We provide a general compilation of the diversity and geographical distribution of Amazonian fishes, updated to the end of 2018. Our database includes documented distributions of 4214 species (both Amazonian and from surrounding basins), compiled from published information plus original data from ichthyological collections. Our results show that the Amazon basin comprises the most diverse regional assemblage of freshwater fishes in the world, with 2716 valid species (1696 of which are endemic) representing 529 genera, 60 families, and 18 orders. These data permit a view of the diversity and distribution of Amazonian fishes on a basinwide scale, which in turn allows the identification of congruent biogeographical patterns, here defined as the overlapping distributions of two or more lineages (species or monophyletic groups). We recognize 20 distinct distributional patterns of Amazonian fishes, which are herein individually delimited, named, and diagnosed. Not all these patterns are associated with identifiable geographical barriers, and some may result from ecological constraints. All the major Amazonian subdrainages fit into more than one biogeographical pattern. This fact reveals the complex history of hydrographical basins and shows that modern basin-defined units contribute relatively little as explanatory factors for the present distributions of Amazonian fishes. An understanding of geomorphological processes and associated paleographic landscape changes provides a far better background for interpreting observed patterns. Our results are expected to provide a framework for future studies on the diversification and historical biogeography of the Amazonian aquatic biota.
INTRODUCTION
The Amazon basin is the largest hydrographic drainage on earth, covering ∼6 × 106 km2 (larger still if estuarine coastal areas are included) (Sioli, 1984; Milliman and Farnsworth, 2011), or about one-third of South America. Its discharge is also the largest in the world, with about one-fifth of the entire freshwater volume on the surface of the planet (Callede et al., 2004). Its vast size is matched by an equally vast fauna and flora (Webb, 1995; Patton et al., 2000; Hoorn and Wesselingh, 2010; Cardoso et al., 2017), comprising the richest ecosystem on earth.
Fishes are one of the faunal elements whose Amazonian biodiversity reaches superlative numbers. Despite such megadiversity and the attention it attracts, knowledge about the diversity and geographical distribution of Amazonian fishes is still not synthesized into a general framework that allows broad generalizations. Most data are scattered within thousands of species descriptions, and attempts to synthesize that information are inherently limited by the incomplete nature of their underlying databases. The Amazon basin still lacks a comprehensive list of its fish species and available estimates suggest a number between 1300 and 3500 species (e.g., Géry, 1969; Lundberg et al., 2000, 2010; Junk et al., 2007; Albert et al., 2011; van der Sleen and Albert, 2017). Some studies have tried to correlate Amazonian fish distributions with underlying causal factors that might form the basis for a historical biogeography of these fishes: Eigenmann (1909), Géry (1969), Vari (1988), Jégu (1992a, 1992b), Hubert and Renno (2006), Ribeiro (2006), Albert et al. (2011), Albert and Carvalho (2011), Lima and Ribeiro (2011), Ribeiro et al. (2013), and Dagosta and Pinna (2017). Some other papers focused on only parts of the Amazon are still relevant for the recognition of distribution patterns, including: Kullander (1986), Jégu and Keith (1999), Pearson (1937), Crampton (2011), López-Fernández and Albert (2011), and Lujan and Armbruster (2011).
The paradigm of vicariant biogeography (Croizat et al., 1974; Rosen, 1978; Nelson and Platnick, 1981) postulates that general distribution patterns demand general explanations, usually by associating lineage splitting with the origin of wide-ranging geographical barriers. Despite such conceptual clarity, in practice, the association of biogeographical patterns with geographical barriers that contributed to taxonomic diversification is difficult because of factors such as dispersal across existing barriers, area coalescence (obliteration of preexisting barriers and resulting biotic dispersal), extinction, and others that make the detection and delimitation of historically cohesive areas (biogeographical units) a complex process (Harold and Mooi, 1994; Szumik et al., 2002). One way to approach the issue is to search for geographical homologies (i.e., biogeographical congruence) in the form of distributional congruence among unrelated taxa (Patterson, 1981; Nelson, 1994; Morrone, 2009) and the application of the notion of primary homology (de Pinna, 1991) as adapted to biogeography (sensu Morrone, 2001, 2009). Despite such caveats, there is ample consensus that the formulation of robust historical biogeographical hypotheses is critically important for understanding biotic diversification and in deciding what conservation policies to adopt (Cracraft, 1994; Crisci, 2000; Whittaker et al., 2005; Guedes et al., 2014).
Analytical considerations aside, the identification of repeated patterns of geographical distribution is the first step toward a formulation of general biogeographical hypotheses. Such a task in itself can be daunting when dealing with large and complex taxonomic groups such as the Amazonian fishes. Basic knowledge about their taxonomy, phylogeny, and distribution has long been so irregularly scattered as to impede proper synthesis. On the other hand, knowledge about the systematics and distribution of Amazonian fish taxa has now accumulated to a degree that synthetic efforts are more enticing than ever before, both at specific and supraspecific levels.
The purpose of this paper is to identify the taxonomic patterns of distribution of Amazonian fishes based on all data currently available in the literature and in some of the largest ichthyological collections with significant Amazonian holdings. The list compiled for this report is an expansion and refinement of the database published in Dagosta and Pinna (2017), the largest previously done on the distribution of Amazonian fishes and provides the first comprehensive list of Amazonian fishes. It permits the identification and delimitation of all repeated patterns of distribution. We also offer a discussion on the possible underlying causes for each of the patterns and on their potential as indicators of a general biogeographical history of the Amazon basin. We expect our work will provide a general framework for the categorization of forthcoming distributional data (new records, new species, and new clades) and facilitate future progress on the biogeography of Amazonian freshwaters.
MATERIALS AND METHODS
Species distributions were compiled from all the information available in the literature, in a total of over 1500 references (see Dagosta and Pinna, 2017: appendices 1–6; and appendix 1, herein, for distributional data on fish species in the Amazon and surrounding basins), including taxonomic revisions, species descriptions, inventories and faunistic lists. Additionally, primary data were obtained from the most relevant (in size and Amazonian coverage) ichthyological collections, namely Instituto Nacional de Pesquisas da Amazônia, Manaus (INPA), Museu Nacional do Rio de Janeiro, Rio de Janeiro (MNRJ), Museu Paraense Emílio Goeldi, Belém (MPEG), Museu de Zoologia da Universidade de São Paulo, São Paulo (MZUSP), and National Museum of Natural History, Washington, DC (USNM). Two smaller collections, LBP (Laboratório de Biologia e Genética de Peixes, Botucatu), and LIRP (Laboratório de Icitiologia de Ribeirão Preto, Ribeirão Preto), were also surveyed because of their unique holdings of material from critical portions of the Brazilian Shield. Examined material is listed in Dagosta and de Pinna, 2017: Supplementary Material 2, and includes all information on source of data (citation, date of publication, and catalog number, when based on collections), institutional catalog, and sample numbers along with published sources for each species. All published information utilized was qualified as to accuracy regarding species identification and locality. Identifications in collections were verified by direct examination of specimens. Doubtful information was discarded. Compilation of the composition and geographical distribution of Amazonian fishes is updated to the end of 2018.
Maps presented are intended to represent general patterns of distribution and individual species plots can vary slightly within those limits. Delimitation of the Purus Arch follows Sacek (2014). Water type of Amazonian Rivers follows Venticique et al. (2016) (SNAPP Western Amazon Group – Amazon Aquatic Ecosystem Spatial Framework Knowledge Network for Biocomplexity. Usage is granted according to a Creative Commons “CC BY 4.0”). Hydrographic shape used in figures 5 through 22 is from Lehner et al. (2008), courtesy from HydroSHEDS (hydrological data and maps based on shuttle elevation derivatives at multiple scales: http://www.worldwildlife.org/hydrosheds). The shaded relief of South America used in figures 5 through 22 is courtesy of NASA/JPL-Caltech. The delimitation of the Amazonian regions follows Dagosta and Pinna (2017).
The taxonomic arrangement in appendix 1 follows Nelson (2006), except for the inclusion of recently described family Tarumaniidae (de Pinna et al., 2017), for considering Arapaimidae and Serrasalmidae as valid families, for including the genus Chalceus in Alestidae, and for adopting Cynolebiidae instead of Rivulidae. Within each family, genera are organized in alphabetical order. Species-level nomenclature follows Fricke et al. (2019), except for the validity of Astyanacinus, which is maintained herein since its synonymization within Astyanax is considered unjustified.
In figure 1, estimates of the number of fish species in the Congo basin is from Snoeks et al. (2011), in the Mekong from Poulsen et al. (2004), in the Nile from Witte et al. (2009), in the Mississippi from Robinson and Buchanan (1988) and in the Ganges from Sarkar et al. (2012). Figures for other basins (Atrato, Capim, Cauca-MagdalenaSinú, Coppename-Suriname, Corentyne-Demerara, Essequibo, Maracaibo, Maroni-Approuague, Oiapoque, Paraná-Paraguay, and Parnaíba) are compiled from the present work. Estimates of fish diversity in continents follow Lévêque et al. (2008), except for Europe, which is based on Kottelat and Freyhof (2007); South America, based on Reis et al. (2016); and Central America, based on Matamoros et al. (2015).
A distributional pattern is herein identified as the overlap (or major overlap) of geographical distributions of at least two species or monophyletic groups. Our definition of biogeographical pattern is deliberately broad and based solely on instantaneous geographical distributions of taxa. Thus, the recognition of a pattern herein is agnostic as to its causal reasons. It also does not preclude the possibility that some of the patterns are hierarchically arranged, i.e., that some patterns are actually subpatterns of larger ones. The selection and the delimitation of the distributional patterns are based on a visual analysis of map distributions of all Amazonian fishes, with consideration for previously published proposals. Repeated geographical distributions may be the result of ecological conditions, historical factors or both, and we consider that the identification and characterization of distribution patterns are relevant regardless of their underlying explanation and even without confirmation by rigorous statistical modeling procedures (which would be impossible at this point because of the heterogeneity of data sources).
Actual geographical distributions are a result of past events and processes in combination with present-day constraints not necessarily related to the vicariant processes that generated that diversity. Many factors are brought into consideration when untangling the nature and meaning of a distribution pattern, and those are discussed separately in the relevant sections. Our notion of taxa assumes that they exist as empirically backed biological entities, corresponding to either species or monophyletic groups. Taxonomic entities that are demonstrably not monophyletic are not considered as valid evidence, even if formally named in classification, because they are not empirically supported as a basis for biogeographical inference.
A chi-square test was used to verify the association of species' distributions with biogeographical patterns proposed (i.e., whether or not the species has a random distribution) and to test whether water type influences the distributions of specific species (appendix 2). The test was applied only to species shown in maps.
RESULTS AND DISCUSSION
Our inventory shows that the Amazonian ichthyofauna is composed of 2716 valid species, included in 529 genera, 60 families, and 18 orders. Such figures make the Amazon drainage, by a wide margin, the basin with the richest fish fauna in the world (fig. 1), with a diversity equivalent to that of some entire continents (fig. 1). As a comparison, the estimated (i.e., not necessarily described) number of species in the second most diverse basin in the world (Congo) is less than half that of the Amazon.
As in other Neotropical, African, and Asian drainages, the majority of Amazonian fishes belongs to the Otophysi (fig. 2), a group representing 80% (2193 spp.) of all Amazonian species. As in other Neotropical basins, the most species-rich orders are Characiformes and Siluriformes. The third largest order in Amazonian waters is the Perciformes, largely due to species of the family Cichlidae. Of the least diverse orders, 10 are from typically marine lineages that secondarily invaded Amazonian waters.
The familial composition also follows that a pattern typical of the majority of continental waters in the neotropics, vastly dominated by small body-size species. Five families (Characidae, Loricariidae, Cichlidae, Cynolebiidae, and Callichthyidae) concentrate most of the diversity (1528 spp. or 56% of Amazonian species), with Characidae alone comprising nearly a quarter of all Amazonian fish species.
The most species-rich genera are those composed of small body-size species. Among the 10 richest genera, half are Characidae (fig. 2C), demonstrating that the diversity of that family in the Amazon is concentrated in few genera. The same happens with Corydoras, the genus with the most Amazonian species, comprising 111 of the 133 species of Callichthyidae in the basin. Among the most diverse Amazonian genera, three were the object of relatively recent revisions (Creagrutus. Cyphocharax, and Sternarchorhynchus). Such revisionary works increased significantly the number of Amazonian species in the respective lineages: 42% of the Amazonian species of Creagrutus were described in Vari and Harold (2001); 28% of Cyphocharax in Vari (1992a), and 72% of Sternarchorhynchus in de Santana and Vari (2010). Thus, the position of those genera among the most diverse in the Amazon, especially in the case of Cyphocharax and Sternarchorhynchus, may be an artifact resulting from the lack of taxonomic revisions in other potentially more diverse yet poorly studied genera, for example, Chaetostoma. Knodus, and Rineloricaria.
Interesting facts appear when the diversity of the richest Amazonian lineages is compared with equivalent data from outside the basin. The Amazonian fish community is formed both by typically Amazonian lineages (i.e., most or all of their diversity is in the Amazon basin) and by lineages that have greater diversity in other South American basins or in the ocean (fig. 3A). At the family level, it is clear that the Amazon dominates, by a wide margin, the alpha-diversity in the vast majority of primary-division lineages sensu Myers (1938), with relatively little diversity in the secondary and peripheral divisions. The only exclusively Amazonian family is the recently described Tarumaniidae (Scoloplacidae and Lepidosirenidae occur also in the Paraguay basin).
Other interesting patterns are revealed by an examination of intrageneric diversity. The first one is that the most diverse genera in the Amazon have the majority of their species in the basin (fig. 3B). Also, some megadiverse Neotropical genera are relatively poorly represented in the Amazon, such as Astyanax. Trichomycterus. Bryconamericus. Characidium, and Hemibrycon. This is probably related to the fact that the species of those genera in the Amazon are concentrated along its outer rims, which indicate that their presence in the region is relatively recent, caused by secondary geological events involving adjacent basins. Two cases that stand out are Astyanax and Trichomycterus, which have their Amazonian diversity concentrated respectively in the upper Tocantins and upper Ucayali, right at the divide with other drainages.
The richest Amazonian tributary is the Rio Madeira (fig. 4), a fact that is attributed not only to its large size, but also to its hybrid nature (Dagosta and Pinna, 2017). The Rio Madeira drains one-third of the Amazonian lowlands and at the same time has tributaries associated with the Andean range and the Brazilian Shield, both areas particularly rich in fish endemics. The Rio Negro is also one of the most diverse Amazonian tributaries (fig. 4), despite its nutrient-poor and extremely acidic waters, which represent ecological barriers to numerous fish species (Goulding et al., 1988; Lima and Ribeiro, 2011). Despite that, the Rio Negro also harbors a large number of endemics (79 spp).
