Epeiric Seas: Definitions and Modern Comparisons

Epeiric seas are shallow bodies of ocean water that have flooded onto continental crust, potentially extending for thousands of kilometers inland while often retaining connections to a larger open ocean. Such seas, which are very broad and shallow (<200 m deep) by comparison to deep oceanic basins (closer to 5000 m deep), may also have substantial freshwater influence (Menard and Smith, 1966; Schlager, 2005). The Trans-Saharan Seaway (fig. 1A) was one of several intracratonic seaways that existed globally during the late Mesozoic and early Cenozoic (Haq et al., 1987). It bisected West Africa and created temporary unions between the Tethys Ocean, entering West Africa from the north, and the Gulf of Guinea, entering from the south (Adeleye, 1975; Reyment, 1980; Kogbe, 1981; Bellion et al., 1990). Sediments deposited by epeiric seas represent some of the best-preserved marine rocks on Earth for the study of paleontology and geologic processes because by being internal to continents they are more removed from destructive tectonic events that could eliminate them (Harries, 2009).

A few modern examples of epeiric seas do exist. The Caspian Sea is a landlocked remnant of the ancient Tethys Ocean (Zonenshain and Pichon, 1986), and the Baltic Sea is another small, inland sea on continental crust where fresh and seawater mix (Winterhalter et al., 1981). The Hudson Bay in Canada is also flooded continental crust, and thus an epeiric sea, but not one on the scale of the Trans-Saharan Seaway (Harries, 2009). Closer to the tropics, the Java Sea, which is surrounded by the islands of Indonesia, is perhaps the best modern analog of a tropical epeiric system (Tjia, 1980). The phenomenon of ancient shallow seas can be particularly striking for its contrast to the modern environment occupying the same geographic location. The area once covered by, and supporting life within, the Trans-Saharan Seaway is now occupied by the Sahara, the world's largest hot desert (Tucker et al., 1991).

The characteristics of shallow marine ecosystems and shelf seas provide an imperfect analog for interpreting the Trans-Saharan Seaway (Harries, 2009). Coastlines can be defined as the “intertidal and subtidal areas on and above the continental shelf (depth of 200 m); areas routinely inundated by saltwater; and adjacent land, within 100 km from the shoreline” (Martinez et al., 2007: 256). The ocean covering of modern continental shelves is typically only 70 km inland, with an average water depth of 133 m, and rarely exceeds 200 m deep (Tyson and Pearson, 1991; Bianconi, 2002). Such shallow ocean covering of continents tends to consist of waters that are well mixed by seasonal winds, which influences the chemistry of the waters (Tyson and Pearson, 1991). Such waters are also impacted by tides and are categorized as part of the neritic zone of the ocean (Bianconi, 2002; James and Jones, 2016). Modern ocean coverage on continental crust constitutes some of the most biologically rich ocean regions, with the exception of hydrothermal vents (Bianconi, 2002; Martinez et al., 2007).

Tectonics, Geography, and Eustasy Impacting the Trans-Saharan Seaway

Tectonic Factors Shaping Geography. Large-scale tectonic activities and Precambrian basement geology in West Africa inform our understanding of why the fossiliferous sedimentary deposits of Mali were deposited where they are. Tectonics are also a major variable controlling sea-level change, defining crustal boundaries and zones of weakness that created low-relief basins. Such basins can receive water as well as sediment infill from the weathering of high-relief Precambrian massifs. During the time of the Trans-Saharan Seaway, three major variables contributed to high sea level and continental inundation. Rapid sea-floor spreading created younger, hotter, and more buoyant oceanic lithosphere, which displaced seawater onto continental crust. Secondly, a lack of continental ice sheets meant more water was in circulation globally, and third, low topography of continents due to tectonic subsidence provided areas for inland seas to collect (Harries, 2009).

Our field localities, and the entire country of Mali, are situated on the African Plate (fig. 1B), which consists of both oceanic and continental crust. Spreading at the African Plate's divergent boundary with the North American Plate as part of the fragmentation of Gondwanaland, brought about the opening of the southern Atlantic Ocean in the late Mesozoic (Burke, 1996). Such crustal movement was consequential for shaping the low-lying basins in West Africa (Guiraud et al., 2005). A part of the African Plate that has been stable for over 2 billion years is the Proterozoic West African Craton, which has a north-south orientation and underlies much of West Africa, including western Mali (fig. 1B; Black and Liegeois, 1993; Guiraud et al., 2005). The eastern border of the West African Craton is sutured to the Pan-African Mobile Belt (alternatively called the Trans-Saharan Belt or Pharusian Belt), an ancient orogenic mobile belt that represents an upwardly folded segment of Earth's continental crust. This belt was produced during the Pan-African orogeny, a Precambrian mountain-building event related to the formation of the Gondwanan supercontinent (Black and Liegeois, 1993; Guiraud et al., 2005: figs. 5, 8). Pan-African orogeny structures that form modern geographic landmarks in our field area include the Adrar des Iforas Precambrian massif, a geographic elevation composed of Precambrian crystalline basement rocks that likely provided sediment supply for surrounding basins (Petters, 1979: fig. 1A).

Recent tectonic plate reconstructions indicate that movement and counterclockwise rotation of the African continent positioned Mali and the rest of West Africa within the equatorial region, between 7°–15° north latitude, from the Late Cretaceous to the middle Eocene, and simultaneously opened a broad Tethyan Ocean to the north (Heine et al., 2013; Matthews et al., 2016; Müller et al., 2016; Ye et al., 2017). Across West Africa several large intracratonic basins, including the Iullemmeden, Taoudenit, and Chad basins, were affected by major Gondwanan break-up events (Petters, 1979: fig. 2B). These long-lived intracratonic basins originally formed due to failed rifting and their shape has been modified by the periodic reactivation of far-field stresses, sag-basin development and compressional tectonics associated with changes in plate motions (Guiraud et al., 2005). Thus, these basins experienced lithospheric attenuation and subsidence during major phases of continental breakup. As they subsided, they became conduits and reservoirs for the Trans-Saharan Sea to infill when it moved onto the West African Craton (fig. 1A–C).

Another determining factor in the geographic position of sedimentary beds was the formation in the Early Cretaceous of the Gao Trench, an event linked to the development of the Central African Rift System during the opening of the South Atlantic (Guiraud et al., 2005). The Gao Trench (fig. 1) is a narrow, east-west trending graben (and may be a half graben in places) that connects the Taoudenit and Iullemmeden basins. The part of the Trans-Saharan Seaway that specifically infilled the Gao Trench between the Late Cretaceous and Eocene has been referred to as the Détroit Soudanais (Radier, 1959). Although not connected to other segments of the Central African Rift Zone, the Gao Trench region holds a deep (>1 km) Cretaceous-Paleogene stratigraphic succession that records epeiric sea sedimentation from water that flowed between the Taoudenit and Iullemmeden basins.

Our research has concentrated on the Gao Trench and the parts of the Iullemmeden and Taoudenit basins that extend into Mali, areas that were in the Late Cretaceous–early Paleogene, and are today, thousands of kilometers from the open ocean (fig. 1). Given the inland location of these areas, the sedimentary rocks now found there were, as we noted above, deposited in reactivated basins undergoing extensional tectonics linked to far-field stresses associated with the breakup of Gondwana (Guiraud et al., 2005: fig. 1b). The 800,000 km² Iullemmeden Basin (fig. 1C, D) extends through modern Algeria, Nigeria, Benin, Chad, Niger, and Mali, where it occupies most of the country east and southeast of the Adrar des Iforas Precambrian massif (Kogbe, 1981; Bellion et al., 1989; Moody and Sutcliffe, 1991; Zaborski and Morris, 1999). The basin has a gentle syncline shape and sedimentary deposits of the Trans-Saharan Seaway collected widely within it due to its low elevation (Radier, 1959; Boudouresque et al., 1982; Kogbe, 1991). The Taoudenit Basin is an even more extensive (2 million km²) structural feature of the West African Craton that comprises the greater part of northern Mali (Bronner et al., 1980; Bassot et al., 1981). Upper Cretaceous–Eocene Trans-Saharan Seaway strata in the syncline known as the Tilemsi Valley (Bellion et al., 1989) are best ascribed to the Taoudenit Basin because they overlie the deformed Pan-African Mobile Belt that defines the eastern boundary of that basin (Guiraud et al., 2005: fig. 5). The more southerly Benue Trough, which is not in Mali but is relevant when discussing the regional continental flooding, also formed as the product of rifting in the Early Cretaceous (Guiraud et al., 2005).

Eustasy and Sedimentary Rock Formation. Tectonic factors induced marine incursions onto West Africa as early as the Aptian-Albian approximately 113 million years ago (Moody and Sutcliffe, 1993). Those transgressions are, however, distinct from the Late Cretaceous ones that started in the Cenomanian-Turonian, which are the focus of our research. An increased rate of sea-floor spreading at midocean ridges, not only in the Atlantic, but globally, was underway in the Late Mesozoic (Hays and Pitman III, 1973; Reyment and Dingle, 1987; Miller et al., 2005). This type of lithospheric activity increases the volume of midocean ridges and adjacent crust due to thermal expansion of sea-floor rocks. As a result, there is less volumetric capacity in the ocean basins, which forces ocean water onto land where it finds its way to low-relief areas. Pulses of ridge growth and the accumulation of newer, hotter lithosphere have been tied to transgressions, whereas intervening periods of suspended growth are linked with regressions (Hays and Pitman III, 1973).

The subject of our work is the series of transgressive cycles that started in the Cenomanian and continued through the middle Eocene, leaving extensive sedimentary deposits in Mali. These strata comprise mixed carbonate deposits generated in situ along with siliciclastic deposits weathered from the Adrar des Iforas massif, all of which settled into the Iullemmeden and Taoudenit basins and the Gao Trench (Petters, 1979; Reyment and Dingle, 1987; Tapanila et al., 2004; Tapanila et al., 2008). The resulting sedimentary units are relatively thin and lie unconformably on one of the following: pre-Cenomanian continental sediments, Jurassic or Cambrian-Permian strata, or, in some cases of maximum transgression, directly on the Precambrian basement strata of the Adrar des Iforas massif (Radier, 1959; Reyment, 1980; Moody and Sutcliffe, 1991; 1993; Tapanila et al., 2004; Tapanila et al., 2008)

FIGURE 2.

Map of northwestern Mali with our localities from the 1999 and 2003 expeditions plotted on a topographic map with exposed geologic formations indicated. The location of the “Ansongo-1” well log (appendix 2) is also indicated (Elf-Aquitaine, 1979). Note the restricted exposure of the Paleocene Teberemt Formation. Gao Trench graben margins indicated by dashed lines. Orange lines demarcate three major depocenters.

At high sea level, the southern embayment of the Tethys Ocean extended inland along the coast of North Africa and the Trans-Saharan Seaway specifically was its north-south trending arm that reached into Libya, Algeria, Mali, Niger, and even northern Nigeria, taking different routes at different times (fig. 1A; Petters, 1977: fig. 9; Boudouresque et al., 1982: fig. 2). Our field localities in Mali are situated to the west, south, and east of the Adrar des Iforas and represent some of the most inland reaches of the ancient sea (fig. 2). In the Taoudenit Basin, Mesozoic rocks are concentrated directly along the western margin of the Adrar des Iforas in the Tilemsi Valley (fig. 2). As we noted above, the thickest succession of Trans-Saharan Seaway deposits lies to the south and east of the Tilemsi Valley in the Gao Trench. At times the Trans-Saharan Seaway mixed with ocean water that had flooded West Africa via the southerly Benue Trough and the Sokoto Basin (Kogbe et al., 1976; Petters, 1977; 1979; Adetunji and Kogbe, 1986; Reyment and Dingle, 1987).

Finally, it is helpful to contextualize how different the Late Cretaceous world was relative to the Recent. Sea level has shown a 400 m range of variation over the Phanerozoic, and extensive geological evidence indicates that it was higher in the Late Cretaceous than at any point over the last 250 my (Tyson and Pearson, 1991; Harries, 2009; Haq, 2014). In the Cretaceous, as much as 40% of current continental land was submerged under 50–100 m of ocean water (Hays and Pitman III, 1973; Miller et al., 2005). Global sea level today is approximately 300 m lower than it was during the Late Cretaceous (Campanian), and is at one of its lowest points in the last 250 my (Haq et al., 1987; Falkowski et al., 2004; Haq, 2014). By comparison, the estimate for climatically mediated sea-level change from human-generated global warming predicts, at maximum, a sea-level elevation of approximately 2 m by the end of the 21st century (NOAA, 2012). While a 2 m sea elevation would be disastrous for current populations of humans (Nicholls et al., 1999), it is noteworthy that such change differs significantly in size and mechanism from the sea-level changes of the Late Cretaceous under investigation here.

Paleoenvironment of the Trans-Saharan Seaway

Temperature and Seasonality. The Malian portion of the Trans-Saharan Seaway was located in the paleotropics, between paleolatitudes 23° 27′ north and south of the equator (Matthews et al., 2016, and references therein), thus geographic position alone suggests that the Trans-Saharan Seaway would have been relatively warm (fig. 1; see also GPlates Reconstruction of the Trans-Saharan Seaway, below). The same tectonic activity described above affecting sea level is also hypothesized to have caused a greenhouse climate in the Late Cretaceous–early Paleogene due to the release of carbon dioxide (Hallam, 1985; Kennett and Stott, 1991; Pearson et al., 2001; Shellito et al., 2003; James and Jones, 2016). Evidence for this hypothesis comes from the study of fossil foraminiferan tests that contain oxygen isotopes that vary depending on the temperature of the water in which the shells formed (Urey, 1947; Pearson et al., 2001; Zachos et al., 2001). Coupled with deep-sea drilling projects that sample such temperature proxies on a large scale, a picture of a Late Cretaceous–early Paleogene “hothouse” interval has emerged (Savin, 1977; Kennett and Stott, 1991), a time when Earth probably lacked polar ice sheets (Markwick, 1998; Pearson et al., 2001; Zachos et al., 2001; Royer, 2006).

In the late Mesozoic, the sea surface temperature of water in the tropics may have been as high as 33°–34° C (Norris et al., 2002) and on the northern part of the African continent the temperature may have averaged ∼30°–36° C (Sellwood and Valdes, 2006). The very presence of enormous bodies of sea water on continents, like the Trans-Saharan Seaway, also is thought to have stabilized global heat and spread it broadly as heat is very effectively stored and transported in water (Hays and Pitman III, 1973; Harries, 2009). Bioproxies for temperature, such as body size in South American fossil snakes and the physiological limitations it imposes, have been used to suggest that the paleotropics may have been even hotter than the present-day tropics (Pearson et al., 2001; Head et al., 2009). Our discovery of relatively large extinct snakes in Mali is another body-size data point suggesting that the African paleotropics in and around the Trans-Saharan Seaway may also have been relatively hot (McCartney et al., 2018).

Modern tropical climates typically lack strong seasonal temperature change and frost (Linacre and Geerts, 2002), characteristics that would have likely described the Trans-Saharan Seaway. An isotopic study of the shell composition of tropical Cretaceous bivalves indicated that seasonal fluctuations in temperature were of relatively low amplitude during the warmer parts of the Cretaceous, such as the Campanian through the Maastrichtian (Steuber et al., 2005). Markwick (1998) also emphasized that the presence of ancient dyrosaurid crocodyliforms in Mali is consistent with the Trans-Saharan Seaway having had a mean annual temperature no colder than 14.2° C.

Important for the time interval under study here is also the occurrence of extreme temperature changes over short durations. A spike in global temperature known as the Paleocene-Eocene Thermal Maximum (PETM) occurred 56 my ago, and was globally among the hottest times of the last 250 million years (Kennett and Stott, 1991; Thomas and Shackleton, 1996; Zachos et al., 2001; Zachos et al., 2008; McInerney and Wing, 2011). During the PETM, climate warmed an estimated average of 5° C over only ∼5 ky, a relatively short interval geologically (Zeebe et al., 2016; Kirtland Turner et al., 2017; Kirtland Turner, 2018), and then cooled to preevent levels over only ∼100 ky (Kennett and Stott, 1991). The PETM has been associated with terrestrial faunal change in North America (Wing et al., 2005; Woodburne et al., 2009), but is only beginning to be investigated in continental African rocks.

Precipitation and Composition of Water: Modern tropical climates often exhibit profound seasonal variation in precipitation and this also may have been the case for the Trans-Saharan Seaway. Hallam (1985: fig. 8), however, specifically reconstructed the northern region of the Trans-Saharan Seaway as lying within an arid belt and the more southerly part as ranging from seasonally wet to consistently wet. He presented ideas that reinforced interpretations of Petters (1977), who argued that, as in the modern Red Sea, there was little circulation in the water of the Trans-Saharan Seaway and the environment was arid. Adetunji and Kogbe (1986), however, based on their studies in the Maastrichtian Dukamaje Formation of the Nigerian part of the Trans-Saharan Seaway, argued instead for significant water circulation. The model of Sellwood and Valdes (2006) also indicated that the Late Cretaceous had humidity that varied with seasons and precipitation that exceeded evaporation. When sea level was high, such as during the interval under study here, oxygen-deficient ocean waters may have been relatively common (Tyson and Pearson, 1991). Tethyan marginal sediments in particular are hypothesized to have been dysoxic at the PETM (Speijer and Wagner, 2002). The fauna we describe below can be considered brackish to normal marine. Intermittent downpours of rain may have periodically and dramatically decreased the salinity of the Trans-Saharan Seaway, quickly killing stenohaline species (Reyment and Dingle, 1987: 105).

