INTRODUCTION

Global aquatic ecosystems have been losing part of their biodiversity in response to increasingly deleterious anthropogenic pressure, caused mainly by rampant population growth, preposterous consumption of natural resources, energy production, and the absence of efficient management and conservation mechanisms (^{Abell et al., 2008}; ^{Barletta et al., 2010}). This scenario is even more worrying in freshwater systems, which, besides having a smaller physical area, have been historically undergoing severe environmental alterations, such as damming, illegal water catchment, introduction of species and contamination by pesticides and effluents (^{L’vovich, White, 1990}; ^{Meybeck, 2003}; ^{Dudgeon, 2019}). All these impacts have a cumulative effect by reducing habitat quality and quantity and increasing extinction risk for freshwater fishes, making those systems into one of the most threatened worldwide (^{Meybeck, 2003}; ^{Revenga et al., 2005}; ^{Abell et al., 2008}; ^{Reis et al., 2016}; ^{Castro, Polaz, 2020}).

One of the main issues related to aquatic ecosystems is water scarcity, especially in arid and semiarid areas. When the demand for water increases and almost reaches the available limit, debates over management and attainment of new resources emerge as an attempt to converge on an integrated management model (^{L’vovich, White, 1990}; ^{Meybeck, 2003}; ^{Gupta, Van der Zaag, 2008}). Some countries have solved this problem by expanding their hydrographic network through aqueducts and canals (^{Van der Zaag, 2007}; ^{Gupta, Van der Zaag, 2008}). Currently, water transfers projects have started or are under discussion in South America, in the São Francisco Interbasin Water Transfer Project (SFR-IWT, in Brazil), Africa, in the Lesotho Highland Water Project (between Lesotho and South Africa) and Asia, in the South to North Water Diversion Project (SNWDP, in China), and in the Par Tapi Narmada Link Project (PTNLP, in India).These projects differ, mainly in the number of connected ecoregions, basins, canal length, and captured water volume (^{Brasil, 2004}; ^{India, 2017}; ^{Lesotho, 2020}; ^{Long et al., 2020}).

However, such projects are considered as imminent threats to freshwater biodiversity (^{Albert et al., 2020}) and face many criticisms and heated debates over the socio-economic benefits versus environmental issues they can cause (^{Meador, 1992}; ^{Muller, 1999}; ^{Dyrnes, Vatn, 2005}; ^{Pittock et al., 2009}). The latter include species introductions, loss of local adaptations and biota homogenization, genetic introgression, competitive exclusion, hybridization, habitat modification (including terrestrial vegetation fragmentation by canals), and water availability and quality (^{Meador, 1992}; ^{Pittock et al., 2009}; ^{Albert et al., 2020}).

Water shortage is one of the main problems in the Brazilian Northeast region, which is mostly inserted in the semiarid Caatinga (^{Lima et al., 2017}). It is a highly impacted area, some of which is in the process of desertification (^{Albuquerque et al., 2012}), and whose aquatic habitats are considered the most modified and threatened in the country (^{Silva et al., 2004}). Recurrent droughts limit water supply and demand huge engineering projects (^{Rosa et al., 2004}).

The idea to transpose waters from the rio São Francisco was introduced in 1847 by the engineer Marcos de Macedo to Emperor Dom Pedro II. In 1913, after the ‘Great Drought’ (1875-1879), the construction of a canal that would connect the São Francisco and Jaguaribe basins was outlined. Nevertheless, it became impracticable due to technical reasons, lack of financial resources and ignorance of its benefits (^{Andrade et al., 2011}). After other frustrated attempts, in 2005 the project was redesigned and started to be known as the São Francisco Interbasin Water Transfer to the Northeastern Hydrographic Basins Project (SFR-IWT, Projeto de Integração do Rio São Francisco com as Bacias Hidrográficas do Nordeste Setentrional, in Portuguese) (^{Brasil, 2004}).

The SFR-IWT aims to transfer 3.5% of the São Francisco basin water to the Jaguaribe, Apodi-Mossoró, Piranhas-Açu, and Paraíba do Norte basins, through a two-canal system (North and East Axis), aqueducts, tunnels, dams, and pumping stations, resulting in 720 km of extension (577 km, excepting the secondary canals, 357 km in the North Axis and 220 km in the East Axis), whose catchment areas are located in the lower-middle stretch of the São Francisco basin (^{Brasil, 2004}). Since its approval in 2007 and the beginning of construction, in 2008 (^{Andrade et al., 2011}), the project has faced many delays but is now almost concluded, with the East Axis in operation since March 2017 and the North Axis partially operating since February 2018 (^{Brasil, 2019}). Besides the SFR-IWT, the São Francisco basin was artificially connected to the upper Paraná basin in the 1960s through the rio Piumhi, during the construction of the Furnas Hydroelectric Power Plant, which resulted in an input of fish fauna from the captured basin (^{Moreira-Filho, Buckup, 2005}).

