UHC DAK: Faculté des sciences, Université Hassan-II, Dakhla collection (Casablanca, Morocco); UM: University of Michigan Museum of Paleontology (Ann Arbor, USA).

The Tarfaya–Laâyoun–Ad-Dakhla Basin (also named Boujdour or Aioun Basin) is one of a series of mature passive margin basins that lie along the Central Atlantic margins of Northwest Africa and North America (Davison and Dailly, 2010, and references therein). It extends for almost 1000 km along the African margin from the Cap Blanc Fracture Zone in northern Mauritania, north through southern Morocco to the intersection of the North Canary Island Fracture Zone and the South Atlas Fault (Fig. 1A, insert). The Tarfaya–Laâyoun–Ad-Dakhla Basin comprises two discrete sub-basins separated by the Dakhla Fracture Zone (Fig. 1A). The northern is the Boujdour sub-basin, and the southern is the Ad-Dakhla sub-basin. The latter is characterized among other things by the absence of Cenomanian-Paleocene sediments onshore (Fig. 1B; von Rad et al., 1982 and von Rad et al., 1979).

Nevertheless, almost in the middle of Ad-Dakhla Basin along the Atlantic coast, the Paleogene Samlat Formation overlies Cretaceous outcrops (von Rad et al., 1979 and Ratschiller, 1967). The Samlat Formation includes three members: •

a Paleocene Itgui Member with well-preserved foraminifera;

The Eocene Guerran Member of the Samlat Formation of interest here is primarily marine siliceous chalk, becoming more clastic farther onshore (Davison, 2005 and Ratschiller, 1967). In the vicinity of Ad-Dakhla the Eocene part of the Samlat Formation is exposed in cliffs along the Atlantic coast from south of El Argoub village to the military position commonly named “Garitas”, but it is more easily accessible on the flat outcrops south of “Garitas” (Fig. 2).

This sequence of strata (Fig. 3), all part of the Guerran Member, was described in detail by Adnet et al. (2010), who divided it into three units. The lower unit, Unit 1 of Adnet et al. (2010), includes some 22 m of rhythmically-bedded, chert-rich marine siltstones and marls. These are overlain by 1–1.5 m of vertebrate-bearing conglomeratic sandstone, another 4–8 m of rhythmically-bedded siltstone and marl, and a second 3–6-m unit of vertebrate-bearing muddy sandstone (Unit 2 of Adnet et al., 2010). Bonebeds B1 and B2 are both located in Unit 2. The highest unit, Unit 3 of Adnet et al. (2010), is a 2–3-m interval of sandy to bioclastic limestone of Neogene age.

Rhythmically-bedded, chert-rich marine siltstones and marls like those in Unit 1 and Unit 2 of the Samlat Formation are generally interpreted to reflect Milankovitch-periodic variation in the production and accumulation of siliceous plankton, diluted by fine-grained, wind-blown, terrigenous clastics and by carbonate ooze in an offshore deep-water environment (Decker, 1991). More massive, coarser-clastic sandstones with abundant shark teeth and vertebrate bone, like those of bonebeds B1 and B2 at the bottom and top of Unit 2, are conspicuously different and represent a higher-energy nearshore environment. Alternation of such different environments can only be explained by changing sea level. The conglomeratic and muddy sandstones of B1 and B2 represent low sea stands in what is otherwise a deeper water sequence.

Adnet et al. (2010) assigned a Late Eocene age to Unit 2 with bonebeds B1 and B2 based on the selachian fauna. Bonebed B1 has provided a large number of vertebrae of cetaceans belonging at least to five different species, with possible rib fragments of sirenians and also some remains of crocodiles, turtles, sea snakes and birds. Upper bonebed B2 yielded cetacean remains attributed to Basilosaurus sp. and common remains of a dugongid.

CETACEA Brisson, 1762

ARCHAEOCETI Flower, 1883

Basilosauridae Cope, 1868

Saghacetus Gingerich, 1992

cf. Saghacetus sp.

(Fig. 4A–D, G)

Referred specimens and measurements: Table 1.

Provenance: All specimens for which the provenance is known come from bed B1 of Adnet et al. (2010).

