Lavocat (1954a, 1954b) referred to the strata in the Kem Kem area of Morocco as a component of the “Continental intercalaire”, a term used for broadly comparable rocks in many other locales in northern Africa that are capped by a distinctive hard limestone platform of Cenomanian-Turonian age (Figs 3, 4). Lavocat (1948, 1954a) and Choubert (1952) also referred to this continental-marine package of rock as the "trilogie mésocrétacée", comprising two successive continental units underlying the marine Cenomanian-Turonian limestone. The continental beds were described by Joly (1962) as the "unité inférieure", or "grès rouges infracénomaniens", and the "unité supérieure", or "marnes versicolores à gypse." This two-part division of the continental facies beneath the Cenomanian-Turonian limestone along the escarpment in the Kem Kem region of Morocco has thus been recognized for nearly 70 years, although no formal geological nomenclature has been proposed for these fossiliferous continental facies.

Sereno et al. (1996) introduced the informal term Kem Kem beds for the two lower units of Choubert’s (1952) “trilogie mésocretacée”, which underlie the Cenomanian-Turonian limestone complex (Fig. 3). These lower beds, composed of sandstone and mudstone, reach a maximum thickness of approximately 200 m. More recently, Ettachfini and Andreu (2004) and Cavin et al. (2010) used formational names originally proposed in notes to a geological map of the High Atlas to the northwest of the Kem Kem by Dubar (1949; sometimes mis-cited as 1948). Dubar’s formational names (Ifezouane and Aoufous) are based on his observations of strata on the eastern flank of the High Atlas Mountains (Tinghir, west of Goulmima). As we discuss below, these rocks are generally un-fossiliferous, have a greater presence of evaporite facies, and lack other features of both terrestrial units of the Kem Kem beds (Sereno et al. 1996). Thus, although we do correlate the two units of the Kem Kem beds with the Ifezouane and Aoufous formations on a continuum of relatively small interconnected basins, we show that the Kem Kem units are mappable, distinctive and deserving of formal recognition, and we designate effective type sections for each. We summarize the latest geological and paleontological evidence that suggests that they accumulated on a continental ramp in the Kem Kem region through to ocean margins to the north and west (Fig. 2).

The age of the “Kem Kem beds” has been regarded variously as mid or early Late Cretaceous (Albian-Cenomanian). Lavocat (1948) described the most common faunal elements and estimated the age as Albian or Cenomanian. Choubert (1952) and Lavocat (1954b) noted similarities with the Bahariya Formation in Egypt (Stromer 1936, Dominik 1985, Soliman and Khalifa 1993), which was regarded as Cenomanian in age. Choubert (1952) and Lavocat (1954a), however, assigned the Kem Kem beds to the "infracénomanien," suggesting a probable Albian age. Although ammonites and other nonvertebrate fossils have established the Late Cenomanian-Turonian age of the overlying limestone complex (Ferrandini et al. 1985; Ettachfini and Andreu 2004; Ettachfini et al. 2005; Essafraoui et al. 2015), no fossils were known at that time from the Kem Kem beds that could be reasonably associated with stage-level temporal resolution.

Seven elasmobranchs and several dinosaur genera (Spinosaurus, Carcharodontosaurus, and Deltadromeus) reported from the Kem Kem beds are shared with the Bahariya Formation in Egypt (Sereno et al. 1996). One of these elasmobranch species (Haimirichia amonensis; Cappetta and Case 1975), in addition, has a broad circum-Mediterranean distribution seemingly restricted to Cenomanian strata. As the fauna from both lower and upper units appears similar, Sereno et al. (1996) inferred a Cenomanian age for these sediments, which has generally been accepted (Wellnhofer and Buffetaut 1999, Cavin et al. 2001, Dal Sasso et al. 2005, Cavin et al. 2010). Martin and de Lapparent de Broin (2016) recently reviewed the geology and age of the Kem Kem beds and proposed a late Albian-early Cenomanian age for the locality that yielded Lavocatchampsa.

Paleoenvironmental interpretations of the Kem Kem beds all agree on their continental status, but differ in regarding either fluvial and floodplain deposits (Russell 1996), or deltaic facies with rare lacustrine environments, as predominant (Sereno et al. 1996). Cavin et al. (2010) described the Kem Kem beds as mostly terrestrial with local brackish and freshwater deposits.

Paleoecological interpretation has centered around the taxonomic and numerical dominance of predators and, more specifically, piscivorous terrestrial and aquatic vertebrates. Based on a collection of commercially acquired fossils, Russell (1996) suggested that the abundance of fish served as the primary resource for a diversity of piscivorous crocodyliforms and theropods, and later authors have added supporting evidence (Sereno et al. 1996, Russell and Paesler 2003, Läng et al. 2013, Ibrahim et al. 2014a). Others have regarded this as an oversimplification of a more complex food web (McGowan and Dyke 2009, Cavin et al. 2010).

A notable feature of the Kem Kem assemblage is the taxonomic, numerical and ichnological dominance of theropods among dinosaurs. Some authors regard this as an accurate reflection of the dominance of theropods in the fauna during the Cenomanian (Russell 1996, Mahler 2005, Läng et al. 2013, Ibrahim et al. 2014a, Ibrahim et al. 2014b, Ibrahim et al. 2016). Others have suggested that the perceived diversity (Dyke 2010) and abundance (McGowan and Dyke 2009, Cavin et al. 2010, Dyke 2010) of theropods is a result of geological or collecting biases. As many of the fossils are based on isolated remains, there are ongoing questions over the taxonomic diversity of some subgroups including theropods (e.g., Evers et al. 2015).

With a similar approach, it has been suggested that Kem Kem fossils as a whole represent a “compound assemblage” derived from two formations (Cavin et al. 2010) or from disparate paleoenvironments (Dyke 2010). The geological and paleontological evidence reviewed in this report bears directly on these controversies.

In the late 1940s, Choubert (1948) described bony fish from the Kem Kem beds exposed on the escarpment of the Guir Hamada along the Morocco-Algeria border. From 1948 to 1951, Lavocat brought to light a range of vertebrate fossils, prospecting and collecting on camelback and vehicle along the escarpment formed by the Guir and Kem Kem Hamadas (Lavocat 1949, 1954a, 1954b, Lapparent 1958). His best-known fossil discovery is the partial skeleton of the diplodocoid sauropod, Rebbachisaurus garasbae (Lavocat 1954b, Wilson and Allain 2015) from the locality Gara Sbaa (Fig. 4C).

During the following 45 years from 1950–1995, only sporadic, small-scale fieldwork was undertaken. In the 1970s a small team led by German scientist Helmut Alberti collected fossil vertebrates near Taouz at the northeastern end of the escarpment of the Kem Kem Hamada (pers. com. M. Reich to NI, 2007). The fossils, housed at the University of Göttingen, include isolated remains of cartilaginous and bony fish, crocodyliforms and non-avian dinosaurs. The coelacanth remains were described by Wenz (1981), and some jaw remains were referred to Spinosaurus (Buffetaut 1989, 1992).

Russell (1996) described a collection of commercially acquired fossils, probably collected from the Kem Kem beds, assigning specimens to Carcharodontosaurus saharicus, a new species, Spinosaurus maroccanus, a new genus and species Sigilmassasaurus brevicollis, and an unnamed abelisaurid. Sauropod bones were referred to Rebbachisaurus garasbae and a family of basal titanosauriforms.

In the last 25 years, paleontologists brought to light a diverse array of new vertebrate fossils (e.g., Sereno et al. 1996, Ibrahim et al. 2010, Cavin et al. 2010, Ibrahim et al. 2014a). Commercial fossil collecting in the Kem Kem beds has also accelerated during this time. Many of the fossils described in recent years were acquired from commercial sources of uncertain geographic origin (e.g., Wellnhofer and Buffetaut 1999, Gaffney et al. 2002, Milner 2003, Evers et al. 2015, Hendrickx et al. 2016). These specimens, with very rare exception, are isolated bones recovered from channel deposits with no locality information or taxonomic association.

The Kem Kem beds have revealed an important and remarkably diverse vertebrate assemblage including elasmobranchs, osteichthyes, and basal sarcopterygians (Choubert 1939, Tabaste 1963, Wenz 1981, Forey 1997, Forey and Grande 1998, Taverne and Maisey 1999, Cavin and Dutheil 1999, Dutheil 1999a, 1999b, Cavin and Brito 2001, Taverne 2000, 2004, Cavin and Forey 2001a, 2001b, Filleul and Dutheil 2001, 2004, Cavin and Forey 2004, Forey and Cavin 2007, Cavin et al. 2010, Martill and Ibrahim 2012), amphibians (Rage and Dutheil 2008), lepidosauromorphs (Rage and Dutheil 2008, Apesteguía et al. 2016a, Klein et al. 2017), turtles (Gmira 1995, Tong and Buffetaut 1996, Gaffney et al. 2002, 2006), crocodyliforms (Buffetaut 1976, 1994, Larsson and Sidor 1999, Larsson and Sues 2007, Sereno and Larsson 2009, Martin and de Lapparent de Broin 2016, Young et al. 2017), pterosaurs (Wellnhofer and Buffetaut 1999, Mader and Kellner 1999, Ibrahim et al. 2010, Rodrigues et al. 2011, Martill and Ibrahim 2015, Martill et al. 2018, 2020, Jacobs et al. 2019, 2020, McPhee et al. 2020), non-avian dinosaurs (Lavocat 1948, 1951, 1952, 1954b, Buffetaut 1989, 1992, Sereno et al. 1996, Russell 1996, Milner 2003, Amiot et al. 2004, Novas et al. 2005a, Mahler 2005, Dal Sasso et al. 2005, Ibrahim at al. 2014a, Ibrahim et al. 2014b, Evers et al. 2015, Ibrahim et al. 2016, Chiarenza and Cau 2016, Ibrahim et al. 2017), and a possible avian (Cavin et al. 2010).

Kem Kem vertebrates are typically preserved in two general taphonomic situations, most commonly in clastic fluvial or deltaic facies or, rarely, within a lake, or pond, facies at the locality Oum Tkout. In the predominant fluvial facies, isolated and transported fossils are the norm. Only three associated partial dinosaur skeletons have been recovered, the diplodocoid sauropod Rebbachisaurus garasbae (Lavocat 1954b, Wilson and Allain 2015) and the theropods Deltadromeus agilis (Sereno et al. 1996) and Spinosaurus aegyptiacus (Ibrahim et al. 2014b). In addition, one associated dinosaur skull pertaining to the large theropod Carcharodontosaurus saharicus has been recovered (Sereno et al. 1996). Because of the prevalence of isolated remains, taxonomic identification of some Kem Kem vertebrates remains controversial, with some authors splitting and others lumping recorded taxonomic diversity (Russell 1996, Sereno et al. 1996, Mahler 2005, Rauhut and López-Arbarello 2006, Averianov et al. 2008, Cavin et al. 2010, Ibrahim et al. 2014b, Evers et al. 2015, Ibrahim et al. 2017).

The lentic waters and fine bottom mud at Oum Tkout preserve leaves, crustaceans (prawn, macruran decapod), and the intact skeletons and scales of bony fish (Dutheil 1999b, Filleul and Dutheil 2001, 2004, Garassino et al. 2006). The specimens are acquired by quarrying and splitting blocks followed by fine preparation. Multiple specimens of a given taxon are often preserved.

In 1995 a joint expedition from the University of Chicago and the Service Géologique du Maroc explored southern outcrops of the Kem Kem beds beween Erfoud and Hassi Zguilma (Fig. 1C). Geological sections were logged at intervals, which comprise the majority of the stratigraphic sections presented in this report (Sereno et al. 1996). Notable paleontological finds included the discovery of a partial postcranial skeleton of the theropod Deltadromeus agilis in the upper part of the lower unit. Near the base of the upper unit, a pond locality named Oum Tkout was discovered with complete remains of decapods and articulated bony fish (Dutheil, 1999b). The upper unit also yielded a partial skull of the theropod dinosaur Carcharodontosaurus saharicus and footprint horizons including the only record to date of an ornithischian dinosaur in the Kem Kem assemblage (Sereno et al. 1996). Dozens of macro- and micro-vertebrate fossil localities were mapped and hundreds of isolated fossils and footprints were collected. These specimens currently reside in the University of Chicago Research Collection.

In 2007 and 2008, joint expeditions from University College Dublin, the University of Portsmouth and the Faculté des Sciences Aïn Chock (Casablanca) collected fossils and recorded ichnological, taphonomic, and sedimentological data, focusing on seven sites between Erfoud and Hassi Zguilma (Tables 1–3, Figs 4, 5; Ibrahim et al. 2014a, 2014b, 2016). Vertebrate fossils include bony fishes, turtles, crocodyliforms, the large azhdarchid Alanqa saharica (Ibrahim et al. 2010, Martill and Ibrahim 2015), and dinosaurs, including a partial humerus of a large titanosaurian sauropod (Ibrahim et al. 2016). These specimens reside in the collections of the Faculté des Sciences Aïn Chock in Casablanca.

In 1999, 2000, 2002, and 2011, field work was undertaken at the pond locality Oum Tkout by a joint team from the Muséum national d’Histoire naturelle, the Service géologique du Maroc, and the University Cadi Ayyad. Quarrying operations resulted in the discovery of numerous nonvertebrate fossils (plants, insects, ostracods, decapods, etc.) as well as articulated elasmobranchs and actinopterygians. Much of this material remains to be thoroughly prepared and studied and likely represents several new taxa.

In 2013, field work was undertaken by a joint team from the University of Chicago, the Museo Civico di Storia Naturale (Milan), and the Faculté des Sciences Aïn Chock (Casablanca) to explore a locality (Zrigat) approximately 20 km north of Erfoud, where a partial skeleton of Spinosaurus aegyptiacus was discovered by a local collector (Ibrahim et al. 2014b: fig. 1A). Additional fragmentary pieces of this specimen were recovered at the site, and the specimen is catalogued in the collections of the Faculté des Sciences Aïn Chock. A geological section was logged across both units of the Kem Kem beds at the nearby Al Gualb Mesa, positioning this locality at the base of the upper unit (Ibrahim et al. 2014b: fig. 1B). Additional fieldwork in the Kem Kem was performed in 2015 on an expedition led by one of us (NI).

In 2018 and 2019, a multidisciplinary team of scientists from the University of Detroit Mercy, the Faculté des Sciences Aïn Chock, the Museo Civico di Storia Naturale (Milan), and the University of Portsmouth, led by NI, explored a number of sites along the Kem Kem escarpment, including several new localities. Finally, regular fieldwork by University of Portsmouth researchers and students has continued to grow the Casablanca collection (Faculté des Sciences Aïn Chock).

Besides major collections in Casablanca (Faculté des Sciences Aïn Chock), Paris (Muséum national d’Histoire naturelle), and Chicago (University of Chicago), additional collections were assembled over the last 50 years from privately acquired specimens collected by locals in villages near the Kem Kem escarpment, although without specific locality data. During a general survey of the Kem Kem beds in 1995, excavation pits in channel sandstones made by local collectors were observed along the entire length of the outcrop. Commercial collecting was therefore already well established by the mid-1990s along most of the available outcrop of the Kem Kem beds with activity concentrated in the northern one-half between Erfoud and Taouz.

The most important collections of commercially collected fossils are in Canada (CMN, Ottawa; ROM, Toronto) and Europe (MNHM, Paris; MPDM, Mèze; BSPG, Munich; MSNM, Milan; NHMUK, London). Several authors (PCS, DMM, DBD, HCEL) have visited some of these collections; one author (NI) has visited all the major collections in the course of doctoral research at University College Dublin, collecting quantitative data on thousands of specimens (Fig. 6, Table 3).

Geographic feature names are often cited or adopted for geographic locations, fossil localities and geological terms. Whereas the Geological Survey maintains an authoritative federal database for geographic feature names in the United States (Geographic Names Information System), no such resource is currently available for Morocco. In the Sahara, geographic feature names often originate in Arabic or Amazigh (Berber) languages and exhibit considerable spelling variation in scholarly papers in German, French or English. Sometimes the meaning of feature names is lost, as may be the case with the most salient feature name in this study, Kem Kem. If a specific meaning to feature names persists, nevertheless, that meaning is often unknown to western scholars. To remedy that situation, we have compiled a list of important fossil localities in the Kem Kem region along with their meanings and spelling variants (Table 4).

Variation in usage can run counter to the original meaning of the name or the feature to which it was originally applied. The term “Tafilalt”, for example, is an Amazigh word referring to a jar made of clay for water and was used as a feature name for the valley of oases south of Errachidia in eastern Morocco. It is used to refer to pre-Mesozoic (mostly Paleozoic) outcrops in the scientific literature (e.g., Wendt et al. 1984, Baidder et al. 2016). It should not be used to describe the Kem Kem escarpment or the location of the Kem Kem vertebrate fauna (Russell 1996), which derives from a larger area extending much farther to the south.

Other variation in geographic names is the result of official name-changing, with older names on geological maps and reports supplanted by newer names. “Ksar es-Souk”, for example, is Arabic meaning “fortified village of the market” and was the longstanding name of a pivotal city in east-central Morocco. In the 1970s it was renamed “Errachidia”. Both names have several spelling variants. Neither should be used with “Province” as the location for the entirety of the Kem Kem outcrop (e.g., Kellner and Mader 1997, Weishampel et al. 2004). The relatively large administrative area of Errachidia Province does not include the southern one-half of the exposures of the Kem Kem beds.

Some geographic names are simple errors that gain traction in secondary citations. In a prominent compilation of dinosaur localities, for example, the term “Tegana Formation” was cited for the “Kem Kem beds” (Weishampel et al. 1990). This may have arisen as a misspelling of the “Tegama Group”, a name for Cretaceous age beds in Niger. Although the error was noted (Sereno et al. 1996), it has reappeared in subsequent publications (e.g., Bailey 1997, Kellner and Mader 1997, Taverne and Masey 1999, Weishampel et al. 2004).

In this report, we formally name several geological units, abiding by the guidelines of the North American Stratigraphic Code (NACSN 2005). Sereno et al. (1996) previously introduced the informal geological term “Kem Kem beds” for the fossiliferous outcrop of the Kem Kem region. Some authors have mistakenly referred to the “Kem Kem Formation” (e.g., McGowan and Dyke 2009, Holliday and Gardner 2012), when no formal geological unit by that name was ever proposed.