High levels of endemism are also seen in the Marañon and Ucayali (25% and 16%, respectively; fig. 4), a likely result of the Andean range in the diversification of the Amazonian biota. The largest absolute number of endemic taxa (87 spp.) is found in the upper Rio Tocantins followed by the Marañon (81 spp.). Other basins draining Amazonian Shield regions, such as Juruena, Teles Pires, Jari, and Madeira Shield tributaries, also show high levels of endemism, although relatively low total numbers of species. Such figures corroborate the hypothesis that highlands have proportionally fewer taxa with broad distributions and fewer cases of sympatry when compared with lowland rivers (Dagosta and Pinna, 2017).
The majority of Amazonian fish diversity is exclusive, with 63% (1696 spp.) of its species found nowhere else (fig. 4). That number alone would place the Amazon as the richest basin in the world in fish species.
PATTERNS OF DISTRIBUTION OF AMAZONIAN FISHES
Below we present the general patterns of distribution herein recognized on the basis of information compiled in appendix 1. Each category listed is described as to its geographical (physical) boundaries and subsequently discussed as to its historical significance, possible causes (if any is identified) and any other relevant points. The taxa supporting each pattern is provided in the text. Patterns are arranged according to logical sections and subsections, with each of them given a name (when a pattern was already recognized in the literature by a widely used name, that name was maintained here) and a number reference, so as to provide an easy reference in future publications.
Broadly Distributed Lineages
This pattern includes lineages which are broadly distributed in major South American drainages such as Amazon, Coastal Atlantic, Parana-Paraguay, São Francisco, Orinoco, and Guianan basins (fig. 5A). This is the most common pattern for the majority of families of fishes in the Amazon and also includes a number of genera such as Acestrorhynchus (see González, 2015), Brycon (see Lima, 2017), Characidium. Crenicichla (see Ploeg, 1991), Corydoras. Eigenmannia. Geophagus. Gymnotus (see Albert et al., 2005), Hoplosternum (see Reis, 1997), Hypostomus. Leporinus. Megaleporinus (see Ramirez et al., 2017), Pimelodella (see Slobodian, 2017), Pimelodus, Prochilodus, Serrasalmus, Schizodon, and Steindachnerina (see Vari, 1991).
As to species-level taxa with such wide distribution, three categories exist: (1) well-studied ones that have been extensively revised and are demonstrably a single taxonomic entity, such as Acestrorhynchus falcatus (see González, 2015) and Callichthys callichthys (fig. 5D; see Lehmann and Reis, 2004); (2) species that have never been the object of detailed revisions but that nonetheless lack published evidence of comprising a complex of species, such as Corydoras aeneus. Leporellus vittatus. Leporinus fasciatus. Pinirampus pirinampu; and (3) taxa recognized or suggested as species complexes but that have not yet been disentangled, such as Eigenmannia virescens (L. Peixoto, personal commun.), Erythrinus erythrinus (Martins et al., 2012), Hoplerythrinus unitaeniatus (Giuliano-Caetano et al., 2001), Hoplias malabaricus (fig. 5B; see Dergam et al., 1998), Leporinus friderici (Silva-Santos et al., 2018), Sternopygus macrurus (Silva et al., 2008), and Synbranchus marmoratus (fig. 5C; see Torres et al., 2005).
Therefore, it is likely that this “pattern” may not actually be a congruent biogeographical pattern among species, but simply the result of a taxonomic artifact. More refined biogeographical patterns may be hidden under incomplete or deficient taxonomic definitions.
It is interesting that none of the migratory fishes with notably vast living ranges, such as Brachyplatystoma spp., Prochilodontidae spp., and Curimatidae spp. include species with such broad ranges as those listed above (see Barthem and Goulding, 2007; Barthem et al., 2017). This is evidence that even species with extreme dispersal abilities meet with effective ecological/geographical barriers that keep them from achieving such broad ranges as defined in this pattern.
Amazon Core
The name given to this pattern has no relation to the origin of its components, but simply alludes to the region of greatest fish diversity in South America. The Amazon Core pattern is formed by fish lineages typical of the Amazonian biota that may be found both in lowlands and highlands of the Amazon basin and in adjacent basins such as Guiana drainages, Parnaíba, Capim, Orinoco, and Essequibo (fig. 6A) and that are absent in coastal drainages of southeastern Brazil, São Francisco, and Paraná-Paraguay. It is interesting to note that there is no species or lineage of fishes that is widely distributed throughout the high and low Amazonian lands not also present in adjacent basins of the Amazon such as Orinoco, Essequibo, or Guiana drainages. This is evidence that the Amazon basin is not an area of endemism, but instead is a historically composite area.
Examples of the Amazon Core pattern are numerous and include supraspecific taxa such as Chilodontidae, Anostominae, genera Argonectes (see Langeani, 1996), Boulengerella (fig. 6B; see Vari, 1995), Cichla (see Kullander and Ferreira, 2006), Jupiaba (see Zanata, 1997; presence of the genus in Paraguay basin is hypothesized to be secondary according to Ribeiro et al., 2013), Mastiglanis. Panaque (see Lujan et al., 2010), and Semaprochilodus (see Castro and Vari, 2004). This pattern of distribution is also reflected in some individual species such as Moenkhausia collettii (fig. 6C), Moenkhausia oligolepis (fig. 6D) and Potamotrygon orbignyi (see Da Silva and De Carvalho, 2015).
South American Lowlands
As the name suggests, taxa following this pattern occupy only the lower altitudes of cis-Andean South America, not occurring in Guiana and Brazilian Shield regions except in lowermost sectors of some large rivers such as Tapajós, Trombetas, and Xingu. Almost invariably, species with this pattern of distributions occur in the main channel of the Amazon and the Rio Madeira but exact limits vary according to taxon. Eigenmann was the first author to recognize that the ichthyofaunistic composition was different between South American high- and lowlands and named part of the latter as the “Amazon Province”: “East of the Cordilleras, and therefore east of the Magdalena basin, is found the most extensive and intricate fresh water system in the world. It forms a network of rivers practically uninterrupted, extending from the mouth of the Orinoco through the Casiquiare, Rio Branco, Rio Negro, Rio Madeira, Rio Guapore, Rio Paraguay, Parana and La Plata to Buenos Aires” (Eigenmann, 1909: 317).
Eigenmann (1909) correctly pointed out that the pattern extrapolates the hydrographic limits of the Amazon basin, and it is possible to list countless other examples of species that occur in the lowlands of the Amazon that are also present in other lowland drainages such as Orinoco, Paraná-Paraguay, Essequibo, and drainages east of the mouth of the Amazonas, from the Capim to the Mearim. Although this pattern is not exclusively Amazonian, the highland/lowland divide is the most widely discussed of all distributional patterns of Amazonian fishes (see Menezes, 1969, 1976; Kullander, 1986; Jégu, 1992a, 1992b; Lima and Ribeiro, 2011).
Lima and Ribeiro (2011) underscored a clear dichotomy between two geomorphological domains in northern cis-Andean South America: lowlands and highlands. Besides differences in historical-geomorphological parameters, high- and lowland regions affect their associated biotas differently (Albert et al., 2011; Lima and Ribeiro, 2011). Lowland drainages are more susceptible to hydrogeological changes and are in general more dynamic than highland drainages. Lowland rivers tend to be more directly interconnected than rivers draining other geomorphological areas, due to the action of meanders, anastomoses, megafans and mouth-position changes during sea-level oscillations (Lundberg et al., 1988). All those factors are less intense or nonexistent in highland rivers, which are typically well fitted in valleys of exposed crystalline rock and do not undergo significant lateral movements (Lima and Ribeiro, 2011). Thus, lowland rivers undergo constant and much faster hydrogeographic changes and, as a consequence, congregate more taxa with broad distributions and more cases of sympatry than highland rivers. In addition, habitat stability provided by an enormous living space for lowland Amazonian fish species seems to be an important factor in decreasing the extinction rate of lineages (Lundberg et al., 1988; 2010), which in turn also results in increased diversity. Therefore, those are the factors that explain why most Amazonian fish species fit a South American lowland pattern.
There are spatial differences among distinct groups of lowland Amazonian fishes, with at least five different subpatterns as explained below. The examples and subpatterns recognized herein have a direct relation with the complex geomorphological history of the Western Amazon.
Amazon and Orinoco Lowlands
Eigenmann (1909) delimited his Amazonian Province from a dispersionist perspective, where present-day connections among drainages provided the explanation for faunal sharing among basins. However, the correct interpretation for most of such massive ichthyofaunal sharing among different lowland South American basins is directly related to a complex historical context that began in the Upper Cretaceous, at least, with the formation of the Sub-Andean Foreland basin (Lundberg et al., 1998) and has little relation to present (and rather ineffective) physical connections (e.g., Casiquiare canal).
The sub-Andean Foreland is a series of retro-arc depressions lying to the east of the Andean Cordilleras that served as the main drainage axis of South America throughout the Upper Cretaceous to the Paleogene (Cooper et al., 1995; DeCelles and Giles, 1996; Lundberg, 1998; DeCelles and Horton, 2003; Albert and Reis, 2011; Lima and Ribeiro, 2011; Wesselingh and Hoorn, 2011). For much of its existence, the Sub-Andean Foreland was drained mostly by the proto-Amazon-Orinoco basin (Lundberg et al., 1998), even though the latter has also drained other areas of the South American platform further east.
Both the pattern described here and the Amazon Province of Eigenmann (1909) match mostly (exclusive of the La Plata basin included in Eigenmann's province) the spatial limits of the proto-Amazon-Orinoco, which was a continuous basin until its fragmentation in the late Miocene (ca. 10 Ma) as a result of the rise of the Vaupes Arch in eastern Colombia that separated the modern Orinoco and Amazon basins (Hoorn, 1994a; Cooper et al., 1995; Harris and Mix, 2002; Albert and Carvalho, 2011; but see Mora et al., 2010, for a more recent estimate). That barrier may have prevented lineages that diversified after its rise from increasing their range throughout all lowland regions and may also have caused the extinction of lineages in some of those basins. Those two factors may explain the absence of some typical Amazonian lowland forms in the Orinoco basin (see examples in Amazon-only Lowland). On the other hand, part of the faunal sharing between the Amazon and Orinoco may result from broad distributions before the modern separation between those basins, i.e., from the proto-Amazon-Orinoco. Still another hypothesis to explain the same pattern is megafan dynamics, geologically more recent (see Wilkinson et al., 2010). As will be discussed in the section Negro and Orinoco, the Canal Casiquiare does not seem to be a relevant dispersal route to explain the extensive list of taxa shared between the Amazon and Orinoco lowlands.
Some examples of exclusive taxon sharing between the Amazon and Orinoco lowlands are: Acanthicus hystrix (see Chamon, 2016), Acestrorhynchus heterolepis (see González, 2015), Brachyhypopomus sullivani (see Crampton et al., 2016), Brycon amazonicus (see Lima, 2017), Boulengerella maculata (see Vari, 1995), Cetopsis coecutiens (see Vari et al., 2005), Cynodon gibbus (see Toledo-Piza, 2000a), Colossoma macropomum, Lasiancistrus schomburgkii (see Armbruster, 2005), Leptodoras paelongus (see Sabaj Pérez, 2005), Metynnis guaporensis and M. luna (see Ota, 2015), Moenkhausia comma, Moenkhausia lepidura (fig. 7B, see Marinho and Langeani, 2016), Mylossoma albiscopum (see Mateussi, 2015), Nemadoras cristinae (see Sabaj Pérez et al., 2014), Paragoniates alburnus (see Quevedo, 2006), Peckoltia bachi (see Armbruster, 2008), Potamorhina altamazonica (fig. 7C, see Vari, 1984), Sorubim elongatus (see Littmann, 2007), Trachydoras brevis. T. gepharti. T. microstomus and T. nattereri (see Sabaj and Arce, 2017), Vandellia cirrhosa (fig. 7D), Adontosternarchus spp. (see Mago-Leccia et al., 1985), Brachyrhamdia spp. (see Slobodian, 2013), Chalceus spp. (see Zanata and Toledo-Piza, 2004), Compsaraia (see Bernt and Albert, 2017), Hassar spp. (see Birindelli et al., 2011), Laemolyta spp. (see Mautari and Menezes, 2006), Liosomadoras spp. (see Birindelli and Zuanon, 2012), Microphilypnus spp. (see Caires and Figueiredo, 2011), Tenellus spp. (sensu Birindelli, 2014; Sabaj Pérez et al., 2014), and Sternarchogiton spp. (see de Santana and Crampton, 2007).
Amazon and Paraguay Lowlands
There are many Amazonian Lowland fish lineages that also occur in the Paraná-Paraguay basin, which was not permanently connected to the proto-Amazon-Orinoco. The location of the watershed divide between the proto-Amazon-Orinoco River basin and the La Plata changed between the end of the Paleogene and the beginning of the Neogene (see Tagliacollo et al., 2015). Initially, it was the Chapare Buttress in the Late Oligocene (ca. 30–20 Ma) (Lundberg, 1998) and subsequently the Michicola Arch, starting during the Late Miocene (ca. 11.8–10 Ma) in the area of modern eastern Bolivia (Lundberg et al., 1998; Montoya-Burgos, 2003; Albert and Carvalho, 2011; Carvalho and Albert, 2011a). Several events may have permitted biotic dispersal between the Amazon and Paraguay: upper Paraguay captures of protoAmazonas-Orinoco headwaters (Lundberg et al., 1998), Amazon capture of upper Paraguay headwaters (Lundberg et al., 1998), river megafans involving the upper Río Mamoré and tributaries of the upper Río Paraguay (Wilkinson et al., 2006, 2010; Ota et al., 2014) and capture of upper Rio Paraguay into the upper Rio Guaporé (Ota et al., 2014). Because all possible connections between the Amazon and the Paraná-Paraguay happened as a result of separate events of different ages, it is very likely that many species shared between those basins, despite their congruent distributions, lack temporal congruence. They correspond instead to cases of pseudocongruence, sensu Donoghue and Moore (2003), and are not biogeographically homologous. Taxa shared between those basins include: Acestrorhynchus abbreviatus (see González, 2015), Acestrorhynchus gr. lacustris (see González, 2015), Brachyhypopomus bombilla (see Crampton et al., 2016), Epapterus dispilurus (see fig. 8B; Vari and Ferraris, 1998), Hemigrammus lunatus (see fig. 8C; Ota et al., 2014), Mesonauta festivus (see fig. 8D; Kullander and Silfvergrip, 1991; Schindler, 2005), Moema spp. (see Costa, 2004), and Prionobrama spp. (see Quevedo, 2006). A complete list of species shared exclusively between the Madeira and the Paraguay is presented in Madeira and Paraguay.
Amazon-only Lowland
Whitewater Amazonian rivers have high sediment and nutrient loads and a neutral pH, draining a relatively young Andean range. Major whitewater tributaries include the Marañón, Purus, Madeira, Juruá, Putumayo, Japurá, and Napo rivers. The whole Rio Amazonas system exhibits whitewater, although it receives other water types from various tributaries. There are few investigations into the impact of such water type changes on the biogeography of Amazonian fishes. Vari (1988) suggested that some curimatids are restricted to whitewater rivers and that their distribution may be more closely linked to ecological rather than historical factors. While the pattern is correct in some cases, we also agree with Lima and Ribeiro (2011: 157) that “some ecological factors that clearly influence fish distribution patterns in northern cis-Andean South America, such as water typology, are, as mentioned previously, a consequence of geomorphological processes and, as such, possess a historical component.” Thus, it is possible that lowland species restricted to the Amazon reached such distribution from different causes and histories, either because they diversified after separation of the Orinoco from the protoAmazon-Orinoco basin or because they are whitewater dependent.
