A preliminary inventory of the catfishes of the lower Rio Nhamundá, Brazil (Ostariophysi, Siluriformes)

Abstract The Rio Nhamundá is a poorly-known clearwater river draining the southern Guiana Shield of Brazil. In this study we report the findings of a preliminary ichthyological survey, focusing on catfishes (Siluriformes). We identify a total of 36 species (31 genera, seven families) from the Nhamundá, including 11 species already recorded from the river. Overall, our survey results show that even rapid surveys can provide important information on Amazon fish biodiversity, suggesting potential new species, providing range extensions for nominal species, and additionally highlighting taxa in need of taxonomic revision and genetic study. As well as the traditional forms of data collected on biodiversity surveys (i.e. preserved specimen vouchers), our study also provides "new" types of data in the form of DNA barcodes and images of fishes exhibiting colouration in life, information that will be invaluable in future work addressing difficult groups. O Rio Nhamundá é um rio de água clara, pouco conhecido, que drena parte do Escudo das Guianas em território brasileiro. Nesse estudo, nós reportamos os resultados de um levantamento ictiofaunístico preliminar dessa área, tendo como foco os bagres (Siluriformes). Nós identificamos um total de 36 espécies (31 gêneros, sete famílias) provenientes de nossa coleta, e adicionamos 11 espécies já conhecidas para o rio. De maneira geral, os resultados de nossa pesquisa mostram que mesmo levantamentos rápidos podem gerar informações importantes sobre a biodiversidade de peixes amazônicos, sugerindo potenciais espécies novas, ampliando a área de distribuição de espécies, além de apontar a necessidade de revisões taxonômicas e estudos genéticos para alguns taxa. Para além das formas tradicionais de dados coletados em pesquisas de biodiversidade (i.e. espécimes preservados), nosso estudo fornece "novas" formas de dados, como DNA barcodes e imagens com o padrão de coloração dos espécimes vivos, informações essas que serão de valor inestimável para futuros estudos que abordem grupos taxonômicos difíceis.


Introduction
The Rio Nhamundá is one of the south-draining Guiana Shield tributaries of the Rio Amazonas, Brazil. The river is located between the Uatumã-Jatapú and Trombetas rivers (Fig. 1), and comprises the administrative border between Amazonas and Pará States. Chemically, the Rio Nhamundá is of the clearwater type (Carvalho de Lima and Araujo-Lima 2004), being poor in sediments, dissolved minerals, and humic compounds (Crampton 2011).
The ichthyofauna of the southern Guiana Shield rivers is poorly known in comparison to the northern Brazilian Shield rivers (Lujan and Armbruster 2011), and few scientific collections appear to have been made in the area (Vari et al. 2009). One exception is the Trombetas river, which has been extensively surveyed over several years (Ferreira 1993). Aquarium hobbyists have, however, explored parts of the Rio Nhamundá in search of new discus (Cichlidae: Symphysodon spp.) variants (Bleher 2006), and ornamental fishermen currently operate on the river and specialise mainly in the catching of these discus (pers. obs.).
The aim of this study is to assess our current knowledge of Siluriformes (catfishes) of the Rio Nhamundá, and to report the results of a preliminary survey of the lower reaches of the river (i.e. downstream of the last major cataracts).

