Microvertebrate fauna from Gadoufaoua (Niger, Aptian, Early Cretaceous)

Cretaceous


INTRODUCTION
Continental microvertebrate faunas are important because they provide abundant data concerning the biodiversity of a fossil site, as well as providing information concerning possible continental faunal interchange. In addition, fossils of some taxa (for example, lissamphibians, some sauropsids and mammals) are only represented by microremains, because of the minute size of many of these taxa during the Mesozoic. Thus, to be complete, the study of a fossil site must include consideration of the "microremains" component. Furthermore, microvertebrate studies make it possible to test palaeobiographical scenarios based mostly on macrofaunas (Grigorescu et al. 1999;van den Berg et al. 2012;Haddoumi et al. 2016;Lasseron 2019;Lasseron et al. 2020).
To date, only a few African Mesozoic continental microvertebrate faunas have been described: these are known from the Middle Jurassic of the Mahajanga Basin, Madagascar (Flynn et al. 2006), Middle and Late Jurassic to Early Cretaceous of Anoual, Morocco (Knoll 2000;Evans & Sigogneau-Russell 2001;Haddoumi et al. 2016;Lasseron 2019;Lasseron et al. 2020), Early Cretaceous of the Douiret Formation, Tunisia (Cuny et al. 2010a;Fanti et al. 2012) and from the Late Cretaceous of Wadi Milk, northern Sudan (Rauhut & Werner 1995).
Gadoufaoua in the remote Ténéré Desert of Niger is one of the most famous palaeontological sites of Africa. Known since the 1960s, this locality has yielded numerous Lower Cretaceous (Aptian-Albian) vertebrate remains (see paragraph History of exploration). However, until this study, there were no data available concerning the microfauna associated with these remains.
We describe here, for the first time, microvertebrate remains from this locality. The objective of this study is to update the faunal list of the Gadoufaoua deposits by considering data provided by the microfauna in order to compare this fauna with those of other localities of similar ages. The purpose is then to highlight the existence of potential palaeobiogeographical affinities between these localities, in a global context preceding the opening of the South Atlantic Ocean, as has previously been done using some groups, such as the crocodilian (Buffetaut & Taquet 1977;Buffetaut 1981) and actinopterygian faunas (Maisey 2000).