Evidence suggests that the interpretation of Vari (1988) may be correct for a set of species showing this pattern of distribution. Some of the exclusively Amazonian lowland species are absent in the Rio Tocantins basin, having their distributions limited to the region of the mouth of the Madeira. This may indicate an association with whitewater since tributaries with that type of water become practically nonexistent downstream of that part of the Amazon river. Some examples of the Amazon-only Lowland pattern are: Adontosternarchus balaenops (see fig. 9B; Mago-Leccia et al., 1985), Agoniates anchovia, Aphanotorulus horridus (see Ray and Armbruster, 2016), Aphanotorulus unicolor (see Ray and Armbruster, 2016), Apionichthys nattereri (see Ramos, 2003), Brycon melanopterus (see Lima, 2017), Chalceus erythrurus (see Zanata and Toledo-Piza, 2004), Cetopsis candiru (see fig. 9C; Vari et al., 2005;), Cetopsis oliveirai (see Vari et al., 2005), Chaetobranchopsis orbicularis, Copella stigmasemion (see Marinho and Menezes, 2017), Crenicara punctulatum, Curimata aspera (see Vari, 1989a), C. kneri (see Vari, 1989a), Curimatella meyeri (see fig. 9D; Vari, 1992b), Cyphocharax spiluropsis (see Vari, 1992b), C. notatus (see Vari, 1992b), C. plumbeus (see Vari, 1992b), Denticetopsis seducta (see Vari et al., 2005), Hydrolycus scomberoides (Toledo-Piza et al., 1999), Hypostomus pyrineusi (see Armbruster, 2003), Leporinus jamesi (see Garavello et al., 2014), Protocheirodon pi (see Vari et al., 2016), Mylossoma aureum (see Mateussi, 2015), Nemadoras elongatus, N. hemipeltis, N. humeralis (see Sabaj Pérez et al., 2014), Potamorhina latior (see Vari, 1984), Prionobrama filigera (see Quevedo, 2006), Psectrogaster amazonica (see Vari, 1989b), Pseudobunocephalus amazonicus (see Cardoso, 2008), P. bifidus (see Cardoso, 2008), Scoloplax dicra (see Schaefer et al., 1989), Sorubim maniradii (see Littmann, 2007), Steindachnerina bimaculata (see Vari, 1991), S. leucisca (see Vari, 1991), Sternarchella calhamazon (see Lundberg et al., 2013), Trachydoras steindachneri (see Sabaj and Arce, 2017), Aphyolebias spp. (see Costa, 2004), and Chaetobranchopsis spp.
The pattern of distribution described herein is repeatedly supported as biogeographically coherent in the analyses of Dagosta and de Pinna (2017).
Amazonas-Paraguay-Orinoco Lowland
This pattern is the least common one among lowland species in the Amazon basin. Most cases are also present in the Tocantins basin, but not in Guianan drainages. Some examples of this pattern are: Abramites hypselonotus (see Vari and Williams, 1987), Curimatella dorsalis (see Vari, 1992a), Hypophthalmus oremaculatus (see fig. 10D; Littmann et al., 2015), Rhaphiodon vulpinus (see fig. 10B; Toledo-Piza, 2000a), Roeboides affinis (see Lucena, 2007), Sorubim lima (see fig. 10C; Littmann, 2007), and Mylossoma spp. (see Mateussi, 2015). The Amazonas-Paraguay-Orinoco Lowland pattern comprises areas from the previously described Amazon and Orinoco Lowlands as well as the Amazon and Paraguay Lowlands, and thus the associated geological processes are the same as discussed in the respective headings.
Amazonas-Guiana-Orinoco Lowland
Species with this pattern of distribution are broadly distributed in the lowlands of the Rio Amazonas and, in some cases, also of the Orinoco and Essequibo, but they are also found in the lowlands of Guiana coastal drainages. The pattern differs from the Amazon Core pattern described above because it is restricted to lowland lineages. Some examples of this pattern here recognized are: Iguanodectinae, Arapaima spp., Brachyplatystoma spp., Copella spp. (see Marinho and Menezes, 2017), Crenuchus spp. (see fig. 11B; Campanario, 2002; Pires et al., 2016), Electrophorus spp., Heros spp., Hypophthalmus spp. (also in Paraguay basin, see Littmann et al., 2015), Mesonauta spp. (also in Paraguay basin, see fig. 11C; Kullander and Silfvergrip, 1991; Schindler, 2003), Osteoglossum spp., Pachypops spp. (see Casatti, 2002), Ageneiosus dentatus (see Ribeiro et al., 2017), Brachyhypopomus beebei. Brachyhypopomus brevirostris and Brachyhypopomus regani (see Crampton et al., 2016), Caenotropus labyrinthicus (see Vari et al., 1995), Gnathocharax steindachneri. Hemigrammus unilineatus (fig. 11D), Hemigrammus ocellifer (fig. 11D), and Serrasalmus rhombeus.
The geomorphological explanations associated with this pattern as the evolution of the sub-Andean Foreland region (discussed in Amazon and Orinoco Lowlands) and the advance of the freshwater plume associated with marine regressions (discussed in Guiana Mangrove Province), which allowed contact the regions of Amazonas-Orinoco Lowland and the coastal drainages of Guiana.
Other Cases of Amazonian Lowland Distribution
Fossils of Lowland Amazonian Lineages
Several Tertiary fossils belonging to typically lowland Amazonian taxa are present in regions currently lacking any Amazonian connection, such as Magdalena and Caribbean coastal rivers from the northern coast of Venezuela: Arapaima. Brachyplatystoma. Colossoma. Doras. Hydrolycus, and Phractocephalus (G. Ballen, personal communication; Lundberg et al., 1986, 1988, 2010; Lundberg, 1997, 2005; Sabaj Pérez et al., 2007). This demonstrates that such regions were in the past also part of some other distributional patterns described herein. If such fossils were not known, our understanding of the biogeographical history of the region would be severely incomplete or incorrect. Besides, the absence of extant representatives of the listed lineages in the Caribbean coastal rivers from northern coastal Venezuela and in Rio Magdalena basin is a clear demonstration of the dynamic nature of biogeographical phenomena, which change drastically over time and may bear little relationship to present-day physical barriers that determine the distribution of recent taxa.
Eastern Lowland Amazon
Some lowland distributional patterns comprise basins east of the mouth of the Rio Tocantins, beyond the eastern limit of the Amazon basin itself (i.e., Capim, Gurupi, Turiaçu, Mearim, Itapecuru, and Parnaíba). Interestingly, each of those basins has fewer and fewer Amazonian lineages as they get progressively farther from the mouth of the Amazon. Some of them are poorly sampled, mainly the Gurupi, Turiaçu, and Mearim, resulting in probably artifactual discontinuous distributions between neighboring basins and blurring details of the pattern. Species shared exclusively between the lower Amazon and above-cited eastern basins include: Brachychalcinus parnaibae (lower Tocantins and Parnaíba, see Reis, 1989), Brachyhypopomus pinnicaudatus (Amazon estuary, Capim, and Mearim, see Crampton et al., 2016), Corydoras jullii (lower Tocantins, Mearim, and Parnaíba, see Dagosta and Pinna, 2017); Apistogramma caetei (lower Tocantins, Capim, and Gurupi; see Dagosta and Pinna, 2017) and Nannostomus nitidus (Amazon estuary and Capim; see Dagosta and Pinna, 2017).
Most examples of Amazonian lineages that occur in drainages to the east of the Amazon basin are also widespread in lowland waters of the Amazon. Some examples are: Anablepsoides urophthalmus (Capim, Gurupi and Mearim, see Costa, 2006), Brachyplatystoma spp. (Capim, cf. Lundberg and Akama, 2005; Mearim and Parnaíba; see Ramos et al. Ramos et al., 2014), Caenotropus labyrinthicus (Capim, see Vari et al., 1995; Parnaíba; see Ramos et al., 2014), Curimata spp. (Itapecuru, see Barros et al., 2011; Parnaíba, see Ramos et al., 2014), Cynodon gibbus (Itapecuru, see Barros et al., 2011; Parnaíba, Toledo-Piza, 2000a), Gymnocorymbus thayeri (Gurupi, see Benine et al., 2015; Parnaíba, see Benine et al., 2015), Jupiaba polylepis (Parnaíba, see Ramos et al., 2014), Poptella compressa (Capim, Mearim, Parnaíba, and Turiaçu, see Reis, 1989; Itapecuru, see Barros et al., 2011), Pseudoplatystoma punctifer (Itapecuru, Barros et al., 2011; Parnaíba, see Buitrago-Suárez and Burr, 2007) and Vandellia cirrhosa (Capim and Turiaçu, fig. 7D).
Marine Derived Lineages
A lowland Amazonian pattern is also seen in typically marine lineages (peripheral division sensu Myers (1938) that invaded Amazonian waters, such as Achiridae (fig. 12A), Batrachoididae, Belonidae (fig. 12B), Clupeidae, Engraulidae (fig. 12C), Gobiidae, Hemirhamphidae, Pristigasteridae (fig. 12D), and Tetraodontidae (fig. 12E) (see Bloom and Lovejoy, 2017, for further details).
Bloom and Lovejoy (2017) convincingly demonstrate that different marine lineages colonized South American rivers at different ages, influenced by separate events of marine transgression. According to those authors, different groups have biogeographical patterns consistent with invasions during the Oligocene, Eocene, or Miocene marine incursions. The Amazonian half-beak is the only lineage younger than the Miocene to have invaded Amazonian freshwaters less than a million years ago. These facts make it clear that the lowland Amazonian pattern of marine-derived lineages is pseudocongruent (sensu Donoghue and Moore, 2003).
None of the marine-derived lineages is found in the highest parts of Guianese and Brazilian shields (see figs. 12A–E). Such fact suggests that the invasion of peripheral groups in the Amazon occurred subsequently to the establishment of the ichthyofauna in those upland regions. The alternative hypothesis that such lineages went extinct in the higher regions of the shields has no evidential support, either paleontological or geomorphological.
Deep Channel Species
Another lowland Amazonian pattern is demonstrated by species restricted to the deep channel of large Amazonian rivers, mainly the Amazonas itself (e.g., Sternarchella duccis, see Evans et al., 2017; Sternarchella rex, see Evans et al., 2017; Sternarchella sima, see Evans et al., 2017; Leptodoras juruensis, see Sabaj Pérez, 2005; Pariosternarchus amazonensis, see Albert and Crampton, 2006). Benthic regions of many of these large rivers contain specialized communities, mostly composed of electric fishes. Some of those species are strictly associated with deepwater environments, though there are records also in flooded beaches. Crampton (2007) listed at least 64 species of the Gymnotiformes that inhabit deep waters in the Amazon. Because deepwater Amazonian samples are still few and have been mostly focused on large Amazonian rivers, it is possible that this apparent pattern is the result of a sampling artifact of species actually belonging to some of the other South American lowlands patterns.
Guiana Mangrove Province
This pattern comprises the lower portions of the Orinoco basin, the Guiana coastal drainages and the lower Amazon (fig. 13A). Its western limit for most taxa is usually the mouth of the Orinoco, but in some cases it extends to the small independent coastal Venezuelan drainages, such as with Polycentrus schomburgkii (see Coutinho and Wosiacki, 2014). The eastern limit of this province is usually the mouth of the Amazon, with some species occurring also in the lower sectors of the Rio Jari, Tocantins, and Xingu and some reaching even further east to the Brazilian State of Maranhão. This pattern was first recognized by Myers (1960), in describing distribution patterns in the subfamily Aspredininae, of Aspredinidae: “They [the Aspredininae] are fishes of the lowland, muddy coast of Guiana and Amazonia, where they occur in the sea, in brackish water, and in the estuaries and tidal portions of rivers. They do not seem to be found far inland anywhere, except in the lower Amazon, where they apparently occur in many (or all) parts of the vast, complicated delta area, where the tides or tidal bores (pororoca) are felt” (Myers, 1960: 133).
Myers (1960) describes the limits of the pattern as from the Orinoco delta in Venezuela into the Brazilian state Maranhão, and calls it the “Guyana Mangrove Province,” a name adopted here. The same pattern was later independently described by Vari (1988: 355) on the basis of data from Curimatidae: “The Atlantic slopes of Guyana, Surinam and French Guiana and Amapá in Brazil are another area of endemism....” Vari (1988) notes that some species in that area also occur in the lower Amazon and in the Rio Tocantins, such as Curimata cyprinoides. This pattern is strongly corroborated as a historical unit in the analyses of Dagosta and de Pinna (2017).
Examples of Guiana Mangrove Province Pattern include: Anableps spp., Anablepsoides urophthalmus group (see Nielsen, 2016), Aspredinichthys filamentosus (see Myers, 1960), Aspredinichthys tibicen (see Myers, 1960), Aspredo aspredo (see Myers, 1960), Copella arnoldi (see Marinho and Menezes, 2017), Curimata cyprinoides (see fig. 13B; Vari, 1989a), Cyphocharax helleri (fig. 13C), Hemigrammus rodway, Hemigrammus guyanensis, Nannostomus beckfordi, Piabucus dentatus, Platystacus cotylephorus (see Myers, 1960), Poecilia parae, P. picta, Poptella longipinnis, Polycentrus schomburgkii (see fig. 13D; Coutinho and Wosiacki, 2014), Pristella maxillaris, Pseudauchenipterus nodosus, Rhinosardina amazonica (see Whitehead, 1985), and Tomeurus gracilis (see Myers, 1960).
It seems unlikely that nonrheophilic species such as those listed above have the pattern of distribution they do as a result of sequential stream capture events or that they represent relictual distributions of ancient and more widely distributed populations. Instead, their distribution pattern seems to be strongly correlated with events of marine transgression and regression.
An immense freshwater plume is formed by the discharge of the Amazon in the Atlantic (Goulding et al., 2003; Rocha, 2003). Such volume of freshwater floats above the heavier saltwater and spreads northwestward carried by the southern equatorial current (Jégu and Keith, 1999; Albert et al., 2006), resulting in turbid waters and largely unconsolidated substrates between the mouths of the Amazon and Orinoco (Curtin, 1986; Rocha, 2003). The effect of that plume varies seasonally and according to changes in sea level over time (Rocha, 2003). The first authors to propose the effect of the Amazonian plume on freshwater fish distribution were Jegú and Keith (1999). In their model, the plume serves as a corridor permitting the advance of species from the lower Amazon toward Guiana coastal drainages, thus explaining the common elements between those regions (see the distribution of Brachyhypopomus pinnicaudatus in Crampton et al., 2016, for another example). It is also likely that events of marine transgression and regression have altered the hydrogeological dynamics of the region affected by the plume, isolating or uniting different coastal drainages.