“Continental intercalaire”. The Kem Kem beds correlate to the top of a package of continental sediments in basins across northern Africa referred to by Kilian (1931) as the “Continental intercalaire”, or the “intercalated continental (deposits)”. In many regions, the beds assigned to the “Continental intercalaire” discordantly overlie Paleozoic marine sediments that Kilian termed the “Continental de base”. They are overlain, in turn, by a prominent limestone plateau and younger sediments he termed the “Continental terminal”. Additional informal subdivisions (between the three outlined by Kilian; Table 5) have been inserted by geologists and paleontologists from time to time, prominent among them Lapparent and Lelubre (1948). Despite its widespread use, “Continental intercalaire” remains an evolving, poorly defined stratigraphic term. Below we question its utility.

The stratotype for the “Continental intercalaire” is located in the Djoua Valley in Algeria. Rocks of comparable age are located in continental basins to the east in Libya and Egypt, to the north and west in Tunisia, Morocco and Mauritania, and to regions south of the Hoggar Mountains in Niger (Lavocat 1954a, Bellion et al. 1990, Lefranc and Guiraud 1990, Sereno et al. 1996, Benton et al. 2000, O’Leary et al. 2004). Lapparent and Lelubre (1948) and Lefranc (1958) described and divided strata of the “Continental intercalaire” across northern Africa, delineating a number of “series” (Table 6).

The lower boundary is problematic. It does not correspond to any geological event across northern Africa. Kilian (1931) suggested the lower boundary should include all marine sediments of Carboniferous age (thus including the Tiguentourine series; Table 6). Later papers inferred a Triassic age for the earliest beds (Kilian 1937, Lefranc 1958, 1963, Busson 1970), whereas Lapparent (1949) included only the Cretaceous outcrops of the Djoua series (Table 5) in the “Continental intercalaire”. In their review, Lefranc and Guiraud (1990) identified a lower boundary that includes Late Paleozoic strata “younger than Namurian”, although many of these largely un-fossiliferous beds are difficult to correlate. Some authors exclude contemporaneous continental rocks in coastal basins from the “Continental intercalaire”. Cavin et al. (2010: 393), for example, exclude the fossiliferous Upper Cretaceous strata in marginal basins along Morocco’s Atlantic coast.

The upper boundary of the “Continental intercalaire”, in contrast, is bounded by a well-dated, fossiliferous Late Cenomanian-Turonian limestone platform (Fig. 7), which may be the reason the term has persisted. The limestone platform records a sudden global transgression (Gale 2000) that is well exposed in Morocco (Kilian 1931, Lefranc and Guiraud 1990, Choubert 1948, Lavocat 1954a, Ferrandini et al. 1985, Ettachfini and Andreu 2004). The nearshore and terrestrial strata underlying the platform in Morocco correspond with the Djoua series in Algeria, although their precise correlation is poorly constrained. These underlying strata include the Kem Kem beds, which rank amongst the most fossiliferous continental beds of the “Continental intercalaire”.

In sum, there is no consensus regarding the lower boundary of the “Continental intercalaire”, which has never been tied to any regional geological event or episode. Some coeval continental rocks of Cretaceous age, in addition, are excluded because of their location in coastal basins. For these reasons, we question the continued use of “Continental intercalaire” as a heuristic term in discussions of northern African continental strata.

Late Cenomanian-Turonian carbonate platform. Choubert (1948) recognized the Late Cenomanian-Turonian platform (“Calcaires cénomano-turoniens”) as the uppermost unit of his “trilogie mésocrétacée”. The carbonate platform was first described in detail by Ferrandini et al. (1985: 561, fig. 2) on the basis of a stratotype section at Akrabou (cited as “Akerboûss”) approximately 45 km north of Erfoud. Ferrandini et al. (1985) and Ferrandini (1988) divided the platform into four successive subunits, the lower three within the Late Cenomanian and the last in the Turonian, with ages based on considerable biostratigraphic evidence. The subunits, which represent successive transgressive facies (coastal, reef, interior platform, and benthic platform), were not formally designated as members.

Additional sections were logged across the Akrabou Formation (at Ziz near Akrabou and at Tadighoust 50 km to the west) by Ettachfini and Andreu (2004), who also identified four subunits. Only the lower two were placed within the Late Cenomanian; the upper two units were regarded as Turonian in age. A little farther west in the High Atlas Mountains, Ettachfini et al. (2005) divide the Cenomanian-Turonian platform into five subunits, describing it as a new formation (Ben Cherrou Formation). Only the lowest subunit is regarded as Late Cenomanian in age, whereas the upper four are assigned to the Turonian. To the east across the Algerian border in the Béchar region, Benyoucef et al. (2012: fig. 2) divide the platform into four subunits dated by ammonites and other nonvertebrate fossils, the lower three of which were regarded as Late Cenomanian in age and only the uppermost Early Turonian.

Correlation between sections of the Cenomanian-Turonian platform in central Morocco is challenging, because of lateral variation in the character and age of platform subunits. Transgressive conditions, nonetheless, characterize all of these sections, which grade upward from shallow-water coastal facies to a more massive, deeper-water (subtidal) carbonate platform (Ferrandini et al. 1985). The platform complex also markedly thickens to the north of what Essafraoui et al. (2015) identify as the “Kem Kem embayment”, on which the fluvio-deltaic Kem Kem beds were deposited. The platform is typically ~ 30 m thick over the Kem Kem beds. It rapidly thickens north of Aoufous to 90 m at Tazouguerte, which is located only 50 km northeast of Aoufous. Far to the north in the Anoual Syncline of the High Atlas, the Cenomanian-Turonian platform is often thinner, measuring ~ 20 m in thickness.

Kem Kem beds and embayment. The Kem Kem beds are exposed along the generally western-facing escarpment of the Guir and Kem Kem Hamadas. Capped by the Late Cenomanian-Turonian platform, the escarpment extends north-south approximately 200 km near the Morocco-Algeria border (Figs 1A, C, 8–11). The northern boundary of the outcrop is located approximately 30 km south of Errachidia (near Zrigat). It extends southeasterly toward Taouz, following the western edge of the Guir Hamada. From there it extends southwest along the western edge of the Kem Kem Hamada, diving under the low outcrop of the platform ca. 30 km north of Mhamid.

The Kem Kem beds rest unconformably on fossiliferous marine Paleozoic rocks of Silurian, Devonian, and Cambrian age, a contact exposed only in few areas (Figs 12, 13A). The escarpment, generated by Cenozoic erosion (Schmidt 1992), varies in topographic height from ca. 50 to 200 m (Figs 10–12). As much as 90% of the face of the escarpment is covered by limestone talus from the capping Cenomanian-Turonian platform. Available Kem Kem outcrop, therefore, is considerably less than the relatively continuous band shown on regional geological maps. Exposures most often are isolated patches or on the flanks of ravines. Broader areas of outcrop of the Kem Kem beds occur only in regions where the escarpment is prominent, such as near Zrigat, Gara Sbaa and Iferda N’Ahouar.

The Kem Kem beds comprise two units representing predominantly deltaic deposition on a northeast-southwest trending ramp recently termed the “Kem Kem embayment” (Essafraoui et al. 2015: 140, fig. 2). The Kem Kem embayment is bounded to the west by the Anti-Atlas and to the east by the Bechar promontory (Essafraoui et al. 2015). The “Taouz Basin” (Cavin et al. 2010) is another name applied to the structural depression that received sediments in the Kem Kem region. Deposition within this depression constitutes reactivation of the Paleozoic “Tafilalt Basin” in the same area (Wendt 1985, Bockwinkel et al. 2013).

Essafraoui et al. (2015: 162) used published sections (Cavin et al. 2010) and satellite images to conclude that the Kem Kem beds “thin to the south” and were laid down on a “South to North dipping (Tethyan-oriented), low-gradient ramp”. The stratigraphic sections and paleocurrent data we logged during fieldwork in 1995, 2008, and 2013 confirm the main thrust of this interpretation. The exact thickness of the lower unit of the Kem Kem beds often cannot be established, because the contact with the Paleozoic is usually covered. The upper unit and the overlying limestone, nevertheless, thin to the south toward Hassi Zguilma and to the east toward Taouz. The Kem Kem beds, thus, do appear to have been deposited on a low-gradient ramp opening to the north, with sediments derived from the Anti Atlas Massif to the west and lowland regions to the south and east in southern Algeria (Busson 1970).

Northeastern Kem Kem equivalents. Continental beds underlying the Cenomanian-Turonian platform are also exposed in the Béchar area of western Algeria northeast of the Kem Kem embayment. Lapparent (1960) reported on isolated bones and teeth of fossil vertebrates from these beds that are similar to those of the Kem Kem fauna. More recently, Benyoucef et al. (2012, 2014, 2015) documented the presence under the carbonate platform of coastal sabkha deposits in the Guir Basin of western Algeria. These deposits of Cenomanian age, although much thinner (~ 10 m) than the Kem Kem beds, have yielded a fauna broadly comparable to that reported here from the Kem Kem beds (Benyoucef et al. 2015). These data led Essafraoui et al. (2015) to conclude that a north-dipping, low-gradient ramp also existed during the Cenomanian on the Algerian side of the western Saharan craton.

Northern Kem Kem equivalents. The so called “Sillon Préafrican”, or Pre-African Trough, trends southeast-northwest between the Anti-Atlas Massif to the south and the Central and Eastern High Atlas to the north. The trough follows a major rift system (South Atlas Fault or Front) that separates the northern Atlas and Rif regions from the African craton proper to the south (Fig. 13A). Possible equivalents of the Kem Kem beds are preserved on the shoulders, or margins, of the trough and taper in outcrop width from Meski and Aoufous in the west to Goulmima and Tinghir in the east, where they pinch out.

Just a few kilometers north of Zrigat near Aoufous, there are two units of red beds underlying the limestone platform along the Meski-Aoufous Hamada. These beds lack the well-developed cross-bedding common in the lower unit of the Kem Kem beds and are considerably thinner, markedly gypsiferous, and barren of footprints or fossils (Fig. 14). Less than 10 km to the southeast near Zrigat, by contrast, both units of the Kem Kem beds are thick and well developed, as fully exposed and measured on the side of Al Gualb Mesa (Ibrahim et al. 2014b: supplementary Information). Not far from that mesa, a typical Kem Kem fauna was recovered, including a partial skeleton of the dinosaur Spinosaurus aegyptiacus (Ibrahim et al. 2014b).

Farther east near Goulmima, a generalized geological section shows the carbonate platform and underlying marl sediments that could correlate with the upper unit of the Kem Kem beds (Cavin et al. 2010: fig. 3). The published section, however, did not include significant sandy deposits down-section that might correspond with the lower unit of the Kem Kem beds. Confidence in correlation with units of the Kem Kem beds was further weakened by the absence of fossils.

One of us (DMM), however, discovered a better outcrop approximately 10 km to the north of Goulmima at a locality near Tadighoust known as Asfla. Both units comparable to those of the Kem Kem beds are exposed, although the facies differ from those typical of the Kem Kem beds. The lower unit has mudstones that pass upward into more typical channel sandstones, although commonly these are more cemented than in the Kem Kem region. The upper unit is finer-grained, as in the Kem Kem beds, but significant gypsiferous evaporates are more common. Fossils appear to be absent from the upper unit but are present, if rare, in the cemented sandstones of the lower unit. Vertebrate teeth recovered pertain to the sawfish Onchopristis, the theropod cf. Carcharodontosaurus, and a large, subconical tooth pertaining to either a crocodylomorph or spinosaurid.

Farther to the west on the flanks of the High Atlas near Tinghir, a similarly more complete section is present and was visited by one of us (DBD). Two units are present under the carbonate platform that correspond well with the two units of the Kem Kem beds (Gauthier 1960, Essafraoui et al. 2015). The finer-grained upper unit, nevertheless, has more marine marls, marine sandstones, a beach sandstone, and gypsiferous evaporate beds than typical in the corresponding Kem Kem unit (Essafraoui et al. 2015: fig. 16). The lower sandy unit shows ripple marks and occasional bioturbation as in the Kem Kem beds, but lacks the tiered cross-bedding common in the Kem Kem beds. Unlike the Kem Kem beds, vertebrate fossils are very rare and none have been figured (Gauthier 1960).

Choubert (1948), who described the Kem Kem beds, was aware of these infra-platform continental deposits but thought they were laid down in an Atlasic basin separate from the Kem Kem embayment. Using satellite images, Essafraoui et al. (2015) described Tinghir-like infra-platform deposits around the margins of the Anti Atlas Massif, which they regarded as the likely source of terrigenous input for these beds during the Cenomanian. Perhaps that accounts, in part, for the difference in lithology from comparable units in the Kem Kem embayment, which also appear to have had terrigenous input from the south. No matter their sediment source, we agree with the overall Cenomanian correlation outlined by Essafraoui et al. (2015), linking more coastal (Tinghir) and inland (Kem Kem) regions regarding the carbonate platform and two underlying units. Marked lithologic, ichnological and paleontological differences, however, suggest differing depositional conditions and favor the distinction of these units as lateral equivalents.

The two units in the Tinghir region have come to be called, in succession, the Ifezouane and Aoufous formations. Ettachfini and Andreu (2004: 278) and Cavin et al. (2010: 393) attributed authorship of the Ifezouane, Aoufous, and Akrabou formations to Dubar (1949; cited as “Dubar 1948”). Although it may have been Dubar’s intention to replace Choubert’s (1948) “trilogie mésocrétacée” with formations, no formation names, stratotypes or type localities are mentioned in this publication, which comprises notes accompanying a geological map of the High Atlas in the region of Midelt. These formation names appear to have been adopted in 1997 in notes accompanying a more recent geological map of the region around Tinejdad to the south of Goulmima (Ettachfini and Adreu 2004: 279). The next citation of these formation names appears to have been in a doctoral dissertation (Rahlmi 2000).

Given that notes to maps and dissertations are often difficult to obtain and of limited distribution, neither are considered “adequate publication” by the North American Stratigraphic Code (NACSN 2005). Here we suggest that stratotypes be recognized for the Ifezouane, Aoufous, and Akrabou formations based on the most detailed published sections, which are located at Tinghir (Essafraoui et al. 2015: fig. 16), Goulmima (Ettachfini and Andreu 2004: fig. 12), and Akrabou (Ferrandini et al. 1985: fig. 2), respectively.

Far north Kem Kem equivalents. Red bed units that may also correspond to the Kem Kem beds are situated some 50 km north in the Anoual Syncline in the eastern High Atlas, a basin known for yielding a diverse Lower Cretaceous (Berriasian) vertebrate assemblage including mammals and dinosaurs (Hahn and Hahn 2003, Knoll and Ruiz-Omeñaca 2009, Lasseron et al. 2019). The section extends upwards from these fossiliferous horizons to units immediately below a capping Cenomanian-Turonian platform. In 1995 fossils were collected from localities in the Lower Cretaceous beds and a section was logged across well-exposed Cretaceous outcrop immediately north of the well at Aïn Mellouk. No fossils were found in the two units immediately below the Cenomanian-Turonian platform, which in this region has a depth of ~ 20 m.

More recent geological and paleontological work in the Anoual region has resulted in the naming of several formations and the establishment of a better temporal framework based on recovered fossils (Haddoumi et al. 1998). The coarser-grained Dekkar 2 and finer-grained Dekkar 3 formations were named for the beds underlying the Cenomano-Turonian platform (Haddoumi et al. 2016: fig. 2). These formations bear a strong general resemblance to the lower and upper units of the Kem Kem beds, located some 500 km to the south. The Dekkar 2 Formation, for example, is composed of fine-grained marls interleaved with sand- and silt-stones, calcarenitic lenses, and gypsum evaporates. The overlying Dekkar 3 Formation, in contrast, is finer-grained and composed principally of marls interleaved with thin-bedded carbonate and gypsum evaporites (Haddoumi et al. 2016).

Haddoumi et al. (2016) tentatively regarded the Dekkar 2 Formation as Aptian in age, although the only fossils on which the date could have been established are charophytes, ostracods and bivalves near the base of the unit. The Dekkar 3 Formation was regarded as Cenomanian in age; it must predate the Late Cenomanian age of the base of the overlying limestone platform. These ages, therefore, broadly correspond with the age of the Kem Kem beds.

Dekkar formations 2 and 3 are preceded by Dekkar Formation 1, a coarse-grained unit lacking vertebrate remains and tentatively regarded as Barremian to Aptian in age. These Dekkar formations compose the Dekkar Group, bounded below by an unconformity and above by the Cenomanian-Turonian platform.

Here we recognize the Hamadian Supergroup for a package of continental rocks across north Africa (Table 7), exemplified in central and eastern Morocco where they were first identified as the “trilogie mésocrétacée” (Choubert 1948). The name derives from the Arabic word “hamada”, or rocky plateau, which refers to the resistant Upper Cenomanian-Turonian carbonate platform that forms barren plateaus across broad areas of northern Africa. Softer underlying sediments of sandstone, mudstone and marl are exposed on the edges of these resistant plateaus. The carbonate platform has yielded vertebrate fossils as well as a rich nonvertebrate fauna of Late Cenomanian and Early Turonian age. The softer underlying deltaic and marginal marine sediments are the source of most of what is known about terrestrial vertebrate life on Africa during the Early and Middle Cenomanian.

The Hamadian Supergroup includes strata laid down during a sustained, stepwise transgressive trend during the Cenomanian and Early Turonian. The rocks record a global eustatic second-order stratigraphic cycle that generated a tripartite rock record in continental areas, with two rock units underlying a carbonate platform. The underlying formations are deltaic or near shore, the first coarser-grained than the second, the base of which can often be recognized by sustained and significant mudstone deposition. Several pairs of formations of this general description are recognized below as regional manifestations of the Kem Kem sequence.

The “Hamadian series” or “Continental hammadien” (Lefranc and Guiraud 1990) are little used terms coined by Kilian (1931, 1937) for marine strata across northern Africa situated between the “Continental intercalaire” and “Continental terminal” (Tables 5, 6). They included the Late Cenomanian-Turonian carbonate platform and various younger strata of the early Paleogene (Danian). Unfortunately, the upper boundary of this package of units is ill-defined, much like the lower boundary of the “Continental intercalaire”. The problematic upper boundary is limited only by the overlying, poorly dated, terrestrial rocks of the “Continental terminal”.