Description: The most interesting specimen is DAK-1 (Fig. 4A–B), a well-preserved juvenile right maxilla with dP^2-4 representing a very small basilosaurid. The teeth are all long, narrow, and low-crowned relative to adult teeth at the same position. Deciduous dP² is simple with anterior and posterior but no accessory roots; dP³ has anterior and posterior roots, but also a small medial root visible on the buccal side of the tooth and a prominent protocone swelling of the crown and root visible on the lingual side; dP⁴ lacks the medial root but has a prominent protocone swelling and root like that on dP³. There are deep embrasure pits located posteromedial to all three teeth (the latter to receive the apical cusp of M¹). The midline suture joining the maxillae is preserved, indicating that the rostrum was narrow. A suture for the premaxilla is preserved along the dorsal margin of the maxilla, and an infraorbital canal opens in via a 9 mm maxillary foramen above the anterior root of dP³.

The best preserved thoracic element is a simple centrum, DAK-20, that is ‘D-shaped’ in transverse section, with the flat part of the D, the dorsal surface, flooring a broad neural canal. Facets for rib capitula are small and high on the centrum, indicating an anterior thoracic. The best preserved lumbar centrum is DAK-7 (Fig. 4C–D), which is elliptical in transverse section. Left and right pedicles of the neural arch are well separated, indicating that this is an anterior lumbar. The better of the two distal humeri is DAK-31 (Fig. 4G). This has a prominent deltopectoral crest on the anterior margin of the diaphysis, rising 45 mm above the distal end of the bone. The curved distal articular surface is 24 mm in diameter, but the articular surface itself is damaged, as are the lateral and posterior surfaces of what remains of the diaphysis.

Discussion: Archaeocete teeth form in the jaws before they erupt, and these teeth do not continue to grow after eruption. The definitive lengths of deciduous and permanent cheek teeth (premolars and molars) in juvenile and adult Zygorhiza and Dorudon (Kellogg, 1936 and Uhen, 2004) can be used to enable prediction of M¹ length in DAK-1: four specimens with dP³and M¹ yield a prediction of 15.0 mm, and three specimens with dP⁴ and M¹ yield a prediction of 18.5 mm for M¹ length in DAK-1. These predicted M¹ lengths indicate that DAK-1 represents an archaeocete the size of Saghacetus osiris from the Middle Priabonian, Late Eocene of Egypt, which is distinctly small for a basilosaurid. There are no dental characters on deciduous teeth that enable one basilosaurid to be distinguished from another, but the sizes of deciduous and permanent teeth limit possible identifications. We identify the taxon represented here as cf. Saghacetus sp. based on size, pending recovery of more complete specimens. No other basilosaurid known is this small.

The vertebrae and humeri included in cf. Saghacetus sp. are considered to represent the same taxon as the maxilla because they are similar in size, but none of the elements described here were found associated and more than one taxon may be represented.

Stromerius Gingerich, 2007

cf. Stromerius sp.

(Fig. 4E–F, H–J)

Referred specimens and measurements: Table 2.

Provenance: All specimens for which the provenance is known come from bed B1 of Adnet et al. (2010).

Description: Specimens described here are distinctly larger than those included in cf. Saghacetus sp. They are referred to cf. Stromerius sp. on the basis of size. DAK-10 (Fig. 4E–F) is an anterior lumbar with a centrum more circular in transverse section than is typical for Stromerius. DAK 21 is a posterior lumbar vertebra, and DAK 35 (Fig. 4I–J) is a posterior lumbar or anterior caudal vertebra. Both are more elongated anteroposteriorly than is typical for Stromerius. The vertebral centra of cf. Stromerius sp. described here are comparable in length to those of Dorudon cf. D. atrox, but they are conspicuously smaller and more similar to Stromerius in width and height.

DAK-36 (Fig. 4H) is a left humerus of an adult that is complete except for a missing proximal epiphysis, with a large bite mark perforating the lateral surface of the midshaft. There is no cancellous bone that would indicate this to be immature or subadult. The curved distal articular surface is 33 mm in diameter. The distal end of DAK-36 is about 30% wider than that of humeri here attributed to cf. Saghacetus sp., and it is about 33% narrower than the distal ends of humeri of Dorudon atrox from Egypt, making it likely that DAK-36 belongs with other specimens grouped here as cf. Stromerius sp. DAK-36 is smaller but has a very prominent deltopectoral crest like that of Dorudon atrox from Egypt (Uhen, 2004).