It is not yet known whether the matching distributions of strictly freshwater (as listed above) and marine-tolerant species (e.g., Aspredinichthys filamentosus. Aspredinichthys tibicen, Aspredo aspredo, Platystacus cotylephorus, Pseudauchenipterus nodosus, Rhinosardina amazonica, and Stictorhinus potamius) are congruent or pseudocongruent (sensu Donoghue and Moore, 2003). Myers (1960) notes that Tomeurus and Anableps have the same distribution pattern as Aspredininae, even though the two former taxa are not as tolerant to saltwater.
The salt-tolerant species are restricted to lower portions of rivers, close to their mouths. Their tolerance to marine water may have allowed their range expansions without the need of transgression-regression events. Population phylogeographical studies may bring light to this question, by comparing divergence times between populations in the lower Amazon and those in the Guiana coastal drainages, both in strictly freshwater and marine-tolerant lineages. Regardless of salt-tolerance considerations, it is expected that the biogeographical pattern described above is at most ∼11 Ma, i.e., as old as the age of the present connection between the Amazonas and the Atlantic (Hoorn, 1994a, 1994b, 1996; Potter, 1997) and also coinciding with the corresponding Andean uplift at that latitude (Hoorn et al., 1995).
Eastern Amazon (East of the Purus Arch)
Structural arches like the Purus Arch are basement structures located under sediments of different ages that are not exposed superficially in the eastern Amazon (Rossetti et al., 2005). As a consequence, such structures cannot have acted as biogeographical barriers from the end of the Miocene to the present (Campbell et al., 2006; Lima and Ribeiro, 2011). The Purus Arch is thought to have acted as a crucial barrier until the late Miocene, dividing the Eastern and Western Amazon (Figueiredo et al., 2009; Hoorn et al., 2017), although its role in the formation of the Amazon and its influence in the distribution of the biota remain controversial (see Wesselingh and Salo, 2006). Why and how two previously disconnected aquatic systems, the eastern and the western Amazon, merged is still unclear (Hoorn, 1994a, 1994b; Wesselingh, 2006; Figueiredo et al., 2009; Hoorn et al., 2017). Such uncertainty impedes proper understanding of the consequences of the event for the biogeography of Amazonian fishes. It is certain nonetheless that several lineages of fishes display distributions spatially congruent with a western/ eastern Amazon divide, with their limit coinciding exactly with the Purus Arch.
The Eastern Amazon pattern may represent the distribution of species historically associated with the region lying east to the Purus Arch, for the most part cratonic and draining clear or black waters (Harris and Mix, 2002; Wesselingh and Hoorn, 2011). This pattern comprises the drainages of the Rio Negro, Orinoco, Essequibo, and Amazonian versants of the Brazilian and Guianan shields (fig. 14A). Some examples of this pattern are: Aphanotorulus emarginatus (see fig. 14B; Ray and Armbruster, 2016), Baryancistrus spp., Bivibranchia fowleri. Colomesus tocantinensis (see Ruiz, 2015), Caquetaia spectabilis. Geophagus altifrons. Gnathodolus bidens. Hydrolycus tatauaia (see Toledo-Piza et al., 1999), Leporinus brunneus (see Lima and Ribeiro, 2011), Pachyurus junki (fig. 14C) and Synaptolaemus latofasciatus (see fig. 14D; Britski et al., 2011). There are at least two examples of lineages distributed also in parts of the Atlantic drainages in the Guiana Shield, Brycon gr. pesu (see Planquette et al., 1996) and Hoplias curupira (see Oyakawa and Mattox, 2009).
Water type and drainage relief are not enough to explain why the Rio Negro basin, for example, shares such a great number of lineages with highland drainages, despite the Negro's small shield coverage. The existence of so many shared taxa exclusively between the Negro and shield rivers indicates a shared history.
The East of the Purus Arch pattern may be directly related to the hydrogeological dynamics of the Amazon basin during the Miocene. The uplift of the central and north portions of the Andes created an overload on the South American plate that caused a lithospheric flex-ion, which in turn opened space for the formation of a sub-Andean Foreland basin (Sacek, 2014). Between ∼24 and 16 Ma, this foreland basin received sediments from rivers draining west of the Purus Arch and east of the Andes, carrying them northward toward the Caribbean (Crampton, 2011; Sacek, 2014). At least since the Eocene, the Purus Arch (fig. 14A) was a divide between the basins draining east (eastern Amazon basin) and those draining westward (Lundberg et al., 1998; Costa et al., 2001; Crampton, 2011; Lujan and Armbruster, 2011). At that time, the eastern Amazon basin was formed by rivers draining shield areas, sediment poor and probably clear- or blackwater (Harris and Mix, 2002; Wesselingh and Hoorn, 2011). That phase was followed by the formation of an immense lacustrine system known as Pebas (∼16 to 10.5 Ma), probably separated from the eastern Amazon system also by the Purus Arch (Figueiredo et al., 2009; Crampton, 2011; Sacek, 2014). The accumulation of sediments, mostly Andean in origin, in the foreland basin and the continuing Andean uplift (Crampton, 2011; Lima and Ribeiro, 2011) resulted in a breaching of the Purus Arch and a connection between that drainage and the eastern Amazon, forming a transcontinental basin and starting the Andean sedimentary deposition in the Brazilian equatorial margin, which extends to the present (Figueiredo et al., 2009; Sacek, 2014).
Andean sediments in large amounts in the mouth of the Amazon begin approximately by 7 Ma, indicating that the west-east water divide was effective until that date (Hoorn, 1994b; Figueiredo et al., 2009; Crampton, 2011; for more recent estimates, see Roddaz et al., 2005; Rossetti et al., 2005; Campbell et al., 2006; Espurt et al., 2007). As a consequence, the pattern East of the Purus Arch pattern described herein is at least between 2.6 and 7 Ma, but in reality, it is probably far older because that time interval marks only the last instant before lineages south and north of the Amazonian tributaries were separated. A biogeographical pattern similar to the one described here was mentioned by both Eigenmann (1909) and Lima and Ribeiro (2011) under the highlands designation. However, for both authors relief is the decisive factor explaining the spatial distribution of fish species, rather than the past influence of the Purus Arch as proposed here. It is important to note that the Purus Arch today has no influence as a barrier on species distributions. Its role is relevant as a past barrier, when it formed the water divide between eastern and western Amazon. The reason why most lineages on each side of the divide do not expand their distributions is a mystery, perhaps related to historical-ecological factors independent of any present-day physical remains of the Purus Arch. The fish lineages east of the Purus arch are mostly ecologically restricted to fast-flowing and sediment-poor tributaries, not entering the main Amazonian channel. The westward dispersion of those species was once limited by the Purus Arch when it was an effective barrier. The demise of the Purus Arch as a significant barrier is synchronous with the formation of the main channel of the Amazon. Slightly upstream of the mouth of the Rio Negro, approximately at the site of the ancient Purus Arch barrier, the main channel of the Amazon becomes significantly deeper (Geritana and Paiva, 2013). That factor, in combination with the simultaneous massive input of acidic waters from the Rio Negro, probably makes the region impervious to many taxa narrowly adapted to conditions of western Amazon waters.
This region then started acting as an ecological barrier to those species west of the former Purus Arch, effectively replacing it. This provides an example that distributional patterns may have been determined by past barriers having no relation to current geographical boundaries, but nonetheless linked by a causal chain of different yet overlapping barriers. Of course, the number of fish species showing this pattern may seem small in view of the potential importance of the Purus Arch. However, the congruent distributions of unrelated lineages despite the absence of any apparent physical or ecological barriers cannot be ignored and may represent the last remnants of a common biogeographical history.
Amazon-Core Uplands
This pattern comprises species endemic to basins that drain the Brazilian and Guiana shields both in Atlantic and Amazonian versants. As mentioned above in South American Lowlands, Eigenmann (1909) was the first author to identify faunistic differences between South American lowlands and highlands. That author also inferred ages for those regions, implying that they have distinct biogeographical histories. Eigenmann (1909: 318) correctly proposes that both the Guianan and Brazilian shields are older than lowland regions: “The parts that first arose out of the sea and became populated with freshwater fishes were probably two land areas. The one embraces the highlands of Guiana and Northern Brazil, the other the highlands of Brazil east of the Araguay and south of the falls of the Tapajos.” The pattern described here is very similar to the one described by Eigenmann (1909) and differs from the Eastern Amazon pattern in excluding predominantly lowland drainages such as the Rio Negro and by including Guiana coastal basins (fig. 15A). Fishes displaying the Amazon-Core Highlands pattern are in general rheophilic: Acnodon spp., Anostomus ternetzi (see Lima and Ribeiro, 2011), Cetopsidium spp. (also in upper Rio Negro, see fig. 15B; Vari et al., 2005), Centromochlus schultzi. Hemibrycon surinamensis (see Bertaco and Malabarba, 2010), Hemigrammus ora (see fig. 15C; Jerep et al., 2011), Hoplias aimara (see fig. 15D; Mattox et al., 2006), Jupiaba essequibensis. J. gr. meunieri and J. polylepis. Krobia spp., Leporinus maculatus. Leporinus gr. granti. Moenkhausia grandisquamis. Mylesinus spp., Petulanos spp., Retroculus spp., Roeboexodon guianensis (see Lima and Ribeiro, 2011), and Tometes spp. (see Andrade, 2013; Andrade et al., 2016). The five categories described below are sub-patterns within the larger Highland Amazon Core pattern.
Amazonian Uplands
This pattern comprises exclusively Amazonian rivers draining both shields, Brazilian and Guianan (fig. 16A). Exact limits of this pattern are yet somewhat vague because known examples are species or lineages that occur in very narrow sectors of rivers, forming fragmented distributions based on sparse records. Species following this pattern are typically rheophilic and include: Archolaemus luciae (see Vari et al., 2012), Baryancistrus niveatus, Cetopsidium orientale (see Vari et al., 2005), Doras higuchii (see Sabaj Pérez et al., 2008), Hypomasticus julii. Leporinus britskii (see Feitosa et al., 2011), Leporinus microphysus (see Birindelli and Britski, 2013), Leporinus pachycheilus (also in Rio Araguari basin, see Santos et al., 1996), Metynnis anisurus (also in upper Rio Paraná basin, see Ota, 2015), Moenkhausia celibela (see Marinho and Langeani, 2010), Mylesinus schomburgkii. Sartor spp. (fig. 16B), and Teleocichla spp. (fig. 16C) and Tocantinsia piresi (fig. 16D).
Guiana Shield (Atlantic and Amazonian Versants)
This pattern includes lineages shared exclusively between Amazonian and Atlantic versants of rivers draining the Guiana Shield (fig. 17A). It possibly results from ichthyofaunistic exchange caused by stream capture events (see Cardoso and Montoya-Burgos, 2009). The pattern as a whole is probably not temporally congruent, but instead formed by independent events that caused faunistic mixing in the region, a common phenomenon between neighboring headwaters in shield rivers. Nijssen (1970) was the first author to propose that the headwater regions of north and south Guianan rivers might serve as a corridor for fish distribution. Subsequent authors, such as Cardoso and Montoya-Burgos (2009) and Lujan and Armbruster (2011), proposed additional examples of this pattern and its role as a faunistic connection between the Guianas and the Amazon.
There are few examples of this pattern, in part as a result of the yet incipient knowledge on the fish fauna of upper reaches of Amazonian versants of the Guiana Shield. Some examples include: Corydoras baderi (Paru do Oeste and Maroni, see Nijssen and Isbrücker, 1980), Cteniloricaria spp. (Paru do Oeste, Maroni, Suriname, Corentyne, and Essequibo, see Covain et al., 2012), Hypomasticus megalepis (Trombetas, Uatumã and Guianese rivers, see Mol et al., 2012; J. Birindelli, personal commun.), Lithoxus spp. (Fisch-Muller, 2003), Microglanis secundus (Trombetas and Saramacca, see Ruiz and Shibatta, 2010), Parodon guyanensis (Paru do Oeste, Maroni, Suriname, Corentyne, and Essequibo), Parotocinclus halbothi (Trombetas and Maroni, see Lehmann et al., 2014), Pseudancistrus brevispinis (Paru do Oeste, Jari and Guianese rivers, see fig. 17B; Cardoso and Montoya-Burgos, 2009), Stenolicmus ix (Curuá and Maroni, see Wosiacki et al., 2011; G. Dutra, personal commun.), and the clade Hypomasticus despaxi + H. lineomaculatus (Paru, Jari, and Maroni, see Birindelli et al., 2013).
Longitudinal Shield Correspondence among Amazonian Shield Versants
This pattern is characterized by lineages that are present in both shields and follow a longitudinal correspondence among basins (fig. 18A). The pattern is expressed as lineages shared among the westernmost and easternmost parts of the cratonic region. In the western basins (Trombetas and Tapajós) examples include Sartor gr. elongatus (fig. 18B), Bryconexodon spp. (fig. 18C), Laimosemion dibaphus (see Costa, 2006) and Hypoptopoma elongatum (see Aquino and Schaefer, 2010). In the eastern basins, as the Jari, Xingu, and Tocantins, some examples are Acnodon spp., Anablepsoides urophthalmus (see Costa, 2006), Bivibranchia velox (Fig 18D), Hypomasticus multimaculatus (see Birindelli et al., 2016), and Sternarchella sima (Ivanyisky III and Albert, 2014). The first author to recognize this pattern was Jégu (1992a), on the basis of some shared characiform taxa.
Brazilian Shield
This distribution pattern is defined by lineages occurring exclusively in the area corresponding to Amazon-draining Brazilian Shield rivers, formed by the Tocantins, Xingu, Tapajós basins, and some shield tributaries of the Rio Madeira (fig. 19A). Those are all highland rivers draining the ancient crystalline basement of the Brazilian Shield and most of them possess major rapids and/or waterfalls (Innocencio, 1989; Lima and Ribeiro, 2011). This pattern is recovered, in part, in the analyses of Dagosta and de Pinna (2017).
Known examples of this pattern are typically rheophilic species. Probably the ecological conditions prevailing in lowland Amazonian environments act as barriers to their distributions (Géry, 1969). Géry (1962) proposes the circumferential pattern (lateral interbasin migration) for some species, suggesting that this pattern (encircling lowland South American lands, but never entering them) results from the ecological limitations of taxa restricted to fast-flowing rivers with high oxygen levels. This seems to be the explanation for the distributions of many Amazonian taxa restricted to the Brazilian Shield.
Géry (1962) used a dispersalist paradigm to explain lateral movements between basins. The author, however, actually adopted the notion of biotic dispersal (sensu Platnick and Nelson, 1978) rather than a true dispersalist framework. In that sense, his argument was essentially correct, because elements shared between neighboring basins have been associated with rearrangements of the hydrographic network (e.g., stream capture) resulting from neotectonic activity (see Lima and Ribeiro, 2011; Ribeiro et al., 2013). This interpretation of Géry's hypothesis is clear in the following passage, where he proposes that the suppression of a barrier, even if momentarily, would have allowed the spread of a lineage: “Characids show a tendency to invade laterally their adjacent basins (by means of these temporary or permanent connections)” (Géry, 1962: 68). Another instance that demonstrates that the author did not follow pure dispersalism is: “The speciation (or subspeciation) occurred after the passage of the forms from one basin to another in ‘circumferential’ progression, rather than after having propagated along each great river” (Géry, 1962: 78).