The opposite condition, hypersalinity, of the Trans-Saharan Seaway has also been proposed, a condition that would be linked to periods of aridity (Reyment, 1980). The widespread existence of shallow seas is hypothesized as having caused a substantial increase in seawater evaporation and the creation of highly saline ocean bottom waters in general (Brass et al., 1982). In Mali, gypsum and salt deposits occur in the very earliest part of the section we examined, the Cenomanian-Turonian sediments of the Tilemsi Valley of the Taoudenit Basin, which Reyment (1980: 317) interpreted as recording extremely arid time intervals. The late Cenomanian deposits, but not subsequent ones, were also characterized by black shales that Reyment and Dingle (1987: 102) interpreted as pockets of “stagnation.” Moreover, there were quite possibly intervals during which the sea, or parts of it, was not connected to the Tethys and any embayments may have become shallow, warm and hypersaline (Petters, 1979; Reyment, 1980).

Tidal Activity: Given their broad and shallow geographic expanse, epeiric seas are thought to have been microtidal, or to have had a tidal range of approximately 2 m (James and Jones, 2016: 254). Petters (1979: 753) wrote that there are few “current-induced sedimentary structures” in the Upper Cretaceous-Paleogene rocks of the Malian region, and that the predominance of sediments with high mud content suggests that in the most internal reaches of marine embayments, the water was “low energy.” As we discuss below, this generalization is not supported by our observations of the Upper Cretaceous rocks because we identified sandstones with abundant delta foresets, a signature indicative of high coastal energy. We do not see evidence of cross-laminated sandbodies, an indicator of strong currents. Thus, the generalization of Petters (1979) is a reasonable summary of the sedimentary signal in many aspects of the younger rocks of the region, except the phosphate conglomerates interleaved throughout the three younger formations, that indicate storm-related winnowing activity and persistent fair-weather wave action during deposition (Tapanila et al., 2008). Thus, fair weather and storm waves would have been the major means by which the Trans-Saharan Seaway water mixed (James and Jones, 2016). Without other forms of water mixing, stratification of the epeiric sea may have occurred, which would have limited the circulation of nutrients and possibly affected primary producers like phytoplankton.

Repository and Institutional Abbreviations

Repository: The specimens collected belong to the Republic of Mali. From 1999–2019 they have been on loan for scientific research to the laboratory of M.A. O'Leary in the Department of Anatomical Sciences, Renaissance School of Medicine, Stony Brook University, Stony Brook, New York. The specimens were originally cataloged with the prefix “CNRST-SUNY” (defined below). With permission from the Republic of Mali, the specimens are being transferred to the Department of Paleontology, American Museum of Natural History in 2019. Specimens in this paper carry the original catalog numbers and appendix 4 lists their newly assigned AMNH numbers. As of this writing the collection contains three holo-types. CT scans and other digital media associated with the paper are deposited in MorphoBank, Project 2735 ( www.morphobank.org).

Abbreviations: ANSP, Academy of Natural Sciences, Philadelphia; CNRST-SUNY, joint collection of the Centre National de la Recherche Scientifique et Technologique, Bamako, Republic of Mali and Stony Brook University, Stony Brook, New York; MUVP, Mansoura University, Vertebrate Paleontology collections, Mansoura, Egypt; NHMUK PB V, paleobotany collection and PV P, R, and M, paleovertebrate collections of fish, reptiles, and mammals, Natural History Museum London; TGE, ensemble divisions for Muséum National d'Histoire Naturelle specimens from Morocco, Algeria, and/or Mali from Martin (1995). Superscripts in the Systematic Ichnology and Systematic Paleontology sections indicate locality numbers.

Subdivision and Proposed Nomenclature for the Upper Cretaceous-Lower Paleogene Stratigraphy in Northeastern Mali

For nearly a century, geologists have recognized that Malian deposits of the Trans-Saharan Seaway have similarity to and continuity with better-studied sediments in Nigeria and Niger. As a result, gross-scale stratigraphic correlations have been proposed by many authors for Malian deposits throughout the Taoudenit and Iullemmeden basins and the Gao Trench (e.g., Radier, 1953; Krasheninnikov and Trofimov, 1969; Berggren, 1974; Petters, 1977; Bassot et al., 1981; Reyment and Schöbel, 1983; Moody and Sutcliffe, 1991; Mateer et al., 1992; Tintant et al., 2001; Swezey, 2009). The importance of these attempts at broad geographic correlation notwithstanding, such efforts have lacked precision regarding Malian rocks, and were often based on more extensive investigation of deposits in Nigeria and Niger than in Mali. As a result, such correlations have been difficult to apply on the ground in Mali because much of the prior work to date had been based on regional-scale mapping or well-logging in the Gao Trench. Aside from the seminal fieldwork by Radier (1959), little attention has focused on correlation of outcrops and stratigraphies among the three different depocenters in Mali.

Synthesis has been further complicated by a lack of formal or consistent stratigraphic nomenclature associated with the deposits in Mali. Many workers have simply referred to units in Mali by formation names described from type sections in Nigeria or Niger (e.g., Dange Formation, Wurno Formation sensu Petters, 1979) without supplying or referring to detailed lithostratigraphic descriptions for the sections in Mali. However, as our work attests, the stratigraphy and sedimentology of the deposits in Mali, while generally similar to other deposits across the Trans-Saharan Seaway region, are sufficiently different to prohibit direct correlation based on lithostratigraphic definitions alone.

The development of a formal lithostratigraphic framework for the Mesozoic-Cenozoic deposits in Mali is critical for continued progress in the correlation of strata across and within national borders. In addition, such a framework represents a foundation for studying the complexities and variations of sequence stratigraphy and paleoenvironments in the ancient seaway as a whole. Toward this end, Moody and Sutcliffe (1991, 1993) provided an initial lithostratigraphic nomenclature for the Maastrichtian-Eocene deposits in Mali. They did not, however, describe any of the units using formal type sections and boundary definitions, critical information necessary for subsequent identification of rock units as outcrops. As such, the published names and definitions of these units have minimal utility.

Using our own field-based stratigraphic work and two decades of research on the new localities we established in Mali (plate 1; table 1), we provide here formal definitions of stratigraphic units and designate type or lectostratotype sections for these rocks. Because the names proposed by Moody and Sutcliffe (1991, 1993) have found their way into the literature, we have attempted to repurpose them in our formal designations, rather than abandon the terms at risk of causing further confusion. We are aided in this effort by access to an unpublished well log, “Ansongo-1” (fig. 2; appendix 2), that provides critical new details for correlation (Elf-Aquitaine, 1979).

We propose formal subdivisions and definitions of a number of Upper Cretaceous–Paleogene sedimentary deposits exposed within and across Mali, and correlate measured sections through these units among the Iullemmeden and Taoudenit basins and the Gao Trench study areas (plate 1). Subdivision is strongly warranted based on distinctive lithological and depositional variations, as well as temporal differences among these units. Subdivision is also necessary to provide context for describing the floras and faunas from the region as a whole. Where formation names have already been informally applied to units, we have endeavored to retain these names and to define lectostratotype sections. For the pre-Campanian sedimentary units in northeastern Mali, however, we do not have sufficient new information to define lectostratotype sections or to formalize these units. Thus, we simply provide a brief review of such units and show how they correlate with others in the basin (fig. 3). Facies analysis of Upper Cretaceous–Paleogene units across Mali has been presented in detail previously by our team (Tapanila et al., 2008), and the five repeated facies described for the rocks initially in that paper are again referenced in the sections below. We stress that further fieldwork is required to produce a more in-depth regional correlation among the rocks of Mali, Niger, and Nigeria.

FIGURE 3.

Composite of formation names, several formally diagnosed here for the first time, through the Trans-Saharan Seaway of Mali and equivalent rocks of Niger and Nigeria. There is a sedimentary hiatus at the base of the Paleogene and the Danian is not preserved (Dikouma et al., 1993). Formation names for Niger and Nigeria from several publications (Kogbe et al., 1976; Reyment, 1980; Reyment and Schöbel, 1983; Adetunji and Kogbe, 1986; Mateer et al., 1992; Moody and Sutcliffe, 1993). Unit ages from Cohen et al. (2013).

Biostratigraphy has historically been and continues to be the basis for dating these Upper Cretaceous–Paleogene sediments (e.g., Radier, 1959; Krasheninnikov and Trofimov, 1969; Berggren, 1974; Petters, 1977; Reyment, 1979; Boudouresque et al., 1982; Bellion et al., 1990; O'Leary et al., 2006; Claeson et al., 2010) because the field area lacks volcanic sediments for radiometric dating. Thus, ongoing identification of index fossils, taxa that are restricted in time, abundant, and widely distributed geographically (Levin, 2003), has been essential. We denote our fossil discoveries relevant to age determinations with each formation below. Microfossil-based biostratigraphy, including very important contributions on Foraminifera (Krasheninnikov and Trofimov, 1969; Berggren, 1974), Ostracoda (Damotte, 1991; Damotte, 1995), Ostracoda and Foraminifera (Bellion et al., 1989), and fossil pollen (Boudouresque et al., 1982), forms the primary temporal structure for the correlation of these Malian rocks.

TABLE 1

Localities from the 1999 and 2003 expeditions that yielded fossils resulting in publications.

Geographic coordinates, age range and fossils identified to date are listed (higher clades provided here, more specific taxonomy within the text or other papers). See plate 1 for lithology by locality except for Mali-6 and -12, which and are older than the Late Cretaceous–Eocene focus of this paper (see Bamford et al., 2002; and O'Leary et al., 2004a). The asterisk (*) indicates a type specimen is part of the fauna. Appendix 3 lists localities examined by us for microfossils.

Continued

An examination of sediments we collected by L. LeVay and L. Edwards did not yield significant biostratigraphically useful microfossils, such as calcareous nannoplankton, dinocysts, or fossil pollen (appendix 3). The only identifiable findings were from rocks of localities Mali-11, which yielded nonspecific palynodebris, and Mali-17 unit 3, which yielded an indeterminate species of the dinocyst Spiniferites. The latter ranges from the Cretaceous through the present (Paleobiology Database, 2018) and is simply consistent with other, more refined, age constraints we discuss below. E. Theriot also examined our sediment samples for diatoms and found no informative results. Thus, in the absence of more biostratigraphically informative microfossil discoveries, our biostratigraphic contributions are based exclusively on macrofossils.

Continental Intercalaire: Kilian (1931) first used the term Continental Intercalaire to refer to the extensive succession of poorly dated nonmarine deposits exposed across much of West Africa (see also Lefranc and Guiraud, 1990; Mateer et al., 1992). These beds typically lie above Carboniferous marine deposits and below Cenomanian marine strata associated with flooding of the Trans-Saharan Seaway (hence the name, which describes an interleaving or “intercalation” of nonmarine strata between marine beds). Bellion et al. (1989) identified two distinct stratigraphic units that compose the Continental Intercalaire in northern Mali (see fig. 3): the red conglomerate of Tezzoufi (Permian–Middle Jurassic) and an unnamed overlying complex of multicolored clays and limey sandstones (Upper Jurassic–Lower Cretaceous). Those authors also identified a disconformable contact between the top of the Continental Intercalaire and the overlying sequence of Upper Cretaceous marine units. Bellion et al. (1989) placed the contact between these two units at the first appearance of gypsum horizons.

These deposits are, however, poorly exposed in Mali, and limited to the northern portions of the Taoudenit and Iullemmeden basins. Continental Intercalaire rocks were documented in the well log “Ansongo-1” (Elf-Aquitaine, 1979), which bisected the sediments of the central part of the Gao Trench (appendix 2). In this log, the Continental Intercalaire was estimated to be as much as 1500 m thick, and of similar character to the deposits observed elsewhere in outcrop. We described an extensive petrified forest from the upper part of this succession based on our collections at locality Mali-6 (Bamford et al., 2002), as well as new titanosaurian dinosaurs from locality Mali-12 (O'Leary et al., 2004a). Additional details about the sedimentology of this unit are available in those publications. Petters (1977) discussed the equivalents of these beds in Nigeria, synthesizing work by Kogbe et al. (1976), as the Illo and Gundumi formations, noting that the former also contains fossil wood (see also Mateer et al. (1992)). There is currently very poor age control on the Continental Intercalaire rocks in Mali and further refinement of its age is out of the scope of this paper.

Unnamed Upper Cretaceous Continental Deposits: Radier (1959) described a series of Upper Cretaceous continental deposits dominated by ferruginous sandstones and shales, minor bedded gypsum units, thin siliceous limestones with gastropods, and sandy mudstones with large vertebrate bones near the Iullemmeden Basin to the northeast of Ménaka. Similar deposits were also identified by Bellion et al. (1989), and briefly by our team in the northwestern part of the study area in the Taoudenit Basin. These beds were also encountered in several water wells and in the “Ansongo-1” well (Elf-Aquitaine, 1979) where they reach up to 250 m thick. The water well cuttings have been identified as unfossiliferous, and have not been described in detail.

In our examination of outcrops, we did not identify the contact between these beds and the base of the overlying formation (described below), nor was the contact observed in the “Ansongo-1” well report or other water well logs (Elf-Aquitaine, 1979). However, based on the “Ansongo-1” well report and brief notes made by Radier (1959: 349), we interpret a conformable relationship between these beds and the overlying formation. The term “Continental Hamadian” has been used (e.g., Mateer et al., 1992: fig. 8) to describe these rocks, but the term is so poorly defined that we do not advocate for its continued use and do not apply it here. This formation is poorly dated, with an age constraint derived largely from superposition and broad regional correlation. Further refinement of its age was out of the scope of our work.

Tichet Formation: We propose the name Tichet Formation (fig. 4) for the sandstone unit that makes up the base of what has been considered a succession of Campanian-Maastrichtian rocks across Mali (e.g., Moody and Sutcliffe, 1993; figs. 3–5). These deposits were originally described by Monod (1939) and Radier (1959) and later mapped as part of the late Mesozoic “PGi” by Bassot et al. (1981). Bassot et al. (1981) mapped these deposits as a continuous swath along the flanks of the Adrar des Iforas massif across the Taoudenit and Iullemmeden basins through the Gao Trench. The name “Tichet” refers to the well and village located on the northwestern side of the Tilemsi Valley on the far eastern margin of the Taoudenit Basin, which is ∼40 km north of the proposed type locality (fig. 1D).

The Tichet Formation is defined herein as the sequence of buff to locally green and pale orange-colored, cross-bedded sandstones typified by large-scale inclined foresets located between the upper Cenomanian gypsiferous deposits and the base of the overlying Ménaka Formation (defined below). The basal contact of the Tichet Formation has not been directly observed anywhere in the field area but is placed above the highest-bedded gypsum unit of putative Cenomanian-Senonian age (sensu Bellion et al., 1989). The bedded gypsum unit crops out above the Continental Intercalaire. The upper contact of the Tichet Formation is defined herein as the top of the uppermost sandstone bed (or siltstone locally) below the thinly bedded succession of paper shales and carbonates of the Ménaka Formation. The contact is sharp and distinctive in outcrop, where it is commonly characterized by an intensely bioturbated, deeply oxidized contact.

Type locality and stratotype section. The type section is located at the base of the large escarpment ∼40 km north of Tichet and another reference section was measured at a locality that we termed Mali-1 (fig. 2; plate 1), located at 19° 28.918′N / 0° 19.499′E. The lower contact of the Tichet Formation cannot be observed in the type area and has not been observed directly anywhere by the authors. Three reference sections for the Tichet Formation have been measured and described: from a nearby locality in the Tilemsi Valley at Mali-3/4; and at two localities in the Iullemmeden Basin (Mali-7 and -8) to the west of Ménaka, all of which expose the top of the formation (fig. 2; plate 1).

Description of the sedimentology. The Tichet Formation is characterized by fine-grained, cross-bedded sandstones and siltstones of facies 1 and 2 of Tapanila et al. (2008). In three out of the four areas where the Tichet Formation was studied (Mali-3/4, -7, -8) the basal portion of the formation is dominated by a succession 5–15 m thick of distinctive, large-scale (∼5–7 m), gently dipping (5°–8°) sandstone beds (Facies 1). These beds dip to the south at Mali-3/4, and to the northeast at Mali-7 and Mali-8. Laterally, the large-scale inclined sandstone beds transition along strike, as well as stratigraphically upward into horizontally stratified sandstone beds with trough cross-stratification and herringbone cross-stratification (also Facies 1). The facies has abundant impressions of the ichnofossil Teredolites (fig. 5), a type of fossilized boring formed in wood. Skolithos burrows are also present in this unit and most abundant toward the base of the gently inclined sandstones (fig. 4D). Complex root traces become more common toward the top of the unit, along with indistinct burrows. In the type locality, the horizontally bedded sandstones of Facies 1 are conformably overlain by a 20–25 m thick, upward-coarsening succession of shales (Facies 2), followed by siltstones and sandstones of Facies 1. These units are also pervasively bioturbated with evidence of oxidized burrow and root networks. Bedding is commonly lenticular, with alternating mudstone and siltstone couplets. These uppermost facies are not observed in the Iullemmeden Basin sections, where the total thickness of the formation is less than half that observed in the Taoudenit Basin section, the latter reaching a maximum exposed thickness of 30 m. The Tichet Formation is completely absent from the Gao Trench. In all studied sections, the top of the formation is characterized by a deeply oxidized, heavily bioturbated capping sandstone/siltstone, which sits sharply beneath the first paper shale or thin limestone bed of the Ménaka Formation.

FIGURE 4.

