African fossil fish

Olga Otero, Alison Murray, Lionel Cavin, Gaël Clément, Aurélie Pinton et Kathlyn Stewart

Résumé

This chapter illustrates how palaeontology helps to retrace the evolution of African fish, and what can be learned about their palaeobiogeography, adaptation and ancient environment. The first part (1. Africa and fish through geological time) is a presentation of the fish fossil record within the environmental contexts that prevailed in continental Africa through geological time, followed by a focus on the fossils themselves. Using the case study of African characiform fossils, the second part (2. Which fossils? Case study of the African characiforms) illustrates the various kinds of fossils and fossil remains that can be recovered and explains what information fossils can give us. The third section (3. Old groups, old cradles!) focuses on biogeographical information, using the example of one emblematic archaic fish: the lungfish. The fourth part (4. When a marine fish adapts to freshwater) tracks the invasion and extinction of stingrays in the Turkana basin and discusses the still-debated origin of the Nile perch in Africa, to demonstrate the links between fossil fish and their environment, thus giving us information on long-term environmental change. The last section (5. Neogene changes in the African ichthyofauna) is a conclusion concerning the impact of the most recent long-term environmental changes which deeply modified Africa fish faunas, notably the uplift in Eastern Africa and the Saharan aridification.

Africa and fish through geological time

1The history of life on Earth started about 3.8 billion years ago. The metazoans (multicellular animals) appeared 545 million years ago, initiating the Phanerozoic eon which literally means the times of visible animals (figure 3.1). Metazoans diversified first in marine waters and then in both marine and continental environments, following the colonization of fresh waters and land by plants. Among these metazoans, vertebrates occupied various available habitats. Their evolutionary history is shaped by the geological and climatic context prevailing on the continent, and on the processes of diversification and extinction.

2Here, we present the main issues of the geological and environmental background in which the evolutionary history of fish took place in Africa throughout time (figures 3.2, 3.4, 3.5). The tectonic history reviews the periods of connection and isolation of the continent as well as the latitudinal position of the African plate. Combined with continental topography and global climate, this explains the main features of the environmental conditions that prevailed at a regional scale. We use one emblematic ichthyofauna to illustrate the main features of African fish fauna for each era.

FIGURE 3.1. Timetable resuming the main aspects of geological modifications and biological evolution through the Pre-Cambrian time.

FIGURE 3.1. Timetable resuming the main aspects of geological modifications and biological evolution through the Pre-Cambrian time.

The Palaeozoic

3Although some rare isolated vertebrate remains are known in the Early Palaeozoic of Africa, the rise and diversification of major groups of vertebrates occurred during the Middle Palaeozoic era (Silurian and Devonian; figure 3.2). Some of them, such as the armoured agnathan fishes (heterostracans and osteostracans for instance) and the placoderms (armoured jawed fishes), disappeared at the end of the Devonian. Some others, such as the “naked” agnathans (lampreys), chondrichthyans (sharks, rays and chimaeroids), actinopterygians (ray-finned fishes), actinistians (coelacanths), dipnoans (lungfishes) and tetrapods (“terrestrial” vertebrates), had long and flourishing evolutionary histories. These groups now make up the modern vertebrate faunas. Although the actinopterygians currently represent the most diversified group of vertebrates with about 32,000 species distributed in both marine (58%) and freshwater (41%) environments (Nelson, 2006), Middle Palaeozoic actinopterygians were neither diversified nor abundant, and originally were limited to marine waters. So far, the oldest-known putative actinopterygian comes from the Upper Silurian marine deposits of Sweden and China. The earliest presumed freshwater actinopterygians are known from mid-Devonian deposits of Scotland, Australia and Antarctica (see Friedman & Blom, 2006 for a review of the first-known actinopterygians).

FIGURE 3.2. Timetable resuming the main aspects of geological modifications and biological evolution through the Palaeozoic era.

FIGURE 3.2. Timetable resuming the main aspects of geological modifications and biological evolution through the Palaeozoic era.

4During the Middle and Late Palaeozoic, present-day Africa was part of the Gondwana landmass, which itself was part of the supercontinent Pangea (figure 3.2). The movements of land masses have been reconstructed on the basis of palaeomagnetic, biogeographic and palaeoclimatic data. This land mass slowly moved in the southern hemisphere and, during the Devonian, central Africa was located near the South Pole (figure 3.2). The polar position of Africa during the Middle Palaeozoic and the climatic conditions related to that position explain the scarcity of the fossil fish record in general and of freshwater fishes in particular during those times. However, some African sites document this astonishing fish fauna which exhibits the last groups of the earliest vertebrates and the early actinopterygian fishes (figure 3.3). A few heterostracans and acanthodians (“spiny sharks”) have been recorded in the Silurian of Algeria. Placoderms, acanthodians, chondrichthyans (sharks), actinopterygians and sarcopterygians (lobe-finned fishes) have been found in the Devonian of South Africa (see box “A Late Devonian African fish fauna in South Africa”), Morocco, Algeria, and Libya (e.g., Lelièvre et al., 1993, Murray, 2000).

FIGURE 3.3. Map of the main sites that have yielded Palaeozoic lower vertebrates with outlines of the animal reconstructions (from Anderson et al., 1999b, modified). Their distribution at the southern and northern borders of the African continent is to be related with the polar position of the continent and the climate induced.

FIGURE 3.3. Map of the main sites that have yielded Palaeozoic lower vertebrates with outlines of the animal reconstructions (from Anderson et al., 1999b, modified). Their distribution at the southern and northern borders of the African continent is to be related with the polar position of the continent and the climate induced.

5During the Late Palaeozoic (Carboniferous, Permian), no major diversification is reflected in the fish faunas. The actinopterygians that inhabit the African fresh waters are primitive taxa, notably palaeoniscoid fishes known from the lacustrine and fluviatile deposits of Carboniferous sediments of South Africa (Jubb, 1965a; Gardiner, 1969; Jubb & Gardiner, 1975) and of Permian sediments of South Africa, Namibia, and Zimbabwe (Evans & Bender 1999; for a review see Murray, 2000).

The Mesozoic

6During the Triassic and the Jurassic, Pangea started to split apart as it rotated in a global northward movement (figure 3.4). At the end of the Jurassic and first part of the Cretaceous, this resulted in a configuration of the world land masses that would eventually result in the isolation of the Afro-Arabian plate that was to last throughout the Late Cretaceous and most of the Tertiary (figure 3.4). These changes in land and marine connections, and the associated changes in oceanic circulation pattern and climate, produced new conditions for the organisms living at that time, but did not induce any great biotic extinction/radiation events. Lungfishes and other sarcopterygians show a greater diversity of species throughout the Mesozoic era than today, including in African fresh waters. In addition, the diversification of basal teleost fish was rather slow during most of the Mesozoic. Despite the absence of any remarkable biotic crisis at the Jurassic/Cretaceous boundary, the Late Jurassic/Early Cretaceous was a significant time with the initial diversification of crown group teleosts occurring (Cavin et al., 2007a).

FIGURE 3.4. Timetable resuming the main aspects of geological modifications and biological evolution through the Mesozoic era.

FIGURE 3.4. Timetable resuming the main aspects of geological modifications and biological evolution through the Mesozoic era.

7By the Late Jurassic, fishes had been colonizing the waters of the Earth for well over 300 million years. Although they had already evolved greatly from their early ancestors, the Jurassic fish fauna still would be unfamiliar to people today. Many of the Jurassic fishes were heavy bodied forms with thick scales covered by enamel (ganoin), large mouths, paired fins placed well posteriorly on the body, and heterocercal tails (see box “The Stanleyville Beds assemblage, a typical ‘ganoid fish fauna’ ”). They formed a diverse fauna of ganoid fishes in most regions of the world including non-teleost ray-finned fishes (basal Actinopterygii), essentially the chondrosteans and neopterygians with common semionotiforms, amiiforms, pycnodonts, macrosemiiforms, and lepisosteiforms. Today, only some of them are still represented by a few species: chondrosteans by five genera including Acipenser and Polyodon; lepisosteiforms by Lepisosteus and Actractosteus, and amiiforms by Amia (Nelson, 2006). The teleosts arose in the Late Triassic and Early Jurassic with stem-group forms of uncertain relationships such as the “palaeonisciforms” and “pholidophoriforms” that formed a dominant component of the African fauna at the end of the Jurassic only.

8The “pholidophoriforms” became extinct in the Late Jurassic, and the diverse fauna of ganoid fishes disappeared and was replaced during the Cretaceous by early forms of various teleost lineages including among others clupeomorphs, salmoniforms, osteoglossomorphs, elopomorphs and gonorynchiforms (Murray, 2000), and also probable osteoglossiforms (Chanopsis) and lepisosteids (Paralepisosteus?). By the Late Cretaceous, most of the modern orders of African freshwater fish diversify in the Afro-Arabian plate, which has become isolated from all the other land masses (figure 3.4). This early stage of the modern ichthyofauna is documented in a few outcrops in northern Africa, including in Morocco which offers one of the richest fish fossil records for this time period, including fossil species of extant freshwater fish orders such as Lepidosireniformes, Polypteriformes and Osteoglossiformes (see Cavin et al., 2010, for a review).