Examples of this pattern include: Acestrocephalus nigrofasciatus (Xingu, Juruena, and Jamanxim), Acestrocephalus stigmatus (Tocantins, Xingu, and Tapajós), Ancistrus ranunculus (Tocantins, Xingu, and Tapajós), Baryancistrus longipinnis (Tocantins, Xingu, and Tapajós), Bryconadenos tanaothoros (Xingu, Teles Pires, and Juruena), Caiapobrycon spp. (Tocantins, Xingu, and Tapajós, fig. 19B), Crenicichla acutirostris (Xingu, Tapajós, and shield portions of the Madeira; see Ploeg, 1991), Jupiaba apenima (Tocantins, Xingu, and Tapajós, fig. 19C), Jupiaba iasy (Xingu, Tapajós, and shield portions of the Madeira), Leporinus tristriatus (Tocantins, Xingu, and Tapajós; see Birindelli and Britski, 2013), Moenkhausia gr. pankilopteryx/pirauba (Tocantins, Xingu, Tapajós, and shield tributaries of the Rio Madeira, fig. 19D), Panaque armbrusteri (Tocantins, Xingu, and Tapajós; see Lujan et al., 2010), Petulanos intermedius (Xingu, Tapajós, and shield portions of the Madeira), Thayeria boehlkei (Tocantins, Xingu and, Tapajós; see Lima and Ribeiro, 2011), Rhinopetitia spp. (Tocantins, Xingu, and Tapajós) and Scobinancistrus spp. (Tocantins, Xingu, and Tapajós). Some species occur both in the Amazonian sector of the Brazilian Shield and in the headwaters of the Rio Paraguay, a pattern discussed by Ribeiro et al. (2013) for Jupiaba acanthogaster. Additional examples include: Hyphessobrycon gr. vilmae. Moenkhausia gr. phaeonota. Moenkhausia gr. lopesi, and the genus Utiaritichthys.
Barring the unlikely possibility that all the taxa listed above became extinct in the Guiana Shield, then their age of diversification is maximally ∼12–10 Ma. (Dobson et al., 2001; Figueiredo et al., 2009; Mora et al., 2010), when the Amazon river began depositing sediments on the Brazilian equatorial margin (Sacek, 2014), thus impeding rheophilic lineages from spreading their ranges to Guiana Shield regions.
Extreme Shield: Chapada dos Parecis
The Chapada dos Parecis is an elevated geomorphological formation located in the western portion of the Brazilian Shield, in central South America. It includes headwaters of various drainages, such as Rio Machado, Rio Guaporé, Paraguay, and mostly the Juruena. Many papers have proposed the Chapada dos Parecis as an area of endemism (Carvalho and Bertaco, 2006; Britski and Lima, 2007; Lima et al., 2007; Pastana and Dagosta, 2014; Ohara and Lima, 2015a). Yet, there are other noteworthy characteristics that must be noted for the ichthyofauna in that region. The portion of the Chapada dos Parecis drained (mostly) by the Juruena is the extreme case of the pattern expected for shield composition, with an extremely high level of endemism (Carvalho and Bertaco, 2006; Britski and Lima, 2007). Cases of closely related lineages coexisting are rare, diversity is low and there are very few taxa broadly distributed in the rest of the Amazon.
East of the Rio Juruena in the Chapada dos Parecis, through the basins of the Rio Arinos, Rio Teles Pires, Rio Xingu, and Rio Tocantins-Araguaia, there is a trend toward reduction in endemism and an increase in the number of sympatric congeneric species, in species diversity, and in widely distributed species. The Rio Juruena, like other basins in that formation, contains no members of typically marine lineages (Myers' 1938 peripheral division) (figs. 12A–E) and very few lowland Amazonian components. For example, Arapaima. Osteoglossum. Colossoma, and large pimelodids (Brachyplatystoma, see Barthem et al., 2017; Phractocephalus) are all absent. More relevant still is the fact that dozens of lineages present in other Brazilian Shield drainages are absent in the Rio Juruena at Chapada dos Parecis, such as Acestrocephalus acutus. Acestrorhynchus micropelis (see González, 2015), Anostomoides passionis. Archolaemus luciae (see Vari et al., 2012), Argonectes robertsi. Astyanax multidens (see Marinho and Birindelli, 2013), Bivibranchia notata. Bryconadenos tanaothoros (present only in Rio Arinos basin), Cyphocharax stilbolepis. Electrophorus electricus. Harttia dissidens. Hemigrammus levis. H. ora (see Jerep et al., 2011), Hyphessobrycon loweae + H. pegeouti clade (see Ingenito et al., 2013), H. moniliger. H. vilmae. H. pulchripinnis. Jupiaba apenima. J. anteroides. J. apenima. J. iasy. J. paranatinga. J. polylepis. Laetacara araguaiae. Leporinus britskii. L. julii. L. microphysus. Leptodoras oyakawai. Leptorhamdia schultzi. Macropsobr ycon xinguensis. Megadontognathus kaitukaensis. Moenkhausia celibela. M. collettii. Otocinclus hasemani. Panaque armbrusteri. Petulanos intermedius. Pseudanos spp. (see Birindelli et al., 2012), Pimelodus tetramerus. Rhinopetitia spp., Roeboexodon guyanensis. Serrasalmus rhombeus. Sorubim trigonocephalus. Spectracanthicus murinus. Teleocichla spp. (fig. 16C), and Tocantinsia piresi (fig. 16D).
Among all basins of the Amazonian versant of the Brazilian Shield, the portion of the Rio Juruena draining the Chapada dos Parecis has the most rapids and waterfalls. Britski and Lima (2007) suggest this factor as the reason for the high endemism in the region. We add that the same factor may serve as barriers in the opposite direction and explains also the absence of many lineages common in other Brazilian Shield basins. Thus, the abundance of rapids and waterfalls provides a threefold explanation: for the lack of specific lineages, for reduced sympatry and for decreased species richness. The Rio Iriri and Rio Teles Pires draining the Serra do Cachimbo and the upper Tocantins at the Chapada dos Veadeiros are two additional regions that can be classified as Extreme Shield. Both of them also drain the ancient crystalline basement of the Brazilian Shield and are dotted with rapids and waterfalls. Thus, they show pronounced faunal regionalization and are very poor in diversity when compared to other sectors of the Tapajós, Xingu e Tocantins basins.
Exclusive Faunal Sharing between Neighboring Basins
The sharing of exclusive faunal elements between two basins does not imply that such lineages are broadly distributed in both basins. This fact is evident in the eastern Amazon basins. Geographical distributions tend to be more restricted in highlands (see Albert and Crampton, 2005; Ribeiro, 2006; Maxime and Albert, 2009), where species have smaller ranges and most cases of broader distributions involve species or clades shared with neighboring basins. Such faunal similarities are in most instances associated with river captures caused by reactivation of faults or headward erosion (Ribeiro, 2006; Lima, 2017). Below we list and discuss stereotypical cases of exclusive faunal sharing between neighboring basins in the Amazon:
Tapajós and Paraguay
The fish fauna shared between Tapajós and Paraguay basins has been repeatedly recognized in the literature (see Lima et al., 2007; Carvalho and Albert, 2011a; Ribeiro et al., 2013) and some examples include: Aequidens rondoni (see Lima et al., 2007), Leporinus octomaculatus (fig. 20, see Birindelli and Britski, 2009), and Crenicichla ploegi. Additional examples are: Moenkhausia cosmops (also present in Guaporé basin, fig. 20), Moenkhausia gr. lopesi (also present in Rio Araguaia basin) and the genus Utiaritichthys (also present in shield tributaries of the Madeira).
Tapajós and Xingu
Taxa shared between the Tapajós and Xingu basins are: Anostomoides passionis. Archolaemus janeae (see Vari et al., 2012), Bryconadenos spp. (fig. 20, see Menezes et al., 2009), Cichla mirianae (fig. 20, see Kullander and Ferreira, 2006), Creagrutus cracentis (see Dagosta and Pastana, 2014), Hopliancistrus spp., Hyphessobrycon cachimbensis (fig. 20), Hyphessobrycon cyanotaenia (fig. 20, see Dagosta et al., 2016; also in Guaporé basin), Leptodoras oyakawai (see Birindelli et al., 2008), Lebiasina melanoguttata (fig. 20), Leporinus villasboasorum (see Burns et al., 2017), Leptorhamdia schultzi. Megadontognathus kaitukaensis (see Campos-da-Paz, 1999), Peckoltia feldbergae. Pyrrhulina marilynae (fig. 20, see Netto-Ferreira and Marinho, 2013), Retroculus xinguensis. Spatuloricaria tuira (see Fichberg et al., 2014), and Teleocichla prionogenys.
Tapajós and Madeira
Different subdrainages that compose the Rio Madeira basin variably share exclusive ichthyofaunistic elements with the Rio Tapajós. Most shared elements are between the Aripuanã and the Juruena, such as Ancistrus parecis (see De Oliveira et al., 2016), Hemigrammus silimoni (fig. 20, see Dagosta, 2016), Inpaichthys spp. (fig. 20, see Dagosta, 2016), Moenkhausia levidorsa (fig. 20, see Dagosta, 2016), the clade Crenicichla chicha + C. hemera (see Varella et al., 2012), and genus Utiaritichthys (also present in upper Paraguay, fig. 20). There are also at least four examples of exclusive sharing between the Rio Juruena and the Rio Guaporé basin (Hyphessobrycon psittacus, fig. 20; Hyphessobrycon hexastichos, fig. 20; Moenkhausia rubra, fig. 20; Moenkhausia uirapuru, fig. 20) and three with the Rio Machado: Bryconops piracolina and Hyphessobrycon melanostichos (see Dagosta, 2016), and Corydoras hephaestus. Some taxa have wider distributions in the Tapajós and the Madeira, but are shared exclusively between the two basins (Steindachnerina fasciata, see Netto-Ferreira and Vari, 2011).
Most importantly, virtually all cases listed above involve shield tributaries of the Madeira, and no known case of a species or clade that occurs in the main channel of the Madeira that is also shared exclusively with the Rio Tapajós basin.
Recently, Tencatt and Ohara (2016) proposed a distribution pattern of Amazonian fishes delimited by interfluvial region between the Rio Madeira and the Rio Tapajós. Their arguments on fish species distributed in both systems are the same examples previously listed in Dagosta (2016) as evidence for the historical connections between the Tapajós and Madeira basins. However, Tencatt and Ohara claim the existence of congruence between the distribution of freshwater fishes and terrestrial organisms (birds, butterflies, primates and vascular plants) in the region between the Rio Madeira and the Rio Tapajós. However, freshwater fish distributions are limited by land tracts (save rare exceptions, e.g., Géry, 1964; 1969, for the Rio Amazonas and Goulding et al., 1988, for the Rio Negro). The ichthyofaunal sharing between the Madeira and Tapajós results from a recent and localized history, influenced by geomorphological processes that resulted in stream capture events across the region that separates those basins and which caused biotic dispersal. Stream capture is a phenomenon entirely independent of the geographical isolation of terrestrial animals as inferred by Tencatt and Ohara. The rivers Tapajós and Madeira are the obvious barriers for the distribution of other terrestrial animals (e.g., birds, see Fernandes et al., 2014; Oppenheimer and Silveira, 2009). Species of fish are limited by waterfalls and land tracts. The patterns result from entirely different biogeographical phenomena and we believe there is no spatial or temporal homology between such apparent coincidences.
Xingu and Paraguay
There are few examples of species or clades shared exclusively between these basins: Hypoptopoma inexspectatum (see Aquino and Schaefer, 2010), Steindachnerina brevipinna (see Netto-Ferreira and Vari, 2011), and the clade Characidium nupelia + C. xavante (see da Graça et al., 2008).
Xingu and Tocantins
Examples of this pattern here recognized are: Acnodon normani. Aspidoras poecilus (see Nijssen and Isbrücker, 1976), Astyanax argyrimarginatus. Bivibranchia velox (see Langeani, 1996), Centromochlus simplex. Creagrutus britskii (see Meza-Vargas, 2015), Creagrutus mucipu (see Meza-Vargas, 2015), Hemiancistrus spilomma. Hemiodus tocantinensis. Hyphessobrycon loweae (see Ingenito et al., 2013), Hypostomus faveolus (see Zawadzki et al., 2008), Laemolyta fernandezi. Melanocharacidium auroradiatum. Mesonauta acora (see Kullander and Silfvergrip, 1991), Moenkhausia loweae (see Marinho, 2009), Moenkhausia pyrophthalma. Rhynchodoras xingui (see Birindelli et al., 2007), Semaprochilodus brama (see Castro and Vari, 2004), Sternopygus xingu, and Tometes ancylorhynchus (see Andrade et al., 2016).
Tocantins and Paraguay
Only Cyphocharax vanderi (see Claro-García and Shibatta, 2013), Hasemania hanseni, and Knodus chapadae (see Ferreira, 2007) are exclusively shared between these basins.
Tocantins and São Francisco
Although examples of taxa shared exclusively between the Tocantins and São Francisco are few, they have received considerable attention in the literature (see Lima and Caires, 2011; Dagosta et al., 2014). The cases recognized here are: Cichlasoma sanctifranciscense (see Lima and Caires, 2011; Dagosta et al., 2014), Hyphessobrycon diastatos (see Dagosta et al., 2014), and some lineages of Cynolebias and Hypsolebias (see Costa, 2010). The species Astyanax novae was previously considered as one more example of this pattern (Garutti and Venere, 2009; Lima and Caires, 2011; Dagosta et al., 2014), but in reality has a wider distribution (see Freitas et al., 2015).
Tocantins and Upper Paraná
Some confirmed examples of lineages shared between those regions are: Characidium xanthopterum (see Silveira et al., 2008), Corumbataia spp. (see Britski, 1997; Carvalho, 2008), Hasemania crenuchoides (see Serra and Langeani, 2015), and Rhinolekos spp. (see Martins and Langeani, 2011; Roxo et al., 2015). Additional species are shared exclusively between the Tocantins and upper Paraná plus the São Francisco: Brycon nattereri (see Lima, 2017), Moenkhausia aurantia, Hyphessobrycon coelestinus (see Aquino and Carvalho, 2014), and Cetopsorhamdia iheringi.