The SFR-IWT installation license was granted in 2007 based on an environmental impact study (^{Brasil, 2004}). This report, written at the same time as publication of the seminal studies on Caatinga fish (^{Rosa et al., 2003}, 2004; ^{Rosa, 2004}), came across taxonomic issues and species distribution inaccuracies, reflecting the incipient knowledge on this ichthyofauna. Recently, a global database of freshwater fish species listed by drainage channels included data from more than 3,000 basins; however, the Northeast of Brazil was indicated as spatial gap, with data available from few basins (^{Tedesco et al., 2017}). This database is useful to evaluate non-native species’ influence on the native ichthyofauna, which are under homogenization processes (^{Tedesco et al., 2017}). The limited information on the distributional ranges of the species also hampers evaluation of the conservation status of the species (^{Dias et al., 2016}).

Due to climatic and edaphic conditions, which resulted in temporary rivers, the fish diversity of the Caatinga was underestimated until recently (^{Rosa et al., 2003}). ^{Lima et al. (2017)} updated this list, increasing it to 386 species, 203 of them endemic to a single freshwater ecoregion (sensu^{Abell et al., 2008}), and 33 threatened species. However, the biological diversity and the conservation status of its biota are still poorly known, and inventories should be performed on a regional scale to support public conservation policies (^{Rosa et al., 2004}; ^{Lévêque et al., 2005}, 2008; ^{Tedesco et al., 2017}). The Caatinga’s hydrography is composed of four freshwater ecoregions: Maranhão-Piauí (MAPE), Mid-Northeastern Caatinga (MNCE), São Francisco (SFRE), and Northeastern Atlantic Forest (NAFE) (^{Rosa et al., 2003}; ^{Albert et al., 2011}; ^{Lima et al., 2017}). Most MNCE basins exhibit an intermittent regime and small to medium extensions (^{Rosa et al., 2004}). They also have lower fish species richness and endemism (88 and 38, respectively) when compared to the adjacent SFRE (211 and 135) and to the MAPE (151 and 54) (^{Albert et al., 2011}; ^{Ramos et al., 2014}; ^{Silva et al., 2015}; ^{Reis et al., 2016}). The high endemism level of these ecoregions suggests ancient separation and low connectivity among the basins (^{Albert, Carvalho, 2011}), but these numbers are still partial, especially for the MNCE, since the regional collections are scarcely digitalized, and only a few species of this ecoregion, are available (^{Buckup et al., 2007}; ^{Lévêque et al., 2008}; ^{Langeani et al., 2009}).

Regarding the SFR-IWT basins, there is a huge data deficiency, and comparative studies are paramount to inform which species are endemic, naturally shared and potentially invasive before the artificial connection (^{Langeani et al., 2009}). This urgency encouraged the present comprehensive study of the SFR-IWT’s freshwater fish fauna. Therefore, the purpose of this study was to elaborate an original ichthyofaunal species list of the basins encompassed in the SFR-IWT, proposing a taxonomic nomenclatural standardization, and indicating which species are endemic, endangered, non-native, and naturally shared between basins and ecoregions. Notwithstanding, this baseline is not definitive and should be refined and updated regularly. Besides that, it may help in the detection of possible changes in fish composition resulting from the SFR-IWT and the proposition of public policies for the conservation of Caatinga ichthyofauna.

MATERIAL AND METHODS

Information source. The freshwater fish species occurrence data comprised the rio São Francisco lower-middle stretch, the SFR-IWT catchment area in SFRE; and receptor’s drainages of the Jaguaribe (JAG), Apodi-Mossoró (APO), Piranhas-Açu (PIA), and Paraíba do Norte (PAR) river basins in the MNCE. The map was produced with QuantumGIS free program (www.qgis.org). Calculations of the drainage area and number of municipalities were done using georeferenced shapefiles of the National Water Agency (^{Brasil, 2014a}) and Ministry of the Environment (^{Brasil, 2014b}), while the São Francisco Hydrographic Basin Committee provided the SFR-IWT canal trails (^{CBHSF, 2014}) (Fig. 1).

FIGURE 1 |  Sampling sites of the freshwater fish species in the São Francisco Interbasin Water Transfer Project basins in the Brazilian semiarid.  