The Cenozoic

9The Afro-Arabian plate remained isolated from other continents during most of the Cenozoic, until it finally collided with Eurasia about 18 Myrs ago in the early Neogene (figure 3.5). The Palaeogene freshwater fish fossil record is relatively limited and is concentrated mostly in a few regions, notably the Lower Nile outcrops of the Fayum of Egypt, and in Niger and Nigeria (Otero, 2010). In these areas, outcrops yielded fossils assigned to extant families, and the earliest members attributed to modern genera were identified in late Eocene deposits from Egypt and Libya dated 40 and 34 Myrs ago (Murray et al., 2010). Among a few others, two rich and diverse ichthyofaunas from late Miocene sites have been extensively described: one from the deposits of the ancient Lake Turkana in Kenya (Stewart, 2003a) and one from the Toros-Menalla fossiliferous area of the Djurab desert in Northern Chad (Otero et al., 2010a). These faunas show the relative homogeneity of the African ichthyofauna that existed during the Miocene, at least in the north equatorial zone. The Chadian ichthyofauna resembles the modern Nilo-Sudanian fauna (see box “Toros-Menalla, a Cenozoic freshwater fish fauna 7 Myrs ago”). It includes members of modern genera that belong to lineages that evolved in Afro-Arabia since the Cretaceous, and also the Eurasian fish lineages, such as the cyprinids, that have entered the continent by virtue of the tectonic connection of the two land masses. Some other extinct fish were also present and give a peculiar cachet to the Neogene fish fauna when compared to the modern. Finally, the other remarkable feature of the African fish during the Neogene is probably their distribution that was deeply modified notably at the northern margin of the Afro-Arabian plate, including in the Maghreb and into the Arabian Peninsula (see section How the Neogene times shaped the African fish fauna).

FIGURE 3.5. Timetable resuming the main aspects of geological modifications and biological evolution through the Cenozoic era.

FIGURE 3.5. Timetable resuming the main aspects of geological modifications and biological evolution through the Cenozoic era.

10In any fossil assemblage, the diversity recorded is very fragmentary when compared with a living assemblage. Because of their strong resemblance with the modern fauna, the Neogene assemblages allow an evaluation of what information has been lost. First, due to sedimentary and taphonomical processes, the smallest taxa (such as cyprinodonts) and fishes with fragile bones and minute teeth (like mormyrids) are usually not recorded. This is the case in most of the Chadian outcrops that correspond to fluvio-lacustrine deposits (Otero et al., 2009, 2010a, b). Moreover, the fishes from aquatic environments without a sedimentary record also have no chance to show (or preserve) a fossil record. This is the case for cave fish and taxa living in upstream environments. Meanwhile, fishes with strong ossification or differentiated large teeth and living in the main streams and water bodies have a better chance to be recorded, and their fossil record informs us of their distribution and evolution through time. For instance, this is true of several catfishes and of the Nile perch Lates, which are known by skeletal elements, and also of the tiger fish Hydrocynus and of the puffer fish Tetraodon known by their teeth and tooth plates respectively (Otero et al., 2010a). However, some outcrops with exceptional preservation provide sporadic information on some taxa. This is, for example, the case of fishes described in an extinct family of killifish (Cyprinodontiformes) endemic in late Miocene deposits of the Lukeino Formation in Kenya (Altner & Reichenbacher, 2015). Indeed, the information available on extinct animals is highly dependent on the geological context and on the ecology and evolution of the animal (Otero, 2010). The case study of the characiform fossils from Africa shows the various kinds of fossils that are available depending on preservation, and how they demonstrate characiform evolution.

A Late Devonian African fish fauna in South Africa Gaël Clément
The earliest probable freshwater fish fauna in Africa is the Late Devonian assemblage of the Witpoort Formation in South Africa, where various fishes have been found in black shales interpreted as having been laid down in a stagnant back barrier lagoon, an environment receiving both freshwater and marine sediments.
The fish fauna (figure 3.6) consists of placoderms, acanthodians, unidentified palaeoniscoid actinopterygians, and three sarcopterygian groups, i.e., osteolepiform, coelacanth and lungfish (Anderson et al., 1994; Long et al., 1997). Exquisitely preserved fossil shark and lamprey fossils have also been described from this formation (Anderson et al., 1999a; Gess et al., 2006).
According to Young (1987) and Anderson et al. (1999b), this fish fauna, although partly composed of endemic forms, shows more affinities with those of East Gondwana (Australia, Iran) than West and North Gondwana (South America, Maghreb).

FIGURE 3.6. The African lower vertebrate fauna at the Witpoort Formation, Witteberg Group, South Africa: drawings of some fossils and reconstructed outline of the animals (modified from Anderson et al., 1999b).

FIGURE 3.6. The African lower vertebrate fauna at the Witpoort Formation, Witteberg Group, South Africa: drawings of some fossils and reconstructed outline of the animals (modified from Anderson et al., 1999b).

The Stanleyville Beds assemblage, a typical “ganoid fish fauna”
Alison Murray & Olga Otero
The Stanleyville Beds in the Democratic Republic of the Congo have been assigned a Middle Jurassic age (Aalenian-Bathonian; Colin, 1994). They have yielded one of the better documented Jurassic freshwater ichthyofaunas in Africa with about twenty species (Saint-Seine, 1955). The majority of the fish remains from the Stanleyville Beds are “pholidophoriforms” (figure 3.7) represented by four families and 11 species.
Fishes of the genus Lepidotes, once included in the family Semionotidae, but now considered to be a lepisosteiform, two species of coelacanths, and one ionoscopiform have also been recovered (Arratia & Schultz, 1999; Arratia, 2004; Lopez-Arbarello, 2012).
The other fish species from the Stanleyville Beds have been more difficult to place taxonomically. Macrosemius maeseni was originally considered to be a macrosemiid, but has since been removed from the group and left in an unknown position, tentatively into the genus Tanaocrossus (Bartram, 1977), and Leptolepis caheni, originally described as Paraclupavus caheni, is also of unknown relationships within the basal teleost lineage.

FIGURE 3.7 One Jurassic fish (Catervariolus hornemani) from Stanleyville Formation in the Democratic Republic of the Congo. Note the stout scales covered by shiny enamel (photograph by Lilian Cazes, MNHN/CNRS, Paris).

FIGURE 3.7 One Jurassic fish (Catervariolus hornemani) from Stanleyville Formation in the Democratic Republic of the Congo. Note the stout scales covered by shiny enamel (photograph by Lilian Cazes, MNHN/CNRS, Paris).

Toros-Menalla, a Cenozoic freshwater fish fauna 7 Myrs ago
Olga Otero
This Chadian ichthyofauna includes over 20 species (figure 3.8) (Otero et al., 2006, 2007, 2008a, 2010a; Pinton et al., 2011).
Three extinct species belong to the extinct genera Sindacharax, Bunocharax (characiforms) and Semlikiichthys, this last known locally by the species S. darsao (acanthomorph).
Four extinct species belong to extant genera, among which the catfish Auchenoglanis soye and Mochokus gigas and the bichir Polypterus faraou, are all known only from Toros-Menalla so far. The other fish are either assigned to indeterminate species in extant genera, or are attributed to modern species.
They are bichirs (Polypteridae: Polypterus sp.), bony tongue fish (Arapaimidae: cf. Heterotis niloticus), aba fish (Gymnarchidae: cf. Gymnarchus niloticus), mud fish (Cyprinidae: cf. Labeo sp.), tiger and tetra fish (Alestidae: Hydrocynus sp. and Alestes/Brycinus sp.), catfishes among which bagrids (cf. Bagrus sp.), claroteids (cf. Clarotes sp.), clariids (two Clarias spp.), mochokids (two Synodontis spp.), and some acanthomorph fish such as the emblematic Nile perch (Latidae: Lates niloticus), two puffer fishes (Tetraodon sp.), and also indeterminate cichlids.

FIGURE 3.8. Some of the fish fossils from the site TM266 in Toros-Menalla (7Ma, Djurab, northern Chad), with outlines of the body of a modern relative, from Otero et al., 2010a.

FIGURE 3.8. Some of the fish fossils from the site TM266 in Toros-Menalla (7Ma, Djurab, northern Chad), with outlines of the body of a modern relative, from Otero et al., 2010a.

Which fossils? Case study of the African characiforms

11The order Characiformes is probably best known for containing the voracious South American piranha or the large robust pacu, but it also includes a wide variety of fishes found in Africa, notably the tetras and tiger fish. This group of exclusively freshwater fish arose during the split of western Gondwana and its separation into the land masses of South America and Africa. The relationships between the African and the South American characiform families are still under debate. However, most specialists agree in the composition of four monophyletic African families (Hepsetidae, Citharinidae, Distichodontidae, Alestidae) that evolved and diversified in the fresh waters of the newly isolated African plate during the Late Cretaceous and the Cenozoic, when rifting and volcanism drove the formation of large water basins. When preserved, the sedimentary deposits in intra-cratonic and coastal basins constitute the fossiliferous archives that are sporadically exposed today. In certain cases, they yield freshwater fish assemblages including various types of characiform remains (see box “African freshwater fish localities since the Late Cretaceous: a review”). The fossils are like incomplete messages to decode so they can tell us about fish evolution and the freshwater environment that prevailed or the depositional context that allow their preservation. Most of the past diversity of characiforms is known through their teeth, emphasizing the importance of screening sediments to collect minute remains. It is easy then to realize why taxa with skeletons of similar robustness, but with no differentiated teeth, would have a very reduced, if not entirely lacking, fossil record.