Basal deposits of the Upper Cretaceous–Campanian Tichet Formation near the base of the Mali-1 and Mali-3/4 sections in the Tilemsi Valley, Taoudenit Basin: A, 5 m thick alternating succession of rhizoturbated sandstones (yellow arrows) and unmodified, horizontally laminated and ripple cross-laminated sandstones (black arrow) interpreted as estuarine to tidal flats with evidence of mangrove roots; B–C, close-up photos of intensely rhizoturbated sandstone beds from A showing fossil mangrove root networks (red arrows) among indeterminate invertebrate burrows; D, horizontally laminated to ripple cross-laminated sandstone with climbing ripples (red brackets) and the ichnofossil Skolithos sp. (black arrow) represented by an isolated vertical burrow.

FIGURE 5.

Example of intensely bored fossil wood fragment from the Upper Cretaceous–Campanian Tichet Formation collected at locality Mali-4; the clavate (sock-shaped) borings are Teredolites clavatus (CNRST-SUNY 134) . A, external view of wood with borings mostly filled in with medium-grained sand, and B, internal view of bored wood after sectioning. Scale bar = 1 cm.

Interpretation of the sedimentology. The base of the Tichet Formation is interpreted to represent a deltaic system, with the large-scale, gently inclined sandstone beds indicative of delta front depocenters. Laterally and stratigraphically overlying these deposits are the horizontally bedded sandstones with trough cross-stratification and herringbone cross-stratification, as well as overlying mudstone and siltstones with tidal couplets and complex root networks, which we interpret as lower delta plain deposits. The lenticular bedding and paired sand-silt couplets are interpreted as low-energy tidal deposits consistent with a lower delta plain setting; and the complex, oxidized, root networks are interpreted as mangrove root systems similar to those reported from Paleogene deposits from the Fayum region in Egypt (Bown and Kraus, 1988). We interpret this series to represent a single continuous delta system of freshwater that drained into the Trans-Saharan Seaway because there is a north to south along-strike facies change from lower delta plain deposits at Mali-1 to delta front deposits at localities Mali-3/4.

The point source for the drainage was to the north, most likely out of the Adrar des Iforas massif. Interestingly, a similar pattern is observed at localities Mali-7 and -8, where the basal, gently inclined sandstones are interpreted as delta front deposits and the overlying horizontally bedded sandstones and siltstones are interpreted to be south-southwest directed, pro-grading upper delta plain deposits. We also interpret the more western localities of Mali-7 and -8 to have been connected to the same delta system, but as laterally correlative delta lobes connecting to a large, bird-mouth type delta system that was most likely sourced from a second river system draining out of the Adrar des Iforas Mountains to the south. We can not, however, rule out that the Tichet Formation deposits observed at all four localities could be part of a single, progradational delta system. The distinctive oxidized, intensely bioturbated top of the formation, which is sharply overlain by paper shales or thin limestone beds, was interpreted by Tapanila et al. (2008) as a sequence boundary and superimposed marine flooding (transgressive) surface.

Discussion of the sedimentology. Petters (1977) observed that the Taloka Sandstone, the Nigerian equivalent of the Tichet Formation, also contained Skolithos trace fossils and worm burrows as well as Thalassinoides, bivalves, and vertebrate fossils. Petters (1979: 755) interpreted the Taloka Formation as “coastal plain–beach–shallow sublittoral” environment. Mateer et al. (1992: fig. 8), in an overview paper, interpreted the Continental Hamadian, a poorly defined group of continental sediments (see also Greigert, 1966), as extending through the Campanian in Mali and contacting a Taloka Formation at the start of the Maastrichtian. We do not follow that terminology here because of the insufficient precision associated with the term “Continental Hamadian.”

Age based on biostratigraphy. The oldest rocks in our composite section, and the only ones that we interpret as Cretaceous, are found in both the Taoudenit and Iullemmeden basins. To date, we have identified no index fossils and no vertebrate or invertebrate macrofossils in the Campanian-Maastrichtian rocks of the older Tichet Formation. The fossilized burrows, both crustacean and otherwise, and root casts have yet to provide temporal information. Our correlation of the continuity of Tichet Formation beds suggests that it is one of the formations in our study area that is time transgressive. Its age assessment is based on a combination of constraints placed by fossils from units above and our sequence stratigraphic correlation (i.e., the sandstones vary in time and are represented at the margins of the basin during sea-level regression), and on comparisons with published sequences from Nigeria and Niger (Kogbe, 1991; Moody and Sutcliffe, 1991; 1993). Lithostratigraphy indicates that this formation lies between the highest gypsiferous facies of the Continental Intercalaire (upper Cenomanian) and the lowest paper shales of the Maastrichtian Ménaka Formation. Comparisons with published ages on correlative sequences from Nigeria and Niger suggest a Campanian-Maastrichtian age for these rocks in Mali.

Ménaka Formation: The name “In-Fargas Formation” was originally used by Moody and Sutcliffe (1991) to describe the Upper Cretaceous succession across Mali, however, those authors later referred to this set of rocks as the “Cheit Keini Formation” (Moody and Sutcliffe, 1993). Although Moody and Sutcliffe (1993: figs. 3, 5) suggested that they intended to restrict the Cheit Keini Formation to the uppermost Maastrichtian, it was not in the end further specified in their paper and it remained unclear whether Moody and Sutcliffe (1991, 1993) were referring to all Upper Cretaceous strata in Mali or whether they were specifically referring to only the uppermost Cretaceous deposits. They only provided an idealized graphic log for these deposits across the combined Mali and Sokoto (Nigeria) regions, which suggested that they may have been referring only to the uppermost deposits (because it does not match what we have observed in Mali). However, no detailed type section was included. The details on the idealized graphic log for this unit could not be matched to the observed stratigraphy anywhere in the study area our team examined. The graphic log is also difficult to reconcile with measured stratigraphic sections and descriptions for the uppermost Cretaceous strata provided by Radier (1959) and Bellion et al. (1989). Moody and Sutcliffe (1993) described the Cheit Keini Formation as mostly sandstones of shallow marine and deltaic influence, which is more consistent with what is observed lower in the Upper Cretaceous stratigraphy (what we have described above as the Tichet Formation). Adding to the confusion, the Tichet Formation, or indeed any deposit fitting a similar description, does not crop out near In Fargas, closest to this study's localities Mali-18 and -10.

Thus, to avoid confusion and perhaps error, we suggest that the terms In Fargas Formation and Cheit Keini Formation are to be avoided and propose that both terms be abandoned as they have not been formally defined and do not match the observed stratigraphy. Hence, we propose the name Ménaka Formation for the uppermost Cretaceous deposits in northern Mali (figs. 2, 6; plate 1). These deposits were originally described by Radier (1953) and mapped as the “PGi” by Bassot et al. (1981), who demonstrated their continuous distribution across the Taoudenit and Iullemmeden basins through the Gao Trench along the flanks of the Adrar des Iforas massif. The name “Ménaka” refers to the large village to the southeast of the Adrar des Iforas mountains, on the far western margin of the Iullemmeden Basin, to the south of localities Mali-7 and Mali-8 (figs. 1D, 2). The Ménaka Formation is defined herein as the sequence of shales and marly limestones, with minor sandstones and phosphate conglomerates, that overlies the Tichet Formation and is overlain by the Teberemt Formation (see below; fig. 6). The basal contact of the Ménaka Formation area typically overlies the large-scale delta foresets of the Tichet Formation, but locally sits above the highest sandstone unit of the Tichet Formation. The top contact of the Ménaka Formation is easily recognized across most of the study area, as it is placed below a silicified limestone bed with distinctive karstic weathering that is regionally traceable. This bed was termed the “Terrecht I” limestone bed by Radier (1959), who also used it as a marker horizon. Using Foraminifera and Ostracoda in the Tichet region of Mali, Bellion et al. (1989), consistent with Radier (1959), placed the K-Pg boundary at the top of what we are now calling the Ménaka Formation, specifically, at the top of the karstic “Terrecht I” limestone.

FIGURE 6.

Outcrops in the northern Iullemmeden Basin illustrating the sedimentology through the middle part of the Maastrichtian Ménaka Formation, locality Mali-7. The 40 m thick succession has yielded ammonite and shark specimens that are critical for biostratigraphy. A, upper part of section with thick bioturbated siltstones (unit 7), and richly fossiliferous, thin, blocky marls and interbedded shale units (units 8–11); B, lower part of the section showing: pronounced, dipping delta foresets, overlain by horizontally bedded delta topsets, then capped by a sequence boundary, and followed by overlying transgressive, deeper water shales and marls.

In the field we observed a weakly karstic to silicified capping limestone bed that we consider to be the “Terrecht I” bed. We have marked the top of “Terrecht I” as the approximate location of the K-Pg boundary in our correlated section (plate 1) following Bellion et al. (1989). The karstic nature of the bed suggests a prior subaerial exposure possibly due to sea-level fall, a characteristic consistent with an unconformity. We have no further data to confirm or refine this hypothesis for the position of this boundary other than the biostratigraphy we discuss below.

Type locality and stratotype section. The type section is located at the base of the large escarpment ∼25 km north of Ménaka at the locality know as Mali-7, which is located at 16° 44.837′N/2° 31.471′E. The lower contact of the Ménaka Formation lies directly above a deeply oxidized, highly bioturbated, horizontally bedded sandstone at the top of the Tichet Formation, and extends to the top of a distinct orange paper shale unit just below the base of the Teberemt Formation. Three reference sections (plate 1) for the Ménaka Formation have been measured and described, including a nearby section (locality Mali-8) a few kilometers to the south of type locality, and two more in the Tilemsi Valley (localities Mali-1 and -3/4).

Description of the sedimentology. The Ménaka Formation is characterized by a heterogeneous succession of paper shales, thin, interbedded shales and marly limestones, thicker limestones, and siltstones and fine sandstones, corresponding respectively to facies 1–4 of Tapanila et al. (2008). The entire formation tends to be recessive in outcrop, particularly the paper shales, with the thin marly limestones and thicker limestones forming benches. Limestones become more abundant and thicker upsection. The paper shales are extremely fissile and preserve abundant mollusk shells (fig. 7), foraminifera, ostracods, shark and other fish teeth, and ray dental plates. The most unusual unit is a thin phosphate conglomerate (fig. 8) with abundant vertebrate bones and coprolites (for further detail see Hill et al., 2008; Claeson et al., 2010). The interbedded shales and marly limestones (mollusk and echinoid packstones; Facies 3 of Tapanila et al., 2008) are often pale yellow in color and highly fossiliferous, although moldic preservation is quite common. Some beds preserve extensive carpets of aligned high-spire turritellid gastropods, which were commonly observed to have their long axes oriented in a subparallel N-S orientation (fig. 9). Large oysters, some forming small reefs, are common, as are more isolated bivalves, echinoderms, and ammonites. Distinctive Thalassinoides burrow networks are common in these facies (examples of this trace fossil from the Teberemt Formation are figured below), along with a range of other indistinct burrows. Significantly, although there are phosphatic marls and shales in the Ménaka Formation, the phosphate conglomerate is the only one observed anywhere in the study area below the K-Pg boundary. A single, well-preserved ammonite was found in place from a thick limestone unit (locality Mali-8) above this bed (Claeson et al., 2010). This limestone has a karstic weathering pattern suggesting subaerial exposure. The Ménaka Formation is nearly twice as thick in the Iullemmeden Basin as it is in the Tilemsi Valley. In the Tilemsi Valley, the section appears to be a bit more condensed (i.e., characterized by slower depositional rates), but there is more limestone and less shale. This formation was not observed in the Gao Trench area. A pale yellow sandy limestone at locality Mali-5, which produced an exquisitely preserved dyrosaurid skull, may be part of the Ménaka Formation. However, this locality displayed little vertical relief making it difficult to correlate laterally with the rest of our section. Because the locality is geographically mapped among the “Pgi” sediments of Bassot et al. (1981), we currently consider it to be age-bracketed as Maastrichtian-Paleocene in age.

FIGURE 7.

Example of a nautiloid steinkern from a limestone unit (locality Mali-8 unit 12, Facies 3) at the top of the Maastrichtian Ménaka Formation.

FIGURE 8.

Fossiliferous phosphate conglomerate bed of the Maastrichtian Ménaka Formation, locality Mali-8, showing a typical assemblage of dyrosaurid crocodyliform and turtle fossils. The fossiliferous phosphate conglomerate rests directly on a well-lithified gastropod coquina (see also CT scan, fig. 9). This is the only such phosphate bed we have identified in the Mesozoic part of our section; others are present in the Paleogene.

Interpretation of the sedimentology. We interpret the Ménaka Formation as representing shallow, normal-to-restricted marine deposits of the Trans-Saharan Seaway, which formed during the T5 cyclothem (the latter is discussed below). We consider most facies to have been formed in a sublittoral environment, with the fossiliferous limestone beds interpreted to have been rapidly buried due to storm reworking, while the phosphate conglomerate bed likely formed due to storm reworking and stratigraphic condensation (Tapanila et al., 2008). We interpret the base of the formation to have been deposited during a period of rapid marine transgression. The phosphate conglomerate bed just noted as concentrated by storm reworking, represents a condensed section, and is associated with a maximum-flooding surface. The upper part of the formation is interpreted to represent a highstand systems tract (fig. 6). We see a distinctive pattern of deepening upward, indicating a sea-level rise and deepening seawaters because we find deltaic facies (of the underlying Tichet Formation) overlain by more offshore muds. As noted above, the gastropod assemblage captures fossil shells oriented in a formation that is subparallel and opposed. Our interpretation of this finding is that waves in water, which flowed roughly perpendicular to the long axis of the shells, created transport alignment of the shells. Our paleocurrent measurements on the shells indicate that at locality Mali-8, waves were moving in N-S or S-N orientation. These types of gastropods are often, although not exclusively, indicators of nearshore, shallow water environments, such as the upper intertidal zone. The presence of in situ oysters also supports this suggestion.

This formation ranges from over 20 m of strata dating from what may be the latest Campanian–early Maastrichtian in the Iullemmeden Basin, to only 10 m of deposits restricted to the late Maastrichtian in the Taoudenit Basin. We interpreted the time-transgressive nature of these deposits based on facies stacking patterns described above. The Gao Trench contains no surface-exposed Ménaka Formation rocks based on our mapping to date.

Discussion of the regional correlations. Moody and Sutcliffe (1993: table 1) summarized different naming schemes for Upper Cretaceous rocks in the Sokoto Basin, Nigeria, and noted that the “Mosasaurus shales” of Niger are the equivalent of the Dukamaje Formation of Nigeria. The Dukamaje Formation of Kogbe et al. (1976: 248) is ascribed to the Rima Group, which we consider a lateral equivalent of the Ménaka Formation, and to have been deposited in swampy waters that were brackish, with occasional interruptions of fully marine water. Petters (1977) described the laterally equivalent Dukamaje Formation of Nigeria as representing two distinct environments: a hypersaline lagoon or estuary and a system of ponds. Petters (1977: 756) also mentioned that the Wurno Formation, also a Ménaka Formation equivalent in Nigeria, showed signs of greater marine influence and might also contain the K-Pg boundary, however, he noted that there was insufficient evidence for this conclusion. Mateer et al. (1992) made important inroads correlating Cretaceous terrestrial rocks in West Africa. They used two sets of terminologies: Terme 1–3 (from Radier, 1959; see also Moody and Sutcliffe, 1993), as well as Taloka, Dukamaje, and Cheit Keini (which we replace herein with the Ménaka Formation) formations from oldest to youngest for what we consider to be Ménaka Formation rocks. We have not recovered mosasaurs in our field area.

FIGURE 9.

CT scan-based reconstruction of fossil shells of Gastropoda, Turritellidae (CNRST-SUNY 114), Turritellinae indet. “C” in a limestone from the Maastrichtian locality, Mali-7 unit 12, of the Ménaka Formation. Specimens present in lithified beds as extensive carpets of aligned high-spire forms that appear to have been rapidly buried, oriented with their long axes in a subparallel north-south orientation. Technical details on the scan are in text. Scale bar = 1 cm and the rock specimen is 2 cm deep (into page).

Age based on biostratigraphy. As discussed by Bellion et al. (1989) the presence of the ammonite Libycoceras crossense in the uppermost beds of what we now call the Ménaka Formation supports the assignment of a Maastrichtian age. Integrating information on foraminiferans from the Tilemsi Valley with these ammonite discoveries, Bellion et al. (1989) and Damotte (1991) noted specifically that the limestone layer called “Terrecht I” by Radier (1953, 1959) is in the Maastrichtian. Additionally, the presence of the elasmobranch Schizorhiza stromeri, which we recovered low in the section, is further indication of a Maastrichtian age for the Ménaka Formation.

Ménaka Formation sediments in the Iullemmeden Basin have yielded particularly numerous index fossils. As noted, the oldest index fossils come from paper shales of Facies 2 at the base of the Ménaka Formation at locality Mali-7 unit 10, and represent sawfish shark teeth identified as Schizorhiza stromeri (Claeson et al., 2015). This shark species occurs widely in, and is restricted to, Maastrichtian deposits in Africa, Iraq, and North and South America (Cappetta, 1987; Wueringer et al., 2009; Claeson et al., 2010). Below the occurrence of Schizorhiza stromeri (see locality Mali-7 in plate 1) we tentatively propose that the beds may extend into the Campanian following Moody and Sutcliffe (1991) and lithostratigraphic correlations with Nigeria (Petters, 1979; Moody and Sutcliffe, 1993).