Discussion: cf. Stromerius sp. almost certainly represents a genus and species new to science, but known specimens are inadequate to justify naming a new taxon.

Stromerius nidensis is known from Middle Priabonian, Late Eocene, strata in Egypt, as is Saghacetus osiris. Identification of Dakhla specimens from bed B1 as cf. Saghacetus sp. and cf. Stromerius sp. is consistent with a Priabonian age for the bed. Basilosaurids are rare before the Priabonian and not known after the Priabonian.

Dorudon Gibbes, 1845

cf. Dorudon atrox (Andrews, 1906)

(Fig. 5A–E)

Referred specimens and measurements: Table 3.

Provenance: All specimens for which the provenance is known come from bed B1 of Adnet et al. (2010). Some specimens received from local collectors may have come from bed B2.

Description: Cranial fragments are known, but none is complete enough to describe here. The most informative elements are vertebrae. These include DAK-37 with pieces of cervicals C1 (atlas), C2 (axis), and C7, and the capitular process of thoracic T1 or T2. DAK-18 is the centrum of a middle cervical, probably C4; DAK-16 is the centrum of cervical C7; and DAK-17 is the centrum of a first thoracic T1. DAK-14 is the centrum of a middle thoracic vertebra, and DAK-28 is the centrum of a posterior thoracic with the capitular facets on short but massive transverse processes.

DAK-2 (Fig. 5A–E) is an associated series of nearly identical posterior lumbar or anterior caudal vertebrae. One vertebra has the anterior epiphysis fused to the centrum, showing that the series came from an adult specimen even though epiphyses are missing on the other vertebrae. These we excavated from sand in bed B1, where they were found associated within a few centimeters of each other but not articulated (meaning we do not know their sequence in life). The centra are large for their length, cylindrical, and match the size and form of Dorudon atrox. Each includes at least one of its two transverse processes. These are broad anteroposteriorly, flat dorsoventrally, and angled slightly downward and slightly anteriorly relative to the centrum axis. The neural arch is well preserved on one lumbar in the series (Fig. 5A–B), and this specimen has a neural spine that is elongated anteroposteriorly and compressed bilaterally. It rises to a moderate height dorsally above the centrum.

The other lumbar and anterior caudal centra listed in Table 3 are a little smaller or larger than the centra of DAK-2, but lie within the range of variation expected for Dorudon atrox. DAK-23 is a relatively small near-terminal posterior caudal with a centrum tapering posteriorly. This too is similar to terminal caudals found in D. atrox in being compressed dorsoventrally relative to its width, indicating that it was in the caudal fluke in life.

Finally, DAK-37 includes the distal end of a rib R1 measuring 26.2 × 35.9 mm in anteroposterior and transverse diameter, respectively, that is the right size to belong to Dorudon atrox.

Discussion: Each of the elements described here is indistinguishable in size and shape from the homologous element in a sample of Dorudon atrox (Uhen, 2004). We identify the Ad-Dakhla specimens as African D. atrox (Andrews) rather than North American D. serratus Gibbes because of their provenance and because D. atrox is much better known anatomically. The two species D. atrox and D. serratus are the same size and only questionably distinguishable.

Dorudon atrox is abundant through all Early Priabonian strata in Egypt, and identification of Ad-Dakhla specimens from bed B1 as cf. D. atrox is consistent with an Early Priabonian age for the bed. The type specimen of D. serratus is also Early Priabonian in age (Uhen, 2013).

cf. Dorudon sp.

(Fig. 5F–I)

Referred specimens and measurements: Table 4.

Provenance: Both specimens were found in bed B1 of Adnet et al. (2010).

Description: DAK-42 (Fig. 5F-G) is the centrum of a lumbar vertebra similar in proportion to those of Dorudon but substantially larger. The centrum is massive and distinctly amphicoelous in having slightly concave anterior and posterior epiphyses, but little more can be said due to its poor preservation.