GEOGRAPHIC, GEOLOGICAL AND HISTORICAL SETTINGS
GeoGraphical and GeoloGical context The Gadoufaoua deposits are located southeast of the Aïr massif in the South-central Sahara Desert. The outcrop of fossiliferous strata is about 2 km wide over a length of approximately 180 km with an area of 360 km 2 (Fig. 1;Taquet 1976).
Gadoufaoua is located in the Iullemmeden Basin, which extends from Algeria in the north over Mali and Benin in the west, Chad to the east, and northern Nigeria to the south (Taquet 1976). The continental Mesozoic deposits of the Iullemmeden Basin are currently divided into the Agadez, Irhazer, and Tegama groups (Moody & Sutcliffe 1991; Fig. 1). The fossiliferous beds in the region of Gadoufaoua are part of the Tegama Group. They were subdivided into eight units, numbered from 1 to 8 and bearing the abbreviation GAD by the French geologists of the Commissariat à l'Énergie Atomique (CEA 1965 ; Fig. 1). The studied material comes from unit GAD 5, which also yielded numerous macroremains of vertebrates (see below) and corresponds to the top of the Elrhaz Formation (Faure 1966;Taquet 1976). This stratigraphic unit comprises an alternation of clay and sandstone banks, often channalized, with a thickness of up to c. 60 m. The site of Gadoufaoua was dated by geological correlation with other Saharan localities and utilising its vertebrate associations (Taquet 1976). Currently, it is dated as Aptian (c. 125 to 113 Ma). Evidence for a more precise range within the Aptian is lacking.  (Sankey & Baszio 2006). The benefits of studying such assemblages are numerous. Given the large sample size, the probability of obtaining "rare" taxa (otherwise not often present in the macrofauna) is much higher. By analyzing the data produced, it is possible therefore to focus on many topics, including taphonomy, the reconstruction of palaeoenvironments and palaeobiogeography (Sankey & Baszio 2006). Although microremains are often highly abraded and fragmented, rendering their identification problematic, their number compensates for this and it has been proven that they can be considered reliably representative of local palaeobiodiversity (Rogers & Brady 2010). identification of the material The material studied here comes from processing a small fraction (less than 100 kg) of the six tons of sedimentary material obtained from the GAD 5 unit in 1970 by Taquet and his team and now housed at the MNHN. Only this small sample was studied for reasons detailed in the discussion (Comparison of the GAD and GADb sites).
After being separated from the lithoclasts by acid digestion and sieving, as well as visual inspection, the fossils from this sample were sorted and identified with a binocular microscope, and then counted and catalogued.
The fossils were examined using a Leica MEB115 binocular stereomicroscope, equipped with auxiliary fiber-optic lighting (Fiber-Lite MI-150) and a Zeiss Discovery V8 stereomicroscope, equipped with Leica LED1000 ring light. The observations and image acquisitions were made using an optical platform (Nikon SMZ1270 and Photonic Optics auxiliary lighting) under Archimed, then processed using Photofiltre 7. Drawings were made through a camera lucida associated with the binocular microscope. Diversity calculations and statistical analyses were performed using Past (version 3.25, Hammer et al. 2001) and RStudio (version 1.1.463, R Core Team 2014).
countinG methods All specimens were counted. When the number of specimens was relatively small (n < 100), counting was done manually, but in the case of larger batches, it was performed using the Python script "Count Image Elements", developed by Auréliane Gailliègue (PhD student, Institut des Sciences du Calcul et des Données). The principle is quite simple and based on image segmentation. The material is spread on a uniform black or white background and photographed (Fig. 2), after ensuring that none of the fossils were in contact with each other. This image is then imported into the Python environment, where the script counts the number of elements based on the differences in contrast between the fossils and the background. This script has already been tested and used several times (Lasseron et al. 2020;Allain et al. 2022) and has always given statistically acceptable results (at a 95% threshold). It was therefore considered sufficiently reliable to allow exhaustive counting with a relatively small and acceptable margin of error. Some 3703 microvertebrate specimens were obtained from the Gadoufaoua sample forming the basis of this study and were counted using the above methods. The specimens were then identified (whenever possible). At least 20 species are present in the assemblage.

separation of sites
We do not have information concerning the exact position of sites from which the material was collected. However, some of this material was stored at the MNHN in a separate box. During sorting and identification of samples, we found this material to have a lighter color and a better quality of fossil preservation than other seen in other samples. This suggested a different depositional environment to that of other samples so, as a result, this sample was separated from the rest.
It is referred to below as "site GADb". Fossils from this sample were then compared to those obtained from the other lithology in order to determine whether the potential difference in environment of deposition is reflected in faunal compositions. This permitted a determination as to whether or not a distinction has to be made between the two samples.  description Forty tooth crowns are tentatively referred to this taxon, but all lack roots. The crowns are between 2 and 4 mm long. Crowns are 0.5 to 1 millimeter in height. They are about one millimeter thick, longer than wide and have a flat diamond or hexagon outline in occlusal view (Fig. 4C). This is a diagnostic character for the genus Tribodus. The occlusal surface is almost flat but bears a slight mesiodistal depression. All specimens are probably anterolateral teeth (Brito 1992

description
The teeth are between 2 and 5 mm high. They are convex labially. Roots are not preserved. The main cusp is not very high but wide and slightly inclined lingually and the accessory cusps are also wider than those of Egertonodus Maisey, 1987 and