Madeira and Paraguay
As discussed above in South American Lowlands, there are many events of biotic dispersal between the Amazon and the upper Paraguay, with all cases involving only part of the Rio Madeira basin. It is therefore not surprising that several taxa are shared between the Madeira and the Paraguay and that such congruent distributions are for the most part temporally decoupled, i.e., pseudocongruences (sensu Donoghue and Moore, 2003). Different subbasins of the Rio Madeira drainage share taxa exclusively with the Paraná-Paraguay, with most of such cases being from the Rio Guaporé and the Rio Mamoré. Many studies have discussed a common biogeographical history between those two regions (see Pearson, 1937; Hubert and Renno, 2006; Carvalho and Albert, 2011b; Ota et al., 2014). Some examples are: Aequidens plagiozonatus, Aphyocharax anisitsi (see Souza-Lima, 2003), Apistogramma trifasciata (see Kullander, 2003), Astyanacinus moorii, Astyanax lineatus, Cetopsis starnesi (see Vari et al., 2005), Corydoras polystictus, Gymnogeophagus balzanii (see Reis and Malabarba, 1988), Hemigrammus machadoi (see Ota et al., 2014), H. mahnerti (see Ota, 2010), H. tridens, Hyphessobrycon elachys, H. megalopterus (see Lima and Malabarba, 2003), Imparfinis guttatus (see Queiroz et al., 2013), Laetacara dorsigera (see Linke and Staeck, 1994), Markiana nigripinnis, Megalonema platanum (see Queiroz et al., 2013), Odontostilbe paraguayensis (see Bührnheim, 2006), Oligosarcus pintoi (see Ribeiro and Menezes, 2015), Parodon carrikeri (see Schaefer, 2011), Piabucus melanostomus (see Britski et al., 1999; Queiroz et al., 2013), Pimelodella mucosa (see Queiroz et al., 2013), Psectrogaster curviventris (see Vari, 1989b), Rineloricaria aurata (see Vera-Alcaraz et al., 2012), Scoloplax empousa (see Schaefer et al., 1989), and Trachydoras paraguayensis (see Sabaj and Arce, 2017).
Branco and Essequibo
Those two basins have a common geomorphological history resulting from a series of capture events of the proto-Berbice by the Rio Branco drainage during the Pleistocene (Crawford et al., 1985; Gibbs and Barron, 1993; Souza et al., 2012). Such events may account for the conspicuous elements shared between the two basins (see Lujan and Armbruster, 2011; Souza et al., 2012). Some examples include: Apistogramma rupununi (see Kullander, 2003), Astyanax rupununi (see Souza et al., 2012), Cetopsidium roae (see Souza et al., 2012), Denticetopsis iwokrama (see Souza et al., 2012), Guianacara dacrya (Arbour and López-Fernández, 2011), Hypostomus macushi (see Armbruster and Souza, 2005), Parodon bifasciatus (see Souza et al., 2012), Pseudancistrus nigrescens (see Souza et al., 2012), Rhinodoras armbrusteri (Sabaj Pérez et al., 2008), and Sturisoma monopelte (see Souza et al., 2012).
Negro and Orinoco
A number of contributions have explored the common biogeographical history of these two basins (see Winemiller et al., 2008; Willis et al., 2010; Winemiller and Willis, 2011). Examples of fish species shared exclusively between the Negro and the Orinoco are numerous and include: Acestridium dichromum (see Retzer et al., 1999), Acestridium martini (see Retzer et al., 1999), Creagrutus phasma (see Vari and Harold, 2001), Creagrutus runa, C. vexillapinnus and C. zephyrus (see Vari and Harold, 2001), Geophagus abalios and G. dicrozoster (see López-Fernández and Taphorn, 2004), Hemiancistrus subviridis (see Wernecke et al., 2005), Hemigrammus barrigonae. Hemigrammus bleheri (see Géry and Mahnert, 1986), Heterocharax leptogrammus (see Toledo-Piza, 2000b), Hoplarchus psittacus. Hyphessobrycon epicharis (see Weitzman and Palmer, 1997), Hypostomus sculpodon (see Armbruster, 2003), Laetacara fulvipinnis (see Staeck and Schindler, 2007), Leporinus enyae (see Burns et al., 2017), Microcharacidium gnomus (see Buckup, 1993), Neblinichthys pilosus (see Ferraris et al., 1986), Odontostilbe pulchra (see Bührnheim and Malabarba, 2007), Phenacogaster prolatus (see Lucena and Malabarba, 2010), Prochilodus mariae (see Castro and Vari, 2004), Pseudancistrus pectegenitor (see Lujan et al., 2007), Pseudancistrus sidereus (see Armbruster, 2004), Pseudanos varii (see Birindelli et al., 2012), Pseudolithoxus nicoi (see Lujan and Birindelli, 2011), Pterophyllum altum (see Schultz, 1967), Ptychocharax rhyacophila (see Weitzman et al., 1994), Racenisia fimbriipinna (see Mago-Leccia, 1994), Rhinobrycon negrensis (see Lasso et al., 2004), Serrabrycon magoi (see Lasso et al., 2004), and Tometes makue (see Andrade, 2013).
The Casiquiare Canal is a portion of the Rio Orinoco that was redirected to flow part of the year to the Rio Negro basin (Albert and Carvalho, 2011) and that now connects the two drainages by a permanent waterway with minimal gradient. Such a connection was mentioned by Eigenmann (1909) in his description of his Amazon Province in a dispersalist context and later proposed by Vari (1988) as the factor responsible for some curimatid species shared between the Amazon and Orinoco. Albert et al. (2006) and Winemiller et al. (2008) questioned the relevance of the physical Casiquiare connection as a species-dispersion route, because there are rapids on both sides of the divide (e.g., in Porto Ayacucho and in São Gabriel da Cachoeira) and possible chemical barriers (pH, temperature, and conductivity). It is possible that part of the fish fauna shared between the Orinoco and Negro is in fact derived from the proto-Amazon-Orinoco and predates their hydrological separation.
Negro and Branco
The Rio Negro basin provides a clear example that hydrographic limits do not necessarily imply historical connections. Although the Rio Branco is the largest tributary of the Negro and the two are not separated by any physical barriers, each of them shares more species with adjoining non-Amazonian basins than with each other (with the Orinoco in the case of the Negro and with the Essequibo in the case of the Branco). In addition to the different geomorphological history of each basin, markedly different physicochemical parameters may also in part explain the small number of taxa exclusively shared between them. As pointed out by Ferreira et al. (2006), the Rio Negro predominantly drains lowland soils poor in cations with exceptionally low rates of mineral erosion, while the Rio Branco drains highland soils of an older landscape, richer in cations derived from the erosion of relatively stable igneous rocky beds. There are very few examples of species reliably restricted to the Negro and Branco: Physopyxis cristata (see Sousa and Py-Daniel, 2005), Apistogramma gibbiceps (see Kullander, 1980), and Crenicichla virgatula (see Ito, 2013).
Negro to Trombetas
This pattern refers to the fish fauna common to the left-bank Amazonian tributaries east of the Rio Negro: Urubu, Uatumã, Nhamundá, and Trombetas. Together, they share some taxa exclusively with the Rio Negro or with the Negro-Orinoco: Acestridium discus (Negro, Branco, and Trombetas), Ageneiosus polystictus (Negro, Urubu, and Trombetas; see Ribeiro et al., 2017), Asterophysus batrachus (Orinoco-Negro and Uatumã), Anduzedoras oxyrhynchus (Orinoco-Negro, Branco, Urubu, and Trombetas), Auchenipterichthys punctatus (Negro, Branco, and Urubu), Nemuroglanis pauciradiatus (Negro, Branco, Urubu, and Trombetas), Pygidianops amphioxus (Negro and Nhamundá; see de Pinna and Kirovsky, 2011), and Poecilocharax weitzmani (Orinoco-Negro, Branco, Urubu).
Poorly Sampled Neighboring Basins in the Guiana Shield
Some basins draining the Guiana Shield into the Amazon, such as the Urubu, Uatumã, Trombetas, and Paru, are relatively poorly known as to their ichthyofaunal composition, with comparatively few reported species, rare cases of endemism and few species shared among each other. Such precarious knowledge precludes a clear understanding of the connections of the fish faunas in those basins and few relevant examples deserve note: the Uatumã and Trombetas exclusively share Cetopsidium ferreirai (see Vari et al., 2005) and Cichla vazzoleri (see Kullander and Ferreira, 2006), while the Paru and Jari exclusively have Hypomasticus lineomaculatus (see Birindelli et al., 2013).
Cis-Andean Foothills
Another pattern related with the circumferential pattern of Géry (1962) is the cis-Andean Foothills distribution. The name refers to the highland region surrounding the cis-Andean lowlands, mostly around the Western Amazon (fig. 21A). As in the Brazilian Shield pattern, the present one comprises rheophilic species, restricted to fast-flowing, highly oxygenated waters. This pattern was first identified by Vari (1988: 360): “Other species ranges appear to be associated with the more swiftly flowing piedmont streams of the western margins of the Amazon basin, and those species extend north into the western and northern margins of the Río Orinoco system.” Shortly thereafter, a similar pattern was described by Ibarra and Stewart (1989) for the Rio Napo, where the altitudinal gradient decisively influenced species composition (see Lujan et al., 2013, for a more complex scenario). The pattern described here differs from the shield patterns in being not only wider, but also associated with rivers draining the eastern versant of the Andean range, and sometimes the western versant as well. Because the examples known are absent in the Amazonian versants of the Brazilian and Guiana shields, this pattern seems to be strictly associated with the history of the foreland basin and with the Andean uplift.
Some of the known examples include species with both narrow and wide distributions. Examples in the former category include Acrobrycon ipanquianus (see Arcila et al., 2013), Attonitus (see Vari and Ortega, 2000), Creagrutus flavescens. C. gephyrus. C. kunturus, and C. muelleri (see Vari and Harold, 2001). Cases of wide distributions in the Andean Foothills pattern comprise the family Astroblepidae (see Schaefer and Arroyave, 2010), Astyanacinus spp. (see fig. 21B; Dagosta, 2011), Ernstichthys spp. (see Stewart, 1985), Rhyacoglanis (see Shibatta and Vari, 2017), Xyliphius spp. (see Carvalho et al., 2017), Leporinus striatus (see fig. 21C; Birindelli and Britski, 2013), Steindachnerina dobula (fig. 21D) and S. guentheri (see Vari, 1991), a clade composed of Creagrutus muelleri. C. ouranonastes, and C. peruanus (see Vari and Harold, 2001), and putative sister relationship between Brycon hilarii and B. whitei (see Lima, 2017). Another notable example is the entire genus Hemibrycon (excepting H. surinamensis, sole species in the genus with an Amazon-core Highlands pattern; see Bertaco and Malabarba, 2010).
Lima and Ribeiro (2011) discuss a pattern similar to the one described here, in which lineages are restricted to upper portions of the foreland basin due to ecological requirements. As done here, those authors also distinguish this highland of the Foreland Basin pattern from that of the Brazilian Shield highland pattern. Wilkinson et al. (2010), in a discussion of the action of the megafans, also propose a pattern similar to the one advanced here, although not distinguishing shield highlands from the foreland-basin highlands.
Central Blackwater Amazon
The name of this pattern refers to the most common (although by no means exclusive) water type of the rivers within its limits. Its position is approximately at the central portion of the Amazon (fig. 22A), although its western limits are not precisely defined. The distribution of most examples extends to the mouth of the Rio Negro, with some going farther, to the lower Japurá, lake Tefé, or into Peru. To the east, the pattern is almost always delimited by the mouth of the Rio Tapajós. Northward, most examples are restricted to the Negro/Orinoco, with some lineages found also in the Essequibo. To the south, species extend to the middle portion of the Tapajós, but may be more broadly distributed in the Rio Madeira, to tributaries of Mamoré/Guaporé.
The first author to propose this pattern of distribution was Kullander (1986), in discussing congruent areas between species of cichlids and characids of the genus Paracheirodon (see Kullander, 1986: figs. 5, 6). Independently, Vari (1988: fig. 7) inferred that a then-undescribed species of Curimata had a pattern of distribution indicative of a preference for acidic waters, not exclusively in the Rio Negro basin, but also in other Central Amazonian localities.
This biogeographical pattern also has surfaced occasionally in the literature, where it has been indicative of possible taxonomic problems. The first paper to notice something noteworthy in such distributions was Vari and Harold (2001), in the redescription of Creagrutus maxillaris. That species is broadly distributed in the Orinoco and the upper Rio Negro. The authors then had only a single lot with few poorly preserved specimens from the Rio Madeira (AMNH 39855) and stated that the presence of C. maxillaris in that basin required confirmation by additional material (later reported by Queiroz et al., 2013).
A similar situation happened with Chalceus macrolepidotus in Zanata and Toledo-Piza (2004), whose sole sample from the Madeira basin was considered questionable because of its location widely disjunct from that of other known lots of the species. Likewise, Kullander and Ferreira (2006: 377) disregarded two samples of Cichla temensis from the Rio Madeira, not including them in the map or material examined of the species because “there is nearly no other Cichla material available from the Brazilian portion of the Madeira drainage to permit an understanding of the distribution of C. temensis in this region.” Clearly, in all examples the odd disjunct nature of such distributions influenced the respective authors' hesitation about their own results. Our recognized pattern, however, shows that such distributions joining the Negro and Madeira are not at all abnormal. This pattern of distribution is recovered, in part, in the analyses of Dagosta and de Pinna (2017).