At the same locality that yielded Schizorhiza stromeri, in the overlying succession of interbedded marls and shales (Unit 11, which represents Facies 4), we collected samples of the index fossil mollusk, Trigonarca sp. The genus Trigonarca is known only from the Cretaceous (Newell, 1969). Slightly higher upsection from a chalky marl at the top of unit 11 at locality Mali-7, we also collected the regular echinoid, Echinotiara perebaskinei. Lambert and Perebaskine (1930) previously reported this taxon from the Gao region, and more recently Smith and Jeffery (2000) described it as distributed throughout Africa and the Arabian Peninsula during the Maastrichtian only.

The next-youngest index fossil, the ammonite Libycoceras, comes from Ménaka Formation rocks in both the Taoudenit Basin, locality Mali-1, again, from a paper shale of Facies 2 near the top of the section, and from the Iullemmeden Basin localities Mali-7 and -8. The specimen, which was collected in situ at Mali-8, from a massive limestone (Facies 3, unit 12), is the only one with sutures well preserved enough to make a species-level identification as Libycoceras crossense. This specimen is illustrated in Claeson et al. (2010: fig. 2 [the taxonomic name in that paper was misspelled]). The genus Libycoceras ranges from the Campanian through the Maastrichtian, and the species Libycoceras crossense ranges from the late Campanian through middle–?late Maastrichtian (Zaborski and Morris, 1999). Finally, the Ménaka Formation in the Iullemmeden Basin section also yielded a tooth from the vertebrate Maastrichtian index fossil, Cretalamna (Serratolamna. maroccana, a lamniform shark (Case and Cappetta, 1997; Shimada, 2007; Claeson et al., 2010) that we collected from shale beds (Facies 3) very high in the section (Mali-7 unit 14). Cretalamna is found in both Maastrichtian and Danian rocks, but is much rarer from Maastrichtian ones (Belben et al., 2017). Its presence in the section at Mali-7 suggests the rocks are Maastrichtian in age or possibly younger. In previous publications (e.g., Hill et al., 2008), we identified “Turritella sp.” as an index fossil of the Ménaka Formation, however, based on our taxonomic revisions here we now recognize these specimens only as Turritellidae indet. with several genera likely represented (see Systematic Paleontology). The presence of Turritellidae, a clade that appeared in the Cretaceous and remains extant today (Allmon, 2011), provides only the broadest biostratigraphic indicator that the Ménaka Formation is at most Cretaceous in age. Although the local section at Mali-3/4 in the Taoudenit Basin has 30 m of Ménaka Formation rocks, it yielded neither index fossils nor other invertebrate or vertebrate macrofossils.

Teberemt Formation: The name Teberemt Formation was suggested by Moody and Sutcliffe (1991) for the heterogeneous succession of open marine shales and carbonates that comprises the Paleocene across Mali (fig. 10). These deposits were originally described by Radier (1953) and mapped as part of the “PGs” Paleogene by Bassot et al. (1981) as a continuous swath across the Taoudenit and Iullemmeden basins, through the Gao Trench, along the flanks of the Adrar des Iforas massif. Moody and Sutcliffe (1993) provided a composite graphic log for Mali and northern Nigeria showing the basic stratigraphy of the Teberemt Formation, but neither the rationale for the name “Teberemt” nor a type locality was provided in either of their publications (Moody and Sutcliffe, 1991, 1993). The stratigraphy they described is generally consistent with what we observed for this interval; however, Moody and Sutcliffe (1991, 1993) did not adequately demonstrate the range of lithofacies and the lithological complexity associated with this unit. Rather than establish a new name for this stratigraphic unit, we have opted to designate and describe a lectostratotype section and retain the existing name. The Teberemt Formation is defined herein as the ∼12–17 m thick heterogeneous succession of carbonates, shales, and phosphates that overlie the top of the Ménaka Formation and sit below the overlying Tamaguélelt Formation (defined below). The basal contact of the Teberemt Formation is defined by an extensive 1–5 m thick limestone bed that lies disconformably atop the Ménaka Formation. The upper contact of the formation is distinguished by a similar 1–5 m thick limestone bed.

FIGURE 10.

Paleocene Teberemt Formation near Samit at the Mali-17 locality, which contains numerous vertebrate fossils. Exposure of the upper part of the Teberemt Formation (∼10 m thick) capped by indurated limestone (also known as “Terrecht II”; here formally considered part of the Teberemt Formation) that lies just below the Paleocene-Eocene boundary. Black arrow indicates approximate stratigraphic location of turtle fossil shown in figure 11. Person in image denotes scale.

Type locality and lectostratotype section. We have documented a complete lectostratotype section through the Teberemt Formation at Mali-18 in the Gao Trench, which we herein designate as the type area. In addition, we have identified and measured a second reference section in the Gao Trench at locality Mali-19, along with two partial reference sections in the Iullemmeden Basin at Mali-7 and Mali-10, and one additional reference section in the Taoudenit Basin at Mali-3/4. The lectostratotype section at Mali-18 can be observed at N 16°49′36.0″, E 1°04′14.9″.

Description of the sedimentology. The Teberemt Formation is observed across the entire study area, characterized by a high level of facies change in and between sections. The base of the formation is easily identified by the karstic, weathering limestone bed referred to as the “Terrecht I” bed by Radier (1959). This bed preserves abundant echinoids and oysters. As noted above, an exquisitely preserved dyrosaurid crocodyliform skull found at Mali-5 (Brochu et al., 2002) may be from this bed but is now of uncertain placement in the section and is possibly from the Maastrichtian Ménaka Formation. Above this bed, the section is dominated by interbedded shales, carbonates, and phosphate conglomerates, with the latter facies limited to the Gao Trench area. The Teberemt Formation forms a series of recessive slopes made up of shales and distinctive benches, which are capped by limestones. An incredibly abundant fauna has been recovered from this formation, including numerous vertebrates, such as fish, crocodyliforms, and snakes from the phosphate conglomerates and an abundant invertebrate fauna from the shales and carbonates, dominated by echinoids, oysters, nautiloids, gastropods, and foraminiferans (Gaffney et al., 2007; Hill et al., 2008; McCartney et al., 2018). Interesting taphonomic situations are preserved, including turtle skeletons encrusted with ostreidan bivalves (fig. 11). Fossil wood is also present and trace fossils abound, particularly well-developed Thalassinoides burrow networks. At the top of the section, and included in the Teberemt Formation, is a distinctive karstic limestone bed, which Radier (1959) referred to as the Terrecht II bed. This bed forms a capping bench across much of the region.

Interpretation of the sedimentology. The Teberemt Formation represents shallow, normal to restricted marine deposition within the Trans-Saharan Seaway. Deposition possibly transpired during the sea level change of the T6 Cyclothem (discussed below). The cyclic nature of the deposits and repeating nature of the phosphate conglomerate beds in this interval was interpreted by Tapanila et al. (2008: see also Sequence Stratigraphy section below) to represent portions of three retrogradational parasequences. As with the Ménaka Formation, the phosphate conglomerate beds are considered to have formed as storm-reworked condensed sections associated with periods of maximum transgression of the seaway.

Discussion of the sedimentology. A depositional hiatus has been suggested by the lack of Danian taxa found in Nigeria and elsewhere at the base of the Teberemt Formation (Bellion et al., 1989; Damotte, 1991). Petters (1977: 756) described the Dange, Kalambaina, and Gamba formations of Nigeria (sometimes called the “Sokoto Group”), which we interpret as the lateral equivalents of the Ménaka Formation, as indicative of a marine transgressive environment. Regarding the Dange Formation in particular, Petters (1977) interpreted it as estuarine, citing as evidence a faunal mixture of marine and freshwater fishes. He interpreted the Kalambaina rocks as indicating a warm, more open, shallow sea, and that the Gamba Formation represented a return to more estuarine conditions.

Age based on biostratigraphy. As noted above, our own microfossil analysis yielded the dinocyst Spiniferites indicating that the Teberemt Formation is at most Cretaceous. The exact transition between the Ménaka Formation and the younger Teberemt Formation is imprecise in our field area and likely represents an unconformity. Four of our local sections, Mali-1 and Mali-3/4 in the Iullemmeden Basin, and Mali-7 and Mali-8 in the Taoudenit Basin, nonetheless, extend through this boundary and contain rocks of both the Ménaka Formation and the overlying Teberemt Formation.

FIGURE 11.

Fossilized bothremydid turtle skeleton, disarticulated, but in situ, from the Paleocene Teberemt Formation at the Mali-17 locality near Samit, Mali. Note encrustation of vertebrae and other bones by ostreidan bivalves. Color and contrast of bones has been enhanced.

Perhaps the most complex issue pertaining to the Teberemt Formation geology is the placement of the K-Pg boundary and assessment of the hypothesis that Danian rocks are absent in this section. Bellion et al. (1989) reported, based on work in the Malian region of Tichet, that their stratigraphically lowest Paleocene microfossil samples yielded two species of the planktonic foraminiferan Globorotalia, indicating the beds were Paleocene and specifically Montian (“Zone P2”). These Montian units are stratigraphically superposed on the “Terrecht I” limestone (discussed above as Maastrichtian) below and extend all the way to a capping “Terrecht II” limestone (Radier, 1953; 1959). Bellion et al. (1989; see also Damotte, 1991) concluded that Danian deposits are missing and that there is a regional gap in sedimentation in the earliest Paleocene (also referred to as a “Lower Paleocene gap” [Dikouma et al., 1993: 105]), as evidenced in part by desiccation cracks at the top of Terrecht I, which, as noted above, we corroborated. The bed called “Terrecht II” by Radier (1953, 1959), which is part of the Teberemt Formation, has been dated as Paleocene by the benthic foraminiferan Ranikothalia bermudezi (Bellion et al., 1989, 1990; Damotte, 1991; also discussed in O'Leary et al., 2006).

The Teberemt Formation rocks, including those at the base of the sections in the Gao Trench, contain index fossils in their lowest beds. The limestone beds (Facies 3) at the base of the Teberemt Formation in all three field areas contain specimens of the echinoid Oriolampas michelini. In their review, Smith and Jeffery (2000) documented O . michelini from Paleocene (typically Thanetian) deposits throughout West Africa (i.e., Nigeria, Niger, Sudan, Algeria, and Mali) and from the French Pyrenees. At Mali-7 this echinoid species is found approximately 10 m higher than the specimen of Cretalamna maroccana noted above. Oriolampas michelini also persists in younger beds, again in limestones (Facies 3), in the Teberemt Formation. The Gao Trench contains the most diverse set of index fossils for the Teberemt Formation. Two nautiloids contribute biostratigraphic information: Cimomia reymenti from Mali-17 shales (Unit 3, Facies 2) and Cimomia ogbei from younger Mali-18 limestones (Unit 8, Facies 3). These species have Maastrichtian and Paleocene ranges (Adegoke, 1977). Linthia sudanensis, which is found in limestones of both the Gao Trench (at Mali-16, -17, -18 and -19) and the Iullemmeden Basin (at Mali-7), has been attributed an early-late Paleocene age with a questionable Maastrichtian occurrence (Smith and Jeffery, 2000). Although in our prior work (e.g., Hill et al., 2008) we identified specimens of the mollusk Ostrea multicostata from limestones (Facies 3) of several Gao Trench localities (Mali-16, -17, and -18), in our revised identifications here we find that these specimens are actually of more ambiguous taxonomic attribution (i.e., no more specific than Ostreidae) and do not contribute to the biostratigraphy we propose.

Our observations and research are consistent with the inference of an unconformity, if only due to the negative evidence that we have recovered no macrofossils that are clear biostratigraphic indicators of the Danian. We were not, however, able to collect mollusks in a sufficiently systematic way to perform a thorough search for a faunal signature indicative of the Danian. One such pattern was reported by Hansen (1988: 40), albeit in relatively offshore environments, who observed that “Early Paleocene faunas have a much higher percentage of species in the Carditidae, Ostreidae and Turritellidae than is seen in the Late Cretaceous and the remainder of the Paleocene.” Nor are our mollusk species identifications refined enough to compare with species-level studies of Danian mollusks at other latitudes such as that of Cope et al. (2005) in rocks of southern Illinois or that of Río and Martínez (2015) in Patagonia, Argentina.

Likewise, for echinoids, Smith and Jeffery (1998) reported that a number of echinoid clades made their last appearance in the Danian before extinction. Danian occurrences of echinoids are generally rare from North Africa. However, the foraminiferan-dated Danian Upper Tar Member of the Zimam Formation in the Hamadah Basin of Libya preserves echinoids Hemiaster nucleus. H . frontacutus, and Echinobrissus pannonicus in shallow marine, carbonate ramp facies (Megerisi and Mamgain, 1980; Hallett and Clark-Lowes, 2016). In Egypt, Danian shales and marls that compose the upper Dakhla Formation (previously called the Blätterthone Formation) preserve the echinoids ?Coelopleurus fourtaui, ?Micraster sp ., Hemiaster chargensis, and Cidaris (Tantawy et al., 2001; Quaas, 1902). We can report that none of these genera, let alone species, are found in the Malian sections despite generally good preservation and abundance of echinoids collected during our expeditions. Finally, no mammal fossils have been recovered from Paleocene rocks of Mali, and no other vertebrate fossils currently serve as indices for the Paleocene in Malian sediments.

Tamaguélelt Formation: Radier (1959) originally referred to the deposits overlying the Paleocene carbonates (Teberemt Formation) as the “Schistes Papyreuses”; however, Moody and Sutcliffe (1991) expanded on the description of these deposits to include the phosphates and siltstones that accompany the paper shales in many places, calling this succession the Tamaguélelt Formation. The origin of the name “Tamaguélelt” was not provided, nor was a type locality designated. Moody and Sutcliffe (1993) provided a composite graphic log for Mali and northern Nigeria showing the basic stratigraphy of the Tamaguélelt Formation. The stratigraphy those authors described is generally consistent with what has been observed by us for this interval; however, their work does not adequately demonstrate the range of lithofacies and the lithological complexity associated with this unit. Rather than establish a new name for this stratigraphic unit, we have opted to designate and describe a lectostratotype section. The Tamaguélelt Formation is defined herein as the ∼12–17 m thick heterogeneous succession of carbonates, shales, and phosphates that lies above the top of the Teberemt Formation and below the overlying Continental Terminal (fig. 12). The basal contact of the Tamaguélelt Formation is easily recognized as it sits just above the karstic, weathered “Terrecht II” limestone unit. The upper contact of the formation is distinguished by ferruginized sandstones at the base of what are informally called the Continental Terminal sediments.

Type locality and lectostratotype section. We have documented a complete lectostratotype section through the Tamaguélelt Formation at Mali-20 in the Tilemsi Valley (Taoudenit Basin), which we herein designate as the type area. In addition, we have identified and measured a second reference section in the Gao Trench at Mali-16. The lectostratotype section at Mali-20 can be observed at N 17°35′35.3″, E 0°14′59.7″.

Description of the sedimentology. The Tamaguélelt Formation is dominated by a succession of paper shales and siltstones, punctuated by several phosphate conglomerates of widely varying thickness, as well as by rare sandstone beds (fig. 12). The paper shales are mostly unfossiliferous, although Radier (1959) did report a bivalve from these beds. The paper shales also have a much more orange to yellow color than those of the Ménaka or Teberemt formations, and often contain iron oxide nodules. Bioturbation in the formation is largely limited to the thin siltstone beds, which are abundantly bioturbated. The top of the formation is characteristically black in color with heavy iron- and manganese-oxide concretion development. The lowest phosphate conglomerate in the formation is the thickest and most richly fossiliferous unit in the study area, preserving a remarkable abundance of isolated vertebrate bones, including rare mammal fossils. This bed was described in detail in several papers by our team (Tapanila et al., 2004; O'Leary et al., 2006; Tapanila et al., 2008; McCartney et al., 2018).

Interpretation and discussion of the sedimentology. The Tamaguélelt Formation appears to rest conformably on the Teberemt Formation based on our observations, and we placed the contact at the top of the regionally distinctive limestone marker horizon known as the “Terrecht II” bed. We interpret the Tamaguélelt Formation as having been deposited in a shallow, and slowly filling, marine to possibly brackish embayment of the Trans-Saharan Seaway during the early to middle Eocene, and to lie conformably over the Teberemt Formation. Tapanila et al. (2008) interpreted this unit to represent the final pulse of the Trans-Saharan Seaway in Mali during the highstand-systems tract of the T6 cyclothem (discussed below). The phosphate beds present in this formation, which are geographically widespread and remain to be mapped in full, may be among the most extensive macrofossil bone beds known on Earth based on our comparisons with other deposits (Tapanila et al., 2008). The presence of mammals that may have had an amphibious but not fully marine lifestyle (O'Leary et al., 2006; Sanders et al., 2010) attests to the nearshore nature of these deposits, while the range of taxa present in the phosphate conglomerates encompasses marine to possibly brackish water settings.

Age based on biostratigraphy. The phosphate deposits of the Tamaguélelt Formation do not preserve calcareous shells (discussed more below in the section “Fossiliferous Phosphate Facies”), and invertebrates from other layers have not been informative for biostratigraphy. The Tamaguélelt Formation has been somewhat tenuously dated as Eocene based on vertebrate fossils. We consider the phosphate conglomerates 3 m above “Terrecht II” limestones the oldest Eocene beds in our section (Tapanila et al., 2008).