The characiform fish fossil record in Africa

12The great differences in the richness of the fossil record of the four African characiform families (Hepsetidae, Citharinidae, Distichodontidae, Alestidae) is not only due to the differences in the diversity of each family, but is also attributable to their differing ecology and skeletal characteristics.

13With 119 species, the family Alestidae comprises about half of the modern diversity of African characiforms. This family includes a number of variouslysized fishes from the giant tigerfish, Hydrocynus goliath, which often reaches over one metre long, to the smaller “dwarf African tetras”, including the robber fish, Micralestes. Most of the characiform fossils found in Africa belong to the Alestidae which also has the highest known fossil diversity among characiforms. With the exception of jaw fragments, their bones are generally not preserved and the alestid fossil record consists of numerous isolated teeth and some scarce jaws. Indeed, all the Cretaceous and Tertiary molariform and multicuspidate teeth collected in Africa are assigned to this family even if they cannot be related to any modern genus. The fossil record for the extant genera consists of Hydrocynus and Alestes or Brycinus teeth. The record starts in the late Eocene in Egypt and Libya (Murray et al., 2010; Otero et al. 2015), and they are present in most Neogene outcrops, although in many cases their sampling may depend on using screening mesh (e.g., Otero et al., 2010a).

African freshwater fish localities since the Late Cretaceous: a review
Olga Otero
Continental sedimentation and preservation of deposits depend on environmental conditions such as topography, eustatism (sea level), and climate. The fossiliferous sites do not cover all the different time periods, nor do they cover the continent uniformly in terms of geographical location, leading to many gaps in our knowledge. The various sites also sample different environments, including brackish lagoons and coastal areas as well as freshwater rivers and lakes. These different environmental settings also lead to gaps in the fossil record, as the fishes in each site are most likely to be quite different. Because of these environmental, temporal and geographical differences, there is no continuous record of any particular fish lineage in Africa.
Indeed, the available fish fossil record is generally the exception, and is highly dependent on the fish itself and on the geology and climate (Otero, 2010). The oldest Cretaceous freshwater localities with fish occur in southern Morocco and Niger. They are probably of Aptian or Albian age and represent a variety of environments including streams and lakes as well as deltaic floodplains. The overlying Cenomanian deposits, both marine and fresh water, are associated with a major marine transgression of the Tethys Sea.
They are predominantly located in North Africa. At some point in the Cenomanian, Morocco, Tunisia, and most of Algeria, Libya and Egypt were under water, and the Tethys Sea reached into Sudan, Mali and Niger
through the central area of North Africa (figure 3.9). Late Cretaceous deposits are also known from West Africa, in the Democratic Republic of the Congo and probably correspond to a shallow lagoonal area with fresh water input. Marine waters again flooded the lowland areas through North and West Africa, creating a shallow seaway connecting the Tethys to the Atlantic during the Campano-Maastrichtian and again in the Palaeocene (figure 3.9).
These transgressive cycles are documented by many marine deposits and some freshwater fish localities in Niger, Cameroon, Congo and Togo. Also during the early Tertiary, freshwater localities were mostly located in North Africa and Tanzania.
Definite Eocene freshwater localities are found in Egypt, Libya and Algeria with riverine and lacustrine deposits, and in Tanzania, represented by a crater lake deposit (figure 3.9).
Egyptian, Kenyan, Tanzanian and Libyan localities have also yielded Oligocene freshwater fish, along with a few sites in Somalia, Oman and Kenya (figure 3.9, Ducrocq et al., 2010).
Neogene sites are far more numerous and more broadly distributed than the Palaeogene sites in the north equatorial part of Africa, notably in the Great Lakes region southwards to Malawi, in the lower Nile River valley, in Chad, Egypt, and in the Maghreb and Maschrek (figure 3.9).

FIGURE 3.9. African Cretaceous and Tertiary outcrops that yielded freshwater fish sites (most data in Otero, 2010), and the variation of sea level (in meters), relative to modern level, (from Haq et al., 1987).

FIGURE 3.9. African Cretaceous and Tertiary outcrops that yielded freshwater fish sites (most data in Otero, 2010), and the variation of sea level (in meters), relative to modern level, (from Haq et al., 1987).

14Moreover, several species have been described in two extinct Neogene genera, Sindacharax and Bunocharax. The assignment to the family is more tentative/difficult for a few Oligocene teeth from Oman and for the multicuspidate teeth sampled in Cretaceous and Tertiary deposits in Europe (see box “On the tracks of evaders?”).

15The two families Distichodontidae and Citharinidae number 101 and eight extant species respectively and together form the suborder Citharinoidei. Contrasting with this rich modern diversity, they have a poor fossil record. A small articulated fossil from the Eocene of Tanzania and two incomplete neurocrania from the early Pleistocene of Kenya have been assigned to Citharinoidei. The Tanzanian fossil is the anterior portion of a fish skeleton. It was collected in the site of Mahenge (ca. 45 Myrs) and was named Eocitharinus macrognathus (figure 3.10; Murray, 2003a). The Kenyan fossils have been referred to the modern genus Citharinus. Finally, a few isolated elements, predominantly teeth, from middle Miocene to Pleistocene deposits in Libya, Chad, Kenya, Uganda and the Democratic Republic of the Congo, have been referred to Distichodus sp. (figure 3.10; Stewart, 2001; Otero et al., 2010a; Argyriou et al., 2014). Although the Eocene articulated fossil is the oldest probable citharinoid known, it is at least 50 million years younger than the expected age of the families.

FIGURE 3.10. Citharinoid fossils and fossil record: Distichodus tooth (left) from the Late Miocene of Toros-Menalla (photograph by OO) and holotype specimen of Eocitharinus macrognathus (right). Note the discrepancy between the fossil record of the two sister families, notably the age of the first occurrence and the distribution (data as cited in the text).

FIGURE 3.10. Citharinoid fossils and fossil record: Distichodus tooth (left) from the Late Miocene of Toros-Menalla (photograph by OO) and holotype specimen of Eocitharinus macrognathus (right). Note the discrepancy between the fossil record of the two sister families, notably the age of the first occurrence and the distribution (data as cited in the text).

16The final family, Hepsetidae, is known today by a single species (Hepsetus odoe, the African pike or Kafue pike). Despite a modern distribution close to that of Hydrocynus and characteristic conical teeth, Hepsetus has not been reported in the fossil record at all. This is explained by their radically different habitat: Hydrocynus occupies larger streams, lakes and flood plains, whereas Hepsetus is observed in swift tributaries in the same hydrographical systems. The former group of environments, notably flood plains, is largely susceptible to preservation in the sedimentary record with reduced transport of teeth and bones, whereas the latter environment corresponds to an erosive sedimentary context, in which remains will not be preserved.

Alestidae: the highest fossil diversity observed for an African fish family is revealed by teeth

17The fossil record of the tiger fishes, genus Hydrocynus, is limited to isolated caniniform teeth and a few jaw elements known from Mio-Pliocene deposits, mostly from the Nilo-Sudanian region (Stewart, 2001). These remains do not bear specific characters and the teeth are attributed to Hydrocynus sp. Meanwhile, the remarkable molariform and multicuspidate teeth that characterize the extant genera Alestes and Brycinus, and the fossil genera Sindacharax and Bunocharax exhibit various morphologies which reveal a high specific fossil diversity, the highest observed for fish in Africa.

18The oldest-known molariform teeth were signaled from Upper Cretaceous deposits in Morocco (Dutheil, 1999a) and from Sudan (Werner, 1993) and from marine deposits in the Ouarzazate Basin in Morocco, considered probably Lower Palaeocene (Cappetta et al., 1978). A number of multicuspid and molariform jaw teeth (figure 3.11) of probably several species referable to extinct genera in Alestidae were collected in late Eocene deposits in Egypt and Libya (Murray et al., 2010, Otero et al., 2015) and in the Oligocene in the Fayum, Egypt (Murray, 2004). Other teeth collected from the Oligocene in Oman exhibit different morphologies, including typically elongated teeth with cusps aligned in one row forming a blade (Otero & Gayet, 2001). A few minute teeth from the Eocene of Egypt and Libya (Murray et al., 2010, Otero et al., 2015) and younger remains have been more narrowly related with the Alestes/Brycinus species complex. Van Couvering (1977) reported isolated teeth of an “Alestes-like” fish from the Daban Formation (possibly Oligocene) of Somalia, from the early Miocene in Uganda, and from middle and late Miocene deposits in Kenya. Moreover Alestes/Brycinus teeth have been collected in many Neogene sites by screening (Stewart, 2001; Otero et al., 2009, 2010a). The monotypic fossil genus Bunocharax was erected by Van Neer (1994) for robust round or elliptical teeth and jaws recovered from middle Miocene to Pleistocene deposits near Lake Albert, Uganda. Finally, many molariform alestid teeth in Mio-Pliocene deposits have been referred to the extinct genus Sindacharax in which several nominative species have been identified.