A B
convex than the labial one. These teeth lack the more numerous wrinkles occuring on the teeth of Egertonodus

description
The teeth are about 5 mm in diameter with a crown height of 1 to 3 mm. Roots are not present or poorly preserved. Some teeth have a circular acrodine cap in occlusal view, others have a horizontal wear facet. The crown is low and rounded. Neither ornamentation nor carinae are present. This morphology is reminiscent of some Ginglymodi (Cavin 2010; Grande 2010). Such teeth are very common in Jurassic and Cretaceous deposits and are often referred to as oral teeth of Lepidotes Agassiz, 1832 (Pouech et al. 2015). Gadoufaoua is the type-locality of another ginglymodian, Pliodetes nigeriensis Wenz, 1999(Wenz 1999Lopez-Arbarello 2012), but this taxon is edentulous so it is unlikely that these teeth belong to this genus. The ganoid scales are thin and diamond-shaped, with "peg and socket" joints as is the case in P. nigeriensis specimen MNHN-GDF-1275. However, as for the teeth, they do not bear enough diagnostic characters to attribute them with confidence to this taxon. As a result, both the teeth and scales are attributed to Ginglymodi gen. sp. indet.

Order †PYCNODONTIFORMES Berg, 1937
Pycnodontiformes gen. et sp. indet. description MHNH-GDF-m 19 is a prearticular or vomer fragment with a surface area of 6 mm 2 . Exact determination is problematic due to the fragmentary nature of the specimen. It bears two rows of three teeth all of which are 1 mm in diameter. They are circular in occlusal view and all have a central depression. Pochat-Cottilloux Y. et al.
The vomer fragments have molariform teeth symmetrically distributed along the antero-posterior axis of the bone. Vomerine teeth are ovoid in occlusal view and about 2 mm long. They are low in labial view. The occlusal surface is flat and has a marked mesiodistal edge, from which about twenty vertically oriented wrinkles terminate.
Incisiform teeth have unadorned surfaces and are flattened labiolingually. In labial view, they are wider than high (2 to 3 mm high) and slightly convex in lateral view. There is a clearly noticeable edge in occlusal view. This edge is almost horizontal in labial view.
The molariform teeth are rounded or ovoid in shape in occlusal view and the crest and wrinkles that are present on some in occlusal view are strongly reminiscent of pycnodontiforms (Kriwet 2002;Pouech et al. 2015;Szabó et al. 2016;Cooper & Martill 2020