We do not repeat here the extensive discussion by Patterson and Longbottom (1989) of the challenges of dating this formation using biostratigraphy. Those authors reached the conclusion that the formation is either Ypresian or Lutetian in age. In Longbottom's (1984: 23) review of Malian pycnodonts she cautioned that the clade is “not useful for dating purposes.” Longbottom (1984: 23) also wrote that Pycnodus bowerbanki, and two other pycnodont species she identified in Mali– P . variabilis and P . munieri, “are known elsewhere from the Landenian to the Lutetian,” suggesting the faunal signature was consistent with an early Eocene age for the phosphates, but one that brought no great precision to the age of these beds. This pycnodont-based evidence for a Ypresian age for the Tilemsi Valley Tamaguélelt sediments was repeated by Moody and Sutcliffe (1993).

FIGURE 12.

Type section of the Tamaguélelt Formation at locality Mali-20, southeastern edge of Tilemsi Valley, near the transition between the Taoudenit Basin and the Gao Trench. A, view from the top of the formation to the field camp down the largely recessive, shale-dominated succession (vehicles indicate scale); B, multiple layers of phosphatic shales and phosphate conglomerates (black arrows) interbedded with marls and shale (yellow arrows); and C, deflation surface at the top of a phosphate conglomerate revealing rich assemblage of disarticulated phosphatic pebbles, vertebrate bones and coprolites.

Although mammals from this formation are certainly too fragmentary to serve as index fossils, we note here that Sanders et al. (2010) assigned the proboscidean fossils from Mali (more discussion on this below) to “Plesielephantiformes” incertae sedis and noted that the Malian fossils resemble phosphatheres and daouitheres, both of which are early Eocene clades. O'Leary et al. (2006) reported that the lowest bone-producing bed in this formation is only 3 m above the Paleocene “Terrecht II” limestone layer (described above in the Teberemt Formation) and concluded that the combination of sediment thickness and vertebrate fossils found continue to suggest an early Eocene age.

Post-Eocene “Continental Terminal” Deposits: We did not collect thoroughly in sediments that were younger than Eocene, as our reconnaissance in such rocks yielded only a few invertebrate fossils. The informal term “Continental Terminal” was proposed by Kilian (1931) and reviewed by Lang et al. (1990) who emphasized that this detrital unit of Cenozoic rocks is often poorly dated. The deposits consist of ferruginous oolitic sandstones, oxidized siltstones, and dark red mudstones that span the Eocene to Pliocene and contain few fossils (Kilian, 1931; Lang et al., 1990). We have observed that these deposits are significantly altered by laterite pedogenesis, obfuscating original sedimentary structures and lithology, and resulting in probable leaching of biominerals (carbonate) from fossil shells.

Synthesis of Sedimentology and Sequence Stratigraphy

Trans-Saharan Seaway sediments were deposited as part of a carbonate platform with rocks that precipitated in situ, as well as significant siliciclastic sediments that eroded from adjacent uplifted rocks. Sedimentary sequences across the entire area the sea covered are sands, clays, shales, and limestones characterized by broad similarities in regional stratigraphy and sedimentology (see Radier, 1959; Greigert, 1966; Petters, 1977; 1979). Our close inspection has revealed that, nonetheless, significant sedimentological variations in thickness, lateral and vertical facies stacking patterns, and depositional environments exist across the Taoudenit Basin, the Gao Trench, and the Iullemmeden Basin, and that these had not been characterized in detail prior to our work (e.g., O'Leary et al., 2006; Hill et al., 2008; Tapanila et al., 2008).

Fundamental to our correlation among these three geographic regions has been detailed facies analysis (Tapanila et al., 2008) revealing the presence of five repeated facies and strata that occur in fining-upward sequences (table 2; figs. 13, 14) tracking transgressive-regressive cycles of the Trans-Saharan Seaway (Moody and Sutcliffe, 1993). In our model, as seas retreated after a transgression, exposed land was then eroded. The coastal plain expanded and contracted during four to six cyclothems, leaving sequences of marine and nonmarine sediments. We have interpreted these sequences as representing a variety of ancient nearshore continental, marginal marine, and open marine environments (Tapanila et al., 2008), interpretations that have informed our paleobiological interpretations of fossils collected (Tapanila et al., 2004; Gaffney et al., 2007; Hill et al., 2008; Tapanila et al., 2008). These interpretations are consistent with the hypothesis that the Trans-Saharan Seaway was likely never deeper than 50 m (Krasheninnikov and Trofimov, 1969; Berggren, 1974; Tintant et al., 2001).

TABLE 2

Facies observed in Late Cretaceous-Eocene sediments of northern Mali and an interpretation of the ancient nearshore marine environments they represented (see Tapanila et al., 2008).

To discuss our model in detail we use the Iullemmeden Basin section at locality Mali-8 (plate 1) and an idealized version of facies change (fig. 14). The Mali-8 locality consists of approximately 35 m of section deposited along the Trans-Saharan Seaway as part of at least one complete sea-level cycle (T5 Cyclothem of Greigert, 1966), as we have discussed (Tapanila et al., 2008). The base of the section is defined by Facies 1 and its large, inclined, fine-grained sandstone beds with glauconite and petrified wood of delta-front origin (fig. 4). The next 20 m represents deposits of the T5 transgressive systems tract, dominated by thinly laminated shales and marls of facies 2 and 3, which are overlain by phosphatic limestones and the distinctive phosphate conglomerate (Facies 5) representing the maximum flooding surface (condensed section) of the T5 cyclothem. Above that level are open marine, quiet water mudstones and wackestones (Facies 4), suggesting that ammonites found in them are in situ. These beds are followed by storm-generated mollusk and echinoderm packstones (Facies 3) associated with the overlying T5 highstand-systems tract (Tapanila et al., 2008: fig. 3).

This section documents a repeated fining/ deepening-upward sequence overlain by extensive bone/coprolite phosphate conglomerates. Phosphate conglomerates at the base of fining-upward sequences represent syndepositional phosphogenesis tied to marine transgression . We suggest that these represent maximum flooding surfaces, and their presence has been important for defining sea-level cyclicity and reconstructing paleogeography. We have established the genetic nature of these surfaces, i.e., that the surfaces formed due to storm concentration in an epeiric setting rather than by the model of upwelling at a continental margin. They could therefore be used, in conjunction with detailed sedimentological investigation of other surfaces and facies, to understand the nature, frequency, and timing of Paleogene sea-level cyclicity (3rd–5th order) in the Trans-Saharan Seaway in Mali (further discussion on this in the GPlates section below). Finally, identifying genetic surfaces has facilitated the correlation of thin, repeated strata across our field area.

Fossiliferous Phosphate Facies

Many of the vertebrate fossil specimens we have described from sites throughout our study area (e.g., at localities Mali-8, -16, -18, -19, and -20) come from Facies 5, the bone, coprolite, and pebble phosphatic conglomerate that ranges in age from the Maastrichtian through Eocene (O'Leary et al., 2006; Hill et al., 2008; Claeson et al., 2010; Hill et al., 2015). This previously unclassified variety of phosphate conglomerate (P₂O₅ content >18%) has been the most intensively studied by us (Tapanila et al., 2008), due to its important bearing on sedimentological and paleontological questions. These strata are separate, decimeter-scale conglomerates that may extend laterally for kilometers and are most common in the post-Cretaceous section (plate 1). They are clast supported and range from unconsolidated to weakly cemented (plate 1; see also fig. 12). Clasts are predominantly granule to cobble sized, but also include pieces of vertebrate bone as large as 25 cm. Bioclast shape and size analysis demonstrated a sorting in the thickest phosphatic deposit: Mali-20 unit 2 (also called “Mali-20a” [Tapanila et al., 2008; McCartney et al., 2018]).

FIGURE 13.

The Trans-Saharan Seaway in Mali, interpreted using sequence stratigraphy and our generalized facies model. A, repeating facies changes in our composite section and the time span (vertical lines) of particular localities in Mali; and B, geographic distribution of those localities in northwestern Mali. Sandstone and phosphate deposits probably formed at different times. During regressions (low sea level), sandstones were deposited on the coast, and during transgressions (rising sea level) at maximum sea level, phosphate conglomerates formed subaqueously in shallow marine areas. Horizontal dashed lines indicate the tops of cyclothems (sequence boundaries), dashed arrows indicate shallowing upward parasequences, and solid arrows indicate sets of parasequences. Dots indicate likely nondeposition during the Early Paleocene (Danian).

FIGURE 14.

Our carbonate platform model for the Trans-Saharan Seaway in Mali showing one complete and idealized cyclothem through time and the cumulative deposits. Nearshore siliciclastic rocks weathered from the Adrar des Iforas massif at the proximal (shoreward) part of the basin and fine-grained carbonate rocks accumulated in situ primarily from calcareous microorganisms. A, regressive deposits formed during late highstand phase as coastline prograded basinward in large delta complexes, depositing thick siliciclastic strata. A sequence boundary associated with sea level fall marks the top of this systems tract; deposits of lowstand systems tracts are conspicuously absent from the basin. B, early transgressive deposits; sea level rises faster than siliciclastic sediment is produced, coastline moves furthest inland and fine-grained shales, thin carbonates and phosphate deposits accumulate. C, late transgressive deposits; sea-level rise at maximum, shutting off siliciclastic sediments to basin; carbonate production is high, and thick limestone and marl deposits accumulate basinward. In restricted portions of basin, thick phosphate conglomerates form (not shown). D, one complete cyclothem. We do not observe rocks representing environments such as reefs, sand distal from shore or very deep ocean sediments and hypothesize that these were absent in much of the study area. Illustration vertically exaggerated to show facies relations and surfaces.

It is noteworthy that coprolites, bones, and bored phosphatic limestone cobbles are the most common constituents of the conglomerate, but that calcareous shelled invertebrates are almost completely absent as body fossils. In one locality studied in detail (again, Mali-20), clasts over 1 cm varied as follows: 27% bone, 20% identifiable coprolite, and 53% indeterminate phosphatic clasts that are likely weathered coprolites (Tapanila et al., 2008). The vertebrate bone consists entirely of disassociated anatomical elements from a range of fossil species that were likely marine to nearshore marine (in order of frequency of occurrence): siluriform catfish, amiids (Maliamia), rays, pycnodonts, lungfish, and osteoglossiforms (Brychaetus). Mali-20 is also the locality that has yielded rare paenungulate mammal fossils (O'Leary et al., 2006).

Teeth, except myliobatid ray and pycnodont tooth plates and a rare isolated mammal tooth, are also almost completely absent from this deposit, a phenomenon that we have attributed to chemical dissolution during phosphogenesis (Tapanila et al., 2008). Indirect lines of evidence, however, do point to the presence of tooth-bearing vertebrates (e.g., dyrosaurids, amiids) and calcareous invertebrates in the ancient ecosystem. Evidence of toothed vertebrates exists through the presence of numerous jaws with empty tooth sockets (shown below; see also O'Leary et al., 2006; Tapanila et al., 2008). Likewise, evidence of calcareous-shelled invertebrates exists through numerous occurrences of the ichnofossil Gastrochaenolites ornatus (see below), a boring into lithoclasts, coprolites, and vertebrate bone, some of which even preserve the external mold of pholadid bivalves (Tapanila et al., 2008).

We have developed a model describing how this peculiar deposit formed. Typically, phosphates form under select circumstances requiring a high threshold concentration of phosphate minerals in shallow marine sediments. The source of the phosphate is generally bone or feces, particularly from carnivorous vertebrates, or plankton with phosphatic tests. These elements are subject to decay, and high concentrations of bone, feces, and plankton tests are required to generate bedded phosphates over any appreciable stratigraphic scale. Tapanila et al. (2008) hypothesized that the particular Malian conglomerate formed in a depositional environment with shallow marine to brackish water that lay between fair weather and storm-weather wave base. We interpreted our discovery of a pholadid bivalve, a taxon that requires water of normal marine salinity, as paleobiological evidence that this type of water must have characterized the ancient environment (Tapanila et al., 2004). We presented evidence that the conglomerate represents a long-lived sedimentary deposit that formed under unusual chemical circumstances allowing phosphate-enriched materials (e.g., bone and coprolites containing bone fragments) to be preferentially mineralized on the seafloor. Once lithified, the fossil clasts were concentrated by wave action to form a gravel substrate on the seafloor that was colonized occasionally by rock-boring clams. The assemblage of fossils and coprolites in these deposits, therefore, represents a concentrated (i.e., time-averaged) temporal sample of marine life. For the deposition of these phosphate beds to occur at our localities in Mali, we hypothesized that several variables needed to align (Tapanila et al., 2008):

Other Noteworthy Fossiliferous Facies

In addition to the phosphate conglomerates, which have produced the greatest volume and diversity of fossil vertebrates from the Trans-Saharan Seaway region of Mali, a number of other facies have been productive for fossil exploration (plate 1; figs. 13, 14). In the Upper Cretaceous-Paleogene sections of all three study regions in Mali, the limestones and shales of facies 4 and 2, respectively, have produced numerous vertebrate and invertebrate fossils. In contrast to the phosphatic facies, the limestones and shales yield the best-preserved specimens, such as numerous intact vertebrate skulls and associated body fossils. Species preserved from Facies 4 include a range of marine vertebrates, from isolated shark teeth to nearly complete dyrosaurid crocodyliform skulls (Brochu et al., 2002) and multiple associated to completely articulated turtles (Gaffney et al., 2001; Gaffney et al., 2007). Moreover, these units have also yielded diverse and well-preserved invertebrate assemblages across the basins, dominated by echinoderms, nautiloids, ammonites, gastropods, bivalves, and oyster beds.

Notes on Taphonomy and Specimen Collection

Specimen Collection. The expeditionary work was restricted to short stays (2–3 days) at each outcrop due to security and proximity to support services. Major priorities were the removal of significant vertebrate fossils and the immediate documentation of the stratigraphic context, including the collection of samples relevant to biostratigraphy. For the collection of vertebrate fossils, we prioritized rare, unusual, and complete specimens. Specimens were almost exclusively surface collected and in some cases were retrieved still entombed in substantial amounts of rock matrix.

With regard to mollusks, about 350 specimens were collected from 10 of the localities that span the Late Cretaceous to Eocene. This collection consisted of a representative sample of the taxa present at each outcrop (not a bulk or systematic collection), thus the rarity or commonness of taxa cannot be inferred. Generally, we collected the best-preserved specimens we found. The mollusk fauna is often preserved as internal molds, occasionally as external casts and sometimes as the shell itself or a combination of these. Specimens are weathered and in poor condition and occasionally compressed and distorted. Nonetheless, family and sometimes generic and species identifications have been possible (see below). Most bivalves are articulated, suggesting rapid burial after death. Because interior shell structures such as hinges, teeth, and muscle scars have not been examined for articulated specimens, many bivalve identifications are to the family level only.

GPlates Reconstruction of the Trans-Saharan Seaway

The translation of field-based geological observations into maps of the Trans-Saharan Seaway has been attempted for over 50 years by numerous authors mentioned in the Introduction who created these maps by hand (e.g., Adeleye, 1975; Kogbe et al., 1976; Petters, 1977; 1979; Reyment, 1979; 1980; Kogbe, 1981; Boudouresque et al., 1982; Adetunji and Kogbe, 1986; Reyment and Dingle, 1987; Bellion et al., 1989; Kogbe, 1991; Damotte, 1995). Intervals of maximum transgression may have lasted only 10,000 years (Reyment, 1986), and whether the sea constituted a full north-south connection at different times has been debated by those same authors.

In Tapanila et al. (2008) we presented our hypothesis that six cyclothems (T1–T6) characterized the sedimentary record of northern Mali from the Late Cretaceous through the Early Paleogene. A cyclothem is a full cycle of sea-level rise followed by a disconformity marking a rapid sea-level fall. These events would have been third-order eustatic changes that would likely have been global in nature. To create an improved estimate of the extent of the Trans-Saharan Seaway through time we here apply GPlates, a computer-generated plate tectonic model (Boyden et al., 2011), to build maps that show the coverage of the Trans-Saharan Seaway from the Late Cretaceous to the middle Eocene. We manipulated the GPlates models for several time periods to reflect the data we have collected ourselves or integrated from generalized regional literature for central West Africa.

Methods: Within the open-source software GPlates, we consider a dataset of rotation and geometry files derived from information in several integrative research papers on tectonics (Seton et al., 2012; Matthews et al., 2016; Müller et al., 2016) to first establish a robust estimate of the positions of the continents at specific time points. These background data points provide up-to-date global plate tectonic reconstructions for the last 400 million years. With this data we used GPlates to model the positions of the present-day coastline and continental shelves, as well as a rotation spreadsheet that records the position and time-related rotation pole for each feature (e.g., continents) in the shape files (Matthews et al., 2016). Each feature is assigned a plate ID and each plate ID moves at the assigned rotation for a specified period. Any feature created and labeled with the same plate ID as West Africa will travel on the same tectonic rotation as West Africa and there is no relative movement between features with the same plate ID. The location of Africa at each age interval in figure 15 is derived from the research in Müller et al. (2016) on global plate boundaries and movements.

FIGURE 15

Our GPlates-based paleogeographic reconstructions of the Trans-Saharan Seaway in West Africa from the Late Cretaceous–early Paleogene (Boyden et al., 2011). Shorelines during six (T1–T6) cyclothems, or alternating marine and nonmarine sedimentary sequences, as described in the text. (opposite page) A, Turonian, modeled from Kogbe (1981), Reyment and Dingle (1987), and Meister and Piuz (2013); B, Coniacian, modeled from Reyment (1980) and Reyment and Dingle (1987); C, Maastrichtian, modeled from Guiraud et al. (2005) with minor additions from Adetunji and Kogbe (1986); D, Maastrichtian-Danian boundary, modeled from (Kogbe, 1981), Bellion et al. (1989) and Damotte (1995); E, Danian, modeled from Guiraud et al. (2005), with minor additions from Reyment (1980), and our new data; F, Eocene, based on information from Guiraud et al. (2005); and (above) G, Recent. The field area in Mali is situated within the tropics from the Turonian to the present day. Black star = locality Mali-17. SA = South America.