The São Francisco basin and ecoregion has an approximate area of 640,000 km². Its main river, the São Francisco, is the largest in the Northeast region, and the third in Brazil (^{Rosa et al., 2003}). It is subdivided in to four stretches: upper, middle, lower-middle, and lower (^{Brasil, 2014a}). Herein, only records in the lower-middle stretch (112,093.68 km², 17.51% of São Francisco basin and 42.52% of SFR-IWT) were considered, and this stretch is under direct influence of the SFR-IWT, including 98 municipalities, from downstream Sobradinho Dam reservoir (Sobradinho Municipality, Bahia State) to the upstream Lajeadinho Stream (Canindé de São Francisco Municipality, Sergipe State), or approximately 50 km downstream of Paulo Afonso Dam (^{Brasil, 2014a}) (Fig. 1). In this semiarid area, most tributaries of the São Francisco are intermittent (^{Rosa et al., 2003}), and some of them (Moxotó, Pajeú, Terra Nova, and Brígida rivers) will also receive water inflow from the main São Francisco channel (^{Lima, 2005}).

The receptor basins are the four main drainage areas of the Mid-Northeatern Caatinga ecoregion (MNCE), presented in the west-east direction. The Jaguaribe basin supplies water to 88 municipalities of Ceará State and is the largest MNCE basin, with an area of 74,077.01 km² (28.10% of SFR-IWT). Besides the Jaguaribe River (633 km of extension), the basin includes tributaries, such as the Banabuiú and Salgado rivers (^{Brasil, 2014a}). The Apodi-Mossoró basin encompasses 68 municipalities in Rio Grande do Norte State with an area of 14,303.71 km² (5.43% of SFR-IWT), and approximately 200 km of extension (^{Brasil, 2014a}). Extending about 350 km, the Piranhas-Açu basin represents the second largest MNCE basin (43,141.54 km², 16.37% of SFR-IWT), draining 158 municipalities in Paraíba and Rio Grande do Norte states (^{Brasil, 2014a}). These three basins will receive water from the SFR-IWT North Axis (^{Brasil, 2004}) (Fig. 1). Lastly, the Paraíba do Norte basin, with 19,977.48 km² (7.58% of SFR-IWT) and 280 km length in 90 municipalities (^{Brasil, 2014a}) in Paraíba State, is the largest one in the MNCE that flows to the east coast (the other ones drain to the north) (^{Brasil, 2019}) (Fig. 1).

Species records. Species occurrences were obtained through latitude and longitude data, taking into account strictly freshwater fish families (^{Buckup et al., 2007}). In the few cases in which the geographical coordinates were not available in the original source, they were approximated using the GeoNames website (www.geonames.org). All records were made before any water input from the SFR-IWT canals in March 2017.

Primary data were obtained from collected and deposited material in ichthyological collections at the Federal Universities of Paraíba (UFPB, 129 localities - 56.09%) and Rio Grande do Norte (UFRN, 101 - 43.91%). All locality data were listed in the supplementary material (S1). Distinct active (trawls, dip nets, sieves, fishhooks, and cast nets) and passive (gill nets and traps) fishing methods were used to reduce selectivity (^{Uieda, Castro, 1999}).

Secondary data (76 localities) were acquired in two steps: 1) data from digitalized ichthyological collections available in speciesLink, PRONEX/NEODAT, and Portal da Biodiversidade (www.splink.org.br, www.mnrj.ufrj.br/pronex/, and https://portaldabiodiversidade.icmbio.gov.br/portal/), and 2) literature review (e.g.^{Reis et al., 2003}; ^{Buckup et al., 2007}). At this stage, the bibliographic databases Web of Science and Google Scholar were consulted during searches of species records in the SFR-IWT basins (filtering dubious occurrences) after January 2006, the time frame used by ^{Buckup et al. (2007)}, until March 2017, current frame (e.g.^{Luz et al., 2012}; ^{Nascimento et al., 2014}), including new species descriptions (e.g.^{Ribeiro, Lucena, 2006}; ^{Lucena, 2007}; ^{Lima, Britski, 2007}; ^{Ramos et al., 2013}; ^{Zawadzki et al., 2017}). The abbreviation “aff.” was used for species that have affinity with, but should be distinct from, the nominal taxon to which they are currently assigned.

Taxonomic validation. Nominal species validity, synonyms, and systematic classification were carried out according to ^{Fricke et al. (2020)}. Some species with uncertain taxonomic status were considered as species inquirendae, following the recommendation of ^{Lima et al. (2017)}. Specific studies on species introduction in the Northeast region were examined in order to determine non-native species (^{Gurgel, Fernando, 1994}; ^{DNOCS, 2002}; ^{Rosa et al., 2003}; ^{Mattheus, 2005}; ^{Leão et al., 2011}; ^{Paiva, Mesquita, 2013}). The Brazilian list of endangered fish and aquatic invertebrates was consulted to indicate the threatened ones (^{Brasil, 2014c}), given that it follows IUCN parameters. ^{Reis et al. (2003)}, ^{Buckup et al. (2007)}, ^{Fricke et al. (2020)} and ^{Lima et al. (2017)} were examined in order to determine the endemic species, defined herein as those restricted to a single basin or hydrographic ecoregion.