We recommend, by contrast, the aptly named Hamadian Supergroup as a heuristic term that includes rocks related to a sustained transgressive stratigraphic cycle that are well exposed across northern Africa. The age of initial deposition is best estimated as Early Cenomanian in the Kem Kem region (see below), and the upper boundary of deposition of the carbonate platform is well established as the end of the Early Turonian (Ferrandini et al. 1985, Ettachfini and Andreu 2004, Ettachfini et al. 2005). In this report, we do not extend correlations of the Hamadian Supergroup to northern African formations farther afield (i.e., outside Morocco). We anticipate future studies will extend correlations across northern Africa, anchored by the well-exposed Late Cenomanian-Turonian cliff-forming carbonate platform.

Here we recognize the Kem Kem Group for rocks in central and eastern Morocco that comprise the first two non-marine units of the Hamadian Supergroup (Choubert’s “trilogie mésocrétacée”, Table 7). As discussed below, we cannot differentiate the Kem Kem Group strata in the Kem Kem region on the basis of their rich vertebrate remains. For this reason alone, it is heuristic to have an inclusive term for the pair of infra-platform non-marine units under the Late Cenomanian-Turonian platform. The Kem Kem Group appears to extend across much of central and eastern Morocco north and south of the Pre-African Trough. In this report, we do not extend Kem Kem Group correlations to strata west of Morocco, although we anticipate future studies making correlations to the thinner terrestrial beds in western Algeria (Benyoucef et al. 2015) and beds farther east across northern Africa.

The Gara Sbaa Formation, with stratotype at the locality Gara Sbaa (Fig. 15, Table 7), comprises the lower unit of Kem Kem Group rocks in the Kem Kem region of eastern Morocco. Northernmost exposures are located approximately 30 km south of Errachidia (near Zrigat). The formation is exposed as a relatively narrow band that extends southeasterly toward Taouz, following the western edge of the Guir Hamada, and then southwest along the northwestern edge of the Kem Kem Hamada, pinching out ca. 30 km north of Mhamid (Figs 15–17). Much of the outcrop is exposed in isolated patches or on the flanks of ravines, the remainder covered by limestone talus from the resistant Late Cenomanian-Turonian platform that holds the edge of the hamada. Broader areas of outcrop occur where the escarpment is prominent near Zrigat, Gara Sbaa, and Iferda N’Ahouar.

The Gara Sbaa Formation rests unconformably on fossiliferous marine Paleozoic rocks of Silurian, Devonian and Cambrian age (Figs 18, 19), a contact of some relief that is exposed only in few areas (Figs 12A, 13A, 19). The contact is exposed in sections at Taouz and Ouzina (Fig. 16), where the formation thickens from approximately 60 m to 100 m, respectively. It is thicker and well exposed at Gara Sbaa, where a ~ 135 m section is exposed (Fig. 17). Projecting from Paleozoic outcrops nearby, the basalmost 10–20 m of the formation is covered at Gara Sbaa. Maximum thickness of the formation, thus, is ~ 150 m in the region of the stratotyope Gara Sbaa, thinning to ~ 60 m to the northeast at Taouz and ~ 50 m to the south at Hassi Zguilma (Figs 15–17). The basal contact with bluish Ordovician siltstones is clearly exposed at Ouzina, which is here designated as the boundary stratotype for the base of the formation. The formation, often terminating in a cemented sandstone, is conformably overlain by a substantial red mudstone of the Douira Formation (Figs 20, 21). This contact is well exposed at Gara Sbaa and in other sections.

Lithology. The Gara Sbaa Formation is composed predominantly of red-colored, fine- and medium-grained sandstone beds. The basal one-quarter of the formation is characterized by poorly cemented coarse sandstones with minor conglomerate beds. The basal bed at Tikniouine Bou Tazoult consists of a relatively thin (< 1 m thick) conglomerate with sub-rounded pebbles and cobbles derived from the underlying Paleozoic clastics and, rarely, Paleozoic volcanics. A red-colored medium sandstone forms the basal bed in some areas (e.g., Aferdou N’Chaft). The remainder of the bottom one-quarter of the formation is predominantly medium-to-fine, poorly sorted sandstone with rare siltstone and mudstone beds. The sandstone has interstitial mud and some mafic and feldspar clasts. The poorly sorted sandstone is reddish on both weathered and freshly broken surfaces. Sandstone beds are generally thickly bedded (30–100 cm) with occasional erosional channel scours up to 1 m deep. Limited in exposure, the lower one-quarter of the formation did not yield any body or trace fossils.

The remainder of the formation is composed of coarse-to-fine sandstone beds interspersed with pebble lags. Thin fine-grained beds and paleosols occur rarely. The sandstones show greater maturity than in the basal beds and are composed almost exclusively of well-sorted, well-rounded quartz grains. The sandstone weathers buff-to orange-pink-red and is typically lighter in color, tan-to-yellow, on freshly broken surfaces.

Bed thickness ranges from thinner beds 0.3 m to 1.5 m thick to major units 2 to 80 m in thickness. Thicker beds, which occur toward the top of the formation, occasionally preserve lateral accretion surfaces with paleo-relief up to 8 m. Sandstone bodies typically accrete as tabular beds with a horizontal bottom. Tabular and trough cross-stratification is common with foresets dipping at 20–25° (Figs 22–25). Cross-beds frequently exhibit sorting, the coarser beds including coarse to very coarse sand, granules, and small pebbles. Desiccation cracks and ripple marks indicate intervals of very shallow water conditions (Fig. 25D, E).

The beds in the upper three-quarters of the formation often show cyclic deposition. A basal sandstone bed exhibits trough cross-stratification and sometimes thin conglomerates. This is followed by tabular cross-stratified beds that indicate channel infill. These beds are followed by sandstones rich in yellowish clay and bone fragments that may indicate incipient soil development. These cycles comprise the infilling of channels.

Unusual sedimentary structures include disrupted bedding, spherical concretions, and slab-forming carbonate and iron cemented beds. Disrupted bedding (escape structures, sand volcanoes) and overturned cross-bedding (Fig. 25A-C) in some places suggests thixotropy (liquefaction). Dewatering and slumping on gentle slopes may have been triggered by earthquakes. Pinkish spherical concretions, with diameters of 35–150 mm, are common and sometimes are present in cemented aggregations on bedding surfaces (Figs 26, 27). These iron-rich concretions, sometimes called ‘kerkoubs’ (Cailleux and Soleilhavoup 1976), often, but not always, follow bedding planes. Carbonate- and iron-cemented beds that are dark brown or even black in color occur in places at the top of the formation as a resistant ledge.

The sedimentary structures described above and the lack of substantial conglomerates suggest that the Gara Sbaa Formation records large-scale transport of detrital material typical of a deltaic river system. Cross-bed orientation in the lower quarter of the formation strongly favors a northward flow direction parallel to the axis of the Kem Kem embayment. The change in sedimentary style in the upper three-quarters of the formation suggest the infilling of an expansive alluvial basin centered on the low-gradient Kem Kem ramp.

Paleontology. Vertebrate body fossils most frequently occur as isolated elements (especially bone fragments, teeth, scales) in conglomeratic deposits characterized by rip-up clasts and pebbles. Rostral teeth of a large sawfish, Onchopristis numidus, are often the most common taxon among recovered teeth. Other common fossils include a wide range of aquatic and terrestrial species including polypterids, lepisosteids, amiids, bony fishes, dipnoans, turtles, crocodyliforms, and isolated teeth of the theropods Spinosaurus and Carcharodontosaurus. The teeth of terrestrial herbivores are conspicuously rare and only include sauropods. Articulated skeletons are extremely rare and only include the holotypic partial skeletons of Rebbachisaurus garasbae and Deltadromeus agilis (Sereno et al. 1996, Wilson and Allain 2015). Nonvertebrate remains are rare and include unionoid bivalves recovered at Aferdou N’Chaft and Zrigat. Trace fossils include root traces within rare mudstones, straight, vertical burrows averaging 2 cm in diameter and tens of centimeters long within cross-bedded sandstones, and borings in dinosaur bone fragments (Ibrahim et al. 2014a). The borings are curved, unbranched, tubular traces situated most commonly in the outer portion of weathered bone. These may represent the insect ichnotaxon Cubiculum ornatus or another arthropod feeding on subaerially exposed bone (Ibrahim et al. 2014a). Thalassinoides-like burrows occur in a thin limestone in the Gara Sbaa Formation at Zrigat.

The Douira Formation, with stratotype at the locality Douira (Fig. 16, Table 2), comprises the upper unit of Kem Kem Group rocks in the Kem Kem region of eastern Morocco. As with the underlying Gara Sbaa Formation, the northernmost outcrop is located approximately 30 km south of Errachidia (near Zrigat). The formation is exposed as a relatively narrow outcrop that extends southeasterly toward Taouz, following the western edge of the Guir Hamada, and then southwest along the western edge of the Kem Kem Hamada, pinching out ca. 30 km north of Mhamid. Much of the outcrop is exposed in isolated patches or on the flanks of ravines, the remainder covered by limestone talus from the overlying Late Cenomanian-Turonian platform that holds the edge of the Hamada (Figs 28, 29). Areas of outcrop are restricted and sometimes vertical, given the softness of the formation underlying the resistant carbonate escarpment. The Douira locality is well exposed and particularly rich in micro-vertebrate remains. Work in this section resulted in the discovery of a number of the shark teeth that established a Cenomanian age for Kem Kem Group strata (Sereno et al. 1996).

The formation is thickest at Douira, where a 124-m section is exposed. Maximum thickness of this formation is located more distal (north) than the thickest part of the underlying, progradational Gara Sbaa Formation. Like the Gara Sbaa Formation, it thins to the east to 52 m at Taouz and to the south to approximately 100 m at Oum Tkout and Gara Sbaa and to 46 m farther south at Hassi Zguilma (Figs 16, 17). The Douira Formation rests conformably on the Gara Sbaa Formation, a contact that shows very little relief. The basal bed of the Douira Formation is an easily recognizable reddish mudstone averaging ~ 1 m in thickness. This bed is always the first substantial mudstone in Kem Kem Group strata (Fig. 21). The uniform presence of this fine-grained bed over a broad geographic region suggests there was a transgressive pulse of some sort that altered the sedimentary regime for the region. The upper boundary of the Douira Formation, like the lower boundary, shows only minor relief, and is marked by a massive fossiliferous limestone, the basal unit of the overlying Akrabou Formation. The last sediments of the Douira Formation are typically green or gray claystone or mudstone. In northern sections of the Douira Formation, a few thin-bedded marl or limestone units occur some meters below the massive limestone of the Akrabou Formation. These are recorded in the section at Ouzina and become more substantial and spaced in the more distal (north) section at Douira, the first located some 75 m beneath the boundary carbonate (Fig. 16).

Lithology. The Douira Formation is finer-grained and has a greater variety of rock types than the Gara Sbaa Formation. It consists of fining-upwards, coarse-to-fine grained sandstones intercalated with siltstones, variegated mudstones, and occasional thin gypsiferous evaporites (Fig. 17). Like the Gara Sbaa Formation, the Douira Formation fines upward, and red-hued rocks predominate. Sandstones dominate the lower portion, whereas mudstone, claystone and ledge-forming siltstones and fine sandstones dominate the upper portion. Non-sandstone lithologies, which are rare in the Gara Sbaa Formation, comprise approximately two-thirds of the Douira Formation. Rock types include mudstones (64%), lesser amounts of sandstone (25%) and siltstone (10%), and other lithologies (1%) such as rare thin-bedded gastropod-rich limestones. In measured sections, the volume of silt and finer lithologies ranges from 50% at Gara Sbaa to 84% farther distally (north) on the delta at Iferda N’Ahouar.

Five sandstone facies occur within the Douira Formation. Some 60% of the sandstones are fining-up beds that begin with a coarse-grained sandstone that is poorly sorted with pebble-sized lithics. This basal bed contains cobble-to-pebble-sized clay balls, mud rip-up clasts, bone fragments and isolated teeth. The rest of the facies is composed of fine sandstone or siltstone often characterized by trough and planar large-scale cross-bedding. These beds, which range from 5 to 50 cm in thickness, are also characterized by mud drapes, flaser bedding, vertical burrows and soft-sediment deformation. Rock colors include yellow, orange, red, and tan, with yellow beds sometimes weathering red.

A second sandstone facies comprising ca. 23% of sandstones consists of stacks of cross-bedded, fine- to medium-grained beds with sharp lower and upper contacts. Cross-bedding includes both trough and planar varieties commonly 10–20 cm deep, with a maximum depth of 40 cm. This facies is well-sorted, fine-grained, and buff, red, and occasionally white in color.

A third sandstone facies comprising ca. 16% of sandstones are red to orange, fine-grained beds interbedded with siltstones and mudstones. These sandstones, commonly 5–50 cm thick, often have ripple-scale cross-bedding, laminations, and mud drapes and preserve dinosaur footprints and burrows (Belvedere et al. 2013, Ibrahim et al. 2014a). Typically this facies occurs with thinner overall sequences, usually less than 50 cm in thickness. However, some meters-thick inclined heterolithic stratifications were observed outside of the measured sections. Rarer sandstone facies include massive, medium-grained, orange to red-orange sandstones that are 2 m in thickness or less and yellow units. Finally, there are coarsening-up sequences, from one to a few meters thick.

Siltstones and very fine sandstones are persistent but uncommon in the Douira Formation, representing only ca. 10% of the entire stratigraphic thickness. Often these lithologies occur as units less than 50 cm thick, with bed thicknesses more commonly between 5 and 20 cm. Color may vary from bluish white to red-orange. Small-scale cross-bedding, laminae, and nonvertebrate traces are abundant. Occasionally these units also preserve mud-cracks, dinosaur footprints, mud drapes, and flaser bedding. Soft sediment deformation is rare.

Mudstones and claystones dominate the Douira Formation. In the complete and well-exposed sections, these fine-grained units represent an average of 64% of the stratigraphic thickness. Mudstones occur in four facies: reddish-brown massively bedded units (38%), interbedded mudstones (40%), minor laminated beds, and green claystone. Red-brown to red mudstones, which are common in the lower portion of the formation, are characterized by mottling, slickensides, and blocky to crumbly textures. Calcitic nodules, gypsum crystals, root traces, and burrows are rare.

Interbedded mudstone with claystone is the dominant facies of the upper half of the Douira Formation (Fig. 30). This facies typically shows a mix of diffuse bedding and gradational contacts together with sharper boundaries. Color of these mudstones includes red-brown, purple, red, green, and tan. Mottling and slickensides occur commonly. These lithologies exhibit occasional root traces, and rare four-sided mudcracks and are variably calcareous, with massive blocky and crumbly textures.

Laminated mudstones, best developed at Oum Tkout (Figs 17, 31) and composed mainly of illite, are a minor component of the lower portion of the formation. They are variable in color (red-brown, tan, gray, white). At Oum Tkout, this facies represents a pond setting that preserved an array of plants, insects, decapods, elasmobranchs, and actinopterygians (e.g., Dutheil 1999a, 1999b). The last minor mudstone facies is green claystone, which occurs just below the massive limestone of the Cenomanian-Turonian transgression.

Limestones and marls, which comprise a very small fraction (< 1%) of the Douira Formation consist of blueish gray-to-white, ledge-forming beds less than 50 cm thick with substantial clay and silt. This facies occurs in the upper 20 m of the Ouzina section. At Douira calcareous beds are more common and comprise 7% of the section. Beds 20–30 cm in thickness occur inter-bedded with mudstones and claystones. One very thin (1 cm) gypsum horizon, in addition, occurs just below a calcareous mudstone. This facies and more significant evaporite deposits become much more common in beds of the Kem Kem Group outside the Kem Kem Hamada (Table 7).

Paleontology. Fossils are less abundant than in the Gara Sbaa Formation, although the same range of vertebrate taxa are recorded. Associated dinosaur remains from the Douira Formation include the cranium of Carcharodontosaurus and partial skeleton of Spinosaurus (Sereno et al. 1996, Ibrahim et al. 2014b).

Trace fossils are common in many horizons. The finer-grained deposits and soil development facilitated footprint and other trace fossil preservation. Horizons with vertebrate tracks occur throughout the formation from within a few meters of its base to within 6 or 7 m of the overlying limestone platform (Ibrahim et al. 2014a). They are most common in a “footprint zone” from 30 to 10 m below the top of the Douira Formation (Sereno et al. 1996: fig. 1C). Tracks pertain to medium- and large-sized theropods and, much less frequently, large-sized ornithopods and sauropods (Sereno et al. 1996, Belvedere et al. 2013, Ibrahim et al. 2014a). Tracks occur as impressions and natural casts, the latter the more common and usually occurring in the rare siltstone and sandstone beds. Some are particularly deep and show striations from the motion of the toes through the soft sediment. Nonvertebrate traces include Conichnus, a possible resting trace of a sea-anemone, Scolicia, a gastropod trail, worm-like Beaconites horizontal meniscate burrows, and short, sub-vertical burrows that appear to be dwelling traces of a crustacean (Ibrahim et al. 2014a). These last two trace types occur together in abundance and may represent the activity of detritivores on a tidal flat.

Cenomanian paleogeography. Currently, there are no major river drainages from northern Africa into the Mediterranean west of the Nile River. In Morocco uplift during the Cenozoic created the Alpine Belt to the north of the Preafrican Trough, cutting off drainage to the Mediterranean. In Cretaceous time, however, a major drainage existed between eastern Morocco and western Algeria that flowed northward into the Tethys Ocean (Delfaud and Zellouf 1995). Our paleocurrent measurements and those of Cavin et al. (2010) strongly support this drainage direction (contra Russell 1996). The Morocco-Algeria border region, and in particular the Kem Kem embayment, would have provided a ramp for northward drainage from the western Sahara (Fig. 32). Interconnected basins along that trough may also have allowed secondary drainage from the Kem Kem embayment to the west into the central Atlantic Ocean (Delfaud and Zellouf 1995, Essafraoui et al. 2015). The Kem Kem embayment and Kem Kem Group formations, thus, may be envisioned as the headland of a vast river system feeding north to a prograding delta (Fig. 33).