DAK-4 (Fig. 5H–I) is the diaphysis of the proximal phalanx of digit II of a left manus. Proximal epiphysis is missing, but the distal end is well formed with an angled distal articular surface very much like that of D. atrox illustrated by Uhen (2004: p. 118). The phalanx is here referred to cf. Dorudon sp. because the diaphysis is much larger in diameter than its homolog in Dorudon atrox (19.8 and 16.7 mm compared to 14.0 and 14.0 mm), and distinctly shorter than its homolog in Basilosaurus isis (68.0 mm compared to 83.0 mm).

Discussion: Specimens of cf. Dorudon sp. are not very informative, but they do indicate presence of a large Dorudon-like species at Dakhla that is intermediate in size between Dorudon atrox and Basilosaurus isis.

Basilosaurus Harlan, 1834

Basilosaurus isis (Beadnell, in Andrews, 1904)

(Fig. 6)

Referred specimens and measurements: Table 4. Additional specimens were mapped in the field.

Provenance: Both specimens described here came from bed B1 of Adnet et al. (2010), but this or a similar species is also known from bed B2.

Description: DAK-43 is part of an articulated skeleton excavated by commercial collectors. Eight vertebrae representing the posterior lumbar or anterior caudal portion of the skeleton were left in the field. These average 320 mm in centrum length, the neural spine is relatively short anteroposteriorly, and the prezygapophyses end well short of the anterior margin of the centrum.

DAK-44 included several vertebrae that were associated but not articulated. These are similar in size and form to those of DAK-43.

Discussion: Vertebrae of Basilosaurus described here match corresponding elements of Basilosaurus isis from Egypt in size, neural spine length, and prezygapophysis length. The size of DAK-43 and 44 is an important difference from lumbar and anterior caudal vertebrae of the larger species Basilosaurus cetoides (Owen, 1841). The relatively short neural spine (short anteroposteriorly) and the relatively short prezygapophyses ending well short of the anterior margin of the centrum are both important differences from Bartonian Basilosaurus drazindai (Gingerich et al., 1997).

Basilosaurus isis is abundant through all Early Priabonian strata in Egypt and Jordan (Zalmout et al., 2000), but it is not known from Bartonian nor from Middle and Late Priabonian strata. Identification of Dakhla specimens from bed B1 as B. isis is consistent with an Early Priabonian age for the bed.

SIRENIA Illiger, 1811

Family Dugongidae Gray, 1821

Eosiren Andrews, 1902

cf. Eosiren sp.

(Fig. 7)

Referred specimen: UHC DAK-45, a partial skeleton. Additional specimens identifiable as sirenian were mapped in the field.

Provenance: All specimens for which the provenance and identification are known come from bed B2 of Adnet et al. (2010; Fig. 7A), but rib fragments that may be sirenian were also found, rarely, in bed B1.

Description: The best specimen of cf. Eosiren sp. known from Ad-Dakhla is DAK-45, which is illustrated in Fig. 7B. This is an association of left and right ribs. Most are moderate in size and osteosclerotic, but several are also pachyostotic like the anteriormost ribs of Eosiren. The largest of these measure 38 × 34 mm and 40 × 35 mm in transverse diameter. None of the ribs is complete, so it is not possible to give a length for these.

Discussion: DAK-45 is an association of left and right ribs similar in size and form to those of Eosiren libyca from Middle Priabonian deposits of Egypt (Zalmout and Gingerich, 2012). The preserved pieces are insufficient to be certain of their identification.

The ribs of DKA-45 are osteosclerotic throughout, meaning that they fragment into pieces when exposed by erosion. One neural spine was found embedded in sand with the ribs, suggesting that vertebrae may have been part of the specimen when it was buried. If so, then these have either been destroyed by erosion or removed by commercial collectors.

Measurements for Ad-Dakhla archaeocete lumbar and anterior caudal vertebrae given in Table 1, Table 2, Table 3 and Table 4 are plotted on natural logarithmic axes in Fig. 8, where the vertebrae form clusters. Ellipses approximate a 95% confidence range for each cluster. These ellipses span 0.6 units on the length axis and 0.4 units on the width axis, reflecting the scatter observed in 75 posterior thoracic, lumbar, and anterior caudal vertebrae in five specimens of Dorudon atrox from Egypt (measurements in Uhen, 2004). Ellipses represent the scatter to be expected in archaeocete species of any size when represented on logarithmic axes. There are clearly a minimum of five species represented: cf. Saghacetus sp., cf. Stromerius sp., cf. Dorudon sp., and Basilosaurus isis do not overlap at all, and the range of variation in the overlapping cf. Stromerius sp. and cf. Dorudon atrox clusters is too great to represent a single species. All five of these species were found in bed B1, and Basilosaurus isis was found in bed B2 as well.