Desk survey
In order to assess the current knowledge of the catfishes of the Rio Nhamundá we carried out a survey of Web databases as well as published checklists and literature. The Web databases searched were FishBase (Froese and Pauly 2014), Catalog of Fishes (Eschmeyer 2014), and the Global Biodiversity Information Facility (GBIF; http:// www.gbif.org). The FishBase and GBIF searches were conducted using the rfishbase 0. 2-2 (Boettiger et al. 2012) and rgbif 0.5.0 (Chamberlain et al. 2014) packages in the R programming language (R Core Team 2014); a script to repeat these searches is presented in Suppl. material 1. The Catalog of Fishes was searched manually using their Web interface (date 05-03-14) using the terms "Nhamundá" and "Nhamunda" to avoid potential discrepancies due to the accent on the last letter. Local searches were performed on PDF copies of the Checklist of Catfishes (Ferraris 2007), the Check List of the Freshwater Fishes of South and Central America (Reis et al. 2003), and on over 3,500 locally stored articles on Neotropical ichthyology in PDF format (search script is available in Suppl. material 2). We additionally checked records for the neighbouring Trombetas and Uatumã rivers using rfishbase, rgbif, Catalog of Fishes, and Checklist of Catfishes. The database at the fish collection of the Instituto Nacional de Pesquisas da Amazônia (INPA) was also searched for catfish records from the Nhamundá drainage.

Field sampling techniques
Fishes were collected during the November 2013 dry season using a combination of methods mostly targeting larger species: gill netting, rod and line fishing, spear fishing, hand netting, and hand catching using snorkelling equipment. Fishes were photographed alive whenever possible following the recommendations outlined in Sabaj Pérez (2009), and were then euthanised using a eugenol (clove oil) solution. Subsequent image manipulation was carried out using GIMP (http://www.gimp.org/), also following Sabaj Pérez (2009). Tissue samples-usually from muscle and fin membranes on the right-side pectoral-fin base-were taken from a subset of specimens and stored in absolute ethanol in the tissue collection of the Laboratório de Evolução e Genética Animal, Universidade Federal do Amazonas (UFAM:CTGA). Fishes to be retained as vouchers were then fixed in 10% formalin before being transferred to 70% alcohol for long-term storage in the fish collection of INPA. Other institutional abbreviations follow Sabaj Pérez (2013).

Description of collection sites
Our survey of the Rio Nhamundá was undertaken at seven main sampling locations on the lower reaches of the river, and these are shown in Fig. 2. Two distinct environments were observed in our survey area: (1) a narrower, sinuous, and faster flowing upstream section encompassing sampling sites NH05, NH08, and NH04 (Fig. 3a); and (2) a wider, slowmoving, lake-like lower section of the river encompassing sampling sites NH12, NH02, and NH01 (Fig. 3b).
Sampling site NH01 comprised a small rocky promontory of a black lateritic conglomerate layer forming part of the island town of Nhamundá (Fig. 4a). The rocks were broken up into large and small boulders overlying a sand/sediment substrate (Fig. 4b), and numerous holes were present in the rocks. Water temperature was 32.0°C (daytime), and water movement was minimal. This specific habitat did not appear to be common in the immediate area, at least on the basis of the amount of rock exposed above the water. Sampling site NH02 comprised a floating wooden pen made by ornamental fishermen to Location of sampling sites on the Rio Nhamundá. Map was created with the R package ggmap (Kahle and Wickham 2013). house discus before resale, with terrestrial plants overhanging to provide shade for the discus. Sampling site NH12 comprised a rocky outcrop at the base of a hill (Fig. 4c). Rocks at NH12 were of a soft red lateritic type, and formed piles of large and small boulders above and below the water. Substrate at this site otherwise comprised fine sand and sediment, and water movement was minimal. Sampling site NH10 comprised a sandy beach of the Rio Paratucu, a right-hand tributary of the lower Rio Nhamundá.
Sampling site NH04 comprised a fairly typical representation of the upstream section of the river, mainly characterised by the presence of woody debris (Fig. 4d), i.e. fallen trees of varying states of decomposition (Fig. 4e). The river margin substrate was clay/soil, and water movement was moderate to strong depending on how exposed the trunks were to the main river current. Water temperature here was 28.9°C (daytime). Rocky habitats were uncommon in the upstream section of the river, with two exceptions being sampling sites NH05 (Fig. 4f) and NH08. These comprised two shallow areas of fast flowing water over a substrate of eroded bedrock and sand, forming a network of vertical fissures and crevices. Water temperature at NH05 was 28.3°C (night).