FIGURE 3.11. Alestid-like teeth from the Oligocene of Egypt, from Murray (2004) modified.

FIGURE 3.11. Alestid-like teeth from the Oligocene of Egypt, from Murray (2004) modified.

19The first species of Sindacharax (S. deserti), described from Pliocene deposits at Wadi Natrun, Egypt, was first placed in Alestes (Greenwood, 1972). With the description of a second species (S. lepersonnei) from the Miocene and early Pleistocene Lake Albert-Lake Edward Rift, the two were given a new genus based on isolated teeth (Greenwood & Howes, 1975). The genus was described on the presence of one major cusp and two or three cuspidate ridges on some teeth (those of the inner row of the premaxillary bones), together with unicuspidate and tricuspidate teeth. Hundreds of teeth have since been recovered and reported in many late Miocene and Pliocene sites of the Nilo-Sudanian ichthyoprovince. Varying tooth shapes and cusp patterns, and lack of living relatives of the genus with similar tooth morphology, have led to considerable confusion in the literature (see Stewart, 2001), because as in the living Alestes/Brycinus fishes, the tooth morphology not only varies according to the species, but also according to their position in the jaws. However, the recovery of several Sindacharax jaws with in situ teeth from the western Turkana basin sites of Lothagam and Kanapoi (Mio-Pliocene, Kenya), enabled Stewart (2003a, b) to demonstrate the systematic validity of five Sindacharax species based on the anatomy of their molariform teeth. There was considerable overlap between the cusp patterns of the outer teeth and the inner third and fourth teeth, but the patterns of the first and second inner teeth were both distinctive (figure 3.12) and consistently identified with a stratigraphic unit or geographic locale within a unit. The five species present in the late Miocene to Pliocene-aged deposits at Lothagam (Stewart 2003a) were: 1) Sindacharax lothagamensis, associated with the Nawata Member, 2) S. mutetii, associated with the Apak Member deposits; it was larger in size than S. lothagamensis and had greater ridging on its inner premaxillary teeth, 3) S. deserti, associated with the Muruongori Member deposits, 4) S. howesi, associated with the north Kaiyumung Member deposits, and 5) S. greenwoodi, associated with the southern Kaiyumung Member deposits (figure 3.12).

FIGURE 3.12. Teeth of the premaxillae in the Lothagam Sindacharax species: inner teeth 1 (left) and 2 (right) in S. greenwoodi, in S. deserti, in S. howesi, and S. mutetii; and inner teeth 2 of the premaxillae in S. lothagamensis.

FIGURE 3.12. Teeth of the premaxillae in the Lothagam Sindacharax species: inner teeth 1 (left) and 2 (right) in S. greenwoodi, in S. deserti, in S. howesi, and S. mutetii; and inner teeth 2 of the premaxillae in S. lothagamensis.

On the tracks of evaders?
Olga Otero
Despite the lack of geological evidence of freshwater connections between Afro-Arabia and other continents during the Cretaceous and the Palaeogene, there is biological evidence of faunal exchanges, notably with Eurasia (Gheerbrant & Rage, 2006). These authors identify five to seven vertebrate interchanges, mainly involving mammals. Among other ideas, transitory emergent platforms, part of the Mediterranean Tethyan Sill, have been invoked to explain dispersions of terrestrial animals during times of lower sea levels. In contrast with mammals, the evidence of dispersal events of freshwater fish to or from Afro-Arabia is very scarce. One freshwater sarcopterygian fish has crossed the Tethys to Southern France at the Campano-Maastrichtian (Cavin et al., 2005). A clariid catfish crossed the Tethys to reach Afro-Arabia in the early Oligocene (Otero & Gayet, 2001). Finally, multicuspidate and molariform teeth attributed to characiform fish might be evidence of iterative dispersals of African freshwater alestid fish to Europe. This last scenario has been discussed by researchers at various points.
The oldest multicuspidate teeth in Europe were collected in Campano-Maastrichtian deposits of Provence (Southern France) in Vérane and Les Pennes-Mirabeau (Otero et al., 2008b, figure 3.13) and possibly of Rumania (Grigorescu et al., 1985), in several Eocene localities from France, Sardinia and Spain (Cappetta et al., 1972; Cappetta & Thaler, 1974; Patterson, 1975; De la Peña, 1993, 1995). After that time, the Oligocene characiform record in Europe is limited to Eurocharax tourainei (Gaudant, 1979, 1980) known from both the skeleton and the teeth, and the youngest characiforms are known in Miocene deposits in the Iberian peninsula (Gaudant, 1996; Antunes et al., 1995). While most specialists agree on the West-Gondwanan origin of the order, there has been a debate on the type of environment where Characiformes may have diversified.
From a palaeontological perspective, based on fossil evidence, it has been suggested that the order Characiformes arose in the Early Cretaceous, possibly from marine ancestors (see Otero et al., 2008b). However, the hypothesis of a marine origin was based on some marine Tethyan fish that were attributed to Characiformes. This marine origin scenario has been dramatically weakened since a review of the fish resulted in them being excluded from the order (Mayrinck et al., 2015a, b). Following from that, there is no evidence to doubt the archaeolimny (primary freshwater
ecology) of African characiforms. The presence of European characiform fish with multicuspidate and molariform teeth is therefore most parsimonious if it is interpreted as a dispersion event, even in the case of the Cretaceous fish.
Additionally, there are two main hypotheses to explain the European characiform fossil record: either there were successive, multiple invasions into European fresh waters by characiform fish from the African stock, or the European forms belong to a single clade whose ancestor reached Europe in the late Cretaceous, with multiple lineages subsequently descended from that ancestor. The first hypothesis has the advantage in that it explains the gaps in the European fossil record, in both time and space.
Moreover, the ages of the proposed dispersals correspond to certain of the dispersal events posited based on evidence for terrestrial animals, as well as to discontinuous land routes (Dercourt et al., 2000; Gheerbrant & Rage, 2006). A Campano-Maastrichtian pathway was used by both a sarcopterygian fish and by derived tetrapods to reach Africa. Next, the timing of the highest known diversity in European characiform fishes (Eocene) is correlated with the main dispersal phase of mammals from Africa to Laurasia. Additionally, a putative minor dispersal phase of mammals is correlated with the Oligocene record of European characiform fishes. Since no phylogenetic study contradicts this second hypothesis as of yet, it is retained as the most probable.
Finally, one primary assumption remains yet untested: are all the multicuspidate and molariform teeth fossil remains of alestid fish? Indeed, most notably when the morphology does not exactly fit with modern alestids, and where there are long gaps in the fossil record, it is safer to attribute a tooth to a characiform indet., while awaiting information (Otero et al., 2008b).

FIGURE 3.13. Photographs of some of the probable oldest-known characiform multicuspidate alestid-like teeth collected in Europe (Campano-Maastrichtian deposits of Provence, southern France), modified from Otero et al. (2008b).

FIGURE 3.13. Photographs of some of the probable oldest-known characiform multicuspidate alestid-like teeth collected in Europe (Campano-Maastrichtian deposits of Provence, southern France), modified from Otero et al. (2008b).

A few articulated skeletal remains, many further evolutionary study perspectives

20Besides thousands of teeth, the known fossil record of characiforms also includes a few articulated skeletons. One such fossil, Arabocharax baidensis, is known from the Oligocene Baid Formation of Saudi Arabia (Micklich & Roscher, 1990). In Africa, only two fossil species have been reported for the family Alestidae based on articulated remains: the middle Eocene Mahengecharax carrolli from Tanzania, based on four specimens (figure 3.14; Murray, 2003b) and the much younger Alestes junneri based on three specimens, from probable late Pleistocene deposits of Ghana (White, 1937; Greenwood, 1972).

FIGURE 3.14. Photographs and drawings of the fossils with zoom of the caudal zone of the Mahengecharax material.

FIGURE 3.14. Photographs and drawings of the fossils with zoom of the caudal zone of the Mahengecharax material.

21Attribution of articulated fossils might be thought to be easier than attribution of isolated teeth, but debates are greater for these few skeletal fossils because the interpretation of these remains may influence views on the evolution of the group. Articulated fossils are most often preserved as flattened skeletons, sometimes covered with scales, with the bones lying atop one another, obscuring many details. Because of this, a particular problem may be the inability to readily identify the material. In the case of characiform fish, they show a particular organization of the Weberian apparatus, a complex built from the modification of the first four vertebrae and existing in all ostariophysan fish (the group that includes among others catfishes, carps and minnows). If the Weberian apparatus cannot be identified, the identification of material as belonging to a characiform fish can be controversial. However, with a better understanding of their osteology, we will be able to build a more complete picture of the phylogeny of the group, including the fossils, which is vital for interpreting time scales of evolution and biogeographic history of these fishes. With little material to work with, every new fossil may cause a significant change in our view of the evolution of the group. Moreover, increased use of computed tomography (CT) scanning technology and 3D reconstruction opens new perspectives and allows researchers to revisit historical fossils. So far these techniques have not been used for African characiforms, but there is little doubt that in the next years we will see the exploration of these new data sets.

Old groups, old cradles! The case of lungfishes: a Pangean or a Gondwanan history?