An evolving delta. The Gara Sbaa and Douira formations of the Kem Kem Group in Morocco have long been envisioned as deltaic in general character (Sereno et al. 1996). Sedimentary structures of the Gara Sbaa Formation were deposited by an anastomosing fluvial system, starting with conglomeratic beds over eroded Paleozoic strata (Fig. 32, stages 1, 2). Fining-upward sediments and sediment cycles characterize the Gara Sbaa and Douira formations. Prograding delta sediments of the upper Gara Sbaa and lower Douira Formation give way and coastal deposits and sabkas in the upper part of the Douira Formation, prior to inundation by a marine transgression (Guiraud et al. 2005, Figs 34, 35). The Gara Sbaa and Douira formations, thus, capture the transition, likely in the Early and Middle Cenomanian, from fluvial to deltaic to lower-energy coastal, pond and sabkha paleoenvironments. Coastal mangrove deposits, recorded in the likely coeval Bahariya Formation in Egypt (Smith et al. 2001), are not present in the sediments of the Kem Kem embayment.

During deposition of the Gara Sbaa and Douira formations as well as other Kem Kem Group formations recognized to date (Table 7), marine influence increases, sedimentary gradient decreases, and lower energy paleoenvironments predominate. The vertebrate fauna, as best as can be determined from transported fossils, does not change appreciably during this interval. Modern analogs for Kem Kem Group beds in Morocco must include a large-scale river system coursing through arid habitats to a prograding delta within reach of a sea or ocean. On Africa the best present-day analog is the Niger delta (Reijers et al. 1997).

Gara Sbaa sediments and paleoenvironments. The conglomeratic components, locally derived clasts, and mix of smaller sandstones indicates within-basin deposition via small-scale fluvial systems at the base of the Gara Sbaa Formation (Fig. 32, stage 1). The increase in bed thickness and larger-scale cross-bedding of overlying sandstones indicate broader and deeper fluvial channels (Fig. 32, stage 2). The rarity of mudstones, the increased lateral extent of sandstone bodies, and their more mature composition indicate reduced accommodation space and lateral redeposition by channels of earlier channel and floodplain deposits. Reworking may have generated the conglomeratic beds with rip-up clasts and pebbles that contain vertebrate teeth and bone (Fig. 36; Rogers 1993).

Several possible indicators of tidal influence occur within the upper portion of the Gara Sbaa Formation through the Douira Formation. These include mud drapes, flaser and lenticular bedding and inclined heterolithic strata. The maturity of the sandstones within this stratigraphic interval may also reflect tidal influence (Tucker 2011). The larger, possibly tidally influenced, channels of the upper portion of the Gara Sbaa Formation indicate deltaic environments. Large-scale cross-bedding with depths exceeding 6 m may indicate progradation of the delta front (Fig. 24).

Douira sediments and paleoenvironments. Evidence favoring deltaic progradation is limited to the lowermost portion of the Douira Formation. As the transgression continued, the entire Kem Kem fluvial system appears to have stalled. Grain size and channel forms diminish up-section. In the deeper northern region of the ramp, evaporites and limestones become more common as clastic input waned. The lowered gradient of the Douira Formation largely consists of a variety of low-energy depositional environments under tidal, and later more fully marine, influence.

These depositional settings involve smaller fluvial channels, floodplains with some incipient paleosol development with root traces, crevasse splay deposits important for the preservation of dinosaur tracks, and a freshwater pond deposit at Oum Tkout with decapods and small-bodied bony fishes (Fig. 32, stage 3; Dutheil 1999a). Marginal marine environments also occur within the uppermost Douira Formation prior to the deposition of the limestone platform of the Akrabou Formation (Fig. 32, stage 4). These environments consist of shallow water tidal flats represented by mudstones with abundant simple tubular dwelling traces or Conichnus, possible sea-anemone resting traces (Ibrahim et al. 2014a). In some sections, green claystones and marls occur at the top of the formation transitional to the overlying limestone. Arid to semi-arid conditions were common in low-latitude northern Africa during the Cenomanian (Scotese 2002), generating intervals of sabkha-like conditions.

Brackish water. The Kem Kem delta was dominated by rapidly moving freshwater/brackish (lotic) paleoenvironments with water flowing toward the open ocean, as opposed to the much rarer still water (lentic) paleoenvironments represented by ponds. The Oum Tkout locality, interpreted here as a pond paleoenvironment, preserves small-bodied (< 10 cm) osteichthyans, such as polypterids and osteoglossiforms. Extant representatives live either in exclusively or predominantly freshwater habitats.

Within the much more common lotic paleoenvironments, the question arises as to whether these were predominantly freshwater, brackish or fully marine. Cartilaginous and bony fish can provide evidence regarding the nature of the water systems. Available evidence points to both freshwater and brackish conditions. Dipnoan tooth plates are common and support freshwater conditions, as all extant dipnoans occupy freshwater habitats.

A diverse assemblage of lamnifom shark teeth, however, suggests that brackish conditions were common. Although the batoid Onchopristis dunklei (McNulty and Slaughter 1962) is a coastal marine species, it is plausible that O. numidus was adapted to freshwater, as the taxon also occurs in Cenomanian deposits in Niger (Lapparent 1953, DBD pers. obs.) at a considerable distance from any maritime coast. Some extant batoids, such as the South American stingray Potamotrygon reside exclusively in freshwater habitats (Stauch and Blanc 1962) but can adjust if subjected to saline conditions (Thorson 1970). The same argument applies to the large Kem Kem coelacanth Axelrodichthys lavocati (previously referred to Mawsonia, Carnier Fragoso et al. 2019), which has also been found in freshwater deposits in Niger and Brazil.

In summary, the Kem Kem fluvial system shows evidence of both freshwater, and brackish conditions. Up section, in the upper portion of the Gara Sbaa Formation and the Douira Formation, tidal indicators suggest brackish conditions may have become stronger.

Hothouse climate. Hothouse conditions likely prevailed during deposition of Kem Kem Group rocks in much of the area now occupied by the Sahara, with harsh seasonality, arid conditions and strong convective storms predominating (Russell and Paesler 2003, Holz 2015). The Earth’s climate is now widely understood as oscillating on a scale of millions of years between icehouse, greenhouse and hothouse states (Fischer 1982, Kidder and Worsley 2010, 2012, Wendler and Wendler 2016, Hu et al. 2017). During the Cenomanian, sea levels reached their maximum during the Cretaceous (Holz 2015), sea surface temperatures were very elevated (Boucot and Gray 2001), and temperate forests grew within the polar circles (e.g., Herman and Spicer 1996, Falcon-Lang et al. 2001).

The ages of the Gara Sbaa and Douira formations are based on relative dating of a suite of nine elasmobranch genera collected from both the Gara Sbaa and Douira formations (Sereno et al. 1996). The overlying Akrabou Formation offers additional relative dates based on marine nonvertebrates, providing a young age limit for the underlying Kem Kem Group sediments (Ferrandini et al. 1985, Garassino et al. 2008, Martill et al. 2011a).

Nine elasmobranchs were collected in the Gara Sbaa and Douira formations (Sereno et al. 1996). Seven of these (and three theropod genera, Carcharodontosaurus, Spinosaurus, Deltadromeus) are shared with the Bahariya Formation in Egypt (Fig. 1), also regarded as Cenomanian in age. One of these elasmobranch species (Haimrichia amonensis Cappetta & Case, 1975) has a broad circum-Mediterranean distribution and has been found elsewhere in Africa and Asia in Cenomanian-age strata (Vullo et al. 2016). No elasmobranch genera have been recovered with an age range restricted to the Albian or earlier. The evidence from elasmobranchs, thus, suggests a Cenomanian age for both the Gara Sbaa and Douira formations.

The Gara Sbaa, Douira, and Akrabou formations comprise a single, stepped transgressive sequence recording a succession of fluvial, deltaic, low-energy coastal environments, to finally an offshore carbonate platform. The formations do not show any major erosional or hiatal surfaces or incised channels that would argue for a contained regressive phase. Instead, marine influence increases steadily up-section to a comformable and gradational contact with the overlying Akrabou Formation. The contact between the Douira and Akrabou formations is conformable and shows almost no topography, as observed at many places along the Guir and Kem Kem Hamadas. A thin laminated clay layer a few centimeters thick is often present in well exposed sections at the top of the Douira Formation immediately below the first carbonate layers of the platform. This suggests that inundation and development of an initial coastal platform occurred swiftly without a significant temporal hiatus sometime in the Late Cenomanian.

This transgression corresponds in general to rising eustatic sea levels during the Cenomanian, although rising sea levels began during the Albian (Holz 2015). The depositional history of the Kem Kem Group appears to represent a single transgressive sequence leading up to this sea-level maximum. Global sea-level curves would thus suggest that the Kem Kem Formation is Cenomanian in age with a maximum age of approximately 98 or 97 Ma (Haq et al. 1987, Ogg et al. 2016) and a total duration of 3.5 to 4.5 Ma. The uniformity of the Kem Kem vertebrate fauna from the two formations is consistent with this relatively short temporal span.

The age of the boundary between the Gara Sbaa and Douira formations is an open question. That boundary is easily distinguished along the length of the Kem Kem region. It also appears to register as a regional event that occurs in comparable strata to the north and east (Table 7). This boundary, which separates a sandy deltaic unit from a predominantly gypsiferous mud, is linked to an abrupt rise of eustatic sea level. Along the Atlantic coast, there are many observable sea level cycles in sections across the Cenomanian, but these cannot be correlated westward by continuous outcrop to Kem Kem Group strata in central and eastern Morocco (Essafraoui et al. 2015). A marked sea level rise during the Cenomanian, nevertheless, is recognized above all others globally and in the well-studied Tarfaya Basin on the Atlantic coast of Morocco. Called the “Mid Cenomanian Event”, it has been dated to approximately 96.0 Ma (Kuhnt et al. 2009, Holz 2015). Here we suggest that this global rise in sea level may be linked to the distinctive shift to finer-grained sedimentation observed in Kem Kem Group rocks.

Regarding the age of the Akrabou Formation, several studies of the carbonate platform in central Morocco have described the range of ammonites and many other nonvertebrates that correspond with two major transgressive events. The first transgression, located at the base of the Akrabou Formation, has yielded a diverse flora and nonvertebrate (limulid, crustacean, insect), elasmobranch and actinopterygian fauna from localities atop buttes near Gara Sbaa with an estimated age near the end of the Cenomanian (Garassino et al. 2008, Martill et al. 2011a, Vullo et al. 2016, Lamsdell et al. 2020). At approximately 94.5 Ma, eustatic sea level swiftly rose to inundate nearshore environments across northern Africa and Europe (Voigt et al. 2006, Kuhnt et al. 2009, Essafraoui et al. 2015). The second transgression occurs within the platform sequence, when rising sea levels generated benthic conditions across much of northern Africa at the Cenomanian-Turonian boundary at approximately 93.6 Ma (Ferrandini et al. 1985, Parente et al. 2008).

The dates discussed above can be assembled into a chronology for Hamadian Supergroup rocks in central and eastern Morocco (Table 7). Deposition may have commenced on a prograding delta in the latest Albian or earliest Cenomanian approximately 100.0 to 99.0 Ma. Sedimentation shifted to finer-grained coastal sedimentation tied to the Mid Cenomanian Event around 96.0 Ma and later underwent rapid inundation and development of a carbonate platform at approximately 94.5 Ma. The platform waters deepened near the Cenomanian-Turonian boundary at approximately 93.6 Ma, with marine conditions persisting until the Middle Turonian regression around 92.0 Ma. The chronology sketched above suggests that the fauna we review in the pages that follow probably comes from a temporal interval within the Cenomanian of less than 5 Ma from approximately 100 to 95 Ma.

Preservation. Five modes of preservation are possible to distinguish when fossils are found in situ in the Gara Sbaa and Douira formations: (mode 1) channel lags of concentrated resistant material (teeth, vertebrae, etc.); (mode 2) microsite lags of concentrated small material (especially teeth); (mode 3) isolated elements (bone fragments or teeth); (mode 4) associated remains (partial vertebrate skulls or skeletons); and (mode 5) a pond deposit (plants, small-bodied teleost fish and decapods).

Most body fossils in the Kem Kem region were probably preserved in modes 1–3 and are discovered as isolated, fragmentary pieces weathered out from sandstones in both formations. Associated or articulated vertebrate specimens (mode 4) are very rare (Fig. 37A). Fossils tend to collect in erosional lag horizons with other hard rock debris (Figs 36B, 38, 39), and so their precise mode of preservation is difficult to ascertain. Robust fossils, such as teeth or vertebral processes, predominate (Fig. 38C), and these are frequently broken and abraded to some degree from transport (Fig. 38B, D). Some small specimens show no discernable indications of transport (Fig. 38E, F), although their size may have reduced the evidence of abrasion.

The best preserved large-bodied vertebrates are isolated specimens of four dinosaurs from different locations, two in the Gara Sbaa Formation and two in the Douira Formation. The partial skeleton of Rebbachisaurus garasbae appears to be the only associated large vertebrate specimen recovered in the Gara Sbaa Formation. Its stratigraphic position and locality are based on the historical records of R. Lavocat (Wilson and Allain 2015: fig. 2). Also in the Gara Sbaa Formation at the locality Aferdou N’Chaft, a partial skeleton of Deltadromeus agilis was discovered weathering from a coarse-grained sandstone. The skeleton was preserved largely intact, as shown by articulated right and left shoulder girdles and forelimbs, right and left pedes and a section of caudal vertebrae and chevrons. It may have been buried suddenly in a channel deposit, judging from the coarse-grained matrix. In the Douira Formation, a partially articulated cranium of Carcharodontosaurus saharicus was discovered weathering from a fine-grained sandstone bluff several km distant attributed to the same locality. The braincase and several complete cranial bones were preserved including a pair of nasals close to one another and one maxilla with teeth preserved in most of the alveoli. Finally, a partial skeleton of Spinosaurus aegyptiacus was found near Zrigat in the northern exposures of the Kem Kem region. From the preserved position of some adjacent bones, it also appears to have been at least partially articulated. Inspection of the site of collection by several of us confirmed its origin in the Douira Formation and that it was preserved in isolation. Given their partial articulation, these four dinosaur specimens could not have been transported significantly before final burial.

The most complete fossils are from a singular pond deposit, Oum Tkout (mode 5), discovered in 1995 and revisited several times in the ensuing years (Dutheil 1999b). Fossils are recovered by quarrying and splitting small blocks of the fine-grained pond deposit. The locality contains small-bodied teleost fishes, prawns, macruran decapods and other soft-bodied nonvertebrates (Fig. 40). The fossils are not concentrated in a single layer or oriented in any particular direction. The rarity of broken specimens suggests that bottom feeding on the remains of decapods and actinopterygians may not have been common. Plant debris is common. The light color of the illite clays that form the majority of the pond mud suggests that anoxia was not a prevalent condition.

A number of bone elements preserve large numbers of borings (Ibrahim et al. 2014a), providing evidence of unrecorded soft-bodied nonvertebrate diversity (Fig. 41). These borings may have been made by insect larvae similar to those of the carrion beetle, Osteocallis mandibulus. The borings together with weather-induced cracking suggest that some bones were subaerially exposed for significant time prior to transport and burial (Roberts et al. 2007, Ibrahim et al. 2014a, fig. 3c).

Nearly all specimens, except those in the singular pond deposit, are preserved in sandstone varying in grain size and degree of silicate cementation; the mudstones composing portions of the Douira Formation appear to be barren. Poorly cemented sandstone is the most common matrix, which is easily removed (Fig. 42A). Patches of more strongly cemented sediment occur (Fig. 42B), and some specimens have a thin hematitic layer adhering to bone surfaces (e.g., the cranium of Carcharodontosaurus). The matrix varies in color from yellow or beige to orange, pink and dark red (Fig. 42C). Similarly, the fossils exhibit a wide range of colors, although orange and red-brown are most common, with teeth commonly a darker color (Fig. 43). Some localities yield fossils of a particular color. Specimens from the Kouah Trick locality collected in the 1950s by R. Lavocat are very dark (Fig. 43B). Specimens from Douira, on the other hand, are often almost white or cream (Fig. 43A). In general, matrix and fossil color is highly variable and cannot be used to confidently identify place of origin.

Completeness. Quantitative logging of isolated specimens collected from several localities in 2008 shows that the majority (~75%) are too incomplete to estimate the percentage of missing bone. Of the remaining more complete specimens, more than half are less than 50% complete. At one locality, Aferdou N’Chaft, approximately half of the collected specimens are nearly complete, although this may be an artifact of small sample size. Clearly most specimens found in the Kem Kem Group are very fragmentary.

The prevalence of breakage among fossils suggests that they were deposited in a relatively high-energy environment and either reworked or transported a considerable distance. Bone breakage, however, does not appear to be a good proxy for distance of fluvial transport (Behrensmeyer 1991). Insect-boring of bones suggests that some terrestrial vertebrates were exposed subaerially before transport and burial.

Abrasion. Most of the fossils can be assigned to abrasion category 2 of Anderson et al. (2007, Table 3). Relatively few are highly abraded (Fig. 44A). At Boumerade, “Valley near Boumerade”, and Aferdou N’Chaft, abrasion is minimal (Figs 45A, B, 46B). At Gara Sbaa a greater range of abrasion is present (categories 1–3, Fig. 45B). Systematic variation in the degree of abrasion between localities may indicate differences in the distance of fluvial transport prior to burial.

Bones, however, can travel long distances without accumulating signs of abrasion (Behrensmeyer 1982, 1991). Other confounding factors in the interpretation of abrasion include the potential for reworking of sandstone deposits within the Gara Sbaa and Douira formations. The time that fossil material has been subject to subaerial weathering in lag deposits prior to collection may represent another potentially confounding factor when considering breakage and abrasion.