Measurements and meristic techniques
Measurements of preserved material were taken with dial callipers to a precision of 0.1 mm. All measurements and counts for symmetrical features were taken from the left side of the specimen. For loricariids, measurements and meristic counts follow Armbruster (2003), and terminology for lateral plate rows follows Schaefer (1997). Terminology of colour pattern follows Kottelat and Freyhof (2007).

Specimen identification
We used published literature to identify the specimens collected during our fieldwork, and we also provide here the characters found in those references that we used to make each identification. The identifications of the specimens from the INPA collection were also a b Figure 3.
Images of the Rio Nhamundá showing the two distinctive environment types.
a: Narrow upstream section near sample site NH05. b: Lake-like lower section at the town of Nhamundá, near sample site NH01.
cross-checked. When required, we examined photographs of type material on the All Species Catfish Inventory Web page (http://acsi.acnatsci.org).
We try to keep terminology consistent across species, but also attempt to maintain a degree of agreement with original authors' work in order to aid in referring back to their a b c d e f  characters. For example, when reporting the Farlowella, we refer to the "scutes" of Retzer and Page (1997) as "plates", following Schaefer (1997), but maintain reference to the "plates of second lateral row" of Retzer and Page (1997) instead of "mid-dorsal lateral plates" of Schaefer (1997).
We were conservative in applying names to the specimens collected, i.e. individuals were assigned to nominal species wherever possible, especially in cases where no modern literature for the group was available. In order to differentiate undetermined from undescribed species, we use consistent and informative tag-names following Leschen et al. (2009) for the latter, and simply "sp." for the former. If a putatively undescribed species could be determined as being allied to a nominal species, we used the "aff." qualifier in order to link the record to other collection events and species. For species with significant inter-individual variation and uncertainty in the identification, we used the "cf." qualifier, although we tried to avoid that wherever possible.

Molecular methods
DNA barcodes were generated for at least one individual per species. Methodology for DNA extraction, PCR and sequencing follows Colatreli et al. (2012), but using primers FishF1 and FishR2 (Ward et al. 2005). Chromatograms were assembled in Geneious (Biomatters 2014) and aligned manually using the translated amino acid sequence. The nucleotide data was uploaded to GenBank in accordance with their DNA barcode submission tool, and comprises accession numbers KP772569:KP772605.

Inventory of the catfishes of the Lower Rio Nhamundá
Specifically, our specimens differ from T. musaica as described by Royero (1992) (images can be found on the All Species Catfish Inventory Web page; http:// acsi.acnatsci.org/base/image_list.html?mode=genus&genus=Tatia) in at least two aspects: unpigmented area dorso-posteriorly to the orbit (slightly elongated and smaller than orbit diameter); and third nuchal plate larger and unpigmented (including some of the surrounding skin). Our examination of topotypic T. musaica material from the Río Atabapo of Colombia/Venezuela confirm these differences, in addition to a 5.2% COI sequence divergence (32 mutations over 615 bp). Our Nhamundá specimens are also similar to Tatia melanoleuca as described by Vari and Calegari (2014), but differ from this species mainly in respect to the distribution of dark pigmentation on the caudal fin, dorsal-fin spine, and lateral surfaces.
Sixteen individuals were caught either by hand from their lodgements in woody substrates (sampling sites NH04 and NH08), or more effectively using a hand net at the surface after attracting insects-on which they were feeding-with a light (sampling site NH12); under such conditions they appeared abundant on the lower Nhamundá. An example of a live specimen is pictured in Fig. 6. Four individuals were caught by hand from their lodgements in woody substrates at the margin of the main river (sampling site NH04). The species was also observed in rocky habitats (sampling sites NH08 and NH12), but were more difficult to catch in this situation. An example of a live specimen is pictured in Fig. 7.   (2011) based on the following characters: lower jaw prognathus; skull roof covered by thin integument; inner mental barbel reaching base of outer mental barbel; dorsal-fin spine serrated only along anterior margin; and caudal fin emarginate (our specimens had suffered damage to the lobes of the caudal fin, presumably due to piranhas).
Two individuals were caught at night using gill nets set in the margins of the main river. An example of a live specimen is pictured in Fig. 8.