22Today lungfishes, or dipnoans, are known by three genera inhabiting the fresh waters of Australia (Neoceratodus), South America (Lepidosiren) and Africa with four modern species of Protopterus (P. annectens, P. aethiopicus, P. amphibius and P. dolloi). At first sight this distribution looks typically Gondwanan. We might expect that the modern lungfish lineage originated in Gondwana and that the whole evolutionary history and distribution of the group was restricted to Southern continents. Because of its incompleteness, the lungfish fossil record is rather hard to interpret (see box “The inconsistent state of the lungfish fossil record”). However, it provides a more complex pattern of an evolutionary history that starts over 360 million years ago.

23In order to understand the biogeographical evolutionary history of the lungfishes in the context of the break-up of Pangea and then of Gondwana, it is necessary to resolve the phylogenetic relationships between recent and fossil representatives of the group since the Triassic. All studies agree that the South American Lepidosiren and the African Protopterus are sister genera, grouped together within the family Lepidosirenidae. Most of these studies also agree that both genera split during a vicariant event associated with the opening of the South Atlantic in the Cretaceous (figure 3.16A and 16B). What remains controversial is the relationship between lepidosirenids and the Australian Neoceratodus lineage, in Neoceratodontidae. Two main patterns dealing with recent taxa have been proposed, with significant biogeographical implications.

24The traditional hypothesis suggests that neoceratodontids are a sister-group of lepidosirenids (figure 3.16B). This solution suggests that both lineages split rather recently, in the Late Jurassic or Early Cretaceous (Martin, 1984), somewhere on the Gondwanan landmass. It explains why recent lungfishes are found only in Southern continents, because crossing the Tethys would have been impossible for these freshwater fishes.

25However, during the Cretaceous, neoceratodontid remains are known in Australia, but also in Africa (see above) and in South America. These spatial and temporal distributions do not fit with a vicariant event between the western Gondwanan lepidosirenids (Africa and South America) and the Eastern Gondwanan neoceratodontids (Antarctica and Australia) because neoceratodontids are found on all these landmasses (except Antarctica). Instead it implies that neoceratodontids rapidly dispersed toward Australia, while both other recent lineages remained static. The hypothesis of a middle Mesozoic split between lepidosirenids and neoceratodontids is thus not parsimonious. An alternative hypothesis (Cavin et al., 2007b) proposes that the split between both modern lineages is much older than mid-Mesozoic and that several Triassic and Jurassic genera are phylogenetically closer to the Lepidosiren-Protopterus lineage than to the Neoceratodus lineage (figure 3.16C). According to this pattern, the origin of the neoceratodontid lineage occurred at the end of the Palaeozoic or at the very beginning of the Mesozoic. The split is apparently not associated with a vicariant event related to some kind of geographical barrier, but it occurred at a time of rapid diversification of the group. The diversification produced several genera in Pangea (figure 3.16C, Ceratodus, Ptychoceratodus in Europe, Paraceratodus, Arganodus and “Ptychoceratodus philippsii” in Africa, Asiatoceratodus in Asia and Gosfordia in Australia). Then the break-up of Pangea at the end of the Triassic isolated a Northern lineage from a Gondwanan lineage. The northern, or Eurasian lineage, is known by fossil remains from the Jurassic of Kyrgyzstan, China and Thailand (figure 3.16C, Ferganoceratodus). This lineage appears to become extinct during the Early Cretaceous. The southern lineage gave rise to the lepidosirenid lineage, which split between Africa and South America. The old neoceratodontid lineage remained on Southern continents then became extinct except in Australia, where they are known by several genera and species during the Tertiary (figure 3.16C).

The inconsistent state of the Lungfish Fossil Record Lionel Cavin
The lungfish fossil record is odd because it exhibits two very distinctive phases in the evolutionary history of the group. Lungfishes are first known from the Early Devonian, 410 million years ago. They were marine fishes that rapidly diversified worldwide. Their skeleton was heavily ossified and some of them had dermal bones and scales covered with a hypermineralized tissue, cosmine. The function of this tissue, unknown in modern fishes, is still debated. The skull of these first lungfishes was composed of numerous ossifications and their lobed fins were sustained by a central skeletal axis. The latter feature, still present in modern forms, allows lungfishes to be referred to sarcopterygians. Teeth of Devonian lungfishes were arranged in patches of small denticles or fused together forming dental plates.
Few African fossils document this first marine radiation of dipnoans (figure 3.15; Lelièvre, 1995; Campbell et al., 2002). This first marine radiation is of little use to understanding the distribution of modern freshwater Dipnoans.
At the end of the Palaeozoic, in the Carboniferous and Permian, lungfishes were in a way intermediate with those from the second period of their evolution that occurred during the Mesozoic and Cenozoic. This second phase of lungfish history indeed started during the Triassic with forms that were quite different from their Devonian relatives. Skull bones are thinner and develop deeper under the skin, and tooth plates are bigger and more heavily mineralized. As a consequence of these osteological transformations, the fossil record is altered: the fragile ossifications of the skull are rarer as fossils, whilst the tooth plates are more common but are usually found isolated. So, in the Mesozoic, only a dozen taxa are known by both cranial and dental remains and most are Triassic. The African fossil record is also richer in the Triassic whatever few sites are available (figure 3.15). The oldest are an almost complete specimen referred to Microceratodus angolensis from the Early Triassic of Angola (Thomson, 1990) and several taxa based on isolated tooth plates in the Early to Middle Triassic in South Africa (Kemp, 1996) and from the Late Triassic of Morocco with Aranodus atlantis known by tooth plates and skull remains (Martin, 1979, 1981). It also appears that lungfishes were mainly freshwater fishes in the Triassic. They stay in these palaeoenvironments from that time onwards so that the break-up of Pangea will have a marked impact on their evolution. In Africa, after the Triassic times, the fossil record of lungfish is poorly diversified.
There is one occurrence from the Late Jurassic-Early Cretaceous in Ethiopia and one from the Late Jurassic in Algeria (Goodwin et al., 1999; Martin, 1984; figure 3.15), both assigned to the same species (Arganodus tiguidensis). Cretaceous lungfishes occur mainly in the Sahara region (figure 3.15). Among them, another Arganodus sp. close to contemporaneous forms from the Albian–Cenomanian Itapecuru Group in Brazil (Dutra & Malabarba, 2001) indicates that relationships between Africa and South America are still very close at that time. Moreover, two species are referred to the modern Australian genus Neoceratodus (Martin, 1984), two other species belong to the recent African lineage of Protopterus, Lavocatodus protopteroides and probably L. humei (Churcher & Iullis, 2001), and the genus Protopterus itself is reported from the Campanian of Egypt with P. crassidens (Churcher & Iullis, 2001).
After that, the Tertiary record of lungfishes in Africa is limited to the genus Protopterus, and from the Oligocene onwards, species close to, or belonging to modern species are known (figure 3.15, Otero, 2011).

FIGURE 3.15. The African fossil record of lungfishes in Africa (data as cited in the text).

FIGURE 3.15. The African fossil record of lungfishes in Africa (data as cited in the text).

FIGURE 3.16. The break-up of Pangea and the possible scenarios for lungfish biogeographical history (from Cavin et al. 2007b, modified). A. Schematic palaeogeographical evolution and modern distribution of lungfishes; B. model based on a direct sister-group relationship of the lepidosirenids and Neocerotodus (and closely related genera); C. model based on a split between the lepidosirenids and Neocerotodus that occurred deeper in the phylogeny.

FIGURE 3.16. The break-up of Pangea and the possible scenarios for lungfish biogeographical history (from Cavin et al. 2007b, modified). A. Schematic palaeogeographical evolution and modern distribution of lungfishes; B. model based on a direct sister-group relationship of the lepidosirenids and Neocerotodus (and closely related genera); C. model based on a split between the lepidosirenids and Neocerotodus that occurred deeper in the phylogeny.