Specimen size. Although Kem Kem fossils vary across a wide size range from < 1 cm to 2 m (Fig. 47), the vast majority of specimens measure less than 10 cm in length (Fig. 48). Gara Sbaa provides the best estimate of the range in specimen size, given the large number of fossils collected (Fig. 48B). Specimens collected in finer-grained sediment at Boumerade and the “Valley near Boumerade” tend to be smaller, the largest rarely exceeding 4 cm (Figs 48A, 49B). Aferdou N’Chaft and Iferda N’Ahouar localities seem to have preserved a greater number of larger specimens (Figs 48B, 49A). The size differential between these localities may reflect variance in the hydrodynamics of the deposits. Somewhat larger specimens would be expected to be transported and buried in the higher-energy deposits of the Gara Sbaa Formation. It must be noted, however, that specimens eroding from the Douira Formation may well accumulate in lag deposits on the underlying outcrop of the Gara Sbaa Formation. As the majority of specimens are not preserved in situ, some outcrops with considerable section exposed create uncertainty as regarding the formational provenance of specimens.

Very large fossil specimens are rare. Two partial sauropod limb bones were found in place. In 1995 the proximal end of a large sauropod ulna was discovered in the Gara Sbaa Formation, measuring 54 cm across its proximal articular end (see taxonomic section for further details). In 2008 the mid-section of a large titanosaur humerus was also recovered in this formation, measuring approximately 25 cm at the narrowest portion of its shaft and with a reconstructed length of approximately 1.5 m (Ibrahim et al. 2016).

In sum, there appears to be a strong taphonomic bias against very small (< 2 cm), large (> 6 cm), and soft (plant, nonvertebrate) specimens in the majority of localities in both the Gara Sbaa and Douira formations. The largest sample is from the locality Gara Sbaa, where nearly all vertebrate specimens fall into the 2–6 cm size range.

At the pond locality Oum Tkout, thin films are suggestive of bacteria or fungi (Eumycetes) (Fig. 50). Common plant fossils on Kem Kem outcrop include weathered pieces of petrified wood that probably represent araucarian conifers. Other macroplant remains at the pond locality Oum Tkout include other gymnosperms and spermatopsids (Fig. 51) and angiosperms (Garassino et al. 2006).

Body fossils of nonvertebrates are preserved almost exclusively at the pond locality Oum Tkout. The fine-grained mud sediment of the pond floor preserves whole and partial specimens of soft-bodied mollusks, crustaceans (prawn, macruran decapod, Fig. 52), and larval and mature insects (Fig. 53). The prawn Cretapenaeus berberus has been described from the freshwater locality Oum Tkout in the Douira Formation (Garassino et al. 2006) as well as in the overlying Akrabou Formation (Garassino et al. 2008). At least one macruran decapod remains to be described from Oum Tkout.

The hooked rostral teeth of the sclerorhynchid, Onchopristis numidus, are the most common vertebrate fossil in Kem Kem group sediments (Dutheil 1999b, Cavin et al. 2010), readily found on outcrops along the length of the Guir and Kem Kem Hamadas (Fig. 54). This common northern African sclerorhynchid, initially described under the genus Gigantichthys from Algeria (Haug, 1905) and later placed in a new genus Onchopristis (Stromer 1917), has been recorded in Cenomanian-age rocks elsewhere in Algeria (Broin et al. 1971) and at sites across northern Africa, including Niger (Lapparent 1953, Dutheil 2001), Libya (Lefranc 1958) and Egypt (Stromer 1917, Tabaste 1963, Werner 1989). Most of the other elasmobranch genera reported in the Kem Kem Group come from screen-washing sediment for micro-vertebrate sampling (Dutheil 1996, Sereno et al. 1996, Cavin et al. 2010). These genera pertain to two clades of elasmobranchs, Hybodontoidea and Neoselachii (Table 8).

†Hybodontoidea Owen, 1846. The Kem Kem hybodontoids, represented by isolated teeth and fin spines, appear to be attributable to three genera, Bahariyodon bartheli, Distobatus nutiae, and Tribodus sp. (Dutheil 2001), and two indeterminate acrodontids (Sereno et al. 1996, Dutheil 1999a, Dutheil et al. 2001, Table 8). The genus Hybodus has yet to be reliably reported among isolated teeth in Kem Kem sediments. Diagnostic features of the genus and species reside in its acuminate, multicuspid teeth and cranial features (Maisey 1987); its fin spines thus far have not proven to be diagnostic.

Isolated tooth and fin spine specimens, in addition, cannot be paired with confidence. Two fin spine morphotypes occur in Kem Kem sediments, one with longitudinal striations and the other with tubercles (Fig. 55A–D). Longitudinal striations are known on the fin spines in Lissodus and Tribodus as well as in Hybodus, whereas fin spines with tubercules are reported in Asteracanthus. The partial fin spines collected at Aferdou N’Chaft, Boumerade, and Gara Sbaa may pertain to Tribodus or Acrodontidae indet., but this remains to be substantiated on the basis of association.

Rare teeth of Bahariyodon bartheli (Werner 1989, Duffin 2001) have been recovered from Douira and Dar el Karib localities in the Douira Formation. Teeth of Distobatus nutiae, previously known only from the Bahariya Formation (Werner 1989) and from the Cenomanian of the Draa Ubari in Libya (Nessov et al. 1998, Rage and Cappetta 2002), were also collected from these two localities. A rare tooth morphotype, here attributed to Tribodus sp. (Table 8), is rectangular with vertical ridges, resembling the Brazilian species Tribodus lima (Brito and Ferreira 1989) and Egyptian species Tribodus kuehnei from the Bahariya Formation (Werner 1989).

Galea, Wagler, 1851. The pond locality Oum Tkout has yielded isolated teeth of lamniform elasmobranchs (Dutheil 1999a). The remaining diverse ichthyofauna from this locality is freshwater (Dutheil 1999a). One tooth (Fig. 55E, F) pertains to the mackerel shark Haimirichia amonensis and another one to the cretoxyrhinid Cenocarcharias tenuiplicatus (Cappetta and Case 1975).

Batoidea Compagno, 1973. Rostral teeth of the large-bodied, sclerorhynchid batoid, Onchopristis numidus (Haug 1905), are the most abundant vertebrate element in Kem Kem sediments (Martill and Ibrahim 2012). The teeth of the rostrum and centra are occasionally found in place in both formations (Fig. 54A-C). New cranial and axial material of this taxon include a large rostrum (Fig. 54D-G), an articulated series of more than 50 vertebrae (Fig. 56), and an exceptional new specimen under study that preserves portions of the rostrum and skull in association with vertebral centra (Dutheil and Brito 2009).

The rostrum was recovered in two pieces from the Valley near Boumerade locality (Fig. 9, locality 7) near a partial quadrate that may pertain to Spinosaurus (Fig. 38A). With both pieces of the tapering rostrum positioned as they would be in life, it measures more than 40 cm in length (Fig. 54D, E). Anteriorly it tapers in width, and ventrally it is marked by a sharp-edged trough approximately 5 mm deep and inset from the lateral margin (Fig. 54). A linear, straw-like texture is present on both dorsal and ventral surfaces of the calcified rostrum.

The calcified disc-shaped, biconvex vertebrae have concave sides and decrease in diameter toward the distal end of the series (Fig. 56), which places the series in the posterior portion of the axial column. The articulated series measures more than 80 cm and pertains to an individual that was probably several meters in length.

Another specimen of Onchopristis numidus, which was collected commercially from Kem Kem sediments, preserves portions of the cranium and several anterior vertebrae (Dutheil and Brito 2009). It provides direct evidence of association between the rostrum and rostal teeth of O. numidus and its oral teeth, oral osteoderms and vertebral centra. This specimen confirms previous suggestions by Tabaste (1963) that the common discoid centra in Kem Kem sediments initially described as Platyspondylus foureaui (Haug, 1905), pertain to this species.

Two additional Kem Kem batoids have been found in screen-washed sediment at Douira in the Douira Formation (Dutheil 2001, Table 8). Marckgrafia lybica, initially described from the Bahariya Formation of Egypt (Weiler 1935, Table 8) is represented by 13 teeth. The pristid Peyeria libyca was initially described from the Bahariya Formation of Egypt (Weiler 1935) and is represented by three teeth, although new evidence suggests it may comprise non-rostral denticles of Onchopristis numidus (Sternes and Shimada 2019). Two sections of caudal centra recovered from the pond locality of Oum Tkout may pertain to batoids on the basis of their numerical dominance among Kem Kem elasmobranchs.

Actinopterygii are usually recovered as isolated bones except in rare instances and at the pond locality Oum Tkout, which has yielded nearly complete skeletons. Actinopterygii include basal clades, such as polypterids (Cladistia), lepisosteids and seminotiformes (Ginglymodi), Amiiformes, and a range of teleosts (Table 8). The review below adds new information to previous summaries of the Kem Kem ichthyofauna (Dutheil 1999a, Cavin et al. 2015).

Cladistia Cope, 1871. Cladistians are widespread in Africa (Stromer 1936, Stewart 1994, Gayet and Werner 1997) and have also been reported from the Americas (Gayet and Meunier 1991). Isolated ganoid scales are present in both formations of the Kem Kem Group, suggesting that polypterids may have been a common element of the ichthyofauna.

Four genera have been recorded. At the pond locality Oum Tkout in the Douira Formation, several articulated skeletons have been recovered of the small cladistian Serenoichthys kemkemensis (Dutheil 1999b, Fig. 57). At the same locality, isolated pinnulae (the spine that supports each dorsal finlet) are referable to Bartschichthys sp. (Dutheil 1999a), based on similarities to Bartschichthys arnouldi from similar age rocks in Niger (Gayet and Meunier 1996). Dutheil (1999a) referred another isolated pinnula from the same horizon to Sudania sp., as it closely resembled specimens from the Cenomanian Wadi Milk Formation in Sudan (Werner and Gayet 1997).

Large jaw bones with teeth (Cavin et al. 2015) and scales (Meunier et al. 2016) from the Kem Kem Group were recently referred to the Egyptian cladistian Bawitius. An isolated premaxilla (Fig. 58) also may pertain to this large-bodied genus. The bone has a pitted surface texture. The teeth are large, hollow and ankylosed to the bone. These features closely resemble a better-preserved partial skull from beds of Cenomanian age in Niger (Fig. 59). The Niger specimen is associated with scales, allowing positive referral to Cladistia.

Ginglymodi Cope, 1872, Semionotiformes Arambourg & Bertini, 1958. Several authors describe a range of ginglymoid and semionotiform fishes from disarticulated material (Dutheil 1999a, Cavin and Brito 2001, Grande 2010, Cavin et al. 2015). The moderate-sized Oniichthys falipoui and Obaichthys africanus are known from partial skeletons, and Dentilepisosteus kemkemensis is represented by scales (Cavin et al. 2015). Other isolated scales that likely pertain to this group measure more than 50 mm in length and are indicative of large-bodied species (Fig. 60). Pycnodonts have also been identified among disarticulated remains from the Kem Kem Group.

Amiiformes Hay, 1929. Isolated teeth and several dentary fragments (Fig. 61A–D) pertain to amiids, which have been found in both formations of the Kem Kem Group. A partial amiiform skull, described as Calamopleurus africanus (Forey and Grande 1998), differs in skull proportions but otherwise is similar to the Brazilian species C. cylindricus (Cavin et al. 2015).

A curved dentary from Aferdou N’Chaft has at least 20 alveoli for small triangular teeth and lacks interdental plates (Fig. 62E–I). The texture and form of the dentary resembles that in amiids (Martinelli et al. 2013).

Teleostei Müller, 1846. More than a dozen genera of teleost fishes have been described from the Kem Kem Group (Table 8). Several are known from partial or complete skeletons from Oum Tkout. The elongate Diplospondichthys moreaui (Fig. 63) has an unusual combination of features that has left its position uncertain within Teleostei (Filleul and Dutheil 2004). The elongate freshwater acanthomorph Spinocaudichthys oumtkoutensis (Fig. 64) has also been recorded at the same locality (Filleul and Dutheil 2004).

Forey and Cavin (2007) described a well-preserved ichthyodectiform braincase from an unknown locality in eastern Morocco as Cladocyclus pankowskii, which later was placed in the genus Aidachar (Cavin et al. 2015). A well-preserved dentary (Fig. 62A-C) is referable to A. pankowskii and is also similar to the ichthyodectine Xiphactinus (Leidy 1870, Schwimmer et al. 1997, Fig. 62D). Osteoglossiform and notopterid remains including skull fragments were assigned to Palaeonopterus greenwoodi (Forey 1997, Taverne 2000, Cavin and Forey 2001a). The median lingual dental plate of this species is composed of several superimposed layers of adjacent teeth (Meunier et al. 2013), which were previously described in error as Pletodus sp. (Dutheil 1999a). Erfoudichthys rosae is a small-bodied teleost of unknown affinity, known from an isolated skull (Cavin et al. 2015). Previously it was thought to be a gonorynchiform (Pittet et al. 2010).

Actinistia Cope, 1871. Isolated cranial bones pertaining to large-bodied coelacanths are present in both formations of the Kem Kem Group (Figs 65, 66). Numerous isolated cranial bones and scales were collected at Boumerade and Gara Sbaa in the Gara Sbaa Formation. Although the size and ornamentation of dermal cranial bones are easily recognized, their generic and specific assignment remains uncertain.

The genera Mawsonia and Axelrodichthys were originally described from South America on the basis of complete specimens in nodules of late Early Cretaceous age (Maisey 1986). In contrast, the African material largely consists of isolated bones from Morocco and Algeria (Tabaste 1963, Wenz 1981, Cavin and Forey 2004). Cavin et al. (2015) suggested that, in addition to Mawsonia lavocati, a second large coelacanth may be present within the Kem Kem Group, assigning an isolated cranial bone to the South American genus Axelrodichthys. More recently, after a review of actinistians assigned to the genera Mawsonia (Woodward 1907) and Axelrodichthys (Maisey 1986) Mawsonia lavocati was included in Axelrodichthys as Axelrodichthys lavocati by Carnier Fragoso et al. (2019). Yabumoto and Uyeno (2005) described a partial skull with lower jaws from an uncertain locality within the Kem Kem Group. They also provide a revision of the species that serves as a guide for assignment of isolated material.

Measuring approximately 30 cm long, the skull confirms the large size of the genus Axelrodichthys in lake and river deposits on Africa and South America (Carvalho and Maisey 2008, Carnier Fragoso et al. 2019). Axelrodichthys lavocati may have grown to a body length in excess of 4 m.

Dipnoi Müller, 1844. Lungfish tooth plates are common in both formations of the Kem Kem Group (Fig. 67). They vary considerably in size and ornamentation with distinctive morphological differences.

Ceratodontidae Gill, 1872. The generic taxonomy of fossil lungfish, which is based almost exclusively on toothplates, has been unsettled and species have been variously assigned to the extinct genus Ceratodus or to the living Australian genus Protopterus and living African genus Neoceratodus. The most recent assessment (Cavin et al. 2015) attributes Kem Kem lungfish to three ceratodontid genera, Ceratodus, Neoceratodus, and Arganodus (Table 8). Toothplates with deeply incised ridges (Fig. 67A) are the most common and were initially identified as Ceratodus africanus (Haug 1905, Tabaste 1963, Martin 1984). More recently, these are referred to Neoceratodus africanus (Martin 1982, Werner 1995, Cavin et al. 2015), which has been reported across all of northern Africa in similar age beds (Haug 1905, Peyer 1925, Werner 1995, Tabaste 1963, Schlüter and Schwarzhans 1978, Schaal 1984, Cavin et al. 2010). In the Kem Kem Group, the toothplates of N. africanus can measure more than10 cm.

Tabaste (1963) described small toothplates with low ornamentation and only four low ridges as Ceratodus humei. These have also been found across northern Africa in similar age beds (Haug 1905, Werner 1993, Martin 1984). Martin (1984) placed this species in the extant genus Protopterus, but later it was returned to the genus Ceratodus (Churcher and De Iuliis 2001). Recently Cavin et al. (2015) referred small toothplates with a characteristic radiating pattern of ridges to Arganodus tiguidensis, a species described originally from Algeria (Tabaste 1963) and later reported in Niger (Broin et al. 1974) and Brazil (Candeiro et al. 2011).

Caudata Scopoli, 1777

Sirenidae Gray, 1825. Several localities in the finer-grained Douira Formation have yielded two braincases and 38 vertebrae pertaining to salamanders (Rage and Dutheil 2008). Most of the vertebrae come from the pond locality Oum Tkout, whereas the braincases were found at Taouz and Talidat localities. The occipital condyle of each partial braincase is transversely broad (Fig. 68A, B), which resembles the condition in the Sudanese genus Kababisha (Evans et al. 1996). The isolated vertebrae, which are tentatively referred to the same genus, resemble those of Kababisha and the South American genus Noterpeton (Rage and Dutheil 2008). The keeled centra, which are 2–3 mm in length, are flanked to each side by flange-shaped transverse processes. A foramen opens onto the dorsal surface of the flange. Unlike the Sudanese site, only a single sirenian appears to be present in the Douira Formation.

One trunk or anteriormost caudal vertebra of an indeterminate salamander is known from the Algerian locality Oued Bou Seroual in a level comparable to the Douira Formation and approximately 25 km distant from Oum Tkout. The morphology of this procoelous vertebra suggests that it pertains to an elongate, snake-like salamander (Alloul et al. 2018).

Anura Fischer von Waldheim, 1813. A partial braincase, jaw fragments, and procoelous vertebrae probably pertain to several species of non-pipid anurans, but the remains are too fragmentary to assign to particular families (Rage and Dutheil 2008).

Pipidae Gray, 1825. The majority of the anuran material collected in the Douira Formation, as in other Gondwanan localities, is referable to the Pipidae (Rage and Dutheil 2008). The holotype for a new genus and species, Oumtkoutia anae, was found at the pond locality Oumt Tkout (Fig. 68C, D). The subtriangular cranium that narrows anteriorly and the distinctive vertebral morphology distinguish this species from other pipids (Rage and Dutheil 2008). Other localities (Dar el Karib, Taouz) generated additional isolated bones. In all there are 22 partial braincases, five vertebrae, and seven pelvic fragments.

Testudines are common among vertebrate fossils in the Kem Kem Group (Gmira 1995). Isolated shell fragments are the most common (Fig. 69) followed by shell pieces (Fig. 70), vertebrae and limb bones (Fig. 71) and, rarely, partial plastron and carapace (Fig. 72) or skull material (Fig. 73). Named Kem Kem Group testudines, thus far, are based solely on isolated crania (Fig. 73). These were collected commercially from uncertain localities and horizons (Tong and Buffetaut 1996, Gaffney et al. 2002, 2006).