The modeling of the Trans-Saharan Seaway integrated data, such as facies analysis and mapping and localities of marine fossils, based on our own field observations from Mali and on publications that we cite below. We developed different methods for different time intervals to maximize the accuracy of the model at each of six intervals. Different models were necessary because some publications that focused on particular intervals provided better reference frameworks than others, which, in turn, could lead to more precise reconstruction in certain cases.

For example, where possible, the Trans-Saharan Seaway was modeled using data from Guiraud et al. (2005) because the paleogeological maps of West Africa those authors provided included a grid of lines of latitude and longitude every 5°. By matching the graticules, or grid squares, in GPlates to the same grid size of Guiraud et al. (2005), and drawing square by square, we aimed to produce a more accurate representation of the maximum transgression at a given time. This approach was not, however, possible for three of the six time periods. Thus, with an approach similar to the grid method we followed with the Guiraud et al. (2005) data, we created our GPlates model of the Trans-Saharan Seaway for these three periods using the seaway location within the borders of different countries as described by Kogbe (1981) and Damotte (1995). The latter method meant obtaining x-y coordinates of different points describing the seaway based on the adjacent coastline map from Reyment (1980) and Reyment and Dingle (1987). Based on our observations of marine fossils from several locations in northeastern Mali, we made small additions to the reconstruction of the Trans-Saharan Seaway for the Maastrichtian and into the Paleocene. Once the location of the seaway was modeled for each of the six time intervals, the seaway was assigned the same plate ID as West Africa in the GPlates program. Using GPlates, we then took snapshots of each of the different ages (fig. 15) to indicate the location of the seaway superimposed over the paleogeographic location of West Africa. Because the underlying data for this GPlates assessment are scaled to periods and epochs rather than to a numerical timescale, we carry this same temporal framework into our results (i.e., we do not provide numerical ages for the maps).

Results and Discussion: The results from GPlates generally resemble those in previously published maps, however; our new snapshot maps provide improved reconstructions for Mali, one of the most inland reaches of the ancient sea (fig. 15). Our interpretation varies slightly from prior reconstructions (Reyment, 1979; 1980; Boudouresque et al., 1982). Beginning in the Turonian, our maps show the transgressions starting out on a more easterly course through West Africa. Later, due to the impact of tectonics, the seas traveled west of the central Precambrian massifs. Finally, the northeasterly arm of the sea returned in the Paleocene. Our locality Mali-17 (starred on fig. 15) is noteworthy in being one of the most geographically internal marine fossil-producing localities in West Africa at this time. Intervals of maximum transgression are hypothesized to have lasted approximately 10,000 years (Reyment, 1986).

T1 Cyclothem: Cenomanian-Turonian (fig. 15A). Reyment (1979; 1980) argued that the first of the transgressions resulted in a north-south interconnection between the invading seas, but he considered the possibility of a connection through the Tilemsi Valley in Mali unresolved. He also considered this transgression to have been the equivalent of the North American Greenhorn cycle. Bellion et al. (1989) documented evidence of extensive gypsiferous deposits (indicating shallow, hypersaline embayments) and carbonates of Cenomanian-Turonian age in parts of the Tilemsi Valley and areas further west in the Taoudenit Basin, but did not establish whether there was a full link between the Tethys and south Atlantic (Gulf of Guinea) at this time. Kogbe et al. (1976: fig. 4) argued for a full north-south Trans-Saharan Seaway connection between the Tethys and the Gulf of Guinea in the Turonian that went through the “Niger valley” (p. 243) and not via the Benue Trough as was argued by Furon (1935).

Our modeling for the Turonian snapshot is based primarily on Kogbe (1981) and Reyment and Dingle (1987). Importantly, our Turonian snapshot does not show any seas within the Tilemsi Valley or the Gao Trench; north-south continuity was established through the Iullemmeden Basin in the eastern part of Mali as well as in Niger. The sea was mostly concentrated in areas such as Niger, Chad, and Nigeria. This picture differs substantially from that shown in Reyment (1980: fig. 3) where the Tilemsi Valley is partly flooded and the Gao Trench fully flooded.

T2 Cyclothem: Coniacian (fig. 15B). The second of the transgressions, considered to be the equivalent of North America's Niobrara cycle, was hypothesized to have been less extensive such that the embayments from the Tethys and the Gulf of Guinea did not connect (Reyment, 1979; 1980). Our modeling snapshot was based on Reyment (1980) and Reyment and Dingle (1987), and our results very closely resemble those of Reyment (1980: fig. 5). Our snapshot is noteworthy in showing that the seawater entering the Tilemsi Valley and Gao Trench in Mali had come from the south. The model is particularly poorly established for parts of Niger.

T3 Cylclothem – Campanian to Maastrichtian (fig. 15C). As many as three transgressions of the Trans-Saharan Seaway have been hypothesized during the Maastrichtian. Later transgressions, T4 (fig. 15D) and T5 (fig. 15E), have been the primary focus of our work, as these are the sediments that yielded important vertebrate fossils (Hill et al., 2008; Tapanila et al., 2008; Claeson et al., 2010). Because of the closure of the Benue Trough as a potential passage for continental seas, Reyment (1979; 1980) argued that the transgression that started in the Campanian became epicontinental only in the Maastrichtian and necessarily took a relatively western route across West Africa compared to earlier transgressions (Reyment, 1980). Specifically, this meant a route west of the Hoggar Massif (see also fig. 1B, C), an elevation of Precambrian rocks in Algeria (Krasheninnikov and Trofimov, 1969; Reyment, 1979: 32). Reyment (1980: 314) emphasized that it was “structural depressions” that were filled by the Maastrichtian and Paleocene transgressions but that the sea also overlay “stable craton” in places. Reyment (1979) also indicated that even though the Maastrichtian sea was extensive, there was little biostratigraphic evidence to support a full north-south Maastrichtian connection between the Saharan regions like Mali and basins adjacent to the Gulf of Guinea. Other authors suggested, based on fossil invertebrates, including bivalves and “free-swimming” ammonites, that the water could be characterized as fully marine into the Iullemmeden Basin in Niger and far from the open ocean (Moody and Sutcliffe, 1990: 21). Adeleye (1975) argued for a full connection at this time via the Niger Valley through Nigeria, Niger, and Sudan. Adetunji and Kogbe (1986) supported this inference, based on the discovery of the ammonite Libycoceras in Niger and southern Nigeria as well as similarities among ostracods in the two places (citing Kogbe, 1979). This idea stands in contrast to hypotheses of Petters (1977; 1979) who argued that the northern marine incursion terminated as an embayment because the foraminiferans are very different in the Sokoto area than in southern Nigeria. Adetunji and Kogbe (1986), however, suggested that this difference could be simply a local environmental variation. The case for a full Maastrichtian connection was also proposed on the basis of vertebrates, specifically mosasaur species, found to be present in Niger and in Nigeria (Lingham-Soliar, 1998).

Our modeling is based on the work of many previous authors, including Reyment (1980) and Guiraud et al. (2005), with minor additions from Adetunji and Kogbe (1986). Although much fieldwork remains to be done, our tectonically informed Maastrichtian snapshot shows a full sea connection through the Tilemsi Valley, Gao Trench, and Iullemmeden Basin. In contrast to the Coniacian, this sea came into Mali from the north. This sea also linked to a large embayment that spread into neighboring Niger. Our model does not show a full southern sea connection in the Maastrichtian, but instead, several areas of unconfirmed marine coverage with the ocean tapering out just short of a full connection to the Gulf of Guinea, similar to estimates by Guiraud et al. (2005: fig. 49B).

T4 Cyclothem: Maastrichtian-Danian Boundary (fig. 15D). Damotte (1995: fig. 1) studied ostracods from North Africa (northern Algeria and Tunisia), south through Mali (Tilemsi Valley, near Tichet), and continuing south to the Congo and concluded that there was a fully connected Trans-Saharan Seaway during the late Maastrichtian-Paleogene transgression. He stated that this conclusion, supported by the work of Reyment and Dingle (1987) derived from the marine nature of the ostracod fauna. Kogbe (1981), Bellion et al. (1989) and Damotte (1995), and Guiraud et al. (2005), with minor additions from Reyment (1980), were the primary sources for our snapshot at the Maastrichtian-Danian boundary (∼ 66 Ma). Our model shows a full north-south connected seaway close to the K-Pg boundary.

T5 Cyclothem: Danian (fig. 15E). In the Paleogene various authors reconstructed the Trans-Saharan Seaway as fully connected, linking the Tethys and Guinean Seas (e.g., based on ostracod species distribution sensu Reyment, 1980; Kogbe, 1981; Bellion et al., 1990: fig. 1). An alternative model reconstructs two separate, disconnected gulfs: the Tethys Sea inundating West Africa from the north, with shallow, warm, and hypersaline waters, and the Guinean Sea, encroaching from the south, bringing poorly oxygenated waters with low salinity (Petters, 1977: figs. 6–9; 1979; Boudouresque et al., 1982). Important tectonic variables impacting the late Paleocene (and subsequently early Eocene) sea-level rise may have been both higher oceanic crust production and the emergence of the oceanic ridge that formed as Norway separated from Greenland (Miller et al., 2005). Adeleye (1975) argued against a connection through the latest Paleocene, as did Petters (1977) based on differences in the micro-faunas of northwestern Nigeria versus the southern Nigerian Sedimentary Basin. Reyment (1979) rebutted this claim, specifically noting extensive similarities among the Paleocene ostracod faunas from Libya, Mali, Niger, and Nigeria. He argued that the 600 km break in the continuity of outcrops may simply have been due to erosion rather than lack of sedimentation in the sea's absence. Nautiloids from Niger have also been used to support a fully connected Trans-Saharan seaway (Tintant et al., 2001) in the Paleocene. Petters (1977) argued for a maximum extent of the Trans-Saharan Seaway in the latest Paleocene, where he hypothesized that the sea had a southern reach well into northern Nigeria from the north. Boudouresque et al. (1982) corroborated this finding in a study of palynology.

Kogbe (1981), Bellion et al. (1989), Damotte (1995), and Guiraud et al. (2005), with minor additions from Reyment (1980), were the sources for our Danian snapshot. It supports the hypothesis that the Trans-Saharan Seaway formed a continuous link between the Tethys and the South Atlantic at times during the Paleocene. Phosphatic shales and conglomerates become part of the cyclothem-succession package near the top of a transgressive-regressive cycle. Our full north-south connection is qualified by some poorly confirmed areas in Nigeria. We also found a separate marine incursion in the east that would have passed through Niger and Chad, isolating the exposed Precambrian basement rocks of the Hoggar, Aïr, and Adrar des Iforas massifs as an island. The maximum continental flooding of Cyclothem 5 occurred in the mid-Paleocene (Reyment, 1979).

T6 Cyclothem: Eocene (fig. 15F). In their study of neritic ostracods of the Tethys, Speijer and Morsi (2002: fig. 1) showed full connection of the Trans-Saharan seaway at the PETM, an interval during which sea level rose ∼ 20 m in less than 100,000 years (Speijer and Wagner, 2002). Guiraud et al. (2005: fig. 57B) argued that substantial sea coverage of West Africa continued into the early-middle Eocene, with water extending into Mali from the North and flooding the Tilemsi Valley, Gao Trench, and part of the Iullemmeden Basin. In our model, which is based heavily on the work of Guiraud et al. (2005), no connection persists between the Trans-Saharan Seaway and the Gulf of Guinea. The presence of the same species of pycnodont fish (Pycnodus bowerbanki) in rocks of the Iullemmeden Basin and rocks of the London Clay was suggested as a reason to support some migration between ocean waters around Europe and those within west Africa via the Trans-Saharan Seaway as late as the Lower Ypresian part of the Eocene (Moody and Sutcliffe, 1990). Nonetheless, by the Paleocene and especially the early-middle Eocene, the depositional environment in Mali had likely become very shallow and verged on isolated, most likely brackish and even occasional fresh water embayments cut off from the retreating Trans-Saharan Seaway. The seaway that had receded northward by this time had begun to close permanently, and the large embayment seen in Mali could have been entirely cut off from the main Trans-Saharan Seaway. This hypothesis is reinforced by the presence of both freshwater and saltwater fish species present at this time, as we discuss below.

Paleoecological Reconstruction

Descriptions above of the fossils found to date are the basis for our reconstructions of how certain key species from the Trans-Saharan Seaway looked in life (frontispiece). We have also set these species in a schematic ecosystem deriving details from both the sedimentology and the paleobiology (plate 2). In the Trans-Saharan Seaway, mangrove swamps lined shallow seas adjacent to the steep Adrar des Iforas massif from which numerous ancient rivers carried fresh water and sediment into the epeiric sea. The waters were warm and tropical, an inference supported particularly strongly by the mollusk fauna. Apex predators such as large fishes, sea snakes, and crocodyliforms existed in brackish water near shore, part of extensive coastlike ecosystems that lay deep within continental West Africa. The environment consisted of stable climate as indicated by temperature bioproxies such as fossilized wood. Our fieldwork has produced little evidence to support the idea that there was an ancient dry climate in this region of Mali as previously argued (Petters, 1977). We have not observed sedimentary indicators of aridity such as evaporites or aeolian sandstones (McKee, 1964; Hallam, 1985). Within a humid tropical setting, water lost through high evaporative conditions along the broad, shallow coast was likely replenished by shoreward flow of offshore waters. During times of high sea level, this strong evaporative pumping may have contributed to cycling of nutrient rich but oxygen poor offshore waters to deliver the unusual chemistry necessary for phosphogenesis that we have described (Tapanila et al., 2008).

Modern coasts are zones of interaction between terrestrial and aquatic realms, where, in the tropics, erosion from land intersects episodes of dramatic, hurricane-type weather originating at sea. We have identified lithological evidence of storm-generated turbulence in Trans-Saharan Seaway sediments—in the phosphatic conglomerate facies (Facies 5) and in shale/marl facies (Facies 3) (Tapanila et al., 2008)—even though both groups of sediments were far from the open ocean at the time they were deposited. Investigators have speculated about ancient ocean circulation patterns through modeling, but have not demonstrated a clear mechanism to generate cyclonic systems capable of forming major hurricanes inland (Kauffman, 1984; Nicholls and Russell, 1990; Lehman, 1997; Lehman, 2001). Evidence for severe and anomalous storm events has been identified in the sedimentary record (Roberts et al., 2008), however, for the Malian rocks we have yet to identify such event beds.

Modern tropical rainforests and reefs are particularly noteworthy for their biodiversity and species richness (Wilson, 1992; Bianconi, 2002). By analogy, the paleotropical Trans-Saharan Seaway was likely to have been among the more bio-diverse parts of the Late Cretaceous–early Paleogene Earth based on the range of taxa we have described. Given the estimated shallow depth of the Trans-Saharan Seaway (50 m), we believe it effectively consisted of neritic and benthic zones only, such that all the waters would have been penetrated by sunlight. Evidence from the mollusk fauna, particularly taxa such as Ostreida, Venericardia, Turritellidae, Carditidae, Cerithioidea, Cypraeoidea, and Strombidae, point directly to the shallow sea being warm.

As emphasized by Harries (2009), the sedimentation of epeiric seas results from an interplay between siliciclastics from land and limestone creation in situ in marine waters, with the latter being the dominant method of rock formation. The five facies we have described show varying degrees of these two sedimentation types alternated and repeated through time as the sea transgressed and regressed (plate 1; figs. 13, 14). Near shore (proximal) sediments are dominated by sandstones, siliciclastic sediments that would have eroded from the adjacent Adrar des Iforas. The deepest water limestones that formed distal or basinward precipitated in situ, most likely from calcareous plankton (including nanoplankton and other algae, as well as benthic and planktic Foraminifera) and calcareous macroalgae. The tropical paleolatitude of Mali also encouraged carbonate production through algal mineralizations, with mudstones and shales developing under a mixture of these forces.

The ecosystem that we hypothesize was present in and along the Trans-Saharan Seaway included primary producers for which direct fossil evidence has been recovered, such as mangrovelike woody plants as well as species of Ostracoda (e.g., Damotte, 1991; Damotte, 1995; Colin et al., 1997), and Foraminifera (Krasheninnikov and Trofimov, 1969; Berggren, 1974; Bellion et al., 1989; Damotte, 1991; Berggren and Pearson, 2005). The primary consumers were filter-feeding invertebrates, including Ostreida, such as those that formed colonies on carcasses of dead turtles (described above), bivalve mollusks that burrowed into lithified coprolites and bone, and crabs, recognized by us from indirect evidence of fossilized burrows, as well as herbivorous hyracoid and plesielephantiform mammals (De Blieux and Simons, 2002; Clementz et al., 2008). Secondary consumers would again have included Ostracoda and Foraminifera among microscopic taxa, as well as larger species, such as pycnodont, amiid, osteoglossid, and siluriform fishes, and medium-sized snakes (e.g., Amananulam), all of which preyed on invertebrates or smaller vertebrates, and were themselves also consumed by tertiary consumers. Tertiary consumers, also called apex predators, included such species as dyrosaurid crocodyliforms, palaeophiid snakes, and sharks and rays, with direct evidence of their feeding present in the ichnofossils that we have collected (Hill et al., 2015). Remains of leaves, birds, and insects are conspicuously absent from the known fossils of the Malian Trans-Saharan Seaway that we have collected to date.