Bichirs: decline, fall and new rise of a Western Gondwanan fish in Africa
Olga Otero
Polypteriforms are the only living members of a very basal actinopterygian group called Cladistia. Today, they are exclusively African and placed in the single family Polypteridae with 17 Polypterus species and subspecies and the monotypic Erpetoichthys.
Isolated skull-roof bones, scales and spines are
found in most of the freshwater outcrops of Africa from the Late Cretaceous to the modern (see box “African freshwater fish localities since the Late Cretaceous: a review”, section 2). They are recognized to belong to polypteriform fish on the basis of their morphology and histology: the pinnulae are the characteristic spine that only these fish exhibit, in front of each of their dorsal finlets; the vertebrae exhibit a typical morphology with closed neural arch and all dermal bones are covered in a layer of ganoin enamel (figure 3.17). Two extinct species have been described on the basis of articulated fossils: Serenoichthys kemkemensis from the Late Cretaceous of Morocco (Dutheil, 1999b), and Polypterus faraou (figure 3.17A, B) from the late Miocene of Chad (Otero et al. 2006).
Several Cretaceous genera have also been erected based on original morphology of the articular head of pinnulae (e.g. Meunier & Gayet, 2000), of scales and of the ectopterygoid (Grandstaff et al., 2012). Moreover, two other fossil species have been described in Polypterus on the sole morphology of the scales (Gayet & Meunier, 1996; Werner & Gayet, 1997). The numerous other African fossils, from the Cenozoic, are referred to Polypterus sp. Indeed despite a rich fossil record, little is known about polypteriform evolution in Africa since the phylogenetical relationships of extinct forms remain totally unknown. Until the early 1990s, only fossils from Africa were referred to polypteriform fish. Fossil scales and bones covered by enamel recovered from outside Africa were systematically attributed to other ganoid fish (such as semionotids and lepisosteids).
However, pinnulae and characteristic vertebrae were collected and described in Late Cretaceous and Palaeocene deposits from Bolivia demonstrating their ancient presence on the South American Plate (Gayet & Meunier, 1991). The presence of histological features characteristic of polypteriform scales was then sought and found; it therefore participates in demonstrating the hitherto unsuspected presence of polypteriforms in South America during Cretaceaous and Palaeogene times (Gayet & Meunier, 1991). Indeed, the scales of living polypterids originally exhibit dentine (like in a tooth) and a bony basal plate with a perpendicular counterplate-like structure (figure 3.17F).
Finally, the other available information on polypteriform evolution is the result of molecular studies and concerns the phylogeny of modern members with estimates of the diversification time (Suzuki et al., 2010; Near et al., 2014). Combined with the fossil record data, a scenario appears for the order (Near et al, 2014).
First, they had a flourishing diversity during the Late Cretaceous in both South America and Africa, illustrated by the great diversity of morphology of the pinnulae and also by the size of the taxon that might have reached up to three meters with Bawitius (Grandstaff et al., 2012). Then polypteriform diversity declines on both plates and the lineage goes extinct in South America. In Africa, Polypterus and Erpetoichthys diverge in the early Tertiary and Polypterus diversifies during the Miocene, marking a new radiation phase after a long decline.

FIGURE 3.17. Polypteriform fossils: Holotype specimen of the only Polypterus fossil species known from articulated material, Polypterus faraou in A. dorsal and B. lateral views (modified from Otero et al. 2006); disarticulated fossil remains as generally found in Tertiary outcrops (modified from Otero et al. 2015), C. jaw tooth, D. pinnula with the ganoin anterior cover in front, lateral, and posterior views, and E. ganoid scale covered by ganoid and showing the peg (that articulates in a socket in the next scale); F. section in a polypteriform scale to show the location of the characteristical plywood structure (modified from Daget et al. 2001).

FIGURE 3.17. Polypteriform fossils: Holotype specimen of the only Polypterus fossil species known from articulated material, Polypterus faraou in A. dorsal and B. lateral views (modified from Otero et al. 2006); disarticulated fossil remains as generally found in Tertiary outcrops (modified from Otero et al. 2015), C. jaw tooth, D. pinnula with the ganoin anterior cover in front, lateral, and posterior views, and E. ganoid scale covered by ganoid and showing the peg (that articulates in a socket in the next scale); F. section in a polypteriform scale to show the location of the characteristical plywood structure (modified from Daget et al. 2001).

26Finally, an accurate review of Tertiary dipnoans of Africa suggests that an extinction phase took place during the Palaeogene and was followed by a recent diversification from a refuge located in eastern Africa (Otero, 2011). This hypothesis based on the fossil record fits with the molecular data available so far (Tokita et al., 2005). Decline in diversity followed by a new diversification also occurred in the bichirs which form the other African archaic fish lineage (see box “Bichirs: decline, fall and new rise of a Western Gondwanan fish in Africa”). So far, evidence is still lacking to explain the causes of this common global pattern in the evolutionary history of the two archaic African fish lineages. The Neogene climate modification with enhanced seasonality, with an increasing dry season, has been invoked to give advantage to these relictual fish that are also characterized by their ability to breath air (with lungs) and even to aestivate in the case of Protopterus (Otero, 2011).

When a marine fish adapts to freshwater

Tolerance to salt water

27Traditionally, scientists have classified fish into three divisions based on their tolerance of salt water (see chapter General characteristics of ichthyological fauna). Of particular interest here is the peripheral (tertiary) division, specifically the vicarious and complementary groups. Only recently has the fossil record in Africa become detailed enough to identify both the adaptations to fresh waters in these groups, as well as the geological processes which altered their habitats. Two cases illustrate how palaeontological data document the adaptation of marine taxa to fresh waters in Africa. First, the Dasyatidae has a fossil record in the Turkana basin that documents the adaptation of a marine coastal species to freshwater, and later its disappearance, in association with geological and biotic changes in less than one million years. Second, the fossil record of the family Latidae demonstrates how they evolved from a marine to a mostly freshwater group, notably in Africa, over their ca. 40 Myrs evolutionary history.

Dasyatidae in the Plio-Pleistocene Turkana Basin

28Modern freshwater stingrays are assigned to two families: Dasyatidae (stingrays) and Potamotrygonidae (river stingrays). The Dasyatidae is chiefly a marine family; however, it includes some members which inhabit brackish and/or fresh waters (Nelson, 2006). The thorny freshwater stingray Urogymnus ukpam, for example, permanently inhabits a lake in Gabon, about 100 km inland from the Atlantic Ocean, and another dasyatid, Dasyatis garouaensis, is permanently established hundreds of kilometers upstream in the Benue River. Potamotrygonidae is the only batoid family which is restricted to fresh waters. Potamotrygonid species inhabit rivers in eastern South America, although several species were identified in the Benue River in Nigeria. These latter species were later reassigned to the genus Dasyatis, based on the lack of a prepelvic process and the low urea concentration in body fluids, both of which are diagnostic for the family Potamotrygonidae (Thorson & Watson, 1975). Potamotrygonidae are unique among the elasmobranchs in having this osmoregulatory feature, which suggest a long period of evolution in freshwaters, while the Dasyatidae have similar osmoregulatory features to the other elasmobranchs, suggesting that they are fairly recent colonizers of fresh waters. Both the freshwater dasyatids and the potamotrygonids would be classified as peripheral–complementary fish, that is, marine fish which colonized fresh waters.

29The fossil record for stingrays is sporadic in Africa, and mainly consists of marine stingrays, so it is rare to find convincing evidence of the transition of stingray groups from marine to fresh waters. In fossil deposits in and around Lake Turkana, northern Kenya, however, there is evidence of marine stingrays colonizing and reproducing over thousands of years in a freshwater inland lake. The modern Turkana basin (including Lake Turkana and the inflowing Omo River) contains highly fossiliferous deposits which date from the early Miocene to the Holocene and have been well studied by palaeontologists and geologists. In the lower Omo River valley, several stingray caudal spines were recovered in the 1930s from lacustrine deposits, in association with other freshwater polypterids, silurids and Lates, a percoid (Arambourg, 1948). No other material from marine fishes (or other marine vertebrates for that matter) was recovered. Arambourg (1948) assigned the spines to Potamotrygonidae (Potamotrygon africana), the only modern family of stingrays which is completely freshwater, even though it was only known in South America. At least 360 more spines and several teeth were later recovered from deposits in the eastern Turkana basin (Schwartz 1983; Feibel & Brown, 1993). The spines were first found in upper Burgi Member deposits (about 2.0 Myrs), with their last appearance in Okote Member deposits at about 1.5 Myrs (Feibel, 1988; Feibel & Brown, 1993). Feibel & Brown (1993) re-assigned the eastern Turkana and Omo River stingray material to Dasyatis africana (figure 3.18A), based on 1) similarities in spines to dasyatids, and in tooth morphology to the modern Dasyatis species D. centroura and D. guttata; 2) the reassignment of the modern African Potamotrygon species to Dasyatis (discussed above), and 3) the fact that Potamotrygonidae are endemic to South America (Thorson et al., 1983; Thorson & Watson, 1975). Dasyatis africana was originally assigned to stingray remains from Cabinda, near the head of the Congo delta in the Democratic Republic of the Congo (Dartevelle & Casier, 1959), but it is not clear if the Turkana stingray material was compared with the Congo material.

30Feibel & Brown (1993) have reconstructed the Turkana basin environment associated with the stingray immigration as follows (figure 3.18B): the stingray spines first appeared around 2 Myrs when a large, relatively stable lake system occupied the basin, and continued for about 50,000 years. About 100,000 years later, the lake gave way to a fluvial system with a floodplain dotted by smaller lakes. These lake deposits were filled with mollusc remains and Dasyatis spines, and as many modern dasyatids eat molluscs as part of their diet, these were assumed to be a main component of the diet of D. africana. After 1.6 Myrs, the lakes became less numerous and more alkaline, and the molluscs began to disappear, along with the stingrays. The drying of the lakes may have been due to climatic or tectonic events. It seems likely that the stingrays disappeared due to the disappearance of their main dietary item, the molluscs.

FIGURE 3.18. Invasion of the Turkana freshwaters by a stingray species: A: one exemplar of a fossil stingray spine collected in Pliocene deposits from the Turkana basin and its location on a living specimen; B. invasion scenario showing the role of the topographical configuration of the Turkana basin and surrounding areas.

FIGURE 3.18. Invasion of the Turkana freshwaters by a stingray species: A: one exemplar of a fossil stingray spine collected in Pliocene deposits from the Turkana basin and its location on a living specimen; B. invasion scenario showing the role of the topographical configuration of the Turkana basin and surrounding areas.