Family Doradidae
Astrodoras asterifrons ( (2010) based on the following characters: flattened body shape; upper limit of pre-opercular canal not reaching lateral border of the cranium; seven branched rays in the dorsal lobe of the caudal fin; dorsal margin of the orbit high; procurrent rays of the caudal fin expanded into bony shields; and diverticula of the swim bladder simple, with tapered rear end.
Two individuals were caught by hand-net at night over a sandy/silty substrate (sampling site NH12). An example of a live specimen is pictured in Fig. 9.

Notes:
Identification to species level for this specimen could not be adequately made using morphological characters due to the small size of the immature specimen. However, rather than exclude the individual, we compared the DNA barcodes to the COI sequences presented by Arce et al. (2013) in their doradid phylogeny. The specimen clustered with, and was 1.36% divergent from (seven mutations in 516 bp), Scorpiodoras heckelii (GenBank KC555695), and is most likely conspecific given the known distribution of the species and its congeners (Sousa and Birindelli 2011).
One individual was caught by hand-net on the Rio Paratucu (sampling site NH10). This specimen is pictured in Fig. 10.   (1900) and Eigenmann (1912) based on the following characters: rounded caudal fin with larger lower lobe; distinct cranial fontanelle; maxillary barbels long, extending to caudal (extended only to caudal peduncle in our specimen); dorsal spine notched anteriorly; dark stripe along lateral line (in our specimen this comprised just a elongated blotch under the dorsal fin); base of caudal with dark bar; and obliquely angled dark saddle behind head (from dorsal insertion to base of opercle); and body and fins irregularly mottled.
One individual was caught by hand-net on the Rio Paratucu (sampling site NH10), and delivered a painful sting, confirming that many heptapterids are venomous (Wright 2009). This specimen is pictured in Fig. 11. Pimelodella sp. Notes: Tentative identification to genus level follows Eigenmann and Eigenmann (1890) and Eigenmann (1917) based on the following characters: occipital process narrow, reaching dorsal plate; fontanel continued to base of occipital process, with bridge above the posterior margin of the eye; dorsal-fin and pectoral-fin spines strong; humeral process spine-like; and dorsal fin with six branched rays.
Given the large diversity of the group, and the paucity of modern treatments dealing with heptapterids, we are currently unable to identify this fish to species level, and our genus identification is tentative. Important characters include the caudal fin lobes of approximately equal length, maxillary barbels not surpassing anal fin (left barbel is damaged in our specimen), the complete lack of dark longitudinal stripe, the wedgeshaped mark on the dorsal-fin, and the dark saddle anterior to the dorsal fin.
One individual was caught by hand-net on a sandy beach habitat (adjacent to sampling site NH05). This specimen is pictured in Fig. 12.   (2010) based on the following characters: flattened head with eyes placed ventrolaterally and visible from below; laterally expanded nuchal plate; six pairs of lateral abdominal plates posterior to coracoids; thoracic plates present; three midventral plates between cleithral posterior process and first plate of ventral series; anal shield composed of single plate; second infraorbital laterally contacting to two ventral dermal plates; patch of odontodes present on anterolateral aspect of cleithrum at opening to branchial chamber; and caudal fin with series of around three dark bands (irregular in our specimens).
Thirteen individuals were caught by hand from submerged terrestrial vegetation (sampling site NH02). An example of a live specimen is pictured in Fig. 13.   (1854) based on the following characters: three rows of lateral plates on the caudal peduncle; snout naked with fleshy tentacles lacking odontodes; 8-9 branched dorsal-fin rays (three individuals with nine rays and eight individuals with eight rays); and black colour with small white dots (apparent on body and fins in life, but only apparent on abdomen in preserved material). We note that Ancistrus, and particularly the Amazonian species, are a group in dire need of taxonomic revision.