All specimens were deposited in one of these museums: ANSP (The Academy of Natural Sciences, Drexel University); FMNH (Field Museum of Natural History); LIRP (Laboratório de Ictiologia de Ribeirão Preto); MCP (Museu de Ciências e Tecnologia da Pontifícia Universidade Católica do Rio Grande do Sul); MNHN (Muséum National d’Histoire Naturelle); MNRJ (Museu Nacional, Universidade Federal do Rio de Janeiro); MZUEL (Museu de Zoologia, Universidade Estadual de Londrina); MZUSP (Museu de Zoologia, Universidade de São Paulo); NMW (Naturhistorisches Museum, Wien); NUPELIA (Núcleo de Pesquisa em Limnologia, Ictiologia e Aquicultura); UFPB (Universidade Federal da Paraíba); UFRN (Universidade Federal do Rio Grande do Norte); UMMZ (University of Michigan Museum of Zoology); ZMB (Museum für Naturkunde). All lots are listed in the supplementary material (S2).

Spatial patterns of species richness. All species inquirendae, and those of the secondary and vicarious divisions families (e.g. Sciaenidae, Gobiidae, Clupeidae, Engraulidae) (sensu^{Myers, 1937}) were removed from comparisons to avoid inflating the number of shared species due to taxonomic uncertainty or possible marine dispersion. However, the distribution of these species in literature was presented in S3 (available only in the online version) for future research. For the evaluation of spatial patterns of richness, each hydrographic basin had its species-area relationship calculated for native and endemic species (^{Albert et al., 2011}). This relationship is extensively used as a way of comparing distinct and size-diverse ecosystems worldwide (^{Arrhenius, 1921}; ^{Connor, McCoy, 1979}). Thus, in each region, the number of species is correlated with their spatial extent, through a function of type S=S0Ab. Where S is the number of species in area A, S0 is proportional species density (i.e., species richness per unit area), and b is the species-area scaling exponent, often with values in the range 0.25-0.50 (^{Dengler, 2009}). Since the scaling exponent b defines the slope of the species-area regression, it may be interpreted as a measure of gamma diversity among areas, with higher values indicating greater differences in the taxonomic composition of areas. The similarities of native species composition between river basins were calculated using Jaccard’s similarity coefficient index, using the following formula J=(c/(a+b+c)).100, where c is the number of fish species existing at both sampling sites and a and b are the number of fish species at different sampling sites. This index takes into account only the values of presence and absence of species in each basin (^{Lévêque et al., 2013}).

RESULTS

Species richness and species-area relationships. In total, 17,002 lots of fish were compiled, of which 3,331 (2,729 primary data) were selected and recorded in 306 sampling sites (230 primaries), 117 (66 primaries) in SFRE (lower-middle São Francisco basin) and 189 (164 primaries) in the MNCE. In the receptor drainages (MNCE’s localities), there were 56 (50 primaries) in the Jaguaribe (JAG), 30 (22) in the Apodi-Mossoró (APO), 59 (all primaries) in the Piranhas-Açu (PIA), and 44 (all primaries) in the Paraíba do Norte basins (PAR) (Figs. 1-2).

FIGURE 2 |  Sampling sites in the São Francisco Interbasin Water Transfer Project basins in the Brazilian semiarid. A = Canals under construction near the rio São Francisco main channel, and B = near Sertânia, Pernambuco State (PE), C = Rio Pajeú, tributary of the São Francisco basin, PE, D = Rio São Francisco near Petrolina, PE, E = Temporary pool in rio Jaguaribe basin in Russas, Ceará State (CE), F = Rio Jaguaribe in Crato, CE, G = Rio Apodi-Mossoró in Pau dos Ferros, Rio Grande do Norte State (RN), H = Rio Apodi-Mossoró in Pau dos Ferros, RN, I = Rio Seridó, tributary of the Piranhas-Açu basin in Caicó, RN, J = Rio Piranhas-Açu, Jardim de Piranhas, RN, K = Rio Paraíba do Norte in Barra de Santana, Paraíba State (PB), L = Rio Paraíba do Norte in São João do Cariri, PB. 

In this study, 121 valid freshwater fish species were recorded, from 25 families and seven orders, or 111, 23 and six, respectively, considering only the native species (Tab. 1, Fig. 3A-I). Additionally, five species belonging to families of the vicarious divisions were found, but disregarded in the list and analyses: three native, Anchoviella vaillanti (Steindachner, 1908) endemic to SFRE, A. lepidentostole (Fowler, 1911) present in JAG, and Awaous tajasica (Lichtenstein, 1822) recorded in PAR, and two introduced ones, Plagioscion auratus (Castelnau, 1855) in SFRE and P. squamosissimus (Heckel, 1840), recorded in all basins of the SFR-IWT, except in PAR (^{Alves et al., 2011}; ^{Leão et al., 2011}). Furthermore, six species inquirendae were listed, Psalidodon rivularis (Lütken, 1875), Pseudancistrus papariae Fowler, 1941, Pimelodella papariae (Fowler, 1941), P. witmeri Fowler, 1941, Hypostomus carvalhoi (Miranda Ribeiro, 1937), and H. jaguribensis (Fowler, 1915) (S3).