31This reconstruction indicates that stingrays had clearly established a permanent, breeding population for 500,000 years in the fresh waters of the Turkana basin. The question of their origin was an enigma, as the Turkana basin had been in sporadic indirect contact with the Nile River, but no evidence of stingrays was found in the Nile or its tributaries or associated lake basins. However, based on detailed stratigraphic evidence, Brown & Feibel (1991) have postulated that the ancestral Omo River, which traversed the Turkana basin, was occasionally connected with the Indian Ocean, about 700 km distant, via a presently low lying, sometimes marshy region to the southeast of Lake Turkana. It is not known when the connection ceased to flow. A previous connection between the Turkana Basin and the Indian Ocean must have occurred at least once, as the rostral fragment of a late middle Miocene ziphiid whale was reported in the western Turkana basin, at Loperot (Mead, 1975).

32This is some of the first fossil evidence of a chiefly marine family, the dasyatids, moving a considerable distance inland to both colonize a freshwater lake and establish a reproducing, permanent population over a long time span (ca 500,000 years). Either increasing of aridity or tectonic changes caused changes in the lake such that the mollusc population disappeared, followed by the stingrays.

Latid fishes: over 40 Myrs of history from marine to fresh waters told from the phylogeny of modern and extant taxa and the fossil distribution

33Modern latid fishes are unequally distributed in two genera: the monotypic Psammoperca and the relatively speciose Lates genus. Depending on the species, they live either in marine or in fresh waters (figure 3.19). Psammoperca waigiensisis is marine and inhabits the Indo-Pacific coastal waters. It overlaps with the widely distributed Lates calcarifer which is also found in the Eastern Pacific in coastal marine, brackish and fresh waters, where it has been described to have a catadromous lifestyle (Marshall, 2005). A third marine species (L. japonicus) is known from the Japanese coasts and estuaries. In addition to these, nine latid species are known from fresh waters. Two are Asian: L. lakdiva from Sri Lanka and L. uwisara from Myanmar. Seven are from African lakes and streams. Among these, the famous Nile perch, L. niloticus, the biggest in size, shows the widest distribution. It is native in the northern and western regions of tropical Africa including Lake Chad, Lake Albert and Lake Turkana, and was also introduced in Lake Victoria and Lake Kioga. The remaining species are endemic in East-African lakes: L. longispinis in Lake Turkana, L. macrophthalmus in Lake Albert, and L. angustifrons, L. microlepis, L. mariae and L. stappersi in Lake Tanganyika.

FIGURE 3.19. Distribution of the species of the two living genera of the family Latidae in African freshwaters and Indo-Pacific coastal waters.

FIGURE 3.19. Distribution of the species of the two living genera of the family Latidae in African freshwaters and Indo-Pacific coastal waters.

34The fossil record starts with extinct fish found in the fossil and paraphyletic genus Eolates, which includes two species from lower Eocene waters (ca. 45 Ma) of the European Tethyan coast (figure 3.20). The last record of Eolates is known in lacustrine environments from the lower Oligocene (ca. 30 Ma) of South France (Otero, 2004). Extant Psammoperca has no fossil record whereas Lates includes several fossil species from various habitats in Afro-Arabia and Europe (figure 3.20). The oldest Lates species are Oligocene in age and are known from a freshwater environment in Egypt (L. quatrani; Murray & Attia, 2004) and a marine environment in Italy (L. macropterus; Bassani, 1889). Another freshwater species, L. bispinosus (Gaudant & Sen, 1979), is known from Turkey but with an uncertain Tertiary age. During the Miocene, Lates fish were present in Western European fresh waters, in Spain and Italy, while fossils assigned to Lates sp. or L. cf. L. niloticus have been recovered from most Miocene and Pliocene outcrops in the Nilo-Sudan, and also from the Maghreb and Arabian peninsula (figure 3.20). With the exception of L. niloticus, extant latid species have no fossil record.

FIGURE 3.20. Occurence of fossil latid fishes with information of the environment of the deposits (data from the review by Otero, 2004, plus the species described by Murray & Attia, 2004 and by Stewart & Murray, 2008).

FIGURE 3.20. Occurence of fossil latid fishes with information of the environment of the deposits (data from the review by Otero, 2004, plus the species described by Murray & Attia, 2004 and by Stewart & Murray, 2008).

35All clues indicate a marine origin for latid and Lates fish: the potential sister groups are marine, and a number of species are found in marine waters.

36From the fossil record, the living distribution, and the phylogeny, we can build the following scenario for the history of this family (figure 3.21). The latids have a marine early Eocene (55 Myrs – 40 Myrs) origin with Eolates, afishthat inhabited Tethyan waters, at least along the western coasts. A marine origin is also the most parsimonious hypothesis considering that sister taxa are also marine. In the early Oligocene (ca. 30 Ma), an initial freshwater colonization is recorded in Southern France with a species of Eolates. At the same time, the genus Lates appeared at least along the African Tethyan coast, and, although no fossil has been recovered, the sister-genus Psammoperca must have appeared at the same time. Due to its modern distribution and the absence of a fossil record in the Western Tethys, we may hypothesize that Psammoperca was present instead in the Eastern Tethys. Next, Lates clearly diversified, first in the coastal waters of the Tethys before the closure of this seaway. Following this time, since the early Miocene, Lates fishes immigrated into and inhabited the continental freshwaters on both sides of the proto-Mediterranean Sea, in Africa and in Europe (where they may also be occasional invaders), except the L. calcarifer lineage. Depending on the range extension of Lates westward in Tethyan waters, the Lates calcarifer lineage originated in the Indo-Pacific area before the closure of the Tethys, or it invaded the proto-Indo-Pacific waters later, possibly benefiting from the high sea level that linked marine areas creating a connection between the Mediterranean Sea and the Indian Ocean during the Langhian. At the Miocene-Pliocene boundary (ca. 5 Myrs), Lates disappeared from European fresh waters and was reduced to roughly its modern distribution. The last representative known so far in Europe is a fossil Lates niloticus found at Monte Castellaro, in Italy (Otero & Sorbini, 1999). It probably invaded the Italian peninsula from Africa, during the time of low sea level during the Messinian (between about 7 Myrs and 5 Myrs).

FIGURE 3.21. A probable scenario for Lates fish diversification.

FIGURE 3.21. A probable scenario for Lates fish diversification.

Neogene changes in the African ichthyofauna

37At the very beginning of the Miocene (23 Myrs – 5 Myrs), the Afro-Arabian plate was isolated from the other continents for over 100 Myrs. From this isolation, the inherited ichthyofauna shows an original diversity, with many families endemic in different orders. The Neogene (23 Myrs – present) evolution of the African ichthyofauna is related to the palaeogeographical events within continental Africa and environmental change. Of these, we retain here two of the best documented in the fossil record: first, the connection with Eurasia 20 Myrs ago, that enabled faunal dispersion in both ways between the two continents, and second, the Pliocene development of the Sahara that led to isolation of the North African regions. The influence of palaeogeographical events and environmental change can be estimated in a same family (see box “African fish diversification and geologically-driven events”), but also at the level of the whole ichthyofauna by continuance or disappearance of given taxa.

A main palaeogeographical event: the Afro-Asian connection

38The connection of Africa with Eurasia occurred 20 Myrs ago in the early Miocene. It was followed by a disruption of some millions of years due to a marine transgression, and the definitive continental connexion occurred 12 Myrs ago. At each of these events, faunal interchanges occurred between the two plates, notably with mammal dispersal. Among the modern fishes, one can recognize several freshwater families in common on the two continents but a fossil record exists for very few of them. They are the clariid and bagrid catfishes, the cyprinids, and possibly the channids, which are all known to originate in Asia, and thus to conquer the African freshwaters by the presence of a land pathway.

39Studies on cyprinid palaeobiogeography agree in placing the origin of the family in Asia. For a long time, the absence of any well-known Miocene fossil record of African cyprinids led some authors to consider that the family entered Africa during the Plio-Pleistocene. Doadrio (1994) hypothesized a freshwater faunal dispersion prior to the Pliocene, based on the presence of African cyprinids that are well-differentiated from the Asian Pseudophoxinus and “Barbus” (this later taxon includes various monophyletic taxa; the phylogeny has been deeply reassessed during the last decade, e.g. Stiassny & Getahun, 2007, Yang & Mayden, 2010; Zheng et al., 2012;). This fits with the fossil record. Several forms are known in the Upper Miocene including “Barbus” from Egypt (Priem, 1914), Tunisia (Greenwood, 1972), United Arab Emirates (Forey & Young, 1999), and “Barbus” and Labeo from Kenya (Stewart, 2001). Indeed, the finding of an 18 Myr-old cyprinid fish in the early Miocene of Saudi Arabia (Otero, 2001) indicates that cyprinids certainly entered the Afro-Arabic plate from Asia with the first Tertiary continental connection between the two plates, to which a major mammalian migration also corresponds. The fossil teeth resemble “Barbus” and these invaders may be the ancestor of some of the modern African cyprinids.