Ancistrus dolichopterus
Eleven individuals were caught by hand on both woody (sampling site NH04) and rocky substrates (sampling site NH01). The species appeared abundant throughout the river. An example of a live specimen is pictured in Fig. 14 1912, A. macrophthalmus (Pellegrin, 1912, and A. ranunculus Muller, Rapp Py-Daniel & Zuanon, 1994-this fish is most similar in the shape of head and eyes to A. macrophthalmus and A. lithurgicus. However, the fish collected from the lower Nhamundá had just three branched anal-fin rays, compared to four for both of these species. More individuals will need to be collected, and further investigation of available names in Ancistrus carried out in order to discover if this indeed represents an undescribed species. Two individuals were caught by hand at night from rocky substrates (sampling site NH01). An example of a live specimen is pictured in Fig. 15. ( -Daniel (1985) and La Monte (1929) based on the following characters: lateral plates with welldeveloped keels; hypertrophied odontodes present along snout margin; three rows of lateral plates on the caudal peduncle; large interorbital distance; pronounced medial ridge on snout; head plates with sinuous rows of odontodes; interopercular plate with between 15 and 20 strong and distally hooked odontodes; pectoral spine roughly same length as head, with long odontodes; and spots on body roughly same size as those on head.

Dekeyseria scaphirhyncha
Nine individuals were caught by hand from woody substrates (sampling sites NH04 and NH08). In addition to the main river stem, the species was also observed in lake and igarapé habitats, and appeared abundant. It was not found in association with rocks. An example of a live specimen is pictured in Fig. 16.   (2008) we report the following combination of characters allowing a genus-level identification only: five rows of lateral plates on the caudal peduncle; lateral plates not keeled; dentaries forming angle of < 90°; dentary teeth almost twice as long as premaxillary teeth; dentary teeth widely spaced with medial gap between tooth cups as wide as the tooth cups themselves; supraorbital crests very distinct; dark E-shaped pattern on snout (irregular in some individuals); pattern on body of oblique dark bands (almost horizontal wavy stripes in some individuals), with dark bands generally wider than pale bands (i.e. body more dark coloured than pale coloured); dorsal fin with complete bands; and caudal fin with dark vertical bands.
The Hypancistrus from the Nhamundá is ostensibly similar to Hypancistrus furunculus Armbruster, Lujan & Taphorn (2007), but we hypothesise that it represents a distinct species from H. furunculus due to the pronounced supraorbital crests, wide gap between the dentary tooth cups, and colour pattern of wide dark bands and thin pale bands. We await the description of a number of similar species from the Brazilian Shield rivers, some of which may end up being more closely related to this fish than H. furunculus is. The species is known in the aquarium trade as Hypancistrus sp. "L475" (Konn-Vetterlein and Haagensen 2015) Twelve individuals were caught by hand at night from rocky substrates (sampling site NH01). Four examples of live specimens are pictured in Fig. 17  Four individuals were caught by hand from woody substrates at the margin of the main river (sampling sites NH04 and NH08). An example of a live specimen is pictured in Fig. 18.  (2005) based on the following characters: dentaries forming angle of < 80°; spoon-shaped teeth (although not fully formed in this small specimen); widely-spaced large black spots on a light background; and a lack of longitudinal dark stripes.

Hypostomus macushi
One individual was caught by hand from woody substrates at the margin of the main river (sampling site NH08). An example of a live specimen is pictured in Fig. 19.    (2005) based on the following characters: three rows of lateral plates on the caudal peduncle; bar-shaped opercle; snout plates present; lateral plates not keeled; > 30 teeth per jaw ramus; body and fins (except dorsal and caudal) dark with small pale dots; plates not outlined with dark pigment; and caudal fin with darker lower lobe. We note that the whisker-like odontodes characteristic of the genus were not apparent in our juvenile specimens, but we did observe bifurcating tentacules on the interopercular plate, and combined with the other characters, are confident that the specimens belong to Lasiancistrus schomburgkii as proposed by Armbruster (2005).