description
The tooth crowns are between 4 and 6 mm high, labiolingually compressed with a spearhead-like shape. The mesial and distal edges are overdeveloped and fused with the apical cap. The labial side is more convex than the lingual side. When the root is preserved, it is circular with an ornamentation of vertical wrinkles. There is a strong constriction between the root and the crown. This morphotype is often referred to Caturidae Owen, 1860 (Lambers 1994;Forey & Sweetman 2011   description A small tooth plate (3 mm in length) was found among the Gadoufaoua microremains. It is unclear if this plate belongs to the upper or lower dentition. Five distinct straight crests, bearing dentin bumps and radiating from the mesiolingual angle, are preserved. The most mesial crest is oriented almost at right angles to the internal margin of the tooth plate, while the most distal crest is subparallel to it. The size of the crests decreases along the mesiodistal axis. The distance between the ridges increases from the distal side to the mesial side. The plate is ornamented: it has multiple small perforations that are randomly distributed but which are limited to inter-ridges spaces. The right angle of the tooth plate, its length about twice longer than its width, and the radiating straight ridges oriented posteriorly except for the first one are reminiscent of Arganodus tiguidiensis Tabaste, 1963 and permit distinction from species of Ceratodus Agassiz, 1838, in particular C. africanus Haug, 1905(Goodwin et al. 1999Soto & Perea 2010;Alves et al. 2013 (2015), but no specimen with five crests is known for this species. This crest count is known for Neoceratodus africanus Haug, 1905(Tabaste 1963, but the general morphology of the Gadoufaoua specimen does not permit assignment to this species (see Cavin et al. 2015: fig. 2C). Ptychoceratodontidae are characterized by tooth plates bearing four to six radiating crests (Martin 1982;Skrzycki 2015), but no definitive diagnosis of this family has yet been provided (Skrzycki 2015). The genus Ptychoceratodus, in particular, is characterized by the presence of occlusal pits in the inter-ridge furrows, tooth plates of triangular shape and with five to six ridges and slender ridges as well as deep inter-ridge furrows. These characters are reminiscent of the Gadoufaoua specimen (Skrzycki 2015; Bhat & Ray 2020) which is thus assigned to Ptychoceratodontidae, and with less certainty to Ptychoceratodus.  description The scapulae measure about 5 mm long by 3 mm wide.
Although they are fragmented, a flat lateral face and a concave internal face (ventral/dorsal view) can be distinguished. The dorsal part is slightly bifid and more developed than the ventral part. The glenoid cavity, when preserved, is subcircular. The acromial process, noticeable in some specimens, is well developed ventrally. The scapular notch between these two parts is not noticeable. This morphotype is diagnostic of anurans (Vullo 2007).
The ilia measure between a few millimeters and 10 mm in length depending on their state of preservation. Specimens bear a rounded acetabulum, but the ridge defining it is not well defined. The ilium blade has a dorsal ridge (iliac ridge). It appears slightly curved. This elongated ilial morphotype is characteristic of anurans (Sweetman & Evans 2011).
The two urostyles characteristic of this taxon (Rocek & Nessov 1993;Blanco et al. 2016), measure 5 and 10 mm. They are straight and thin. The proximal end is preserved on one specimen, but it is severely damaged and deformed. The articulation is bicondylar and there are no transverse processes. The condyloid pit is ellipsoidal and filled with sediment. On the dorsal surface, a central depression of varying depth between the two specimens is present along the entire length of the bones.
Zeugopods, often fragmented, comprise two fused bones, characteristic of anurans (Rocek 2000). They measure between 5 and 7 mm. Their cross-section is figure of eight-shaped. They are too incomplete to bear diagnostic characteristics that would permit a more accurate anatomical (radioulna, tibiofibula) or taxonomic assignment. description Among the remains collected, plate fragments of this taxon range from 25 to 50 mm². They are wide in lateral view. The fused ribs clearly noticeable on the pleural plates allow them to be identified as chelonian plates. Folding of the shell boundary of the peripheral plate is also characteristic of chelonians.
There is practically no ornamentation on the plates, which, among the chelonians previously observed at Gadoufaoua, is characteristic of Teneremys lapparenti (de Broin 1980).

description
The plate fragments have similar dimensions to those of Teneremys lapparenti. However, they are thinner in lateral view. Ornamentation is clearly visible, consisting of small granules and tight bulges, a diagnostic feature of this Gadoufaoua taxon (de Broin 1980), and which also distinguish it from Francemys gadoufaouaensis Pérez-García, 2019 (Pérez-García 2019b). There are variations between specimens in the regularity of the ornamentation (depending on the position of the plate on the animal). The pleural plates were determined by the presence of fused ribs that are clearly visible. Their low convexity indicates a very low shell (de Broin 1980). description Tooth crowns measure about 5 mm high. They are rather fragmented and subcircular in cross-section, with a basal diameter that varies from 2 to 4 mm. Mesial and distal carinae are clearly visible. When the enamel is preserved, it is unornamented. After comparison with many specimens from the MNHN collections (material not inventoried), they could belong to juvenile individuals of Sarcosuchus imperator de Broin & Taquet, 1966.
description These teeth are 3 to 4 mm high and are triangular in labial and lingual views. They are flattened labio-lingually. Carinae, lacking ornamentation, are clearly discernible. The base of the crown tends to split in two. Apically, there is a pronounced lingual curvature.
description These teeth are very elongated and rather slender. They measure between 5 and 7 mm in height and are highly compressed labio-lingually with unornamented carinae. They are lingually curved from base to apex.
note The identification of the last two morphotypes remains uncertain. They could represent two of the known Gadoufaoua crocodyliform taxa previously described but could also belong   (Fig. 13A, B).