Cranial material pertains, thus far, exclusively to pleurodires; cryptodires have yet to be reported. Three genera of euraxymydid and podocnemidoid pleurodires were described on the basis of isolated crania (Gaffney et al. 2002, 2006); postcranial remains have yet to be definitively associated with any of the three genera Dirqadim, Hamadachelys, and Galianemys. Araripemydid pleurodires are known only from isolated carapace fragments with pitted texture (Fig. 70A-F); no cranial material has been discovered.

Pleurodira Cope, 1865

Araripemydidae Price, 1973. The flattened, fragile-shelled araripemydids are much better known from slightly older (Aptian-Albian) rocks to the south in Niger (Sereno and ElShafie 2013) and contemporary deposits in Brazil, which were located across a then narrower Atlantic Ocean (Gaffney et al. 2006). The araripemydid carapace is composed of thin, flat, densely pitted bones that resemble carapace fragments from the Kem Kem Group (Fig. 70A-F). We tentatively refer this Kem Kem material to the Araripemydidae on the basis of these features and await the recovery of more complete specimens (Table 8).

That decision, to limit referral of partial isolated shell elements of this form to Araripemydidae, is prudent and based on a cladistic diagnosis of the family that specifically cites shell structure and texture among some 20 synapomorphies that unite the two valid genera Araripemys and Laganemys (Sereno and ElShafie 2013: 219). Some authors, in contrast, have attempted to refer isolated shell pieces from South America and Africa to specific araripemydid genera or species, when the material does not exhibit more specific diagnostic features.

Thin ornamented shell material characterizing araripemydids is rare in the Kem Kem Group (Fig. 70A-F). One thin ornamented partial hypoplastron (Fig. 70A, B) was found among more than 400 shell fragments collected by Lavocat in the 1950s and referred by Gmira (1995) to the genus Araripemys as an indeterminate species (also Cavin et al. 2010). This partial plastron element, however, is insufficient for generic or specific assignment within Araripemydidae, as there are no features that allow reference to Araripemys, a genus based on material from earlier Aptian-Albian deposits in Brazil. Specimens referable to Araripemys are limited to those from the Araripe Basin of Brazil (Gaffney et al. 2006).

In a similar manner, Broin (1980) erected a new genus and species, Taquetochelys decorata, on the basis of a hypoplastron fragment from the Aptian-Albian Elrhaz formation of Niger (Sereno and ElShafie 2013: fig. 14 B, C). It is less complete than the hypoplastron from the Kem Kem araripemydid. Broin listed ten additional shell fragments as paratypes and cited (with uncertain status) an additional 31 shell pieces. This material was surface collected from a region known as Gadoufaoua during four expeditions in the 1960s and early 1970s. There is no specific type locality cited, and the material is surely derived from many individuals.

The diagnosis offered by Broin (1980: 42) for Taquetochelys decorata was inadequate at the time it was coined: “Pleurodire à carapace décorée de très petites granulalions et bourrelets, serrés. Petits mésoplastrons latéraux, courts”. She tried to differentiate the species on details of its beaded decorative pattern and the presence of a mesoplastron, which she inferred from the beveled margin of the hypoplastron. Neither are diagnostic at the generic or species level, as pointed out by Gaffney et al. (2006: 111). The general form of decoration is similar between Nigerien and Brazilian specimens, and the presence of a mesoplastron is a primitve feature, its loss diagnostic for Araripemys (Sereno and ElShafie 2013).

A nearly complete, articulated skeleton with a skull was later described from the Elrhaz Formation as Laganemys tenerensis (Sereno and ElShafie 2013), its diagnosis including more than 20 autapomorphies in the skull and postcranial skeleton. These features clearly distinguish this taxon from the contemporary Brazilian genus and species Araripemys barretoi, its diagnosis also revised. The hypoplastron of the holotypic specimen of L. tenerensis also exhibits several shape and ornamentation differences to the hypoplastron originally described as Taquetochelys decorata (Sereno and ElShafie 2013: fig. 14.14). Sereno and ElShafie (2013) came to the same conclusion as Gaffney et al. (2006), that the material upon which Taquetochelys decorata is based is insufficient and should be regarded as a nomen dubium.

More recently Pérez-García (2017, 2018) has attempted to resurrect Taquetochelys decorata and reduce Laganemys tenerensis as a junior synonym by arguing that the differences cited between their hypoplastra fall within an acceptable range of individual variation. Pérez-García stated that “the anterior margin of the hypoplastron” was broken away, rendering ineffective any shape differences based on this bone. But our examination and specimen figures of the element show it as complete anteriorly (Pérez-García 2018: fig. 1O; see also Broin 1980: pl. III, fig. 10), and the anterior margin was cited for evidence of the presence of a mesoplastron.

Fragmentary holotypic specimens without diagnostic features lend themselves to subsequent taxonomic ambiguity. At no point has any author offered a revised diagnosis of T. decorata based solely on the holotypic partial hypoplaston or even on the numerous additional shell pieces referred to this taxon by Broin (1980). The material associated with Taquetochelys decorata consists entirely of isolated shell fragments that lack diagnostic features and specific locality data. Although this material is generally consistent with the nearly complete holotype skeleton of Laganemys tenerensis, there is no solid basis for regarding them as the same taxon. Gaffney et al. (2006: 111) remarked that it is impossible to diagnose T. decorata on the basis of the holotypic plastron piece. We agree and regard Taquetochelys and T. decorata as nomina dubia, the material referable to the Araripemydide on the basis of its ornamentation and several other familial synapomorphies (Sereno and ElShafie 2013: 219).

We anticipate eventual recovery of more complete araripemydid remains from the Kem Kem Group. More complete remains may resolve its taxonomic distinction and its affinities with the slightly older African and South American genera, Laganemys and Araripemys, respectively.

Euraxemydidae Gaffney et al., 2006

Dirqadim schaefferi Gaffney et al., 2006. Dirqadim schaefferi from the Kem Kem Group is closely related to the slightly older and more completely preserved Euraxemys esswini from the Santana Formation in Brazil (Gaffney et al. 2006). D. schaefferi is known from two crania, one of which is nearly complete (Fig. 73H-J), both of which were commercially collected from unknown locations. Currently no postcranial material from the Kem Kem Group can be referred to the family, genus or species. A few features of the shell are diagnostic for Euraxemys esswini (Gaffney et al. 2006: 40) and may eventually allow reference of more complete shell material from the Kem Kem Group to D. schaefferi.

Podocnemidoidea Cope, 1868

Hamadachelys Tong & Buffetaut, 1996. Hamadachelys escuilliei (Tong and Buffetaut 1996) is known from a well-preserved cranium and mandible. A second mandible was referred to this species (Gaffney et al. 2006: fig. 251). All of this material was collected commercially from unknown localities in the Kem Kem Group. The diagnostic features of H. escuilliei are in need of revision in light of the significant cranial material discovered and described since this taxon was named. The narrow interorbital distance in dorsal view of the cranium (Tong and Buffetaut 1996) more closely resembles Euryaxemys from Brazil than the contemporaneous genera Dirqadim and Galianemys. Postcranial material cannot be reliably referred to Hamadachelys escuilliei. Hamadachelys has been positioned phylogenetically between euryaxemydids and Galianemys at the base of the podocnemidoid radiation (Gaffney et al. 2006, 2011).

Galianemys Gaffney et al., 2002. The podocnemidoid Galianemys is the best known turtle in the Kem Kem Group and is represented by several nearly complete and partial crania (Gaffney et al. 2002). Two species were named, G. whitei and G. emringeri, with several specimens referred to one or the other species. We figure two additional crania here (Fig. 73A-G), the more complete of which was commercially collected and is referable to Galianemys whitei (Fig. 73A-D). The prefrontal-frontal suture is straight rather than posteriorly convex (Fig. 73A), the jugal is separated from the posterior orbital margin by significant maxilla-postorbital contact (Fig. 73A), and the jugal contacts the palatine on the posterolateral aspect of the palate (Fig. 73B). These are diagnostic characters that differentiate this species (Gaffney et al. 2006). Unlike Dirqadim, another pleurodire from the Kem Kem Group (Gaffney et al. 2006), the triturating surface on the maxilla expands posteriorly assuming a broad tringular shape in ventral view (Fig. 73B), only minor ventral embayment is present along the jugal-quadratogugal margin in lateral view (Fig. 73D), and the U-shaped temporal emargination along the posterior margin of the skull roof is quite deep in dorsal view (Fig. 73A).

The second, less complete cranium (Fig. 73E-G), discovered in 2008 by a local in the Douira Formation at Aferdou N’Chaft east of Taouz (Fig. 9, locality 14), is referable to the genus Galianemys by the depth of the U-shaped posterior temporal emargination. Missing portions of the cranium, however, prevent assignment to a particular species.

Two very similar shell types, tentatively referred to Galianemys by Gaffney et al. (2006: figs 271–274), are based on nearly complete specimens of large size (carapace length approximately 55 cm) that were collected commercially from the Kem Kem Group. More recently Cavin et al. (2010) reported that these two shells were found at Tizi Tazguart, a locality south of Jbel Zireg (Fig. 9, locality 9) and near another site that yielded some 30 turtle shells that remain at large. Cavin et al. (2010) regarded the two large shell types as confidently referable to Galianemys, and Karl (2010: fig. 1) tentatively referred them to the two species, G. whitei and G. emringeri. No evidence has been forwarded, however, to justify these referrals, as there currently exists no clear association of cranial and postcranial remains for any testudines in the Kem Kem Group.

A partial carapace, complete plastron and associated pectoral and pelvic girdles of a fairly large testudine (Fig. 72) was collected in 1995 from a site near our section at Oum Tkout in the Douira Formation. The Douira Formation is approximately 100 m in thickness in this region, and the associated shell and girdle material was found approximately 28 m above the first substantial mudstone at the base of the formation. It matches one of the shell types described by Gaffney et al. (2006: fig. 274) and Karl (2010: fig. 1.2). On the plastron, the intergular scute is broad and equal in transverse width to adjacent gular scutes, a plastral scute pattern attributed without justification to G. whitei (Karl 2010: fig. 1.2). The carapace measures 31 cm in length (Fig. 72), or approximately 56% the size of the specimens described by Gaffney et al. (2006). These shell types may pertain to Galianemys and its two closely related species, although that needs to be established by specimens associating cranial and shell material. At present the single specimen we figure here (Fig. 72) is the only one that provides any evidence of association for Kem Kem Group testudines, in this case between shell and non-shell postcranial bones.

In recent years, fossil discoveries have brought to light jaw fragments and, more rarely, nearly complete skeletons of extinct genera positioned as stem taxa to extant squamate clades. For Iguania and Ophidia, in particular, new fossils from circum-Tethyan sites have drawn their stem lineages back to the Early Cretaceous and, in some cases, to the Early Jurassic (Evans et al. 2002, Hsiang et al. 2015, Caldwell et al. 2015, Martill et al. 2015, Simões et al. 2017).

From Morocco more specifically, fragmentary squamate material was first reported in abundance from the Early Cretaceous site Anoual (Broschinski and Sigogneau-Russell 1996). The Kem Kem Group, thus far, has yet to yield abundant isolated vertebrae or more complete remains of squamates, respectively, from sediment screening or the pond locality Oum Tkout. Squamate remains consist of rare, isolated jaw fragments and vertebrae pertaining to stem acrodont iguanians, borioteiioids, and ophidians.

Iguania Cope, 1864

Acrodonta Cope, 1864

Jeddaherdan. A jaw fragment with blunt, unornamented, imbricate crowns that are ankylosed to the dentary was described as Jeddaherdan aleadonta, a new acrodont iguanian with potential affinities with uromasticine agamids (Apesteguía et al. 2016a). It was collected approximately 70 years ago by French paleontologist René Lavocat from a locality (Gara Tabroumit) southwest of Gara Sbaa. The horizon remains unknown. The only other fossil discovered so far pertaining to a squamate is a single trunk vertebra from Taouz of indeterminate relationship (Rage and Dutheil 2008).

Ophidia Brongniart, 1800

Norisophis. More than 100 ophidian vertebrae were recovered from field work in the Kem Kem Group between 1995 and 2018. They were found at sites in both formations, with several recovered at the pond locality Oum Tkout in the Douira Formation (Rage and Dutheil 2008). More recently, isolated vertebrae were also found in the Gara Sbaa Formation at Aferdou N’Chaft and nearby localities pertaining to Norisophis begaa, a new basal snake (Klein et al. 2017). Centrum length is a little more than 5 mm.

Lapparentophiidae Hoffstetter, 1959

Lapparentophis . Vullo (2018) described two moderately elongate mid-trunk vertebrae from near Begaa (close to Taouz), naming a new species in the genus Lapparentophis, L. ragei, which was known previously from Algeria. The diameter of the neural canal is much smaller than that of the cotyle, which is slightly broader than high. Centrum length is a little more than 1 cm.

Simoliophiidae Nopcsa, 1925

Simoliophis. Rage and Dutheil (2008) referred isolated vertebrae to Simoliophis cf. lybicus (Fig. 68E, F). They were found in both formations at several localities including Taouz, Oum Tkout, Douira and Dar El Karib. Two simoliophiid sacral vertebrae were also identified from Taouz and Oum Tkout localities that show similarities to the sacral vertebrae of snakes with hind limbs (Caldwell and Lee 1997, Rage and Escuillé 2000, Apesteguía and Zaher 2006).

Nigerophiidae Rage, 1975. Rage and Dutheil (2008) referred 18 dorsal vertebrae to this extinct aquatic snake family, which has been recorded from circum-Tethyan sites of Late Cretaceous and Paleocene age (Rage and Werner 1999). The dorsal centra are elongate and transversely narrow with tall, anteroposteriorly short neural spines.

Madtsoiidae Hoffstetter, 1961. Rage and Dutheil (2008) referred some 20 ophidian vertebrae from Taouz and Oum Tkout to this diverse extinct family, the neural arches characterized by a pair of parazygantral foramina on the posterior aspect of the neural spine and the absence of prezygapophyseal processes (Wilson et al. 2010). Rage and Dutheil (2008) regarded these Kem Kem vertebrae as distinct from the similar-sized matsoiid snake recorded from roughly coeval Cenomanian age rocks in Sudan (Werner and Rage 1994). These snakes grew to modest adult size with dorsal centra approximately 5 mm in length.

The Kem Kem Group has yielded a diverse array of crocodyliforms in size and trophic adaptations, ranging from small insectivorous or herbivorous candidodontid and sphagesaurid notosuchians less than 1 m in body length to large carnivorous neosuchians approaching the 12-meter length of Sarcosuchus (Broin and Taquet 1966, Sereno et al. 2001). Long-snouted crocodyliform fossils discovered in the late 1940s were initially assigned to the genus Thoracosaurus (Lavocat 1955), a eusuchian genus known only from North America and Europe (de Lapparent de Broin 2002). Later it was given a new genus Elosuchus, as E. cherifiensis (de Lapparent de Broin 2002, Young et al. 2017, Meunier and Larsson 2017). A braincase of this genus and species was erroneously referred to Libycosuchus (Buffetaut 1976), a very different short-snouted crocodyliform from the contemporaneous Bahariya Formation of Egypt (Stromer 1914).

By the 1990s, considerable new crocodyliforms fossils came to light in commercial collections (e.g., Russell 1996) and in the course of field work (Sereno and Larsson 2009). Based on fragmentary material, Buffetaut (1994) named Hamadasuchus rebouli, a peirosaurid crocodyliform. Larsson and Sues (2007) later described and attributed a complete cranium to this genus and species. Sereno and Larsson (2009) named a notosuchian, Araripesuchus rattoides, and a stomatosuchid, Laganosuchus maghrebiensis, on the basis of partial dentaries of distinctive form. A single elongate distal caudal vertebra was initially described as pertaining to a new neotheropod dinosaur, Kemkemia auditorei (Cau and Maganuco 2009), which was soon regarded as a nomen dubium after its close resemblance to the distal caudals of extant crocodilians became apparent (Lio et al. 2012). Derived notosuchian teeth came to light in screen-washed sediment (Larsson and Sidor 1999), and more recently a small notosuchid with multicuspid crowns was named Lavocatchampsa sigogneaurussellae (Martin and de Lapparent de Broin 2016). We review Kem Kem Group crocodyliforms below, a group that will likely increase in diversity with the continued recovery of new specimens.

Notosuchia Gasparini, 1971

Uruguaysuchidae Gasparini, 1971

This family of notosuchians, united as a clade in some analyses (e.g., Leardi et al. 2015), includes the genera Anatosuchus, Uruguaysuchus and Araripesuchus.

Araripesuchus. The speciose genus Araripesuchus, known initially from South America and later from Africa and Madagascar, also occurs in the Kem Kem Group as Araripesuchus rattoides (Fig. 74). It was named on the basis of a partial, edentulous dentary with 14 alveoli that was collected commercially and thus comes from an uncertain locality and horizon (NMC 41893, as CMN 41893 in Sereno and Larsson 2009). Referred specimens include partial dentaries from the Douira Formation collected at Dar El Karib (near Erfoud) and another collected commercially from an uncertain locality and horizon (BSPG 2008 I 41, Fig. 74E, F).

BSPG 2008 I 41 preserves dentary teeth 2–6 and 10 and the roots of dentary teeth 11 and 12. The socket for the first dentary tooth projects anteriorly and presumably held a procumbent and slightly enlarged tooth, a diagnostic feature for the species. As in Araripesuchus wegeneri, the fourth dentary tooth is enlarged, whereas dentary teeth 3, 5, and 6 are smaller and subconical. The alveoli for dentary teeth 9 and 10 are incompletely separated, and dentary tooth 10 has a labiolingually compressed crown with its carina and apex rounded by tooth abrasion as occurs in A. wegeneri (Sereno and Larsson 2009).