Notes: Identification to species level follows Armbruster
Two individuals were caught by hand from woody substrates at the margin of the main river (sampling sites NH05 and NH08). An example of a live specimen is pictured in Fig. 20.  Three individuals were caught by hand from woody substrates at the margin of the main river (sampling site NH04). An example of a live specimen is pictured in Fig. 21 Armbruster (2008) based on the following characters: evertible cheek plates with > 10 hypertrophied odontodes; dentaries forming angle of < 90°; deep body, not dorso-ventrally flattened; lips lacking fimbriae; villiform teeth of equal size in premaxilla and dentary; colour pattern of four irregular dorsal saddles (first and second are usually combined in small specimens); dorsal and caudal fins without orange margin (in life); head plates not outlined in black; dark bar between eyes, e-shaped dark blotch on the snout (broken in some specimens); and fins with dark and light bands of approximately equal width (irregular in some specimens).

Leporacanthicus galaxias
An interesting observation is that our specimens exhibited variation in terms of the degree of abdomen plating and markings: adult specimens from collection points near a b c d e f Figure 22.
Peckoltia cf. vittata live colouration and variation in abdominal plating. the mouth of the river (sampling sites NH01, NH02 and NH12) showed complete abdominal plating and a vermiculated pattern (Fig. 22a, b); those from further up the river (sampling sites NH04, NH05 and NH08) lacked abdominal plating and associated colouration (Fig. 22c, d), with the exception of one individual that was partially plated with discrete spots (Fig. 22e, f). In light of this variation we were hesitant in referring all these individuals to Peckoltia vittata, and hence we use the "cf." qualifier. The three DNA barcodes we generated from the different phenotypes/sites, however, were identical.
Twenty-eight individuals were caught by hand from both woody substrates (sampling sites NH02, NH04, NH05, and NH08) and rocky substrates (sampling sites NH01 and NH12  Armbruster and Provenzano (2000) and Lujan and Birindelli (2011) based on the following characters: three rows of lateral plates on the caudal peduncle; lateral plates without pronounced keels; fully plated snout lacking tentacles; hypertrophied odontodes on the snout in both females and males; lack of whisker-like odontodes on the cheek plates; extremely hypertrophied odontodes on an elongated pectoral-fin spine in both females and males; sucking disk without fimbriae; and greater than 20 teeth per jaw ramus.
Using the above references we were unable to identify the specimens to species level, as the individuals differed from the colour patterns diagnostic of the other species in the genus. We hypothesise that these individuals represent a new species of Pseudolithoxus, and intend to document this in more detail in a separate publication.
Seventeen individuals were captured by hand from crevices in rocky habitats exposed to strong current (sampling site NH08). Live colouration is shown in Fig. 23.
The above characters are consistent with F. nattereri, but some key differences in colour pattern are noted. Retzer and Page (1997) report: for most specimens of F. nattereri, the first anal and dorsal fin rays are entirely darkly pigmented (our specimen has spotted rays); a distinct dorso-lateral dark-stripe is present from base of snout to dorsal fin (this stripe was not apparent in the preserved specimen, but was observed in life); and upper and lower caudal fin lobes pigmented with dark stripes of equal size, with stripes often not reaching caudal fin base (the stripes in our specimen reached the caudal base). Retzer and Page (1997) recognise that F. nattereri probably comprises a complex of species.
One individual was caught by hand from shallow, fast flowing water over a rocky/sandy substrate (sampling site NH05). The live specimen is pictured in Fig. 24 (caudal fin in Fig. 25) Figure 24.    Isbrücker (1981), Thomas and Rapp Py-Daniel (2008) and Thomas and Sabaj Pérez (2010) based on the following characters: elongate lip filaments; three premaxillary teeth per ramus; premaxillary teeth approximately twice as long as dentary teeth; developed odontode crests on head and dorsal trunk plates; 34 lateral plates; 19 coalesced lateral plates; dorsal fin spine not elongated (24% of SL); post-orbital notch relatively well developed; abdomen mostly plated (with the exception of an anterior v-shaped naked area over the pectoral girdle); and all fins except anal fin with dark sub-distal bands (most prominent on caudal and dorsal fins). Identification of this species is tentative, as there appears to be considerable variation in the L. cataphracta group. We await the forthcoming systematic revision of the genus (as mentioned in Thomas and Sabaj Pérez 2010).
One individual was caught by hand from shallow, fast flowing water over a rocky/sandy substrate (sampling site NH05). The live specimen is pictured in Fig. 26.

