subsequent cessation of loading. However, a reduction in accommodation space only (without any change to sediment/water input) would result in poor preservation of carbonate facies inconsistent with the sedimentary record. A reduction in subsidence would still need an increase in sediment supply to achieve a balanced-fill lacustrine system with preservation of sediments.
Finally, capture of the basin could also drive the observed changes in isotope composition. The Draa River is known to have captured the Dades River, which is the main drainage system of the Ouarzazate Basin, forming a 300 m deep canyon through the Anti-Atlas to the south. The timing of this event is not well constrained but is generally placed around the Pliocene -Pleistocene boundary, and has been ascribed to both regressive erosion of the Draa and basin overtopping (Stäblein, 1988;Arboleya et al., 2008). Arboleya et al. (2008) reported alluvial deposits of presumed Mio-Pliocene age resting on the crystalline basement of the Anti-Atlas adjacent to the Draa River supporting the overtopping hypothesis for the change from internal to external drainage.
Currently the temporal constraints on the sedimentary succession in the Ouarzazate Basin, the timing of thrust events and regional climatic trends are not well enough known to fully unravel the competing controls on basin sedimentation, but it is likely that higher rates of tectonic subsidence led to initial lacustrine deposition within the developing foreland basin. Over time the basin gradually filled up and possibly overtopped a sill, either due to the available accommodation space being exhausted or increased sediment/water supply. Headward erosion of the Draa River, or another palaeodrainage could also have captured the Ouarzazate Basin perhaps driven by periodic basin overtopping. While the ultimate mechanism for the changes between alluvial fan and lacustrine sedimentation and the change in lacustrine hydrodynamic conditions still needs further investigation, an explanation solely involving a tectonic hiatus during the Middle to Late Miocene is no longer satisfactory.

Conclusions
The