A. rattoides does appear to be distinct from A. wegeneri, which is much better known from complete skulls and skeletons from the older (Aptian-Albian) Elrhaz Formation in Niger. Compared to the latter, many of the dentary alveoli are exposed in lateral view of the best-preserved dentary of A. rattoides (Fig. 74A) with the dentary symphysis held in a vertical plane. In dorsal view with the symphysis in a sagittal plane, the skull appears to be proportionately narrower than a comparable view of the dentary in A. wegeneri (Sereno and Larsson 2009). Specimen BSPG 2008 I 41 (Fig. 74E, F) is identical in form but slightly larger than the holotype of A. rattoides. Differences in the exposure of the alveoli as figured here are due to the orientation of the specimen when photographed (the dentary symphysis tipped from a vertical plane).

Candidodontidae Carvalho et al., 2004

This derived family of small notosuchians, first described by its namesake genus Candidodon from Brazil (Carvalho and Campos 1988), now is well known from Early Cretaceous genera from east Africa, such as Malawisuchus and Pakasuchus (O’Connor et al. 2010). Candidodontidae forms a clade in some phylogenetic analyses of notosuchians (O’Connor et al. 2010). In other analyses Candidodon and allies are positioned as basal outgroups to Notosuchus and other notosuchians including sphagesaurids (e.g., Pol et al. 2014, Leardi et al. 2015).

Lavocatchampsa. Recently a commercially collected partial skull from the Kem Kem Group was described as Lavocatchampsa sigogneaurussellae (Martin and de Lapparent de Broin 2016). Its complex, multicuspid crown morphology includes a labial and lingual cingulum reminiscent of that in the molariform teeth of Cretaceous mammaliaforms. Other derived features include the absence of a caniniform tooth. Unlike Notosuchus and close relatives, masticatory movement appears to have been orthal rather than propalinal. The relationships of Lavocatchampsa within Candidodontidae are uncertain. It has been resolved as the sister group to the somewhat older genera Malawisuchus and Pakasuchus (Martin and de Lapparent de Broin 2016). Lavocatchampsa also resembles Adamantinasuchus (Nobre and Carvalho 2006) from the slightly younger Adamantina Formation of Brazil.

Notosuchia indeterminate. Small multicuspid crocodyliform teeth from the Douira Formation were first reported by Larsson and Sidor (1999). They were discovered in a clay-rich horizon immediately beneath the fossiliferous pond deposit at Oum Tkout. Two tooth forms were described, both of which differ from the crown morphology in Lavocatchampsa (Martin and de Lapparent de Broin 2016). In the first tooth form, the crowns are subtriangular in labiolingual views and have a major row of cusps flanked by lower parallel rows to each side (UCRC VP155, VP156), as figured in Larsson and Sidor (1999: fig. 2, as SGM-Rep 4, -Rep 5). The crown shape, apical row of cusps, and accessory cusp rows resemble the molarifom crowns in Candidodon (Pol et al. 2014), Adamantinasuchus (Nobre and Carvalho 2006) and the Bolivian notosuchian Yacarerani (Novas et al. 2009). The first tooth form, however, is distinctive in the variable size and location of the cusps and its pattern of tooth-to-tooth abrasion. A large, low-angle, planar wear facet truncates one of the crowns (Larsson and Sidor 1999: fig. 2D), which differs from the wear pattern present in Lavocatchampsa (Martin and de Lapparent de Broin 2016).

In the second tooth form (UCRC VP157) described by Larsson and Sidor (1999: fig. 3, as SGM-Rep 6), multiple cusp rows also occur, but the crown has a more rounded profile than in many notosuchians such as Malawisuchus (Gomani 1997). The crown is ovate in occlusal view with a main cusp and several accessory cusps oriented along the major axis. The main cusp is located at the extreme mesial or distal end of the crown, rather than a central location, assuming the major axis of the crown is mesiodistal. Grossly similar to the second tooth form are ovate crowns with few cusps and an unusual texture of vertical striations pertaining to an unnamed notosuchian from Upper Cretaceous rocks in Brazil (Montefeltro et al. 2009). The second tooth form from the Kem Kem Group, however, differs in a number of regards as noted by Montefeltro et al. (2009) and cannot be assigned to any currently known notosuchian. These tooth forms suggest that there exists a greater diversity of small notosuchians in the Kem Kem Group than are currently recognized.

Peirosauridae Gasparini, 1982

Peirosaurids are a diverse and loosely united group of crocodyliforms lying outside Neosuchia. Some authors have united peirosaurids and sebecids as Sebecia outside Notosuchia (Larsson and Sues 2007, Sereno and Larsson 2009, Meunier and Larsson 2017), whereas others have positioned peirosaurids and sebecids as independent clades within Notosuchia (Leardi et al. 2015, Fiorelli et al. 2016). Adding to this lack of resolution are studies that name new taxa on the basis of fragmentary cranial remains from the Kem Kem Group, which include the holotypic specimen of Hamadasuchus rebouli (Buffetaut 1974) and a partial braincase referred initially to the Egyptian genus Lybicosuchus (Buffetaut 1976). Further confusion has ensued with the use of the poorly defined taxon Trematochampsidae, to which peirosaurids have sometimes been assigned. Neither its namesake genus and species Trematochampsa taqueti, which was based on isolated fragments from Niger (Buffetaut 1976), nor the Family Trematochampsidae Buffetaut 1974 appear to be valid (Meunier and Larsson 2017).

In 2007 a suite of commercially collected material was described and referred to H. rebouli, including a nearly perfect adult cranium (ROM 52620, Fig. 75), braincases (ROM 54511, 52059), a dentary (ROM 49282) as well as more fragmentary subadult specimens (Larsson and Sues 2007). The more gracile dentary differs in several regards from MDEC001. In dorsal view, the first three alveoli are more mesially positioned relative to the fourth alveolus given the sharper median convergence of the dentary rami. The fourth and thirteenth alveoli are largest, with a more dramatic increase in diameter from the eleventh alveolus. The splenial bounds the eleventh and more distal alveoli. Hypertrophied alveolar bone is not present either medially or laterally, and no groove is present. These differences render it difficult to refer the dentary or other material to H. rebouli. This assignment also was questioned by Cavin et al. (2010: 398), although no justification was provided. They also questioned referral to H. rebouli by Larsson and Sues (2007) of the braincase (MNHN-MRS 3101, Fig. 76C) initially identified as Libycosuchus by Buffetaut (1974). Yet, this specimen is clearly closer in morphology to the supratemporal fossa of ROM 52620 (Fig. 75; Larsson and Sues 2007) than to the holotypic specimen of the short-snouted Egyptian crocodyliform Libycosuchus brevirostris (BSP 1912 VIII 574-578, Figs 77A, F, 78).

More recently, a nearly complete skull and lower jaws were commercially collected from the Kem Kem Group and have yet to be described in detail (BSPG 2005 I 83, Rauhut and López-Arbarello 2006). At just more than 30 cm in length (Fig. 79), it is almost the exact same size as the previously described perfect cranium (Fig. 75) and, likewise, was referred to H. rebouli. We see little reason to doubt the assignment to H. rebouli of the new skull (BSPG 2005 I 83, Fig. 79) as well as a suite of more fragmentary material from the Kem Kem Group (Figs 80–82). One braincase was collected from the Gara Sbaa Formation at Aferdou N’Chaft and represents the only partial skull of H. rebouli from a known horizon and locality (FSAC-KK 930, Fig. 81A, B). Another specimen that was commercially collected from an unknown horizon and locality preserves the edentulous, fused dentary-splenial symphysis from an adult skull (CMN 41784, Fig. 82D-F). The description of the new complete skull BSPG 2005 I 83 should carefully refine the diagnostic characters listed for this genus and species based on the first complete cranium (Larsson and Sues 2007: 534). With that knowledge, many isolated cranial specimens may be referred with greater justification.

Neosuchia Clark, 1988

Stomatosuchidae Stromer, 1925

This derived family of moderate to large-sized crocodyliforms was first described as Stomatosuchus inermis from the likely Cenomanian-age Bahariya Formation of Egypt (Stromer, 1925). The nearly 2 m long flat cranium of the holotype was found in articulation with a very slender U-shaped mandible (Sereno and Larsson 2009). This iconic crocodyliform became all the more enigmatic when the holotype and only known specimen was destroyed in WWII. The lack of material has rendered uncertain the position of Stomatosuchidae within Crocodyliformes. As discussed below, the possible association of vertebrae with procoelous centra with this skull type suggests that Stomatosuchidae may fall within Neosuchia.

Laganosuchus. Recently remains of a similar crocodyliform surfaced in the Echkar Formation of Niger (Sereno and Larsson 2009). This material was described as Laganosuchus thaumastos (Fig. 83A, B). A similar species, named L. maghrebensis, was described from dentary fragments from the contemporaneous Kem Kem Group (Fig. 83C-K; Sereno and Larsson 2009).

The holotype (UCRC PV2, Fig. 83C-E) preserves the first four alveoli, the first opened to fully expose an erupting crown. The tapering crown lacks recurvature and exhibits light fluting on its lingual side. A referred specimen (NMC 50838, Sereno and Larsson 2009, Fig. 83I-K) preserves a similar portion of the anterior dentary ramus and also includes an erupting tooth in what appears to be the third alveolus. The tooth is approximately 17 mm in height and curves slightly distally at its tip. A second referred specimen with a fully erupted crown may also pertain to L. maghrebensis (NMC 41786, Fig. 84).

A larger dentary piece of L. maghrebensis was figured by Rauhut (2009), comprising the anterior portion of the left dentary and fused symphyseal end of the right dentary (Fig. 83F-H). The first alveolus contains a fully erupted caniniform crown followed by 10 empty alveoli (BSPG 2008 I 62, Fig. 83F-H). The dentary ramus thickens toward the symphysis, unlike the other smaller specimens of the species (Fig. 83D, G, J). A thickened symphyseal shelf was described as diagnostic of L. thaumastos (Sereno and Larssson 2009). Given the larger size of BSPG 2008 I 62, symphyseal thickening may occur with maturity. As in L. thaumastos (Sereno and Larsson 2009), the first, second and fourth alveoli are larger than the others (Fig. 83G). Unlike L. thaumastos, the more distal alveoli are closely spaced (Fig. 83A, G). In L. thaumastos, in contrast, the alveoli distal to the fourth are all well separated, except for the sixth and seventh (Fig. 83A). In addition, the alveoli in L. thaumastos are separated by an undulating margin that exposes the alveolar rim in lateral view (Fig. 83A). The more complete dentary of L. maghrebensis (BSPG 2008 I 62) confirms the original assessment that Moroccan and Nigerien specimens represent distinct species.

Aegisuchus. Aegisuchus witmeri was named on a commercially collected braincase of uncertain locality in the Kem Kem Group (ROM 54530, Holliday and Gardner 2012). This specimen closely resembles the braincase of Aegyptosuchus peyeri (Stromer 1933) from the Bahariya Formation of Egypt (Fig. 85). Although the braincase of Aegyptosuchus peyeri survived WWII, associated procoelous vertebrae suggestive of neosuchian relationships were destroyed (Stromer 1933).

Striking features of the braincase link Aegisuchus and Aegyptosuchus, which Holliday and Gardner (2012) placed in Aegyptosuchidae on the basis of several cranial characters. In dorsal view, the orbits are very closely spaced, and supratemporal fossae are small, widely separated, and displaced anteriorly, and the occipital surface is broadly exposed in dorsal view (Fig. 85). The skull roof between the dorsally facing orbits is particularly narrow, the medial margin of each orbit closer to the midline than the medial rim of the supratemporal fossa. The quadrate shafts are very broad and posteriorly angled, indicating that the skull is quite dorsoventrally compressed. As reconstructed by Holliday and Gardner (2012: fig. 1), the cranium quite possibly was very low and elongate as in Stomatosuchus inermis.

All of these features, nevertheless, appear to be present in Stomatosuchus inermis, the braincase of which was only partially preserved (Stromer 1925, Sereno and Larsson 2009). The Egyptian Aegyptosuchus peyeri, thus, might constitute a junior synonym of Stomatosuchus inermis from the same formation. We are unable to differentiate the two on the basis of available images of Stomatosuchus inermis. The family Aegyptosuchidae, in turn, might well be redundant with Stomatosuchidae. Among the Moroccan specimens, there is no overlap between the braincase of Aegisuchus witmeri and the dentary sections of Laganosuchus maghrebiensis. Given their similar size, it is entirely possible that more complete specimens from the Kem Kem Group will show that Aegisuchus witmeri is a junior synonym of Laganosuchus maghrebiensis.

Pholidosauridae Zittel & Eastman, 1902

The first specimens of a long-snouted crocodyliform discovered in the Kem Kem Group were identified as “Thoracosaurus” cherifiensis (Lavocat 1955). Thoracosaurus, a eusuchian genus from North America and Europe, was abandoned in favor of a new genus Elosuchus (de Lapparent de Broin 2002), which came to be known from nearly complete crania and mandibles from Morocco, Algeria, and Niger. Given that there are several long-snouted clades of crocodyliforms, the taxonomic status and affinity of this material, including the specimens from the Kem Kem Group, has been uncertain.

Recently, the material of Elosuchus from Morocco and Algeria was redescribed as pertaining to two species of pholidosaurids. The Moroccan material, based on specimens from the Kem Kem Group, was attributed to E. cherifiensis, whereas the fossils from the potentially slightly older Albian beds in Algeria (at Gara Samani, Fig. 1) were attributed to a new species E. broinae (Meunier and Larsson 2017). The two species differ only in minor features. The Nigerien material originally described as E. felixi was transferred to a new genus Fortignathus and identified as a dyrosaurid (Young et al. 2017).

Elosuchus . Lavocat (1955) based “Thoracosaurus” cherifiensis on isolated crocodyliform bones from Gara Sbaa that were never illustrated or numbered. As most of this material was subsequently lost, a nearly complete skull was designated as the lectotype for E. cherifiensis (MNHN E 1, Meunier and Larsson 2017). As the skull was commercially collected, its exact horizon and locality within the Kem Kem Group are unknown. Features listed in a revised generic diagnosis include five premaxillary teeth, the first of which is displaced posterior to the second, paddle-shaped lacrimal, strong anterior process of the squamosal that overlaps much of the postorbital laterally, quadratojugal overlapping all of the quadrate near its laterodistal condyle, and a small pit at the anteromedial end of the dentary near the symphysis that receives the tip of the first premaxillary tooth (Meunier and Larsson 2017).

Using the revised diagnosis and lectotype cranium, additional specimens can be referred to E. cherifiensis, which include paired premaxillae and partial rostra (Figs 86, 87), the posterior half of crania (Fig. 88A-E), fragmentary cranial bones (Fig. 89), mandibular rami (Figs 90, 91), and, with less confidence, isolated teeth (Fig. 92). The bones of E. cherifiensis were surface collected from many sites in the Kem Kem Group. A partial premaxilla (FSAC-KK 923, Fig. 89A-E) was surface collected in the Gara Sbaa Formation, and an isolated jugal (FSAC-KK 09, Fig. 88F-H) and mandibular symphysis (FSAC-KK 753, Fig. 91) were collected from the Douira Formation.

De Lapparent de Broin (2002) referred large isolated scutes to Elosuchus, but the association of the skull remains and scutes described is unclear. Large osteoderms have previously been attributed to “Sarcosuchus sp”. (Sereno et al. 1996), although evidence for this specific genus is lacking in the Kem Kem Group. Two morphologically identical isolated dentary fragments show the substantial range in size among material attributed to E. cherifiensis (UCRC PV159, CMN 41785, Fig. 89J, K). Meunier and Larsson (2017) noted some Elosuchus material from the Kem Kem differs from MNHN E 1, notably in the parietal bar separating the supratemporal fenestrae. MNHN E 1 and other material they refer to E. cherifiensis has a waisted bar, whereas other specimens have a broad, sculpted bar, similar to that of E. broinae (CMN 41912, ROM 52586, 64698, 64699). There may be more than a single species of Elosuchus within the Kem Kem Group (Meunier and Larsson 2017).

Neosuchia indeterminate. A large left jugal (FSAC-KK 07, Fig. 93) was found in situ in the Douira Formation at the locality Aferdou N’Chaft (Fig. 9, locality 14). With a preserved length of more than15 cm, the individual is approximately 150% the size of the adult cranium described by Larsson and Sues (2007: fig. 2; see also Fig. 75A). The postorbital bar, which is smooth and inset in neosuchians, appears to be less inset from the sculpted surface of the jugal in FSAC-KK 07 than in H. rebouli (Larsson and Sues 2007). The posterior ramus, in addition, tapers distally and is deflected medially (Fig. 93A, C). In H. rebouli, in contrast, the posterior process of the jugal expands slightly in dorsoventral height and is deflected slightly laterally.

Four foramina are visible on medial aspect of the body of the jugal (Fig. 93D). In extant crocodilians, one to three foramina open into the body of the jugal for the jugal nerve. The condition in H. rebouli is not known. Two other foramina pierce the large jugal near the ventral margin (Fig. 93C).

Although not recognised as pterosaurian at the time, the first pterosaur remains to be recovered from the Kem Kem Group consisted of isolated teeth collected by Lavocat in the late 1940s and early 1950s and now in the MNHN collections. Isolated, recurved, striated pterosaur teeth, likely ornithocheirid in most cases, are fairly common in these deposits. Kellner and Mader (1997: fig. 4) were the first to figure one of these teeth, four distinct morphotypes were described by Wellnhofer and Buffetaut (1999: figs 6–10) and additional examples are figured here (Fig. 94).

The first remains to be confidently identified as pterosaurian, an elongate mid-cervical vertebra referred to the Azhdarchidae, was described in a short abstract by Kellner and Mader (1996) and later figured by Rodrigues et al. (2011: fig 4). The following year, in another abstract, Mader and Kellner (1997) briefly described a jaw fragment with teeth (LINHM 016, Fig. 95A, B). Formally described by these authors in 1999, this specimen formed the holotype of Siroccopteryx moroccensis, a coloborhynchine, and the first pterosaur to be named from the Kem Kem Group. Recently, Jacobs et al. (2019) named a second coloborhynchine from the Kem Kem Group, Coloborhynchus fluviferox, and reported jaw fragments attributable to Anhanguera sp. and Ornithocheirus sp. (Jacobs et al. 2020).