Five endangered species were recorded, three of them in SFRE, Lophiosilurus alexandri Steindachner, 1876 as vulnerable, Conorhynchos conirostris (Valenciennes, 1840) as endangered (EN), and Hypsolebias flavicaudatus (Costa & Brasil, 1990) as critically endangered), and two in MNCE, Apareiodon davisi Fowler, 1941 (EN) and Parotocinclus spilurus (Fowler, 1941) (EN) (Tab. 1 , Fig. 3F). Eleven non-native species were recorded, eight of which occur in SFRE and MNCE receptor basins (Colossoma macropomum, Parachromis managuensis, Poecilia reticulata, Astronotus ocellatus, Cichla monoculus, C. kelberi, Oreochromis niloticus, and Coptodon rendalli); the last six exist in all SFR-IWT basins, and three only in the MNCE (Arapaima gigas, Megaleporinus obtusidens, native in SFRE, and Xiphophorus helleri) (Tab. 1, Fig. 3J-L).

FIGURE 3 |  Freshwater fish species from the São Francisco Interbasin Water Transfer Project basins in the Brazilian semiarid. A = Hemigrammus brevis, endemic species of São Francisco Ecoregion (SFRE); B = Moenkhausia costae and C = Psellogrammus kennedyi, shared species between SFRE and Mid-Northeastern Caatinga Ecoregion (MNCE); D = Aspidoras menezesi, endemic species of Jaguaribe basin (JAG) ; E = Hypostomus sertanejo, endemic species of MNCE; F = Parotocinclus spilurus, endemic and endangered species of JAG; G = Tatia bockmanni, endemic species of SFRE; H = Cichlasoma orientale, shared species from all basins of SFR-IWT; I = Geophagus brasiliensis, shared species between SFRE and MNCE; J = Colossoma macropomum, non-native species shared between SFRE and Piranhas-Açu basin; K = Parachromis managuensis, non-native species shared between SFRE and Paraíba do Norte basin; L = Xiphophorus helleri, non-native species of JAG. 

Regarding the hydrographic basins, considering only native species, the highest richness (78, corresponding to 70.27%) was recorded in the SFRE donor basin, while the MNCE’s receptor basins accounted for 61 species (54.95%). Of those, 50 species (44.05% of total and 81.97% of MNCE) were recorded in JAG, 39 (35.14% and 63.93%) in PIA, 36 (32.43% and 59.02%) in PAR, and 32 (28.83% and 52.46%) in APO (Tab. 2). The species-area relationship calculation indicated that the SFRE (lower-middle stretch) exhibits the highest native species density (1.75), followed by PAR (1.42), APO (1.41), JAG (1.29), MNCE (1.24, all receptor basins) and PIA (1.20) (Tab. 3). Considering only the endemic species, the following values were obtained: SFRE (0.52), MNCE (0.26), JAG (0.05), PAR (0.04), and PIA (0.03). The APO basin did not contain any endemic species, and its higher species-area value might be related to its smaller size in comparison to the other basins (Tab. 3).

Spatial distribution and species sharing. Among the 111 native freshwater fish species recorded, 16 are present in all SFR-IWT basins (Prochilodus brevis, Leporinus piau, Hoplias aff. malabaricus, Compsura heterura, Hemigrammus marginatus, Psalidodon fasciatus, Serrapinnus heterodon, S. piaba, Triportheus signatus, Trachelyopterus galeatus, Rhamdia aff. quelen, Hypostomus pusarum, Poecilia vivipara, Synbranchus aff. marmoratus, Cichlasoma orientale and Crenicichla brasiliensis) (Tab. 1, Fig. 3H). The ichthyofaunal composition similarity between SFRE and MNCE was composed of 28 (25.23%) shared species. Furthermore, a pairwise comparison of the sharing values of SFRE with the four receptor basins of MNCE indicated that the similarity species composition was even lower: between 20 (22.22%, SFRE x APO) and 24 (23.08%, SFRE x JAG). As expected, the percentage of shared species was higher between the receptor basins of the MNCE, varying from 41.67% (20 spp., APO x PAR) to 64.81%, (35, JAG x PIA). It is worth mentioning that 49 (62.82%) of the SFRE native species do not occur in any receptor basin, and 33 (54.10%) species occur in the MNCE basin but are absent in SFRE (Tab. 1).