description
The teeth are between 8 and 12 mm high. They are slender and lingually curved. The root is not preserved. There is a very strong labiolingual compression, which distinguishes them from the teeth of crocodiles. The state of conservation of the teeth does not permit determination of ornamentation, if present. The cross section is oval.
description These thin teeth are not well preserved, most of the apical parts are missing. They have a height between 0.8 and 1.3 mm. There is also very strong labiolingual compression, but they are not curved and there is no ornamentation. The cross-section is ellipsoidal.
note These two morphotypes are characteristic of Ornithocheiroidea sensu Unwin, 2003(Pentland et al. 2019 Other fragments represent only the basal part of the crown. These are highly compressed labiolingually and therefore have an oval cross-section. The carinae are poorly marked but extend along the entire length of both sides of the base. The denticles are poorly preserved and number 4 per mm. For all specimens, the denticles are weakly apically hooked. For both the teeth and the carinae, the shape of the denticles and their number are typical of theropod teeth and could be assignable to Abelisauridae (Sereno & Brusatte 2008;Hendrickx et al. 2015).   (Fig. 14).
description This tooth is very fragmented and poorly preserved. A small part of the crown can be seen together with one root. The molar is 1.5 mm long, 0.25 mm wide and 2 mm high.
Only the protocone, very pinched, is visible, but based on the shape of the crown in occlusal view, it can be assumed that it would be a tribosphenic molar. The trigon basin is well marked. This tooth is highly compressed mesiodistally (clearly visible at the level of the protocone). The crown is higher on the labial side than on the lingual side. All these characteristics combined (especially the pinched protocone) suggest that this tooth belongs to the stem-Boreosphenida ). An updated taxonomic list is available in Appendix 2, including these new taxa together with those identified in previous studies (de Broin & Taquet 1966;Taquet 1976;de Broin 1980;Sereno et al. 1998Sereno et al. , 2001Sereno et al. , 2003Sereno et al. , 2007Wenz 1999  diversity The relative abundance of taxa (Fig. 15A) was determined using MNE (Minimal Number of Elements). The number of individuals cannot be estimated due to limited and biased representation of paired elements (e.g. limbs). Initially, the Gadoufaoua microfauna appears to be dominated by actinopterygians (79%). One might think that there is an over-representation of the latter because of the large number of teeth and scales found (22% and 21% of the sample respectively). A second analysis was therefore carried out excluding them (Fig. 15B) in order to reduce the over-representation bias as much as possible. Only teeth and vertebrae are counted in the second analysis. Here, actinopterygians represent only 33% of the fauna, which is now dominated by sauropsids. These analyses are made without considering unidentified material which represents 36% of the assemblage (percentages are therefore recalculated). Unidentified material consists of 972 fragments of skeletal elements and 25 osteoderms (belonging to two different morphotypes). Other diversity graphs are provided in Appendices 3; 4. DISCUSSION comparison of the Gad and Gadb sites As mentioned above (paragraph Separation of sites), some lightercolored material containing better-preserved specimens was found in the original collection. It was not known whether or not this came from the same horizon as the other bulk samples.
As illustrated in Figure 16, the GADb level rarefaction curve does not show a plateau. The sampling effort is therefore insufficient to perform α (within a sample) and β (between samples) diversity analyses, which would have been useful to interpret the diversity (Sankey & Baszio 2006). However, the relative abundance of taxa from the two sites of collection can be compared and appear to be approximately identical (Fig. 17), except for the teeth of ichthyodectiforms and pterosaurs which have not been found at GADb. A Wilcoxon test was performed for matched samples with the following assumptions: H0, the two levels can be considered identical; H1, the two levels cannot be considered identical. The p-value is 0.059, so at a 95% confidence interval, H0 cannot be rejected. In other words, the palaeontological content of the two sites is similar, and the material from GADb has been integrated with the rest of the material.
Although it is acknowledged that more data could be gathered by studying the rest of the sediments and agreed that an extensive study of the 6 tons of sediments would be ideal, several issues prevent it. First, the rarefaction curve in Figure 16 shows a plateau for the "GAD" sediments (which can now be associated with GADb) meaning that the locality was sampled sufficiently to account for the paleodiversity of the ecosystem, so more extensive studies would bring much less data and few new occurrences (if any). Second, because the preparation and sorting of sediment is very time-consuming, only one sample of the material was processed and studied. Third, the collection was interesting because it was associated with the GADb sample, which in view of its sedimentological and preservational differences to the bulk of the collection, warranted further study. Finally, although the remaining 5900 kg are accessible, unfortunately they are not as well preserved or conserved throughout the years, especially from humidity, which may have resulted in a loss of identifiable fossils affecting perceived abundance and diversity.