Examples of lineages with a Central Blackwater Amazon pattern include: Aequidens mauesanus (Madeira and Tapajós, see Kullander, 2003), Acestridium spp. (Orinoco/Negro, Madeira, and Tapajós), Astyanax ajuricaba (Negro, Tapajós, see Marinho and Lima, 2009), Boulengerella lucius (Orinoco/Negro, Tapajós, and Trombetas, see Vari, 1995), Brachyhypopomus hendersoni (Tefé, Negro, and Essequibo, see Crampton et al., 2016), Bryconops inpai (Orinoco/Negro, Tapajós, Madeira, and Trombetas), Chalceus spilogyros (Madeira, Tapajós, and Trombetas, see Zanata and Toledo-Piza, 2004), C. macrolepidotus (Orinoco/Negro/Essequibo and Madeira, see Zanata and Toledo-Piza, 2004), Charax condei (Negro and Tapajós, see Menezes and Lucena, 2014), Cichla temensis (Orinoco/Negro and Madeira, see Kullander and Ferreira, 2006), clade Creagrutus maxillaris + C. cracentis (Orinoco/Negro, Madeira, and Tapajós, see Vari and Harold, 2001), Copella nattereri (Orinoco/Negro, Tapajós, Madeira, Trombetas, and some records in the Amazon above the mouth of Rio Negro, see Marinho and Menezes, 2017), Cynodon septenarius (Essequibo/Orinoco/Negro, Uatumã, Tapajós, Trombetas, and Tefé), Curimata incompta (Orinoco/Negro, Madeira, see Vari, 1988), Cyphocharax abramoides (Negro, Tapajós, and Trombetas, see Vari, 1992b), C. nigripinnis (Negro, Tapajós, and Amazonas, see Vari, 1992b), Elachocharax junki (Negro and Madeira, see Weitzman and Géry, 1981), Biotoecus spp. (Orinoco/Negro, Uatumã, and Trombetas, fig. 22B; see Kullander, 1989), Dicrossus spp. (Orinoco/ Negro, Madeira, Tapajós, and Trombetas, fig. 22C; see Kullander, 2011), Gnathocharax (Orinoco/Negro/Essequibo, Madeira, Tapajós, and Trombetas), Hemigrammus analis (Purus, Jutaí, Negro, Madeira, Tapajós, and Trombetas, fig. 22D), H. coeruleus (Orinoco/Negro/Essequibo, Madeira, Tapajós, and Trombetas, fig. 22D), H. hyanuary (Orinoco/Negro, Madeira, and Tapajós), H. stictus (Orinoco/Negro, Madeira, and Tapajós, fig. 22D), H. vorderwinkleri (Orinoco/ Negro, Madeira, Tapajós, and Trombetas), Heterocharax virgulatus (Orinoco/Negro, Madeira, and Tapajós, see Toledo-Piza, 2000b), Hoplocharax goethei (Orinoco/Negro, Madeira, Tapajós, and Trombetas), Hyphessobrycon sweglesi (lower Purus, Negro, and Madeira), Iguanodectes geisleri (Orinoco/Negro and Madeira), Jupiaba gr. atypindi (Negro and Madeira), Leporinus altipinnis (Orinoco/Negro, Madeira, and Tapajós, see Britski and Birindelli, 2016), L. aripuanaensis (Branco, Madeira, and Trombetas), Leporinus gomesi (Madeira and Negro), L. klausewitzi (Negro and Madeira), Metynnis hypsauchen (Orinoco/Negro/Essequibo, Madeira, Tapajós, and Trombetas, see Ota, 2015), M. melanogrammus (Orinoco/Negro, Uatumã, Trombetas, Tapajós, and Sucunduri (Madeira), see Ota et al., 2016), Moenkhausia hemigrammoides (Maroni, Suriname, Corentyne, Negro, Madeira, Tapajós, and Trombetas), M. lata (Orinoco/Negro, Madeira, and Tapajós, M. Marinho personal commun.), Nannostomus marilynae (Orinoco/ Negro and Madeira), Oxyropsis acutirostra (Orinoco/Negro and Tapajós), Poecilocharax spp. (Orinoco/Negro and Madeira), Pygidianops spp. (Orinoco/Negro and Madeira), Rhinobrycon negrensis (Orinoco/Negro and Madeira), Satanoperca lilith (Negro, Uatumã, Trombetas, and Madeira, see Ota, 2013), Steindachnerina planiventris (Negro, Japurá, and Madeira, see Vari, 1991), Symphysodon discus (Negro, Madeira, and Trombetas, see Bleher et al., 2007; Farias and Hrbek, 2008; Amado et al., 2011), Taeniacara candidi (Negro, Tapajós, and Trombetas) and Hypoptopomatinae new genus (Negro, Madeira, and Tapajós, see Delapieve, 2014). Other potential examples are Moenkhausia diktyota (Madeira) and Hemigrammus pretoensis (Amazonas and Negro), which despite their current separate generic assignments are actually close relatives, perhaps even synonyms (F.C.P.D., personal obs.).
The sharing of so many lineages clearly indicates strong historical connections among the Orinoco/Negro, Madeira and Tapajós. More importantly, all lineages with this distribution pattern are absent in the Brazilian Shield (except for some rare cases in the middle to lower Tapajós, at the periphery of the Shield, see fig. 22C). Despite such strong signal, no independent geomorphological history was identified that could explain this pattern. While Cretaceous deposits from those regions are well known, the Cenozoic sedimentary history is still very poorly known (Soares, 2007). The lack of such critical data does not allow a more precise evaluation of the biogeographically relevant processes and events in the region. It is clear that the lower sectors of those rivers (Negro, Purus, and Madeira), and even the portion of the Rio Amazonas in that region, underwent course changes during the Pleistocene as demonstrated by paleocanals (Latrubesse and Franzinelli, 2002; Almeida-Filho and Miranda, 2007; Irion et al., 2010; Teixeira and Soares, 2011; Hayakawa and Rossetti, 2015; but see Albert et al., 2018, for reservations about the accuracy of optically stimulated luminescence, or OSL, method for dating sediments of this type and age). However, details about the dynamics and timing of those events are unavailable at present.
One exception is the work of Ruokolainen et al. (2018). The authors present evidence of river captures and avulsions during the late Pleistocene–Holocene in central Amazon, involving rivers Negro, Madeira, Purus and Juruá. Ruokolainen et al. demonstrate that the river network in the region has been anything but stable. According to them, during the past 50,000 years there have been many cases of river avulsions, with consequent changes in the historical connections among major tributary rivers of the central Amazon. The latest major river capture event converted the Japurá from a tributary of the Rio Negro to a tributary of the Amazon, only 1000 years ago. Such broad-scale lability implies that lowland rivers cannot have been efficient biogeographical dispersal barriers to terrestrial biota, and even less so for fishes. In such a scenario, river captures and avulsions in that region may have contributed, at least in part, to the origin of the pattern of distribution discussed here.
As another relevant point, the Central Blackwater Amazon pattern follows remarkably closely the range of blackwater Amazonian rivers recently compiled by Venticinque et al. (2016) (see fig. 20A–D). Those authors demonstrate that there are numerous blackwater rivers scattered throughout the central Amazon, confirming Fink and Fink (1979: 18): “the Rio Negro is the major ‘black’ water river in Amazonia; however, similar conditions have a spotty distribution through much of central Amazonia and many igarapés and rios of the terra firma consist of this type of water.” Such a network provides ample opportunity for species restricted to blackwater to inhabit regions of the upper Amazon, approximately up to the mouth of the Rio Marañon in the Ucayali. A hypothesis that blackwater is the determining factor in the pattern herein described must be tested against a refinement of the species' locality data. The small-scale mosaic physical distribution of blackwater tributaries in that region makes it very difficult to extract such information from usual museum data. For example, the Rio Madeira, although widely recognized as a whitewater river, is abundantly irrigated by tributaries of all water types (fig. 20). Therefore, the provenance of a sample from the Madeira says little about water type preferences unless associated with very precise locality information. Despite such limitations, it is remarkable that many of the species in the Central Blackwater Amazon pattern that occur in the Rio Madeira or Tapajós are restricted to the lower sectors of those basins, exactly where their blackwater tributaries are most abundant. Again, we highlight the fact that water type is not a random variable, but instead closely related to the geological history of the terrain it drains. Therefore, a distribution pattern determined by water type is also indirectly associated with a historical component and cannot be taken at face value as a purely ecological determinant.
Allopatric Branco-Tocantins
There are few examples of Amazonian fish species with disjunct distributions. Five unrelated species display an intriguing pattern of congruent disjunct distributions: Creagrutus menezesi (see Vari and Harold, 2001), Exodon paradoxus, Leporacanthicus galaxias, Leporinus desmotes (see Burns et al., 2017), and Leptorhamdia essequibensis (see Bockmann, 2003). Those species are found in the Branco and Tocantins basins, with some also having records in the Essequibo and Orinoco. All five species are well known in their taxonomy and geographical distribution, thus reducing the possibility of sampling gaps.
The Branco and Tocantins basins are widely separated, making such allopatric disjunctions all the more noteworthy, but no geomorphological evidence has been associated with such pattern (the explanations in Eigenmann's Eastern Highlands [cf. also Albert et al., 2011: 50–52] do not account for the specific pattern discussed here, because in the present case the lineages involved are not present in the rest of the shield, i.e., the Tapajós, Xingu, Madeira, Trombetas, Jari, etc.). The geomorphological history of the Rio Branco is related to the proto-Berbice (Lujan and Armbruster, 2011), where the former had courses preferentially flowing from southwest to northeast toward the Caribbean Sea. Erosion of the rocky basement of the Guiana Shield caused the reorganization of the proto-Berbice drainage network and the southward reversal of its main course, making the Rio Branco a tributary of the Rio Negro (Schaefer and Dalrymple, 1996). The geomorphological history of the Rio Tocantins, in turn, is mostly associated with the geological evolution of the Brazilian Shield and with other large rivers such as the Tapajós, Xingu, Paraná-Paraguay, and São Francisco (Lima and Caires, 2011; Lima and Ribeiro, 2011). Of course, before 10 Ma there were no large whitewater rivers or floodplains separating clear-water tributaries of the Guiana and Brazilian shields, which might seem like a possible explanation. However, we again emphasize that the lineages constituting this pattern are not widely present in shield drainages, thus invalidating this broader paleoscenario as a causal factor. Of course, this scenario would hold in case the present disjunct pattern is a relict of a broader pattern that comprised other shield rivers, a hypothesis for which there is no evidence.
The savannahs of the Rio Branco and Essequibo are biogeographically distinct from those of central Brazil, even though they share some fish species (Ferreira et al., 2006). As noted by López-Fernández and Albert (2011), the importance of savannas for the evolution of the modern fish fauna of the Neotropics cannot be overemphasized. In the absence of any geological evidence that might explain the exclusive sharing of species between the Branco and Tocantins, the presence of savannah systems may offer clues for a possible ecological explanation.
Absence Patterns
Among the most curious distributional phenomena in the Amazon is the absence of some fish taxa in regions where they were expected to occur on the basis of the distribution of their close relatives and higher groups. Those absences are often associated with some clearly identifiable barriers of physical (e.g., waterfalls) or physico-ecological (e.g., water type) nature.
The most conspicuous Absence pattern is seen in the upper Juruena river, a pattern described in detail in Chapada dos Parecis: Extreme Shield. Another remarkable example is the upper Rio Tocantins. Upriver from the region of Imperatriz (in the Brazilian state of Maranhão) and Itaguatins (in the Brazilian state of Tocantins), the channel of the Rio Tocantins has rapids in sectors that may help explain the absence of various groups otherwise distributed in the entire Amazon that are present in the Araguaia or in lower Tocantins basins. Some examples are: Acestrorhynchus falcirostris (see González, 2015), Apistogramma spp., Chaetobranchus spp., Hydrolycus tatauaia (see Toledo-Piza et al., 1999), Hypophthalmus marginatus. Hypselecara spp., Mastiglanis asopos. Megalechis thoracata (see Reis, 1997), Moenkhausia cotinho. Mylossoma spp. (see Mateussi, 2015), Ochmacanthus spp. (see Neto, 2014), Pellona spp. (see Melo, 2001), Potamorrhaphis spp. (see Collette, 1982), and Semaprochilodus brama (see Castro and Vari, 2004). Other examples of biogeographically isolated Amazonian regions are the mid- and upper Rio Madeira, separated by the rapids in the region of Porto Velho, which block the upriver distribution of, for example, Arapaima.
The absence of certain lineages is also influenced by other factors such as tidal effects. Goulding et al. (2003) showed that downstream from the region of Óbidos (in the Brazilian state of Amazonas), the tidal regime starts to influence the circadian rhythm of the Amazon, probably affecting the distribution of fish lineages (Jégu and Keith, 1999; Lima and Ribeiro, 2011), especially those with feeding and breeding periods narrowly associated with drought-flood cycles. Some examples of fishes that do not occur in the lower Amazon are: Acestrorhynchus granducolis (see González, 2015), Brycon amazonicus (see Lima, 2017), Colossoma macropomum (see Araujo-Lima and Goulding, 1997; Lima and Ribeiro, 2011), Copella nattereri (see Marinho and Menezes, 2017), Piaractus brachypomus (see Jégu and Keith, 1999), Potamorhina altamazonica (see Vari, 1984), and Serrasalmus elongatus (see Jégu and Keith, 1999).
As seen above, water type has a major influence on biogeochemical processes and on the distribution and dynamics of aquatic habitats and associated biota (Venticique et al., 2016). Expectedly, it is an important factor in the geographical distribution of Amazonian fish lineages. As widely reported in the literature (see Sioli, 1984; Goulding et al., 2003), Amazonian rivers display enormous differences in pH and concentration of dissolved solutes, according to the type of soil they drain. Wallace (1889) was the first to note that water type influenced the composition of fish assemblages in the Amazon (Dagosta and de Pinna, 2018), an observation repeatedly confirmed in subsequent studies (see Roberts, 1972; Kullander, 1986; Goulding et al., 1988; Vari, 1988; Araujo-Lima and Goulding, 1997; Saint-Paul et al., 2000; Lima and Ribeiro, 2011). The extremely acidic water of the Rio Negro, in particular, may be a deterrent to many fish lineages. Some examples of fishes absent in the Negro, yet present in neighboring basins and widely distributed in the Amazon include: the subfamily Stethaprioninae (see Dagosta and Pinna, 2017; Reis, 1989), the genera Galeocharax (see Giovannetti et al., 2017) and Hypoptopoma (see Aquino and Schaefer, 2010), and several species such as Anostomus ternetzi (see Lima and Ribeiro, 2011), Brachyplatystoma juruense, Cheirocerus goeldii (see Stewart and Pavlik, 1985), Copella stigmasemion (see Marinho and Menezes, 2017), Curimatella dorsalis (see Vari, 1992a), Hemiodus microlepis (see Langeani, 1996), Hemisorubim platyrhynchos. Jupiaba polylepis. Limatulichthys griseus (see Ohara, 2010), Megalodoras uranoscopus. Oxydoras niger. Pimelodus blochii. Prochilodus nigricans (see Castro and Vari, 2004), Pygocentrus nattereri. Semaprochilodus insignis (see Castro and Vari, 2004), and Tympanopleura atronasus (see Walsh et al., 2015).
It is possible to go beyond the mere identification of absence biogeographical patterns. Our earlier biogeographic analyses have demonstrated that some absences are the result of extinctions rather than primitive absences (Dagosta and de Pinna, 2017); moreover, we found that the absence of several lineages in the Rio Negro are autapomorphic for the basin, i.e., their ancestral areas (historically related) have the respective taxa. Therefore, their absence in present-day Rio Negro may be the result of extinctions (discarding cases of pseudo-absences). Recently, Ruokolainen et al. (2018) provided convincing evidence that the Rio Japurá was a tributary to the lower Rio Negro and that a river capture event diverted it to flow into the Amazon (Solimões). The connection between the Rio Japurá and the Rio Negro may have been broken as recently as 1000 years ago. According to these authors, until that time the lower Rio Negro was not a blackwater river, as it presently is, and it carried a much larger load of sediments. Such evidence further corroborates the hypothesis of Dagosta and de Pinna (2017) that the lower Rio Negro basin was not always as hostile to some otherwise ubiquitous Amazonian lineages as it is today and may have had a less extreme type of water earlier in its history. At least for the lower part of its course, the Rio Negro did not have waters as acidic and nutrient poor as today, and did not impede the existence of some lineages that are now absent in the basin.
Endemism
Although a majority of 2716 species of Amazonian fishes examined here occur in more than one subdrainage, there are numerous examples of basin-specific endemics. With the regions defined by Dagosta and de Pinna (2017) as background, at least 831 Amazonian fish species are found in a single drainage or subregion thereof (fig. 4, appendix 1). An additional 196 species are also basin specific, given a wider definition of basin (e.g., species restricted to the Tapajós but occurring in more than one sub-basin therein). The latter data are also included in appendix 1, with indications of their respective basins of endemism.