The basilosaurid partial tooth illustrated by Adnet et al. (2010) is similar in size to P³or P⁴ of Saghacetus osiris, but it is conspicuously lower crowned. It is probably a dP³ or dP⁴ of cf. Stromerius sp. Identification of three basilosaurid species from the Samlat Formation, Basilosaurus sp., Dorudon sp., and a small basilosaurid (Field et al., 2011), is confirmed, with the addition of a fourth small basilosaurid species, a fifth intermediate basilosaurid species, and a sirenian. It is not clear, contrary to Uhen's 2010 entries in the Paleobiology database that the three species of Field et al. (2011) come from bed B1 and from bed B2. The dugongid sirenian cf. Eosiren is known with certainty only in bed B2.

The depositional environment at Ad-Dakhla parallels that near the base of the Qasr el-Sagha Formation in Egypt (Peters et al., 2009) in the sense that a marine sequence (units 1 and 2 of Adnet et al., 2010) is interrupted by a coarser clastic unit (bed B1) that is rich in bones that are abraded and reworked in some cases, but fresh and intact in others. The difference is that the rhythmically bedded siltstone with gray chert and marl of the Samlat Formation on the Dakhla coast was deposited in deeper water (Decker, 1991), but in both cases the coarser clastic unit represents a temporally-condensed interval related to substantial lowering of sea level. Bones that are reworked and broken were eroded from an earlier nearshore shelf environment higher on the shoreline, and bones that are fresh and intact were deposited locally where a nearshore shelf environment replaced deeper water sediment accumulation.

This is important because there are relatively few low sea stands in the Priabonian. Peters et al. (2009) concluded that the low sea stand in Egypt mixing larger and older Dorudon and Basilosaurus archaeocetes with smaller and younger Saghacetus and Stromerius archaeocetes is Pr-2, which separates the Early and the Middle Priabonian (see also Gingerich et al., 2012). Adnet et al. (2010) regarded the interval, unit 2, spanning fossiliferous beds B1 and B2, as probably Late Eocene in age. Here we go farther and, on the basis of the archaeocete association, correlate bed B1 with the Pr-2 low sea stand in Egypt. We cannot prove that two or more marine mammal faunas are mixed in Ad-Dakhla bed B1 but we suspect this from similarity to the taxonomically mixed fauna in Pr-2 in Egypt; from similarity to the taphonomically mixed collection of abraded and reworked vertebrae in Pr-2 in Egypt, with fresher material including partially articulated and associated skeletons; and from the mixture of reworked early Middle Eocene and Late Eocene sharks at Ad-Dakhla reported by Adnet et al. (2010).

Adnet et al. (2010) reported that Ad-Dakhla beds B1 and B2 yield the same species assemblages of sharks. These beds are separated by 2-5 meters of rhythmically-bedded siltstone with gray chert and marl, indicating a return to deeper water deposition farther offshore (Decker, 1991). The rhythmically-bedded unit required a substantial amount of time to accumulate. Gypsum in the Garitas section (Fig. 2B) suggests a lagoonal environment with evaporates. Gypsum is found in the transition from offshore marine strata overlying bone bed B1 to the nearshore sands of bone bed B2. We cannot estimate the time difference between bone beds B1 and B2, but the mammalian fauna of bed B2 is conspicuously different from that of bed B1, being dominated by dugongid sirenians rather than archaeocete cetaceans. Bone bed B2 probably represents a second low sea stand, either a later phase of Pr-2 in the Middle Priabonian, or possibly Pr-3 separating the Middle and the Late Priabonian. Bed B2 does not have the taxonomic mixture of archaeocetes seen in bed B1, nor does it have the same taphonomic mixture of abraded and fresher specimens.