Pseudoloricaria laeviuscula (Valenciennes, 1840)
•  Isbrücker and Nijssen (1976) and Covain and Fisch-Muller (2007) based on the following characters: lower lip bilobate with median furrow; whip on upper caudal spine absent; abdomen covered with small plates lacking organisation; elliptical area of abdominal plates at level of pelvic girdle absent; rostrum not strongly pronounced; pelvic-fin spine longer than last pelvic-fin branched ray; colouration comprising dark dots (except ventral surface and anal fin); lower lobe of caudal darker than upper; and basicaudal spot present in juveniles.
Three adult individuals were caught by hand-net at night over a sandy/silty substrate (sampling site NH12), and four juveniles were caught further upstream on the sandy margins of the river (sampling site NH05). An example of a live adult specimen is pictured in Fig. 27.
One further specimen record of this species was obtained from the fish collection at INPA; this individual had been identified as P. aff. laeviuscula, but we include it here under P. laeviuscula until further information becomes available.

Rineloricaria lanceolata (Günther, 1868)
•  Fichberg and Chamon (2008) based on the following characters: postorbital notch present; inferior lip with short, round papillae; teeth on dentary larger than premaxilla; four rows of lateral plates; all fins with a broad longitudinal dark band parallel to the first rays (fins almost entirely dark in our specimen); lower lip margin with long fringes; and dorsal surface of head and predorsal region with two longitudinal dark bands. Note that the characteristic dorsal breeding odontodes of R. lanceolata were not visible in this single specimen (probably female).
Two individuals were caught by hand from shallow, fast flowing water over a rocky/ sandy substrate on the main river (sampling site NH05). A live specimen is pictured in Fig. 28.

Family Pimelodidae
Calophysus macropterus (Lichtenstein, 1819)  (1890) based on the following characters: two rows of maxillary teeth (posterior row very small and hidden within the skin folds) and one row of dentary teeth; first dorsal and pectoral rays not spinous; adipose fin long; barbels flattened; upper jaw slightly prognathous; and dark spots on flanks and adipose fin.
Five individuals were caught after being attracted to the boat by suspending a deadfish bait in the water (sampling site NH12). An example of a live specimen is pictured in Fig. 29.   (2003) and Mol (2012) based on the following characters: dermal bones of the skull coarsely sculpted with reticulated ridges surrounding rounded pits; supraoccipital process greatly expanded laterally; anterior nuchal plate enlarged; colour pattern with dark upper and white/yellow lower parts of flank; and caudal fin bright red/orange.

Notes: Identification to species level follows Lundberg and Aguilera
One individual was caught by baited hand-line in a deep pool of the main river. The live specimen is pictured in Fig. 30. Four individuals were caught at night using gill nets in a lake connected to the river. An example of two live specimens is pictured in Fig. 31.