CONCLUSIONS
Distribution patterns decay over time as new ones are superimposed (Grande, 1985; Hunn and Upchurch, 2001; Upchurch and Hunn, 2002; Upchurch et al., 2002; Morrone, 2009), making the disentanglement of their history a complex procedure. More studies on Amazonian fishes are necessary, both on phylogeny, paleontology, phylogeography, and molecular dating in order to empirically test the temporal congruences of the distributional patterns described here. New data on geological history are needed to better understand the effect of riverine configurations in the biogeography of fishes in the basin. Our findings support the conclusion that the biogeographical history of a river is associated less with its size than with its stability through geological time. The mosaic of patterns shown herein demonstrates that the river network in the Amazon has been anything but stable, and that this instability has been a major factor in fish distributions. Different overlapping geomorphological processes, at different times, have left diffuse marks on the composition and distribution of the fish fauna and this process continues to the present. The recent work by Stokes et al. (2018) directly demonstrates the intense dynamism of the region, showing that the Amazon river is capturing headwaters of the Río Orinoco, another step in the continuing reorganization of South American river systems.
Freshwater fishes are physically restricted to hydrographical basins, but in the Amazon basin their distributions often transcend modern hydrographical limits. This is a result of a complex and reticulated history of drainages, a view that has been corroborated by several authors (see Lima and Ribeiro, 2011; Dagosta, 2016; Dagosta and Pinna, 2017). This fact in itself does not disqualify basins as historical agents. Rather, rivers are historically bound areas, even though they are far more complex than hydrographically limited units. Data presented in this paper demonstrate that each hydrographic drainage in the Amazon basin participates simultaneously in various biogeographical patterns and that no single basin is a historically cohesive unit. Likewise, the entire Amazon basin itself does not form a single historical unit. All such conclusions corroborate the hypothesis that hydrographical basins should not be considered a priori as historical units. They are demonstrably reticulate areas that received portions of their biotas at different ages, under the influence of disjunct events. Thus, past geomorphological processes are more informative for understanding the distribution of the Amazonian fishes than present-day basin divides.
ACKNOWLEDGMENTS
This paper is dedicated to the memory of Richard P. Vari for his epochal contributions to the systematics and biogeography of Neotropical fishes and his generous devotion to the formation of new generations of ichthyologists, of which the authors of this paper are deeply grateful beneficiaries. We thank Cristiano Moreira and Marcelo Britto (MNRJ), Lúcia Py-Daniel and Renildo Oliveira (INPA); Cláudio Oliveira and Ricardo Benine (LBP); André Esguícero and Flávio Bockmann (LIRP); Izaura Maschio and Wolmar Wosiacki (MPEG); Jeff Clayton, Jeff Williams, Kris Murphy, Lynne Parenti, Richard Vari and Sandra Raredon (NMNH) for their help and hospitality during visits to the collections under their care. We are grateful to Eduardo Venticique for providing data on water types of Amazonian rivers for figures 7B–D and 20A–D. Gilberto N. Salvador and Eduardo Baena helped in the editing of the maps. José Birindelli provided photos of Hyphessobrycon psittacus. Leporinus octomaculatus, and Utiaritichthys sennaebragai. Authors were funded by FAPESP (FCPD, 2011/23419-1; 2016/07246-3; MP, 2015/26804-4), CNPq (MP, 201088/2014-2) and CAPES (MP, BEX 5840/2014-07). Part of this work was prepared during a sabbatical stay of the second author in the Muséum national d'Histoire naturelle and Université Pierre and Marie Curie (Institut de Systematique, Evolution et Biodiversité), Paris, and the support and hospitality of René Zaragueta-Bagils and Guillaume Lecointre is gratefully acknowledged. The authors thank Universidade Federal da Grande Dourados for funding.
REFERENCES
Appendices
APPENDIX 1
Taxonomic List of Amazonian Fish Species
Definition of Amazonian regions follows Dagosta and de Pinna (2017). List updated by the end of 2018. Symbols: *species exclusive to the Amazon basin; ** species poorly known or with vague distribution records.
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
Continued
APPENDIX 2
Statistical Test for Species with Distribution Maps
H0: The species is randomly distributed. H1: The species is not randomly distributed. X2 = Σ [ (O - E)2 / E ], where: E is the expected frequency (species can occur inside or outside of the defined limits of the distribution pattern. Thus, E is the number of records of a species divided by 2 possibilities [inside or outside]); andO is the Observed frequency (observed number of records of the species inside the defined limits of the distribution pattern). Degrees of freedom (2 - 1 =1). Conventionally accepted significance level of 0.05 (Chi square distribution table 3.841).
Test for species with “Broadly distributed lineages”
A, Hoplias malabaricus, 249 records
Expected inside: 124.5; observed inside: 249; expected outside: 124.5; observed outside: 0; X2 = 249. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
B, Erythrinus erythrinus, 41 records
Expected inside: 20.5; observed inside: 41; expected outside: 20.5; observed outside: 0; X2 = 41. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
C, Hoplerythrinus unitaeniatus, 95 records
Expected inside: 47.5; Observed inside: 95; Expected outside: 47.5; Observed outside: 0; X2 = 95. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
D, Synbranchus marmoratus, 144
Expected inside: 72; observed inside: 144; expected outside: 72; observed outside: 0; X2 = 144. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
E, Callichthys callitchthys, 110 records
Expected inside: 55; observed inside: 109; expected outside: 55; observed outside: 1; X2 = 106,03. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
Test for “Amazon Core” distribution
A, Boullengerella spp., 176 records
Expected inside: 88; observed inside: 176; expected outside: 88; observed outside: 0; X2 = 176. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
B, Moenkhausia collettii, 113 records
Expected inside: 56.5; observed inside: 113; expected outside: 56.5; observed outside: 0; X2 = 113. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
C, Moenkhausia oligolepis, 104 records
Expected inside: 52; observed inside: 104; expected outside: 52; observed outside: 0; X2 = 104. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
Test for “Amazon and Orinoco Lowlands” distribution
A, Moenkhausia lepidura, 51 records
Expected inside: 25.5; observed inside: 51; expected outside: 25.5; observed outside: 0; X2 = 51. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
B, Potamorhina altamazonica, 29 records
Expected inside: 14.5; observed inside: 29; expected outside: 14.5; observed outside: 0; X2 = 29. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
C, Vandellia cirrhosa, 75 records
Expected inside: 37.5; observed inside: 74; expected outside: 37.5; observed outside: 1; X2 = 71. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
Test for “Amazon and Paraguay Lowlands” distribution
A, Hemigrammus lunatus, 21 records
Expected inside: 10.5; observed inside: 21; expected outside: 10.5; observed outside: 0; X2 = 21. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
B, Epapterus dispilurus, 12 records
Expected inside: 6; observed inside: 12; expected outside: 6; observed outside: 0; X2 = 12. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
C, Mesonauta festivus, 33 records
Expected inside: 16.5; observed inside: 32; expected outside: 16.5; observed outside: 1; X2 = 29.12. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
Test for “Amazonas-Paraguay-Orinoco Lowland” distribution
A, Rhaphiodon vulpinus, 152 records
Expected inside: 76; observed inside: 143; expected outside: 76; observed outside: 9; X2 = 118.13. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
B, Sorubim lima, 40 records
Expected inside: 20; Observed inside: 39; Expected outside: 20; Observed outside: 1; X2 = 36.1. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
C, Hypophthalmus oremaculatus, 39 records
Expected inside: 19.5; Observed inside: 38; Expected outside: 19.5; Observed outside: 1; X2 = 35.1. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
Test for “Amazonas-Guyana-Orinoco Lowland” distribution
A, Crenuchus spp., 44 records
Expected inside: 22; observed inside: 44; expected outside: 22; observed outside: 0; X2 = 44. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
B, Mesonauta spp., 64 records
Expected inside: 32; observed inside: 55; expected outside: 32; observed outside: 9; X2 = 33. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
C, Hemigrammus unilineatus, 14 records
Expected inside: 7; observed inside: 14; expected outside: 7; observed outside: 0; X2 = 14. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
D, Hemigrammus ocellifer, 36 records
Expected inside: 18; observed inside: 36; expected outside: 18; observed outside: 0; X2 = 36. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
Test for “Guyana Mangrove Province” distribution
A, Curimata cyprinoides, 19 records
Expected inside: 9.5; observed inside: 19; expected outside: 9.5; observed outside: 0; X2 = 19. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
B, Cyphocharax helleri, 14 records
Expected inside: 7; observed inside: 14; expected outside: 7; observed outside: 0; X2 = 14. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
C, Polycentrus schomburgkii, 8 records
Expected inside: 4; observed inside: 8; expected outside: 4; observed outside: 0; X2 = 8. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
Test for “Eastern Amazon” distribution
A, Synaptolaemus latofasciatus, 12 records
Expected inside: 6; observed inside: 12; expected outside: 6; observed outside: 0; X2 = 12. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
B, Aphanothorolus emarginatus, 46 records
Expected inside: 23; observed inside: 42; expected outside: 23; observed outside: 4; X2 = 28. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
C, Pachyurus junki, 12 records
Expected inside: 6; observed inside: 12; expected outside: 6; observed outside: 0; X2 = 12. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
Test for “Amazon-core uplands” distribution
A, Cetopsidium spp., 41 records
Expected inside: 20.5; observed inside: 32; expected outside: 20.5; observed outside: 9; X2 = 12.9. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
B, Hemigrammus ora, 16 records
Expected inside: 8; observed inside: 16; expected outside: 8; observed outside: 0; X2 = 16. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
C, Hoplias aimara, 41 records
Expected inside: 20.5; observed inside: 41; expected outside: 20.5; observed outside: 0; X2 = 41. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
Test for “Amazonian uplands” distribution
A, Sartor spp., 4 records
Expected inside: 2; observed inside: 4; expected outside: 2; observed outside: 0; X2 = 4. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
B, Teleocichla spp., 24 records
Expected inside: 12; observed inside: 24; expected outside: 12; observed outside: 0; X2 = 24. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
C, Tocantinsia piresi, 11 records
Expected inside: 5.5; observed inside: 11; expected outside: 5.5; observed outside: 0; X2 = 11. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
Test for “Guyana Shield” distribution
A, Pseudancistrus brevispinis, 18 records
Expected inside: 9; observed inside: 18; expected outside: 9; observed outside: 0; X2 = 18. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
Test for “Longitudinal correspondence among Amazonian Shield versants” distribution
A, Sartor gr. elongatus, 4 records
Expected inside: 2; observed inside: 4; expected outside: 2; observed outside: 0; X2 = 4. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
B, Bryconexodon spp., 14 records
Expected inside: 7; observed inside: 14; expected outside: 7; observed outside: 0; X2 = 14. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
C, Bivibranchia velox, 18 records
Expected inside: 9; observed inside: 18; expected outside: 9; observed outside: 0; X2 = 18. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
Test for “Brazilian Shield” distribution
A, Caiapobrycon spp., 18 records
Expected inside: 9; observed inside: 18; expected outside: 9; observed outside: 0; X2 = 18. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
B, Jupiaba apenima, 28 records
Expected inside: 14; observed inside: 28; expected outside: 14; observed outside: 0; X2 = 28. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
C, Moenkhausia gr. pankilopteryx/pirauba, 19 records
Expected inside: 9.5; observed inside: 19; expected outside: 9.5; observed outside: 0; X2 = 19. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
Test for “cis-Andean foothills” distribution
A, Astyanacinus spp., 23 records
Expected inside: 11.5; observed inside: 23; expected outside: 11.5; observed outside: 0; X2 = 23. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
B, Leporinus striatus, 71 records
Expected inside: 35.5; observed inside: 50; expected outside: 35.5; observed outside: 21; X2 = 11.8. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
C, Steindachnerina dobula, 25 records
Expected inside: 12.5; observed inside: 25; expected outside: 12.5; observed outside: 0; X2 = 25. X2 >3.841, rejects the null hypothesis, i.e., the species is not randomly distributed.
Test for type of the water affecting distribution
Ho: The distribution of the species is not affected by the type of the water. H1: The distribution of the species is affected by the type of the water. X2 = Σ [ (O - E)2 / E ], where: E is the expected frequency (species can occur in black-, white- or clear water). Thus, E is the number of records of a species divided by 3 possibilities (black-, white- or clear water); O is the observed frequency (number of records of the species observed in a determined water type). Degrees of freedom (3 - 1 = 2). Conventionally accepted significance level of 0.05 (Chi square distribution table 5.991).
Test for “Amazon-only Lowland” distribution
A, Cetopsis candiru, 9 records
Expected black: 3; observed black: 0; expected white: 3; observed white: 8; expected clear: 3; observed clear: 1; X2 = 12.6. X2 > 5.991, rejects the null hypothesis, i.e., the distribution of the species is affected by the type of the water.
B, Curimatella meyeri, 38 records
Expected black: 12; observed black: 2; expected white: 12; observed white: 33; expected clear: 12; observed clear: 3; X2 = 51.8. X2 > 5.991, rejects the null hypothesis, i.e., the distribution of the species is affected by the type of the water.
C, Adontosternarchus balaenops, 21 records
Expected black: 7; observed black: 0; expected white: 7; observed white: 20; expected clear: 7; observed clear: 1; X2 = 21.16. X2 > 5.991, rejects the null hypothesis, i.e., the distribution of the species is affected by the type of the water.
Test for “Central Blackwater Amazon” distribution
A, Biotoecus spp., 19 records
Expected black: 6.33; observed black: 14; expected white: 6.33; observed white: 5; expected clear: 6.33; observed clear: 0; X2 = 8.38. X2 > 5.991, rejects the null hypothesis, i.e., the distribution of the species is affected by the type of the water.
B, Dicrossus spp., 25 records (collecting points in Orinoco were not considered for the water type classification)
Expected black: 8.33; observed black: 16; expected white: 8.33; observed white: 3; expected clear: 8.33; observed clear: 6; X2 = 7.72. X2 > 5.991, rejects the null hypothesis, i.e., the distribution of the species is affected by the type of the water.
C, Hemigrammus analis, 18 records (collecting points in Orinoco were not considered for the water type classification)
Expected black: 6; observed black: 13; expected white: 6; observed white: 1; expected clear: 6; observed clear: 4; X2 = 6.5. X2 >5.991, rejects the null hypothesis, i.e., the distribution of the species is affected by the type of the water.
D, Hemigrammus coeruleus, 15 records
Expected black: 5; observed black: 12; expected white: 5; observed white: 2; expected clear: 5; observed clear: 1; X2 = 6.16. X2 > 5.991, rejects the null hypothesis, i.e., the distribution of the species is affected by the type of the water.
E, Hemigrammus stictus, 18 records (collecting points in Orinoco were not considered for the water type classification)
Expected black: 6; observed black: 16; expected white: 6; observed white: 3; expected clear: 6; observed clear: 6; X2 = 10.16. X2 > 5.991, rejects the null hypothesis, i.e., the distribution of the species is affected by the type of the water.