The apex predators are particularly well represented in the fossil record. We have reconstructed the palaeophiids as living in nearshore environments, where their fossils are typically found, often associated with mangroves and estuaries (Rage et al., 2003; Parmley and DeVore, 2005; Houssaye et al., 2013). They are known, for example, from similar nearshore fossil deposits of the Fayum (Andrews, 1901; 1906; McCartney and Seiffert, 2016), and are comparatively rare in sediments interpreted as pelagic or open ocean (Hutchison, 1985; Holman et al., 1990; McCartney et al., 2018). Our data are consistent with the hypotheses that the genus Palaeophis, known from Mali, was probably not pelagic (Hoch, 1975), and the nigerophiids would also have lived nearshore as they are elsewhere recovered in presumed freshwater or brackish deposits (Rage and Prasad, 1992; Rage et al., 2004; Rage and Dutheil, 2008; LaDuke et al., 2010; Pritchard et al., 2014).

Fishes recovered from the Malian field area underscore the challenge of determining whether an extinct taxon lived in fresh or saltwater, or both. The fossil fish taxa we have collected that were likely to have lived in freshwater in life include the Claroteidae or catfish (Siluriformes), and the Lepidosirenidae (Dipnoi) or lungfish. The fully marine taxa were the Chondrichthyes, the Pycnodontidae, and the Osteoglossidae. The Amiidae were potentially anadromous. The Malian lungfish belong to a clade that has a single living species, which is a generalized predator and omnivore restricted to freshwater (Myers, 1938). Marine records of Dipnoi are now limited to those that predate the Mesozoic (Heinicke et al., 2009; Otero, 2011). Thus, we conclude that the presence of a lepidosirenid in the Malian deposits is a bioproxy indicating the presence or proximity of freshwater sources entering the Trans-Saharan Seaway.

Myliobatiformes (stingrays) and other chondrichthyans (e.g., various sharks) recovered were all likely marine taxa based on comparisons with their living relatives. The extinct stingrays may have traveled in shoals or schools at high speeds, given the widespread nature of this behavior in their living relatives (Nursall, 1996a; Pavlov and Kasumyan, 2000). Torpediniformes (i.e., electric rays) were semipelagic to benthic fish that would have traveled in groups of smaller numbers (Claeson et al., 2014; Marramà et al., 2017), and sawfishes were also open marine to benthic taxa (Wueringer et al., 2009). Closely related extant stingrays, with crushing pavementlike dentition, are known to pursue hard-part prey, whereas torpediniform electric rays and sawfishes primarily consume softer prey (Summers, 2000; Wueringer et al., 2009; Claeson et al., 2014).

The Tethys Sea is considered to have been the center of origin for pycnodont fishes, as well as the site of their major radiation and last appearance (Poyato-Ariza, 2005), with the clade having undergone a multidirectional dispersal between the Tethys and the southern Atlantic (Nursall, 1996a). Pycnodonts were omnivorous benthic foragers and, as compared with halecostomes, are thought to have been capable of biting and/or nipping in combination with suction feeding to acquire a broad range of shelled invertebrates (Kriwet, 2001). The intriguing molariform teeth of these fishes indicate durophagous or ominivorous diets that would have included a range of benthic species such as “sponges, algae, zoanthids, coral, polychaetes, and other invertebrates” (Poyato-Ariza, 2005: 177). The Late Cretaceous and the modern African fish faunas, have both been described as homogenous for marine and freshwater fishes (Murray, 2000; Otero, 2010), a pattern that might be explained by the low topography across much of the African continent favoring movement of fish species among basins. The precise data on fish species by locality that we have supplied here contributes important new information for further continentwide assessments.

Finally, we note that scleractinian corals are absent from the Malian localities described here. While this may be a taphonomic idiosyncrasy, an environmental explanation for this finding is that Trans-Saharan Seaway waters were not adequately saline to support corals. The sea was highly likely to have varied in salinity under the opposing influences of the precipitation and drainage of freshwater, and the influx of open ocean saltwater. Scleractinian corals are typically stenohaline and cannot thrive outside normal marine salinity (Ferrier-Pagès et al., 1999; Hédouin et al., 2015), however, exceptions to this generalization have been reported (Chartrand et al., 2009). Another explanation for the observed lack of corals may be that the sea exhibited high turbidity, especially during the deposition of carbonates (marls), that would have blocked light to and clogged the feeding apparatus of these animals, thereby selecting against their existence. The micritic nature of the carbonates and abundance of interbedded shale facies amongst the carbonate beds lends support to this idea.

Body Size in Certain Extinct Predators

As epeiric seas experienced episodes of geographic isolation with changing sea level, the species they supported may have been under increased allopatric speciation pressures (Harries, 2009). Bodies of water cut off from the Tethys Sea had the potential to become centers of endemism. It is worth considering how such environmental pressures may have impacted the important life history variable of body size in Trans-Saharan Seaway species. Full clade-by-clade consideration of this topic is beyond the scope of this paper, however, we note that body size appears to have been relatively large for several of the recovered fossil predators, including apex predators: sea snakes, amiids, pycnodonts, and siluriforms (fig. 82).

FIGURE 82.

Certain extinct predators from the Trans-Saharan Seaway exhibited relatively large body length for their clades. In five cases the Malian representative is the largest or close to the largest member of the clade. Body-length ranges in a clade are indicated by black bars. Silhouettes compare the relative size of the average clade body length to the body length for the Malian fossil (clades not shown at actual size). The Malian fossils compare as follows: Paleophiidae: small: ∼100 cm, average: ∼500 cm, Mali: 1230 cm (∼2.45 × average and largest in clade); Nigerophiidae: small: ∼ 26 cm, average: ∼120 cm, Mali: 210 cm (∼1.75 × average), large: 310 cm; Vidalamiinae: small: 5.5 cm, average: 45 cm, Mali: 180 cm (∼ 4 × average and largest in clade); Pycnodontidae: small: 15 cm, average: 25 cm, Mali: 122 cm (∼ 4.8 × average), large: 150 cm; and Claroteidae: small: 3.4 cm, average: 30 cm, extant maximum length: 150 cm; Mali: 160 cm (∼5.33 × average and largest in clade). For Pycnodontidae the comparison includes the length of the caudal fin in all species. Measurements and methods of deriving estimated lengths of extinct species provided in the Systematic Paleontology section.

It has been observed that during the Eocene a major proliferation of large pelagic predatory fish occurred in the oceans (Friedman and Sallan, 2012), and our observations indicate this pattern may have extended to this epeiric sea. As the Trans-Saharan Seaway began to recede for a final time during the Eocene, patches of isolated water may have become increasingly common, and the large embayment described within Mali (fig. 15F) could have been cut off from the main Trans-Saharan Seaway during this time. Species found in the Eocene rocks left behind include Maliamia gigas, which is relatively large compared to other amiids (Grande and Bemis, 1998), and the new unnamed Malian Eocene catfish that is larger than (and morphologically distinct from) other reported African giant catfish, including contemporary fossil species from Egypt (El-Sayed et al., 2017) and later-occurring species from the Miocene of Chad (Pinton et al., 2011). The Malian lungfishes are also relatively large compared to close modern relatives (Martin, 1995), as are the pycnodonts based on body-size estimates from teeth supplied above.

Regarding snake body size, McCartney et al. (2018) recently used regression analysis of vertebral size to estimate the body sizes of palaeophiids from Mali. The three known species of snakes from Mali, Palaeophis colossaeus (Palaeophiidae), Amananulam sanogoi (Nigerophiidae), and an indeterminate species, were all estimated to exceed 2 m in total length. The largest of these, P . colossaeus, is among the three largest known species of snakes to have ever lived, and its body size is exceeded significantly only by the boid Titanoboa cerrejonensis from the Paleocene of Colombia (Head et al., 2009). Although an estimate of body size has not been performed for all known nigerophiids, using the regression in McCartney et al. (2018), the vertebral dimensions of A . sanogoi are matched only by “Nessovophis” zhylga from the Eocene of Kazakhstan (Averianov, 1997; McCartney et al., 2018). The indeterminate fossil snake we discovered is of intermediate size between A . sanogoi and P . colossaeus, with an estimated length exceeding 5 m. Given that we have likely underestimated size, this indeterminate snake was also a relatively large species (McCartney et al., 2018).

Although the pattern is complex and it is premature to draw major conclusions, an intriguing hypothesis to consider is that these species may have experienced a phenomenon of regional gigantism in “aquatic islands,” or relatively isolated parts of the epeiric seas. The so-called island rule describes such a phenomenon in terrestrial island settings in which large species tend to become dwarfs and small species become gigantic (Van Valen, 1973; Lomolino, 2005). Lomolino (2005: 1684) described the reasons for the phenomenon of the “island rule” to be “limited insular resources, the paucity of interspecific competition and predation . . . and the challenge of dispersing to, but not emigrating from, islands.” He concluded with a plea to investigate other “island-like ecosystems” for possible discovery of trends in body-size evolution (Lomolino, 2005: 1696). The Trans-Saharan Seaway may have effectively resulted in “aquatic islands” that emerged during geologically short-term shrinkage and expansion of saltwater bodies. Understanding whether our report here constitutes a strongly supported pattern requires further investigation as to whether these species evolved in situ as a response to fragmented aquatic environments. Clade-by-clade phylogenetic comparisons with fossils sampled through their stratigraphic and geologic ranges will also be critical to the evaluation of this hypothesis. We note, for example, that palaeophiid snakes are large bodied across their known distribution, which may or may not have consisted of fragmented nearshore saltwater habitats.

Cretaceous-Paleogene (K-PG) Boundary

At the transition between the Maastrichtian Ménaka Formation and the Paleocene Teberemt Formation our section spans the K-Pg boundary. As described above, there appears to be a hiatus of nondeposition at this transition, which currently limits high-resolution stratigraphy across the boundary (plate 1). Smith and Jeffery (2000) suggested that carbonate platform drowning in the Paleocene might have disrupted sedimentation on a large scale, providing a possible mechanism to explain this hiatus. Alternatively, there is a global sea-level fall at or just below the K-Pg boundary (e.g., Haq, 2014) that could explain this unconformity and that is consistent with the original work of Radier (1959; see also Damotte, 1991; Moody and Sutcliffe, 1991, 1993) and the karstic nature of the uppermost Maastrichtian limestone bed in the region (“Terrecht I”). Lack of this important part of the section currently limits our ability to investigate such phenomena as eustasy-based extinctions, which has been proposed for the Mesozoic of the Western Interior of North America (Eaton et al., 1997).

Nonetheless, by placing extinct species in a phylogenetic and stratigraphic context we have been able to develop an improved picture of clade diversity through time in the Late Cretaceous–early Paleogene of West Africa, an area for which there has been a paucity of vertebrate fossils tied to detailed stratigraphic sections in this interval. We provide here some comments on two related questions: is the global pattern for a given clade one of species change or stability across this boundary, and, does the Malian fauna conform to or diverge from this pattern based on the evidence to date?

Limited inferences can be made about Malian placental mammals, which are very rare. We can report simply that, to date, they are found only on the Paleogene side of the K-Pg boundary and in Eocene rocks. It has been estimated that the placental mammal subclade Afrotheria had a substantial Cretaceous existence (e.g., Meredith et al., 2011), however, our direct paleontological work here, and broader combined data analyses, have not corroborated this estimate (O'Leary et al., 2013). The Malian fieldwork contributes some direct negative evidence, albeit localized, to the hypothesis that placental mammals are absent from Mesozoic rocks, even in Africa.

The K-Pg event had a catastrophic effect on the clade Archosauria, with the extinction of pterosaurs, nonavian dinosaurs, and many clades of crocodyliforms (MacLeod et al., 1997). By contrast, dyrosaurids have long been recognized as having representatives on both sides of the K-Pg boundary (Buffetaut, 1978a; Buffetaut, 1979). Minimally, the genus Hyposaurus occurs in Cretaceous and Paleogene strata (e.g., MacLeod et al., 1997). Jouve et al. (2005b) showed in a phylogenetic analysis of dyrosaurids that the majority of clade diversification occurred in the Maastrichtian or early Paleocene, even as other clades faced mass extinction globally. Our prior research further refined the timing of this diversification, and showed that as many as six dyrosaurid lineages arose in the Cretaceous and survived the K-Pg event (Hill et al., 2008). This result, based directly on fossils of a given taxon being found on either side of the boundary, provides further evidence for robust clade survivorship across this global mass extinction event. Moreover, even more dyrosaurid lineages are inferred to have crossed this boundary based on ghost lineages implied by the phylogenetic hypothesis (Hill et al., 2008: fig. 7).

This pattern prompts the question of why dyrosaurids appear to have survived the K-Pg event when other marine reptiles (e.g., mosasaurs and plesiosaurs) became extinct. (We note here that in one of our prior papers [Hill et al. 2008], we erroneously suggested that other marine crocodyliforms, e.g., metriorhynchids and teleosaurids, also became extinct at this time; these clades were already extinct before the Late Cretaceous [Markwick, 1998]). Several factors may be responsible for such differential survival including geographic range, eurytopy, and specialized feeding or reproductive strategies (Fara, 2000: and references therein). Jouve et al. (2008) proposed that dyrosaurid juveniles may have begun life in freshwater habitats, as do modern saltwater crocodiles, a life history strategy that would have been possible along the ancient Trans-Saharan Seaway. Such behavior may have afforded dyrosaurids a freshwater refuge unavailable to other marine reptiles. Another behavioral advantage for the clade may have related to the mode of locomotion employed by dyrosaurids. Data from CT scans of Malian dyrosaurid skulls (archived in MorphoBank Project 2735;  www.morphobank.org) showed a vestibular morphology similar to that of turtles that preferentially locomote by walking along the floor of an aquatic habitat, rather than swimming in the water column (Georgi, 2006). If this behavior was present in dyrosaurids, it could be a factor in their survival relative to that of other swimming or fully terrestrial forms.

Our GPlates reconstructions of the shape and extent of the Trans-Saharan Seaway (fig. 15) are consistent with the global pattern of dyrosaurid dispersal proposed by Hastings et al. (2011). Those authors postulated an initial, northerly dispersal of dyrosaurids from Africa to the New World during or before the Late Cretaceous, with subsequent dispersal taking place along more southerly routes. Our Trans-Saharan Seaway reconstruction shows an open northerly route of dispersal for dyrosaurids in the Maastrichtian. By the Danian, the Trans-Saharan Seaway was also open to the south, a change that could have facilitated later dispersals of dyrosaurids to South America.

Unlike dyrosaurids, Malian turtle fossils that can be taxonomically assigned are currently found on the Paleogene side of the boundary only. They are members of Pleurodira (side-necked turtles), and within that clade, are part of the clades Pelomedusoides and its subclade Bothremydidae. A comprehensive review of the pelomedusoids organized the Malian specimens further into the tribe, Taphrosphyini, and showed that it contained two unnamed clades: one of Nigeremys and Arenila, and one of all other taphrosphyins, including two Malian specimens: the sister taxa Azabbaremys and Acleistochelys (Gaffney et al., 2006: figs. 1, 288; Gaffney et al., 2007). The species in the clade of Nigeremys + Arenila are found in Upper Cretaceous rocks, but members of the other clade in Taphrosphyini are found in Paleocene or younger rocks. One possible explanation for this pattern is that a single taphrosphyin lineage survived the K-Pg event and underwent significant diversification in the early Paleogene. This hypothesis could be tested by looking for more pelomedusoid specimens on both sides of the boundary in Mali and Niger, given the substantial facies continuity across the boundary.

The global effect of K-Pg extinction on snakes is poorly understood, with some evidence supporting a significant extinction event among terrestrial species (Longrich et al., 2012). The impact of this event on aquatic snake clades, however, is even less well understood. Only three palaeophiid taxa have been reported prior to the K-Pg boundary—all from Africa (Rage and Wouters, 1979; Lingham-Soliar, 1998; Rage and Werner, 1999). Of these, the date of one has been called into question (Head et al., 2005), and another has not been properly described or figured (Lingham-Soliar, 1998). However, if any of these occurrences is accurate, it would suggest that the clade Palaeophiidae survived the K-Pg event and diversified in the Paleogene. More than 20 species are known from the Paleocene and Eocene globally and the Eocene palaeophiid specimens that we have recovered further document the post-K-Pg diversity of this clade (McCartney et al., 2018). Nigerophiids, however, occur as early as the Cenomanian (Rage and Dutheil, 2008), and approximately equal numbers of these rare fossil species have been found on either side of the K-Pg. The Cretaceous records include a number of presumed freshwater species (Rage and Prasad, 1992; Rage and Werner, 1999; Rage and Dutheil, 2008; LaDuke et al., 2010; Pritchard et al., 2014), and later, marine species occur in the Paleocene (Rage, 1975) and Eocene (Rage, 1980; Averianov, 1997; McCartney et al., 2018). Our findings provide an additional Paleocene datapoint for nigerophiids but no further information about the K-Pg survivorship of this clade. Because there is both fresh- and saltwater influence at the locality that produced nigerophiids it is not clear at present whether our Paleogene species preferred a particular environment.

Globally, the K-Pg event did not result in a mass extinction of fishes (Cavin, 2001). Instead, the event appears to have punctuated certain larger trends of slow, background extinction already underway by the Late Cretaceous (Friedman and Sallan, 2012; Sibert and Norris, 2015; Poyato-Ariza and Martín-Abad, 2016). Our findings in Mali show an overall pattern of replacement of cartilaginous fishes by bony fishes at this boundary, which we now discuss in more detail.