Batrochoglanis villosus (Eigenmann, 1912)
•  (2005), Mees (1974) and Eigenmann (1912) based on the following characters: lower jaw not projecting beyond upper jaw; head large and rounded in dorsal view; head and body with numerous small papillae; insertion of pelvic fin though vertical of posterior base of dorsal; short caudal peduncle; rounded caudal fin; post-cleithral process short, not reaching vertical through dorsal fin origin; axillary pore absent; lateral-line canal terminating on caudal peduncle; premaxillary dentigerous plates with lateral margins posteriorly prolonged; colour brown, with dark mottled appearance; and caudal fin with dark dots irregularly distributed.
An important inconsistency should be noted regarding the caudal fin shape. Eigenmann (1912) described the caudal fin of B. villosus as being "notched", but figured a specimen with a rounded caudal fin (Mees 1974). The photograph of the  Four individuals were caught by hand from their lodgements in woody substrates at the margin of the main river (sampling site NH04). An example of a live specimen is pictured in Fig. 32.

Analysis Desk survey
Our survey of the online databases revealed a single species record from the Rio Nhamundá: an unidentified Ossancora (Doradidae) which on further investigation was found to comprise two paratype lots of Ossancora asterophysa Birindelli & Sabaj Pérez, 2011 (ROM 88244, MZUSP 7838  The collection of fishes at INPA held nine lots of catfishes previously collected from the Nhamundá. One of these records (Centromochlus sp., INPA 35087) was also recorded in the literature survey (Sarmento-Soares et al. 2013). As stated previously, because this lot was out on loan at the time of writing this manuscript, we were not able to examine the specimens and have therefore omitted this species from the total species count due its ambiguity.
In addition to the species listed above and available in GenBank, we provide COI data for 17 putative species with no current sequences apparently available in either GenBank or BOLD:

Inventory of Siluriformes
Our desk and museum collection surveys show that few ichthyological surveys of the Rio Nhamundá have taken place, and that our collection is one of the first to be made on the river. Not including the unverified Centromochlus sp., we report three species from the desk survey, eight species from the museum survey, and 28 species from our field survey; three species (Tatia aff. musaica, Peckoltia cf. vittata, Pseudoloricaria laeviuscula) were recorded from both the museum and field surveys. Therefore, a total of 36 siluriform species are currently known from the river. Of these, one we were unable to determine ( Pimelodella sp.), and four we suggest could represent undescribed species ( Tatia aff. musaica, Ancistrus sp. "INPA 43862", Hypancistrus sp. "INPA 43863" and Pseudolithoxus sp. "INPA 43888").
The checklist is far from complete, however, missing entirely families such the callichthyids and aspredinids, although we sampled habitats most likely to yield these groups, such as igarapés, only superficially. Despite this, we feel that rapid publication of even modest datasets and small-scale surveys can make valuable additions to biodiversity science in the Amazon by both presenting the early discovery of new species and the extension of ranges for known species. The study also highlights the paucity of neotropical ichthyological data in GBIF/GenBank, and makes a small contribution.

Biogeography
The biogeographic composition of the collection is interesting. While many of the species recorded were cosmopolitan inhabitants of the lowland Amazon-e.g.  Chamon 2007), and apparently also the Rio Aripuanã, an affluent of the Rio Madeira (Seidel and Evers 2005). Thus, the ichthyofauna of the Rio Nhamundá appears to have some interesting affiliations to several faunal areas. This is perhaps exemplified by the best known fish from the river, the discus (Symphysodon spp.). Specimens exported from this locality are famous in the aquarium trade for their phenotypic variability (Bleher 2006). Genetic groups of both the Heckel (S. discus) and brown (S. aequifasciatus) discus species are found in sympatry in the Rio Nhamundá, an area that may in fact comprise a zone of natural hybridisation between the two species (Amado et al. 2011).
In summary, the Rio Nhamundá may therefore be an important sampling location for uncovering biogeographic patterns of fishes of the Lower Amazon and Guiana Shield. Of more urgent importance, however, is the sensitivity of freshwater ecosystems to anthropogenic activities and the requirement for catchment-specific conservation plans (Castello et al. 2013