Wellnhofer and Buffetaut (1999) published the first direct evidence for edentulous pterosaurs in the Kem Kem group. This included a well preserved but fragmentary rostrum (BSP 1993 IX 338, Fig. 96A-E) identified as pteranodontid, a fragment of a mandibular symphysis bearing a deep ventral crest (BSP 1997 I 67, Fig. 97), identified as tapejarid and an elongate, lance-shaped fragment of a mandibular symphysis (BSP 1996 I 36; Wellnhofer and Buffetaut 1999: fig. 4) identified as azhdarchid. The latter, and BSP 1993 IX 338, were subsequently assigned to Alanqa saharica (Fig. 96; Ibrahim et al. 2010). Recently, however, BSP 1993 IX 338 has been reassigned to a new chaoyangopterid from the Kem Kem Group (McPhee et al. 2020) The identification of BSP 1997 I 67 as tapejarid is supported by the recent discovery of further tapejarid jaw material in the Kem Kem Group (Martill et al. 2020). Kellner et al. (2007) described the fourth example of an edentulous jaw from the Kem Kem Group (MN 7054-V) and tentatively identified is as pteranodontid.

Alanqa saharica, an azhdarchid founded on a well-preserved fragment of a mandibular symphysis (FSAC KK 26; Fig. 98) collected in 2008, and reinterpreted here as part of the rostrum was the first edentulous Kem Kem Group pterosaur to be named (Ibrahim et al. 2010). Additional remains, including fragments of the rostrum and mandibular symphysis (Kellner et al. 2007, Ibrahim et al. 2010, Martill and Ibrahim 2015) and, more tentatively, cervical vertebrae (Ibrahim et al. 2010, Rodrigues et al. 2011) and a partial humerus (Rodrigues et al. 2011) have been assigned to this taxon (Averianov 2014). A second azhdarchid, Xericeps curvirostris, founded on a fragmentary, elongate, curved mandibular symphysis was described by Martill et al. in 2018 (Fig. 99). Following the erection of this taxon, the second azhdarchid from the Kem Kem Group, it can no longer be automatically assumed that azhdarchid postcranial remains from these deposits pertain only to Alanqa.

All of the Kem Kem Group pterosaur material consists of isolated, often fragmentary specimens (Figs 94–102). Some originate from commercial sources and lack precise locality data. In numerical terms, teeth (Fig. 94) far outnumber skeletal remains, with many hundreds, and possibly more than one thousand already recovered. Among skeletal remains, fragments of the rostrum and mandibular symphysis of edentulous pterosaurs (Figs 96–101), seemingly all azhdarchoids, outnumber all other skeletal elements described so far. The relatively common occurrence of azdarchoid jaw remains is in sharp contrast to the comparatively rare incidence of ornithocheirid jaw fragments which, to date, number only six examples (Mader and Kellner 1999, Jacobs et al. 2019, 2020). This disparity is perhaps due to the unusually robust construction of the jaws of azhdarchoids in which the tips of rostra and mandibular symphyses are composed of relatively thick cortical bone with only a small central lumen (Fig. 101D).

Postcranial remains including cervical vertebrae (Fig. 102; Ibrahim et al. 2010, Rodrigues et al. 2011) and limb bones (Rodrigues et al. 2011) are relatively rare and fragmentary, but often well preserved, undistorted and exhibit fine anatomical detail. Recently collected material, seemingly all referable to azhdarchoids and including additional cervicals, forelimb elements (humerus, ulna, wing-metacarpal, wing-phalanges), and hind limb elements (femur, tibia), has yet to be described.

Ornithocheiroidea Seeley, 1891

Ornithocheiridae Seeley, 1870

Siroccopteryx. Siroccopteryx moroccensis Mader & Kellner, 1999, based on the anterior portion of a rostrum that retains teeth (LINHM 016, Fig. 95A, B), was found near Begaa, a village close to the north end of the Kem Kem Hamada, according to the private collector involved in the sale of the material (Mader and Kellner 1999). The holotype, bearing the anteriomost six pairs of teeth, consists of coossified premaxillae and possibly a small portion of the anterior ends of the maxillae. The original description (Mader and Kellner 1999) has been supplemented by additional observations in more recent studies (Unwin 2001, Veldmeijer 2003, Rodrigues and Kellner 2008). The anterior profile measures approximately 30 mm in width at the base and has a height, to the apex, of 43 mm. A rostrum of these general dimensions corresponds in size to large ornithocheirids with wingspans of 3–4 m (Martill and Unwin 2012).

Several authors have suggested that Siroccopteryx may be synonymous with Coloborhynchus (Unwin 2001, Veldmeijer 2003, Ibrahim et al. 2010). Conversely, Rodrigues and Kellner (2008) and Jacobs et al. (2019) have argued that it represents a genus distinct from other ornithocheirids. LINHM 016 exhibits several distinctive features including an unexpanded, near parallel profile, in palatal view (Fig. 95B), a relatively narrow, deep rostrum, a rounded anterior profile of the rostrum in lateral aspect (Fig. 95A) and a posteriorly displaced rostral crest. In combination these characters appear to distinguish Sirocccopteryx from coloborhynchines, typified for example by Coloborhynchus clavirostris (Owen, 1874) and Uktenadactylus (Coloborhynchus) wadleighi (Lee 1994), and ornithocheirids more generally. Pending the discovery of more complete remains, Siroccopteryx is retained here as a distinct taxon.

Coloborhynchus. Coloborhynchus fluviferox Jacobs et al. 2019, based on the anteriormost portion of a rostrum bearing two pairs of teeth (FSAC-KK 10701), likely from Aferdou N’Chaft, appears to represent a second ornithocheirid from the Kem Kem Group. The holotype and only known specimen exhibits typical coloborhynchine characters such as the development of a tall, triangular, vertically reflected palatal surface on the anterior termination of the rostrum, pierced by the first pair of dental alveoli, and a tall, narrow rostral crest that rises directly from the anteriormost tip of this surface. This rostral fragment is somewhat larger than the holotype of Siroccopteryx moroccensis and, based on comparison with other more complete remains of ornithocheirids (Martill and Unwin 2012), likely represents a relatively large individual at least 4 m in wingspan.

Ornithocheiridae indet. A partial mandibular ramus 160 mm in length and bearing four dental alveoli was collected from Aferdou N’Chaft from the Gara Sbaa Formation (FSAC-KK 33, Fig. 95C, D). Caudal to the posteriormost dental alveolus, likely the last in the series, the dorsal surface of the mandibular ramus is smooth and rounded. Small pits may have accommodated the tips of teeth located in the rostrum (cf. Myers 2010), and suggest that the upper tooth row extended further posteriorly than the lower tooth row. Comparison with more complete material of Anhanguera (Wellnhofer 1985, 1991) and Coloborhynchus (Kellner and Tomida 2000, Veldmeijer 2003) suggest that, when complete, the mandibles were approximately 600 mm in length, representative of a large ornithocheirid 4–5 m in wingspan. Until Kem Kem ornithocheirids are better understood it is not possible to determine whether this fragment pertains to the two named coloborhynchines, or a third ornithocheirid. Jacobs et al (2020) recently described a mandible tip that they referred to Anhanguera. In addition, they figured two isolated rostral tips that they referred to, respectively, Ornithocheirus and a species of Coloborhynchus distinct from C. fulviferox.

Isolated, often incomplete, ornithocheirid teeth have been described by Kellner and Mader (1997) and Wellnhofer and Buffetaut (1999). They are present in several collections (MNHN, UCRC, FSAC) and have been recovered by the authors from multiple localities (Boumerade, Gara Sbaa, Zguilma, Aferdou N’Chaft, Taouz, and Jorf). Tooth crowns often show apical wear and the absence, in many cases, of a root suggests that many of these teeth may have been shed after some root resorption (Fig. 94).

All teeth recovered to date can be assigned to one of the four morphotypes recognized by Wellnhofer and Buffetaut (1999). They include: slender and recurved crowns with an oval cross-section near the tip (morphotype I, Fig. 94H); slender, flattened, gently recurved crowns (morphotype II, Fig. 94I); robust, wide-based, nearly straight crowns with two carinae (morphotype III, Fig. 94C, J); and very robust, large, recurved crowns approximately 35 mm in length (morphotype IV, Fig. 94K). Tooth morphology and size can vary quite considerably along the tooth row in ornithocheirids (e.g. Campos and Kellner 1985, Wellnhofer 1985, 1991, Kellner and Tomida 2000, Frey et al. 2003, Veldmeijer 2003, Wang et al. 2012, 2014). Moreover, this variation was likely compounded by allometric shape changes related to ontogeny, with hatchlings of 0.3–0.4 m wingspan (Unwin and Deeming 2019) achieving sizes, at maturity, of 4 m or more, an order of magnitude larger. Even so, it is difficult to accommodate all four tooth morphotypes within the dentition of a single ornithocheirid and it seems likely that multiple species were present in the Kem Kem Group. This is consistent with the recognition of four distinct genera, Siroccopteryx, Coloborhychus, Anhanguera and Ornithocheirus, but assignment of individual teeth to these, or other Kem Kem ornithocheirids (Jacobs et al. 2020) will require further work.

Azhdarchoidea Nessov, 1984

Tapejaridae Kellner, 1989

Tapejaridae indet. The anterior portion of an edentulous mandibular symphysis bearing a large ventral crest (BSP 1997 I 67) was described by Wellnhofer and Buffetaut (1999; Fig. 97). The jaw is Y-shaped in cross-section with rami that diverge only slightly posteriorly. Slit-shaped foramina are present on the external surfaces, and cancellous bone is exposed on broken surfaces. The ventral crest is large and subtriangular in shape, with a slightly concave antero-ventral margin. The fragment measures 118 mm in length and 16 mm in posterior width, with a maximum of 10 mm across the occlusal surface. Based on comparison with more completely known tapejarids such as Sinopterus from the Jiufotang Formation of China (Wang and Zhou 2003, Lü et al. 2006) it seems likely that BSP 1997 I 67 originally had a wingspan of around 3 m.

The morphology of the jaw and deep ventral crest corresponds well to that of tapejarids such as Tapejara (e.g., Vullo et al. 2012: fig. 5C). By contrast other edentulous pterosaurs either have a low mandibular crest (thalassodromeids) or none at all (pteranodontians, chaoyangopterids, azhdarchids) (Witton 2009, Vullo et al. 2012). Identification of BSP 1997 I 67 as tapejarid is supported by the recent discovery, in the Kem Kem group, of additional tapejarid material recently described by Martill et al. (2020). This new material assigned to a new taxon, Afrotapejara zouhrii Martill et al., 2020, suggests that a fragmentary rostrum, MN 7054-V, described by Kellner et al. (2007: fig. 1) is also tapejarid.

? Chaoyangopteridae

Apatorhamphus. A fragment of an edentulous rostrum missing its anterior tip (FSAC-KK 5010) collected at Aferdou N’Chaft has been made the holotype of a third genus and species of azhdarchoid, Apatorhamphus gyrostega, possibly a chaoyangopterid, from the Kem Kem Group (McPhee et al. 2020). McPhee et al. assigned several additional fragmentary rostra to A. gyrostega including FSAC-KK 5011, 5012 and 5013, BSP 1993 IX 338, originally identified as pteranodontid (Wellnhofer and Buffetaut 1999; Fig. 96A-E) and more recently as azhdarchid (Averianov et al. 2008, Ibrahim et al. 2010, Averianov 2014), and CMN 50895, originally determined by Rodrigues et al. (2011, fig1) as possibly a fragment of the mandibular symphysis of a dsungaripteroid. An additional specimen, FSAC-KK 5013, identified as a fragment of the mandibular symphysis, is almost perfectly complimentary to the holotype and tentatively assigned by McPhee et al. to this new taxon. Further jaw fragments including FSA-KK 27 (Fig. 96F-J), FSAC-KK 29 (Fig. 100A-E), FSAC-KK 32 (Fig. 100F-H) and UCRC PV161 (Fig. 100I-K) collected in 1995 near Taouz, likely also pertain to this pterosaur.

The rostrum shows a relatively rapid increase in depth posteriorly and has a slightly concave dorsal profile, which is typical of chaoyangopterids, but not other azhdarchoids. The lateral and palatal surfaces bear prominent foramina and the palatal surface has well developed dental margins, but no median ridge. Posteriorly the bone walls of the rostrum are remarkably thin, but toward the tip they become much more robust enclosing a deep but increasingly narrow central lumen. Unlike azhdarchids, the jaws of which have a ‘Y’ shaped cross-section (Wellnhofer and Buffetaut 1999, fig. 4a; Ibrahim et al. 2010: fig. 2c, d), the jaws of this pterosaur have a rounded, sub-triangular cross-section (e.g., Fig. 100C, D, H, K). FSAC-KK 5010 is 211 mm long. Comparison with more complete remains of chaoyangopterids (e.g., Lü et al. 2008) suggests that the prenarial rostrum of this individual was at least 0.3 m in length, the skull at least 0.6 m long and the wingspan in the region of 3–4 m, or more.

A combination of features including the shape of the rostrum, its unusual cross-sectional profile and the shape and distribution of foramina appear to distinguish FSAC-KK 5010 from other edentulous taxa found in the Kem Kem group, and azhdarchoids more generally although, in the latter case, comparison is often hampered by severe compression of the skull remains, for example in taxa from the Crato and Jiufotang Formations. While FSAC-KK 5010 compares more closely to the rostrum of chaoyangopterids than to other azhdarchoids, the possibility that it might, for example, be thalassodromeid cannot be entirely excluded, hence the caution in assigning this new species to Chaoyangopteridae.

Azhdarchidae Nessov, 1984

Alanqa. Alanqa saharica, Ibrahim et al. 2010, was founded on a partial mandibular symphysis (FSAC-KK 26; Fig. 98), reinterpreted here as a rostrum, collected in situ in 2008 at Aferdou N’Chaft (Ibrahim et al. 2010). Other Kem Kem group remains that can be assigned to this taxon include BSP 1996 I 36 (Wellnhofer and Buffetaut 1999: fig. 4), CMN 50859 (Rodrigues et al. 2011: fig. 1), FSAC-KK 4000 (Martill and Ibrahim 2015, fig. 3), a specimen from a private collection (Martill and Ibrahim 2015: fig. 5) and FSAC-KK 28 (Fig. 101).

Fragments of the rostrum and mandibular symphysis of this pterosaur have been described in detail, and figured by Wellnhofer and Buffetaut (1999), Ibrahim et al. (2010), and Martill and Ibrahim (2015). Alanqa saharica is characterized by remarkably straight jaw margins and a pronounced boss on the palatal surface of the rostrum, which is matched by paired accessory facets on the occlusal surface of the mandibular symphysis (Martill and Ibrahim 2015: fig. 4f). The latter features have not been reported in any other pterosaur, with the exception of Xericeps curvirostris Martill et al. 2018, suggesting a close relationship between these taxa. The holotype, FSAC-KK 26, appears to belong to an individual with an estimated wingspan of 3–4 m (Ibrahim et al. 2010).

The only phylogenetic analysis to include Alanqa to date (Longrich et al. 2018) recovered this pterosaur as a thallassodromeid. However, this was based on the assumption that the holotype represented the mandibular symphysis, rather than the rostrum, as proposed here. Three features: highly elongate slender jaws with remarkably straight margins and an unusual ’Y’ shaped cross-section, as found for example in Quetzalcoatlus (Kellner and Langston 1996: fig. 7), suggest that Alanqa and its close relative, Xericeps, are members of Azhdarchidae.

Xericeps . Xericeps curvirostris Martill et al. 2018, is represented by a single fragment of a mandibular symphysis (FSAC-KK 10700) commercially collected from the Douira Formation at Aferdou N’Chaft (Fig. 99). The holotype, described in detail by Martill et al. 2018, is distinguished by its curvature in lateral view, with markedly concave dorsal and convex ventral margins. Unique to this pterosaur there is a pronounced midline groove on the ventral border of the mandibular symphysis. Comparison with more complete remains of azhdarchoids suggests that FSAC-KK 10700 was a large individual of at least 3–4 m in wingspan.

Among azhdarchoids the long slender mandibular symphysis of Xericeps curvirostris (Fig. 99) is most closely comparable to that of azhdarchids such as Quetzalcoatlus (Kellner and Langston 1996) and Alanqa (Ibrahim et al. 2010). Indeed, Alanqa saharica and Xericeps curvirostris share a seemingly unique feature: well developed paired accessory occlusal surfaces on the posterior portion of the dorsal surface of the mandibular symphysis (Martill and Ibrahim 2015, Martill et al. 2018). These similarities, and the co-occurence of the remains in the same deposit, raise the possibility that A. saharica and X. curvirostris (Table 8) might be synonymous. Jaw morphology can be quite variable in pterosaurs, as a result of ontogeny, sexual dimorphism (reflected in the presence or absence of crests) and natural variation (Averianov 2014). Moreover, the morphological variation subtended by FSAC-KK 26 and FSAC-KK 10700 falls within the range of variation exhibited by, for example, the jaws of Rhamphorhynchus (Bennett 1995). However, additional, more complete remains are needed to demonstrate, or discount, synonymy.

Azhdarchidae indet.

The Kem Kem Group has yielded several well-preserved cervical vertebrae comparable to those described for azhdarchids such as Quetzalcoatlus (Howse 1986), Phosphatodraco (Pereda-Suberbiola et al. 2003) and Cryodrakon (Hone et al. 2019). These include two highly elongate vertebrae (Kellner and Mader 1996, Rodrigues et al. 2011), likely the fourth or fifth in the cervical series, and a fragment of a much larger cervical representing an individual of at least 6 m in wingspan (Ibrahim et al. 2010: fig. 6). A second individual of comparable size is represented by a nearly complete cervical vertebra (FSAC-KK 3088; Fig. 102). This relatively short vertebra, approximately three times longer than wide, is most closely comparable in its proportions to the third cervical of Phosphatodraco (Pereda-Suberbiola et al. 2003). Additional azhdarchid vertebrae, some of very large size, have yet to be described.

Azhdarchoidea indet. Numerous limb bones including the humerus, ulna, wing-metacarpal, wing-phalanges, femur and tibia have been collected in recent years. So far, however, only a single humerus has been described (Rodrigues et al. 2011). The humerus was assigned by these authors to the Azhdarchoidea of which four species (a tapejarid, a ? chaoyangopterid and two azhdarchids) have now been described from the Kem Kem group. It is not clear, at present, to which if any of these taxa the humerus may belong. The same will likely apply to the many, as yet undescribed, limb bones of azhdarchoids.