Generally, Cretaceous rocks preserve more nonbatoid chondrichthyans (e.g., sharks) than osteichthyans, whereas the reverse is generally true for the Paleogene (Cavin, 2001; Friedman and Sallan, 2012; Sibert and Norris, 2015; Poyato-Ariza and Martín-Abad, 2016). For osteichthyans, Poyato-Ariza and Martín-Abad (2016) described a gradual, macroevolutionary pattern of competitive replacement of nonteleostean osteichthyans (e.g., pycnodonts, semionotiforms, stem-group teleosteans) by teleosts from the Mesozoic into the early Cenozoic and argued that the K-Pg event was of no particular consequence to this replacement. Moreover, significant changes to the chondrichthyan fauna also characterized the K-Pg boundary as species of nonbatoid chondrichthyans such as sharks were abundant in the Cretaceous but dropped in numbers relative to osteichtyhans recorded in similar environments in the Cenozoic (Friedman and Sallan, 2012; Sibert and Norris, 2015).

Several authors have reported that the majority of osteichthyan and chondrichthyan fishes that went extinct during the Maastrichtian, or at its end, were marine fishes, while the majority of freshwater- and brackish-environment fish survived the K-Pg boundary (Cavin, 2001; Friedman and Sallan, 2012; Sibert and Norris, 2015). Those authors also indicated that the species most negatively affected by this event were fast-swimming and piscivorous predators (Cavin, 2001; Friedman and Sallan, 2012; Sibert and Norris, 2015).

The aforementioned pattern of replacement (teleosts supplanting large, pelagic, marine fishes, many of which were chondrichthyans) is not entirely reflected in the data from Mali. For example, Cretaceous Malian lungfish and pycnodonts represent clades that survived the K-Pg boundary and are also recovered from the Paleogene rocks of Mali. Additionally, osteichthyans such as claroteid catfishes and vidalamiine bowfins persist into the Eocene of Mali. Globally the fossil record shows that pycnodonts and amiids were declining by the Late Cretaceous, with the diminishing diversity of the clades not particularly impacted by the K-Pg boundary (Poyato-Ariza and Martín-Abad, 2013). In Mali, Pycnodus jonesae and Pycnodus sp. are found in rocks on both sides of the K-Pg boundary and we do not observe a marked diminishment of species post K-Pg. The addition of vidalamiine bowfin and claroteid catfishes to the Paleogene Malian fauna underscores an overall pattern of increasing diversity of osteichthyans without a pronounced faunal turnover. Of note is that these osteichthyans all have a known (from extant relatives), or hypothesized, brackish and/or freshwater life history, a strategy that has been linked with high survivorship across the K-Pg boundary (Cavin 2001; Friedman and Sallan, 2012).

Sibert and Norris (2015) showed a global pattern of faunal turnover around the K-Pg boundary as osteichthyan species filled ecological niches recently vacated by cartilaginous fishes. Given the observed rise in the number of osteichthyan species in Mali, we should expect a corresponding decrease in cartilaginous fish. Our observations, albeit preliminary, support this prediction. In the Cretaceous of Mali, the chondrichthyan fauna is relatively diverse, consisting of two genera of sclerorhynchid sawfish, the myliobatid stingray Myliobatis, and the lamniform shark Serratolamna. In contrast, only Myliobatis survives the K-Pg globally but is not actually represented in any of the Paleogene records of Mali (Claeson et al., 2010). In the Paleogene of Mali, the Eocene torpediniform electric ray Eotorpedo is the only cartilaginous fish that is recovered. Thus, among fishes at the K-Pg transition in Mali, the trend is one of replacement and overall net decrease in chondrichthyans relative to osteichthyans. It is noteworthy that the Cretaceous and Paleogene chondrichthyans were all considered pelagic, fast-swimming, piscivorous predators that may have spent part of their time near the shore.

Of the invertebrate fauna, there are only two mollusk-producing beds in the Upper Cretaceous part of our section, each with a handful of poorly preserved taxa, mostly identified only to the family level or higher. Thus, although we have quite a few Paleogene mollusk specimens and localities, it is premature to draw conclusions about effects of the K-Pg event on molluscan diversity in the Malian Trans-Saharan Seaway, beyond the observation of a general pattern of persistence of higher clades. In the Paleogene Malian fauna, the Ostreida, Venericardia, and Turritellinae mollusks remain common, and we hypothesize that broadly similar environmental and taphonomic conditions persisted in the Trans-Saharan Seaway through the Cretaceous into the Paleogene.

Of the echinoid species we have collected, none occurs on both sides of the K-Pg boundary in our sections. The regular echinoid, Echinotiara occurs in Upper Cretaceous beds, and the irregular echinoids Linthia and Oriolampas occur in Paleocene beds. Jeffery (2001) reported that the spatangoid heart urchins, in general, were less severely affected by the K-Pg extinction than some other echinoid clades. For example, Linthia is recorded from the Maastrichtian of Algeria and Oman (Cotteau et al., 1881; Smith, 1995) before dispersing more broadly in Africa, the Middle East, Europe, and North America in the Paleogene (Smith and Jeffery, 2000). Finally, palynological studies across the K-Pg boundary (Boudouresque et al., 1982) documented an ongoing warm and humid climate, as well as the presence of mangrove forests.

Paleocene-Eocene Boundary and PETM

The Paleocene-Eocene boundary falls between the Teberemt and the Tamaguélelt formations in our section but stratigraphic resolution is currently very coarse. The warm global temperature spike known as the PETM would likely have fallen close to this formational transition but cannot at present be placed with precision. While our PETM interpretations are preliminary and in need of testing by additional collecting and higher resolution stratigraphy, a picture of significant faunal turnover does not emerge conspicuously from the data we have assembled so far. The Paleocene-Eocene boundary has been far less studied for the neritic realm than for the deep sea realm (Speijer et al., 1996), making new contributions on the neritic realm, such as we provide here, a critical new global datapoint for the paleotropics, if only to seed further work. By the Paleocene, and especially by the early-middle Eocene, the depositional environment of the Trans-Saharan Seaway developed an increasingly shallow-water signal. We hypothesize that the remaining saltwater embayment had probably become cut off from the open ocean as the Trans-Saharan Seaway retreated northward. The water in the relict intracratonic sea then likely verged on having a brackish to increasingly freshwater composition.

No Paleocene mammal fossils have been found in Mali, and only a small number of fragmentary paenungulates have been found in the Eocene sediments as noted above. Likewise, the best-preserved turtle specimens from Mali are Paleocene. Those turtles at our single, vertebrate-producing Eocene locality, Tamaguélelt (locality Mali-20), are so fragmentary as to be of uncertain attribution, and thus, little can be said presently about turtle or mammal diversity across the PETM.

Regarding dyrosaurid crocodyliforms, Hill et al. (2008: fig. 7) showed that approximately half as many dyrosaurid lineages survived the Paleocene-Eocene boundary as survived the K-Pg boundary (note, however, that this figure erroneously showed the two species of Rhabdognathus extending into the Eocene; the youngest occurrence of this genus is Paleocene [e.g., Jouve, 2007; Hastings et al., 2011]). By the end of the Eocene, all dyrosaurids had gone extinct, the youngest record being a few fragments from the Upper Eocene of Myanmar (Buffetaut, 1978b). Moody and Sutcliffe (1990: 22) recognized similarities between Eocene (Ypresian) crocodyliform species from the Iullemmeden Basin of Mali and those from South America, raising the possibility that dyrosaurids were still in wide marine circulation globally at least through the early Eocene. This hypothesis is problematic, however, because diagnostic dyrosaurid remains have yet to be found from Eocene rocks of the Americas. A phylogenetic analysis of Dyrosauridae by Hastings et al. (2011) revealed an African origin for the clade, with New World species nested among African taxa. These complex relationships are consistent with as many as three separate westward dispersal events beginning in the Late Cretaceous and ending by the early- to mid-Paleocene (Hastings et al., 2011). Still, transatlantic dispersal during the Eocene has been hypothesized for other vertebrate taxa (Houle, 1999; Vidal et al., 2007) and may also have continued for dyrosaurids.

Palaeophiid snakes demonstrated their highest global diversity in the Eocene with records from localities across the region of the ancient Tethys Sea (summarized by Rage et al., 2003), including Mali (Rage, 1983). However, of the approximately 20 named palaeophiid species, only one, Palaeophis zhylan, may have a Paleocene occurrence (Nessov and Udovitschenko, 1984). Palaeophiids then appear to have become extinct by the beginning of the Oligocene. As such, the clade appears to demonstrate a diversification following the PETM, at which time it became an important example of an apex predator in nearshore faunas. Nigerophiid snakes are far less common than palaeophiids at all times, and in the Eocene are known from only two marine species (Rage, 1980; Averianov, 1997). The general paucity of records of nigerophiids makes it difficult to discern larger patterns with respect to climatic changes and major faunal boundaries; however, we can say that the greatest nigerophiid diversity occurred prior to the PETM (and in fact prior to the K-Pg boundary). Specifically, in Mali, only three snake taxa are recorded: from the Paleocene, a member of Nigerophiidae (Amananulam sanogoi), and from the Eocene a member of Palaeophiidae (Palaeophis colossaeus) and a taxon of uncertain attribution (McCartney et al., 2018). Elsewhere in the Trans-Saharan Seaway, there are two other snakes reported: the nigerophiid Nigerophis mirus from the Paleocene of Niger (Rage, 1975), and the palaeophiid Palaeophis africanus from the middle Eocene of Niger (Andrews, 1924). Taken together, this pattern suggests that smaller nigerophiids occurred in the Paleocene and larger palaeophiids in the Eocene of the Trans-Saharan Seaway, but much further sampling is necessary to truly demonstrate this hypothesis.

In general terms, the fish fauna in Mali was becoming increasingly freshwater through the Eocene, an observation that coincides with the ultimate retreat of the sea from the region. Malian Eocene fish species (pycnodonts, amiids, catfish) tended to have relatively large body size for their respective clades, as noted above. Among the osteichthyans, our section documents that two pycnodontid species, Pycnodus jonesae and Pycnodus sp. survived the PETM because they are found in rocks on both sides of it. The other pycnodontid species, P . zeaformis and P . maliensis, occur exclusively in our Eocene localities. Pycnodontidae as a clade persisted only as late as the Eocene (Nursall, 1996b; Poyato-Ariza, 2005). Thus, the Malian specimens are among the latest-surviving species. Even though some lineages clearly survived the PETM, the clade as a whole did not thrive subsequent to it. Poyato-Ariza (2005: 182) specifically considered the Eocene pycnodonts in Mali to represent a “relict” population of the Atlantic. The Tamaguélelt Formation pycnodonts are most closely related to contemporary taxa from Monte Bolca, Italy (Poyato-Ariza and Martín-Abad, 2013).

Several other osteichthyans in the Malian fauna are found only in Eocene rocks. For example, the amiid Maliamia has only been found in Eocene rocks of the Tamaguélelt Formation, where it represents the last surviving species of Vidalamiinae. As this clade appears to have already been on its way to extinction by the early Eocene, our work provides no new information regarding how it was impacted by the PETM. Similar temporal restriction in the Malian fauna also describes the osteoglossid Brychaetus, already an outlier in its clade for being a marine osteoglossid. However, among the Dipnoi, we have directly documented the species Lavocatodus giganteus from Paleocene and Eocene rocks and no striking differentials of taxon or specimen counts accompany the PETM. Protopterus elongus, however, is known only from Malian Eocene rocks, thus providing no information about this faunal transition. The last appearance of Lavocatodus occurred in the Middle Eocene Tamaguélelt Formation. Finally, the Paleogene fauna is limited to the batoid Eotorpedo, an electric ray that was semipelagic. Thus, no general conclusions can be reached about the response of batoids to the PETM, and sharks were absent in the Paleogene, as discussed above. Malian catfish are known only from Eocene rocks, thus we cannot interpret how the PETM impacted this clade. Overall for fishes, there is a suggested pattern of large body size and freshwater tendencies emerging as the Trans-Saharan Seaway diminished.

Among the invertebrates, we do not recover echinoids that are younger than Paleocene in age. Indeed, Smith and Jeffery (1998) emphasized that after the Danian, low latitude, shallow-water carbonate deposits diminished greatly. Such an environmental shift would have restricted the paleoenvironment that supported many echinoid species. As Malian rocks preserve Paleocene angiosperm macrofossils, further collection up the stratigraphic section has the potential to yield important new data about tropical plant response to the PETM. Jaramillo et al. (2010) reported, based on a study of neotropical fossil pollen, that through this global warming event, plant diversity actually increased even with high temperatures and relatively high levels of atmospheric carbon dioxide. Further testing in other global regions of the tropics such as Mali will be important for evaluating this intriguing hypothesis.

Our examination of the mollusk diversity in the Malian Trans-Saharan Seaway shows no obvious differences between Paleocene and Eocene faunas, a pattern similar to that displayed in other subtropical regions in general (Sessa et al., 2012). Thus, at present, there are no obvious faunal effects of the PETM, there is simply evidence for the persistence of taxa with a preference for warm waters. There are several different Turritelline genera that are moderately abundant within the Paleogene Mali fauna, indicating a warm, shallow water signature (Allmon, 2007; 2011). Other Paleogene taxa within the Mali outcrops also display affinities for warm waters. These include several representatives of the gastropod families Volutidae and Cerithiidae, which today are most common within the tropics. Similarly, the Ostreida are found in modern tropical and temperate seas, and Pholadidae and Spondylidae are most common in these settings as well (Mikkelsen and Bieler, 2008).

Even though there is no striking pattern of change in mollusks through the PETM, we have new observations that indicate some differences between the Malian faunas and those of higher latitudes during the Paleogene. The gastropod families Strombidae and Volutidae, and the superfamily Cypraeoidea, and the bivalve families Pholadidae and Ostreidae are common and diverse components of the Paleogene Mali assemblage, and today these groups have a tropical to subtropical distribution. These groups were relatively less diverse in coeval subtropical regions of the early Paleogene. For example, comparisons with the well-studied faunas of the U.S. Gulf Coastal Plain (Sessa et al., 2012), show that a global greenhouse climate still resulted in biotas within tropical latitudes (like Mali) that were distinct from those in extratropical regions (like the U.S. Gulf Coast Plain), even though the latter experienced tropical temperatures (Sluijs et al., 2007). The Malian fauna instead shows stronger taxonomic affinities with nearby faunas in Nigeria, Egypt, and Saudi Arabia, which together form a truly tropical fauna.

Future Work

Our fieldwork circumscribes several areas of future research that should be a priority once northern Mali is again accessible to scientists. Several Malian fossil-bearing localities are extremely rich and extensive but underworked. Paleocene localities that yielded complete fossil turtles form very extensive outcrops and have potential to yield other key vertebrate species. There is also high probability of future discoveries of fossil mammals from the phosphate conglomerates of Tamaguélelt with further surface collecting alone. This may produce discoveries impacting our understanding of the early diversification of placental mammals and an important paleontological test of model-based diversification estimates. Skull material of Paleogene fossil sea snakes remains elusive worldwide, but, given the large numbers of vertebral specimens in the Tamaguélelt Formation, the rocks there may also ultimately yield cranial material. It is also noteworthy that significant fossil discoveries in the region have been made in older stratigraphic units other than those that are the focus of this monograph. Exposures of the Continental Intercalaire crop out quite extensively in the Iullemmeden Basin. The western side of the Adrar des Iforas has produced a variety of fossils, including abundant petrified logs (Metapodocarpoxylon libanoticum) and a relatively limited continental vertebrate fauna of titanosaurian sauropods, turtles, fish, and crocodyliforms. These Early Cretaceous localities are each dominated by coarse-grained sandstones with less abundant pedogenically modified mudstones that we interpret to be fluvial channel and floodplain deposits. They await further paleobiological exploration.

This project was initiated and funded primarily as a fossil-finding expedition, and there remains a tremendous amount of descriptive and analytical work in sedimentary geology to be done in this area including refined biostratigraphy and exploration. The area presents an important opportunity for future chronostratigraphers to deploy new and creative methods to date this succession of rocks. Such work, however, requires greater security on the ground than existed during the expeditions described here to permit a more protracted period of scientific data collection. Further reconstruction of the ancient freshwater drainage patterns of West Africa using Malian rocks, would also expand our understanding of sedimentary formations, and species habitats and ranges, particularly of ancient fish. Such work could inform our understanding the ancient height of several Precambrian massifs such as the Adrar des Iforas. We have discussed above regional similarities between rocks in Mali and those in other areas covered by the Trans-Saharan Seaway such as Niger and Nigeria that warrant further exploration. There are also intriguing sedimentological and paleobiological similarities between the Malian Trans-Saharan Seaway rocks and more distant deposits such as those of the Umm Himar Formation of Saudi Arabia (Madden et al., 1995) that warrant further comparison.

Given the recurrent deposition of five different facies we have described in the Late Cretaceous–early Paleogene rocks of Mali, establishing the relative ages of these facies through time has always been extremely challenging. Thus, a major outstanding project is the study of onshore-offshore stratigraphy and facies change throughout the Iullemmeden and Taoudenit basins and the Gao Trench such that detailed hypotheses of the timing of transgressions and regressions can be increasingly refined. Coupled with efforts to sample the fauna more densely and precisely, such an undertaking would facilitate a test of such hypotheses as, for example, whether marine regressions tend to cause extinction among terrestrial vertebrates, or whether marine transgressions tend to impact freshwater species most negatively (Archibald, 1996). Finally, the rocks of Niger and Nigeria have been in many ways more extensively studied than those of Mali. Improved correlation among formations regionally and across international boundaries will be fundamental to developing a fully integrated picture of ancient life in the Trans-Saharan Seaway.