There are also some systematic differences in the ichthyofaunal composition of the SFR-IWT ecoregions, mainly referring to the families Loricariidae, Anostomidae, Pimelodidae, and Characidae. Only three loricariid species occur in both ecoregions (Hypostomus pusarum, Parotocinclus cearensis, and P. jumbo), while three are only found in SFRE (Hypostomus macrops, Pterygoplichthys etentaculatus, and Rineloricaria sp.), and seven only in MNCE (Hypostomus sertanejo, Loricariichthys derbyi, Loricariichthys sp., Parotocinclus seridoensis, P. spilosoma, P. spilurus, and Pseudancistrus genisetiger) (Fig. 3E-F). Only two Anostomidae species were recorded in both ecoregions (Leporinus piau and L. taeniatus), while five are exclusive to SFRE (Leporellus vittatus, Leporinus friderici, Megaleporinus obtusidens, M. reinhardti, and Schizodon knerii), and one to MNCE (Schizodon dissimilis). All pimelodid species (five) were only recorded in SFRE, besides Conorhynchos conirostris, previously allocated in this family, and also endemic to the SFRE. Lastly, of the 26 characid species, only seven were found simultaneously in SFRE and MNCE (Compsura heterura, Hemigrammus marginatus, Moenkhausia costae, Psalidodon fasciatus, Psellogrammus kennedyi, Serrapinnus heterodon, and S. piaba), while 11 were recorded only in SFRE, and eight in MNCE (Tab. 1, Fig. 3B-C).

Endemic species. Among the 111 native freshwater fish species recorded in the basins of the SFR-IWT, 36 (32.43%) species were considered endemic of the SFRE or MNCE ecoregions, some of them occurring in a single basin of the latter, 23 in SFRE (Apareiodon hasemani, Prochilodus argenteus, P. costatus, Megaleporinus reinhardti, Schizodon knerii, Hemigrammus brevis, Orthospinus franciscensis, Phenacogaster julliae, Roeboides xenodon, Triportheus guentheri, Tatia bockmanni, Lophiosilurus alexandri, Pimelodella laurenti, Corydoras garbei, Crenicichla sp., Salminus franciscanus, Duopalatinus emarginatus, Pimelodus pohli, Pimelodella robinsoni, Pterygoplichthys etentaculatus, Conorhynchos conirostris, Cynolebias porosus, Hypsolebias flavicaudatus), and 13 in MNCE (Apareiodon davisi, Psectrogaster saguiru, Cheirodon jaguaribensis, Corydoras sp., Hypostomus sertanejo, Pimelodella dorseyi, P. enochi, Cynolebias microphthalmus, Hypsolebias antenori, some of them restricted to a single basin, such as Aspidoras menezesi, and Parotocinclus spilurus, in JAG, P. seridoensis in PIA, and P. spilosoma in PAR) (Fig. 3A, D-G). Some families must be highlighted: Loricariidae and Characidae with five endemic species each (38.46% and 19.23% of the species of this families, respectively), and Heptapteridae with four species (80.00%). Even though the total number of species is smaller, Rivulidae (4) and Parodontidae (2) included only endemic species (Tabs. 1, 3).

DISCUSSION

São Francisco IWT freshwater fish richness.^{Langeani et al. (2009)} mentioned the absence of comparative data between ichthyofauna of donor and receptor basins of the SFR-IWT. According to our updated database, 49 species (62.82%, considering only native and 56.98%, all species) of the SFRE do not occur in any MNCE receptor basin and might be potentially invasive species. It is important to note that, for species shared among donating and receiving basins, many are still in need of taxonomic revisions, which might change this number. Besides the taxonomic composition, the biological diversity expressed through populational variations are, certainly, present and could be addressed through phylogeographic studies (^{Craig et al., 2017}; ^{Berbel-Filho et al., 2018}). Moreover, in 2019 the team responsible for monitoring fish species during implementation of the SFR-IWT had already identified that three species from the SFRE must have arrived in PAR via the SFR-IWT canal. However, the scientific names of these species were not mentioned (^{Domingues, 2019}), corroborating the potential invasion risk of SFRE species in MNCE basins.

The SFR-IWT environmental impacts assessment (EIA) of the project pointed out possible environmental impacts, such as dissimilar fauna, supposedly, with only twenty-eight (23.93%, of 117 species listed in EIA) shared species between SFRE and MNCE, and possible quali-quantitative alterations in the ichthyofaunal composition of the receptor basin. However, the document minimizes the impacts that might be caused by the existence of eight non-native species (Astronotus ocellatus, Cichla monoculus, Lebistes reticulatus (= Poecilia reticulata), Leporinus “elongatus”, Oreochromis niloticus, Plagioscion squamosissimus, Poecilia latipinna, Poecilia vivipara (= considered as native in our study)) in the receptor basins and the low fish diversity (61 species, including non-native) considering its total drainage area (^{Brasil, 2004}). Although the SFR-IWT EIA provides useful information about the fish fauna from the receptor basins, mainly considering the technical and systematics limitations of its time, it did not provide any species comparison among the drainage areas involved, and the taxonomy is outdated (^{Brasil, 2004}).