Channid palaeo-distribution: palaeogeographical or palaeoclimatological constraints?
Alison Murray & Olga Otero
The discovery of fossil snakehead fishes in Eocene deposits of the African continent provides us with a convincing example of faunal exchange via a connection between Asia and Africa in the early Cenozoic, before the Burdigalian landmass connection.
Snakeheads are a small family of freshwater fishes, with two genera and about 30 species living today. They are currently found in Asia, Malaysia and Indonesia (Channa with about 27 species) as well as in Africa (Parachanna, with three species). They are generally piscivorous fish that range in size from about 15 to 180 cm in length (Courtenay & Williams, 2004). Snakeheads have long been known from the Cenozoic fossil record of Asia and are also known from the Miocene of Europe (Gaudant & Reichenbacher, 1998). Experimental evidence suggests marine waters would be lethal to snakeheads, and so their modern distribution was presumed to be the result of a Miocene invasion of the African continent, with the advent of the land connection between Asia and Africa through Arabia during this time period. Moreover, snakeheads are known to be successful invaders of new areas, an ability that is aided by their capacity to breathe air through a specialized respiratory organ, which allows them to travel significant distances across land. Snakeheads are of increasing interest to a wide range of organizations because of their invasive capacities: release of a few fish have allowed breeding populations to become established in the United States. As voracious predators, they have caused great damage to native fish populations, and as air-breathers, they are able to colonize new areas relatively easily by travelling overland (Courtney & Williams, 2004).
The presumed Miocene migration of snakeheads from Asia to Africa has been correlated with increased rain during changing climatic regimes. Indeed, snakeheads are even shown to be sensitive indictors of precipitation, with high rainfall during the warm season being important to their dispersal (Böhme, 2004). While climate change with increase of temperature and waterfall may well have allowed snakeheads to expand their ranges, fossils from the Eocene of Egypt are a conundrum (figure 3.22). These fossils show that snakeheads were in Africa at least 25 million years before the Miocene land route was available to them. How did snakeheads, freshwater fishes that are intolerant of salt waters, become distributed in Asia and Africa long before these two continents were connected with one another in the Miocene?
One answer might be that the group arose much further back in time, when the landmasses were united in the single continent Pangea. But this would require snakeheads to have evolved considerably before any other derived teleost group, and about 200 million years prior to their appearance in the fossil record.
A Pangean origin of snakeheads is thus not likely. A second suggestion might be that the ancestral snakeheads were able to tolerate marine waters, unlike their modern relatives. However, this idea cannot be tested unless marine fossils are found; there is currently no indication that snakeheads may have migrated through saline waters. The third, and most likely, scenario is that they used an earlier bridge between Europe/Asia and Africa, in the early Cenozoic. Land connections between Asia and Africa during this time span have yet to be documented by geologists. However, the biological indications are that such a route existed. More and more precise data on the exchanges would allow testing the various hypotheses: via movement of microplates (McKenzie, 1970); by means of a temporary connection between the Indian subcontinent and East Africa (Krause & Maas, 1990); or by sea-level-controlled discontinuous routes across the Tethys Sea (Gheerbrant & Rage, 2006), which could be combined by rafting under the control of marine palaeo-currents.

FIGURE 3.22. The skull on the left is the holotype of Parachanna fayumensis from Egypt and the skull on the right is Channa striata (recent), with outline of a modern channid.

FIGURE 3.22. The skull on the left is the holotype of Parachanna fayumensis from Egypt and the skull on the right is Channa striata (recent), with outline of a modern channid.

40Finally, in the case of the cyprinid fish, the data are congruent with the geodynamic history of the region. However, in other cases, other environmental parameters, notably the climate, seem to have been the main factor that favoured trans-continental exchanges, before the complete land connection (see box “Channid palaeo-distribution: palaeogeographical or palaeoclimatological constraints?”). This is notably the case of the aridification trend that developed in northern Africa in the last 7 Myrs and led to the Sahara desert.

A main palaeoenvironmental change: the Saharan extension

41Climate fluctuation and its environmental consequences can be observed at different time scales. For instance when considering the Saharan desert, it was much greener some thousands of years earlier. However, considering a geological time scale, we observe a general aridification trend that started at least at the end of the Miocene with the oldest-known aeolian dunes deposit in the Djurab desert (Schuster et al., 2006, Otero et al., 2011). While short-term variations explain the distribution of some fishes in Saharan oases (chapter Fish communities in small aquatic ecosystems: caves, gueltas, crater and salt lakes), long-term ones may explain the distribution of fishes at the scale of the Saharan belt.

42Indeed, until the late Miocene, we know that a homogenous ichthyofauna that resembles the modern Nilo-Sudan one inhabited Africa, at least in the North equatorial zone (Stewart, 2001; Otero & Gayet, 2001). This palaeontological fish continuum extended onto the Arabian plate in the early Miocene (Otero & Gayet, 2001). It finds its roots in the Palaeogene and was enriched by arrivals of fishes from Asia. The systematic similarities between the fossil ichthyofauna and the extant Nilo-Sudan one can be explained because of an ecological continuum and the opportunism of some taxa in conquering new areas by the way of the African hydrographical network (Greenwood, 1983; Roberts, 1975). These differences are largely explained by the geological context that deeply structured the modern distribution of African fish (Pinton et al., 2013).

43This ichthyological continuum saw reductions. For instance, the Arabian plate appears to lose its African fish fauna during the Miocene. The comparison of Arabic fish assemblages in the localities of As-Sarrar (Lower Miocene), and Abu Dhabi (Upper Miocene) suggests the loss of Lates, an emblematic species that is linked to the existence of a wide hydrographic net (Van Neer, 1994). This could be related to the Saharo-Arabian desert development and/or to the disruption of the drainage network by the East African and Red Sea orogenic phases. Nowadays, Lates and Clarias are both absent from Arabia and the North of Africa (except along the Nile system); the other taxa typical of the pandemic African fauna are also absent (Synodontis, Bagrus, alestids, etc.). Furthermore, the extant cyprinids in the Maghreb seem to be of European origin.

44The preliminary proposal that desertification or a strong change in the environment was the limiting factor of the continuum, which started its effects during the Miocene, fits with some observations made on mammalian faunas (Thomas, 1979). Nevertheless, the palaeogeographical separation of the African and Arabian plates during the Miocene should have enhanced the process or at least made it very patent, because of the disruption of the hydrographical network between the two plates.

African fish diversification and geologically-driven events
Olga Otero & Aurélie Pinton
To test in which respect the evolution of freshwater fishes of Africa is dependent on physical modification of landscapes, i.e., connections or disruption between drainage basins that reflects underlying geological and climatic change, we correlate results of phylogeographical works and fossil fish distribution with the current knowledge of African geology chronology (Otero et al., 2009; Argyriou et al., 2012; Pinton et al., 2013). Indeed, we assume that for freshwater fish, basin boundaries constitute barriers to dispersal: they constrain their distribution and strongly influence spatial patterns of fish assemblage composition.
In our studies we focus on three taxa: the catfish genera Synodontis and Carlarius and the perciform fish Semlikiichthys, all present in Afro-Arabia during the Neogene as evidenced by their fossil record: Synodontis has existed for at least 18 Myrs and is one of the most widespread catfish within today’s African continent (Pinton et al., 2013); and Carlarius gigas is restricted to the Niger-inner delta whereas it had a wider past distribution
including Chad and Libya, while Semlikiichthys is an extinct Neogene genus with two species with disjoint distribution, one in the Chad basin and one in Sirt Basin (Libya), Eastern Africa and along the Nile (Otero et al., 2009; Argyriou et al., 2012). We also include the results of another phylogeographical study that provide results at the continental scale (Brown et al., 2010). Distribution of fossil fish identified at specific level and modern fish phylogeny notably suggest (figure 3.23): 1) the primary role of the Central African Shear Zone that has existed since the Cretaceous and acts all along the Tertiary time period as a main East-West barrier; 2) strong East/West connections between the formatting basins north to the Central African Shear Zone during the early Neogene; 3) sporadic disruptions between Eastern Africa, Syrt Basin and Chad that notably lead to temporary endemism in the later basin; and 4) the probable recent dominant role of the climatic control that drives the modern heterogeneity between the basins during the late Pliocene.

FIGURE 3.23. Some major African structures that settled in the last 100 Million years and related impact on fish diversity. Colours inform on the time period when each geological structure modified the regional geomorphology and the fish palaeogeography.

FIGURE 3.23. Some major African structures that settled in the last 100 Million years and related impact on fish diversity. Colours inform on the time period when each geological structure modified the regional geomorphology and the fish palaeogeography.

Mountains of the Alps orogenesis: together with the Late Neogene climatic change and the Saharan desert extension, they notably depauperate the whole northern zone of the continent and cut exchanges with Asia

45During the late Miocene and the Pliocene, endemism and exchanges occur between the sub-basins of the Nilo-Sudan depending on their palaeogeography, but a rather homogenous fauna is maintained.

Auteurs

Olga Otero

A professorial lecturer at the University of Poitiers (Institute of Paleoprimatology, Human Paleontology: Evolution and Paleoenvironments). She studies the evolution and diversification of African fishes to reconstruct the relationships between their spread, diversification, and the paleoenvironment.

Alison Murray

Lionel Cavin

Gaël Clément

Aurélie Pinton

Kathlyn Stewart