depositional environment
The microvertebrates found at Gadoufaoua are mainly aquatic and some exclusively fresh water (exclusively terrestrial taxa  The two faunas are linked, but they are not identical representing different palaeoenvironmental influences. The Santana Formation was deposited in a lagoonal environment with stenohaline taxa (Araripichthys Silva Santos, 1985, Tharrias Jordan & Branner, 1908, Axelrodichthys Maisey, 1986Maisey 1991) while Gadoufaoua sometimes represents a much quieter continental environment, where the preservation of typically terrestrial taxa is possible (dinosaurs, mammals, etc.) but where a strong aquatic influence is still felt.

palaeobioGeoGraphic implications
The similarity between the fauna of the Santana Formation and that of Gadoufaoua (although based on preliminary data) is consistent with palaeogeography. At this time (Aptian-Albian), Gondwana had already broken-up and the South Atlantic Ocean was opening. South America and Africa were separated but this separation was quite recent having taken place during the Barremian to Middle Albian (Pletsch et al. 2001 The taxonomic distribution of these taxa and others from Gadoufaoua is also shown in Table 1.

CONCLUSION
The study of the microvertebrate fauna from Gadoufaoua has resulted in the identification of new taxa previously not reported from the region (see Appendix 2). Among these new forms, the presence of a tribosphenic mammalian molar is noteworthy as the first occurrence of a stem-boreosphenidan in Africa in the Early Cretaceous. This demonstrates a more global distribution of this taxon than previously thought at this time. The presence of the genus Tribodus in Gadoufaoua also suggests, like Sarcosuchus and Araripesuchus that there were still very close links between the continental faunas of South America and Africa and would indicate that the fauna of Gadoufaoua dates back to the opening of the South Atlantic. The presence of numerous amphibious or euryhaline taxa together with those of terrestrial affinity supports a channeltype environment with strong hydrodynamism, as evidenced by the many unidentifiable fragments recovered, which were not produced as the result of predation. This is at odds with conclusions drawn from a study of the macrofauna (Taquet 1976) but can be explained by the large size of the deposit and differences in environments of deposition within it.
Study of the Gadoufaoua microvertebrate fauna has permitted an important revision of the faunal list to include new taxa. It has also highlighted the need to reconsider some of the macro material. It appears that some taxa have been identified without full consideration. For example, Platycheloides Haughton, 1928(de Broin 1980 has recently been revised to Francemys gadoufaouaensis (Pérez-García 2019b) and the debate remains open on the taxonomic placement of Spinosauridae Stromer, 1915(Taquet & Russell 1999Sereno et al. 1998;Sues et al. 2002;Allain 2002;Carrano et al. 2012) and Araripemydidae (de Broin 1980;Sereno & ElShafie 2013;Pérez-García 2019a;Ibrahim et al. 2020).
The study of microremains deposits associated with macroremains needs to be further developed. Data obtained from them are complementary to a "classical" macroscopic palaeontological study and provides insights concerning community structure, palaeoenvironment, palaeogeography, etc., that cannot be obtained from study of the macrofauna in isolation.
A similar study with the same protocol involving the Santana Formation of Brazil would permit collection of data allowing for a more robust comparison on the microfauna. This could also be extended to other known deposits of Late Jurassic and Early Cretaceous age, such as those of the Douiret Forma-