Freshwater fish fauna basin surveys provide valuable information to detect changes in the species composition, mainly the input of species from the donor basin, such as those present in the lower-middle São Francisco basin in the SFR-IWT. The advances in the knowledge of SFR-IWT freshwater fish diversity were primarily due to the extensive surveys in the last decade, mainly conducted by regional research groups, whose fish collections are not yet online (^{Lima et al., 2017}). To make this database available, this systematic and geographic reference work on the freshwater fishes of the SFR-IWT, including sheets and shapefiles, will be available to fill this spatial gap in future regional or global analysis. This baseline might be useful to detect the establishment of non-native species, including those native to the SFRE via SFR-IWT canals.

^{Rosa (2004)} conducted a comprehensive study of the Caatinga fishes based on literature data and showed species richness information for the main drainage areas of this biome, including three receptor ones from SFR-IWT: JAG (45 species), PIA (24), and PAR (23). These numbers included introduced species, but no species lists were provided for each basin. Another document that reports species richness in the SFR-IWT receptor basins is the SFR-IWT EIA (^{Brasil, 2004}). The ichthyological list included primary data of 50 localities in the receptor basins surveyed between October and December 1998, secondary data (literature) from the donor basin, and interviews with riverine communities. It listed 86 species assigned to the lower-middle São Francisco basin, 52 to JAG, 21 to APO, 29 to PIA, and 34 to PAR, but considering only the native species (except for Poecilia vivipara, herein considered as native), they reduced it to 78 species in SFR, 47 in JAG, 17 in APO, 25 in PIA and 30 in PAR, numbers that are up to 46.88% lower than those currently listed (78 in SFR, 50 in JAG, 32 in APO, 39 in PIA and 36 in PAR) (Tab. 1). The updated values are also higher than those presented by a recent global database of freshwater fishes by basins (^{Tedesco et al., 2017}), which listed 31 species in JAG, 14 in APO, 17 in PIA and 18 in PAR. Recently, ^{Ramos et al. (2018)} and ^{Rodrigues-Filho et al. (2019)} provided inventories of continental fish species from PAR and JAG, respectively. The first, including vicarious division families, listed 47 species, but all freshwater species were also recorded in our study. The second, focused on the organization of fish metacommunities, lists 34 species, but contains misidentifications. Additionally, another 24 species were added in JAG. Considering the SFRE, in our study only the species from the lower-middle stretch were listed, resulting in 86 (78 native), which corresponds to 28.28% of the 304 species (241 native) of the whole ecoregion (^{Barbosa et al., 2017}). The upper and middle SFRE is highly different from the downstream stretches, with several highland taxa not present in the lowlands and semiarid.

Endemic and endangered species. Studies on the distribution of endemic species have important conservationist aims, mainly in the definition of protected areas, including their biotypes (^{Myers et al., 2000}). The presence of 36 endemic species (32.43% of natives) in the studied basins corroborates the high endemism of Caatinga fishes (^{Lima et al., 2017}). The presence of endemic and endangered species in the receptor basins is worrying and deserves special attention, considering the ecological impacts caused by species introduction, which might range from behavioral shifts to complete alterations in the food web, extirpation of native species, and biotic homogenization (^{Kadye, Booth, 2012}; ^{Brito et al., 2020}). Habitat changes, added to invasion of predator or competitor non-native species, may reduce their abundance or distribution range (^{Shelton et al., 2017}). Even endemic and threatened species from the São Francisco basin might suffer from lower hydrological levels, benefiting generalist species over specialists, and might also affect the seasonal killifish species inhabiting the floodplains of the basin (^{Costa, 2002}).

Leporinus taeniatus is referred in the literature as endemic to the SFRE (^{Reis et al., 2003}; ^{Buckup et al., 2007}; ^{Fricke et al., 2020}). However, this species was collected in two out of the four MNCE receptor basins (JAG and APO). Even though there are no records of introduction of this species in the consulted source’s reports (^{Mattheus, 2005}; ^{Leão et al., 2011}; ^{Paiva, Mesquita, 2013}), the hypothesis of introduction cannot be disregarded, given that it is of commercial interest in SFRE (^{Bazzoli, 2003}), and that Megaleporinus obtusidens had already been introduced to the MNCE (^{Gurgel, Oliveira, 1987}).

Three Parotocinclus species are endemic to SFR-IWT receptor basins: P. spilurus in JAG, P. seridoensis in PIA, and P. spilosoma in PAR, and this number could be higher considering putative undescribed species (^{Lima et al., 2017}) (Fig. 3F). This might also be the case of some species not identified at species level (Loricariichthys sp. in JAG, Corydoras sp. in PIA, and Aspidoras sp. in SFRE) which can be putative undescribed species, confirming that ichthyofaunal taxonomic knowledge from the region is still incomplete, and strengthening the necessity for systematic research (^{Lima et al., 2017}). The Parotocinclus and Aspidoras catfishes are usually restricted to intermittent headwaters and could also be impacted by accumulation of fine sedimentary particles carried through the canals.

Final considerations and prospects. IWT projects are considered as one of the main threats to freshwater biodiversity, mainly to fish (^{Dudgeon, 2019}). Furthermore, the modifications of the hydrological regime of the SFR-IWT basins can influence the seasonality, flow, availability and quality of the Caatinga rivers (^{Jardine et al., 2015}). Consequently, these changes can lead to a reduction in fish diversity and inland fishery production (^{Sabo et al., 2017}).

Deleterious anthropogenic alterations to the aquatic environment caused by large engineering projects are indisputable worldwide; however, obtaining reliable species richness lists before artificial IWT canals are implemented is difficult, albeit essential to detect species invasions (^{Johnson et al., 2008}). Deficiency in the pre-construction fish species baseline of the Lesotho Highland Water Project in Africa must have contributed to the disappearance of the endangered minnow Pseudobarbus quathlambae (Barnard 1938) in some areas, which was the only endemic species of the Lesotho highlands (^{Shelton et al., 2017}). In the SFR-IWT basins, it was aggravated by the lack of an updated fish list by basins, which might hamper detection of the changes in richness due to the arrival of invasive species by the artificial canals. This study provides original lists of the freshwater fish species in each SFR-IWT basin, which should not considered as a final list, but as a baseline to detect changes after its operation.

This baseline of species richness and endemism (including endangered species) presents values higher than those previously stated (^{Rosa, 2004}; ^{Brasil, 2004}), as well as making the first comparisons of shared species among donor and receptor SFR-IWT basins. These results cause concern, due to the potential risk of non-native species invading the receptor basins. To the extent that invasive species settle down, the faunal homogenization process will begin (^{Brito et al., 2020}); and possible local and even total extinctions of endemic species might occur in this ichthyofauna adapted to the temporary hydrological regimen (^{Silva, 2017}).

The SFR-IWT canals are already functioning, and special attention should be given to implementing ichthyofaunal protection measures, such as aquatic biota block barriers, mainly to avoid the faunal exchange between basins. The priority actions for the SFR-IWT basins should include the (1) sustainable limits of freshwater withdrawal, and (2) halting expansion of IWT projects, and (3) constant monitoring, especially of threatened and endemic freshwater fish species of the semiarid drainage areas.

To avoid invasive species, it is necessary to build efficient barriers to restrain faunal exchange among the SFR-IWT basins (^{Brasil, 2020}). Behavioral and physical barriers should be adequate to block migrant and carnivore fish that occur in the catchment area (e.g. Leporinus, Megaleporinus, Pygocentrus, and Plagioscion species), which may see their chance of establishment in modified environments rise up to eight times in relation to natural conditions (^{Johnson et al., 2008}). Of the 11 non-native species listed herein, Xiphophorus helleri and Parachromis managuensis are considered as recently introduced (^{Rodrigues-Filho et al., 2016}; ^{França et al., 2017}). However, the second, an exotic predator cichlid, was recorded in the SFRE and in PAR among the receptor basins, and could invade the other basins throughout the canals (^{França et al., 2017}; ^{Ramos et al., 2018}) (Fig. 3K-L).

Environmental changes can force the substitution of native species populations by those from the donor basin, which might be more adapted to the perennial regime resulting from water transfer (^{Shelton et al., 2017}). Nevertheless, ^{Lynch et al. (2011)} recognized the higher possibility of faunal composition changes among distinct hydrographic basins in the Par Tapi Narmada Link Project canals in India, and that in long canals (over 550 km) there is a reduction in the probability of species invasion, as only some can run those long distances. In the case of SFR-IWT, it is worrying taking into account the canals’ intermediate size (357 km in North Axis and 220 in East Axis), in which it is possible that migratory species such as Prochilodus, Pimelodus, Megaleporinus and other medium and large-sized-species from the SFRE may reach the receptor basins. Other studies verified the establishment of invasive species in rivers with altered abiotic conditions and river flow, including IWT canals (^{Kadye, Booth, 2012}). We hope, however, that this baseline is a starting point for predictive assessments of non-native species (e.g. ecological niche modeling) in the SFR-IWT receptor basins.