Fishes of the Mitaraka Mountains (French Guiana)

ABSTRACT The “Our planet reviewed” expedition allowed to investigate the fish fauna from the Mitaraka Mountain Range (French Guiana). We sampled fishes at 14 sites using complementary sampling methods including rotenone, underwater observation, seine, cast nets, traps and hooks. We present the first detailed fish inventories from this region. Thirty eight species belonging to 16 families were observed, and at least one unknown species belonging to the genus Jupiaba Zanata, 1997 was collected. At two torrential sites we recorded an unusual fish assemblage made of a single Trichomycteridae species, Ituglanis nebulosus de Pinna & Keith, 2003. A few rare species were also recorded such as Pimelodella procera Mees, 1983 hereby extending its known distribution 100 km Southward. Anablepsoides gaucheri (Keith, Nandrin & Le Bail, 2006) was collected for the first time since its description in 2006. This species was collected in a stream located 15 km eastern from its type location, revealing its distribution might span over the entire Mitaraka range. Overall, the species turnover between sites was high even between sites with similar environmental conditions, suggesting that dispersal limitation is playing a significant role in fish assemblage composition. This indicates that complementary inventories in nearby areas may still provide new species and original information on the fish fauna inhabiting small streams of Southern French Guiana.


INTRODUCTION
Our knowledge of the freshwater fi sh fauna of French Guiana has benefi tted from intensive fi sh inventories for more than 50 years. Th e seminal works (e.g. Puyo 1949;Planquette et al. 1996;Boujard et al. 1997;Keith et al. 2000;Le Bail et al. 2000), were recently updated (e.g. Le Melki 2016;Covain et al. 2012;Fisch-Muller et al. 2018) and the fi sh fauna of French Guiana is among the best known in the Guiana Shield. Such taxonomic knowledge, together with intensive fi eld sampling occasions, led to the development of ecological approaches on the structure of fi sh communities (Cilleros et al. 2016(Cilleros et al. , 2017 and their responses to anthropogenic stressors (Brosse et al. 2011;Allard et al. 2016). However, gaps in our knowledge remain since access to the southern half of the territory remain particularly diffi cult due to the absence of roads or tracks. Th is is an area of c. 30 000 km² of hilly and mountainous terrain covered by almost undisturbed tropical rainforest. Access to this zone is only possible by boat, and the fi sh fauna from most of the small streams of southern French Guiana, particularly those located far from navigable waterways, remains largely unknown. To remedy this defi ciency, several research projects were initiated during the last decade to inventory the fi sh fauna of streams throughout French Guiana, leading to set an upstream-downstream fi sh typology for small streams (Cilleros et al. 2017). Th is typology identifi es fi ve successive types of stream fi sh assemblages from the source to the confl uence with a larger river (i.e. more than ten meters wide and one metre deep). In this typology, the upstream zone account for the mountainous streams hosting only a few specialised fi sh species. Studies focussing on this particular zone revealed a strong fi sh micro-endemism Brosse et al. 2013), but were limited to a few mountains due to access limitations. In this context, acquiring data from the Mitaraka Moutains, located in the southernmost region of French Guiana has a particular interest since: 1) this remote area of steep sloped granitic hills (also called inselbergs) has never been sampled for entire fi sh communities; and 2) studies on fi sh assemblages from Guiana Shield mountains remain particularly scarce, and the works of Mol et al. (2007) and Fisch-Muller et al. (2018) on Lely Nassau and Brownsberg Mountains (Suriname) and Brosse et al. (2013) on Itoupé mountain (French Guiana) remain, to our knowledge the only fi sh inventories on such kind of environment. Here we report on the fi sh assemblages from 14 environmentally contrasted sites in the Mitaraka Mountains and analyse faunistic diff erences between sites in relation to the environmental characteristics of the streams.

MATERIAL AND METHODS
Th e Mitaraka Mountains is a series of low altitude hills (not exceeding 600 meters a.s.l.), which constitute a part of the Tumuc-Humac mountain range (Hurault 1973). Th ey constitute the Southeastern boundary between French Guiana and Brazil, and therefore the watershed divide between the Amazon Drainage in Brazil and the Maroni Drainage in French Guiana. Th ese mountains are covered by dense rainforest, with the exception of their summit, where the granitic substratum is apparent, resulting in rocky islands emerging from the rainforest, also called inselbergs (Fig. 1). Th e southern slope of the Mitaraka Mountains is drained by several small streams fl owing to the Alama River, a tributary of the Maroni. Due to the remoteness of the area, access to the base-camp (located close to site S11, Fig. 1) was only possible by helicopter, and Fishes of the Mitaraka Mountains all the equipment was hand carried from the base-camp to the sampling sites, a reason why we were unable to sample more sites or cover a longer distance to fi nd distant sampling sites. Th e fi sh samples were collected in March 2015. We sampled 14 sites ( Fig. 1) belonging to the Alama main stream (site S11) and its tributaries (the remaining 13 sites). All the sites were GPS referenced and altitude was measured using a Garmin GPSMAP62. We measured the site physical characteristics by measuring the stream width, water depth and current velocity along three transects. Current velocity was estimated by measuring the time necessary for a fl oating object to cover a distance of one meter. Th e roughness of the stream granulometry was visually estimated as the percentage of large-sized pebbles and boulders (above 10 cm diameter). We also measured pH and water temperature with a WTW pH 3110 fi tted with a WTW pH-Electrode SenTix 41. Turbidity was measured with an Eutech Instruments Tubidimeter TN-100, and water conductivity was measured with a WTW 3110 conductometer fi tted with a tetraCon 325 sensor.
Fishes were collected using various sampling techniques in relation to the stream/site environmental characteristics, to maximize the probability to collect the entire fauna. At all stream sites with the exception of the Alama main stream (site S11) fi shes were sampled using rotenone. At each site, a river section was isolated using two fi ne mesh (4 mm) stop nets. Th e length of each section was proportional to stream width and was on average 16.35 ± 5.85 m (mean ± SD). Fishes were collected after releasing a small quantity of rotenone (PREDA-TOX®: a 6.6% emulsifi able solution of rotenone extracted from Derris elliptica (Wallich) Benth. by Saphyr, Antibes, France) a few meters upstream of the fi rst net. Fishes were then collected with fi ne meshed dip nets (2-mm mesh). At the end of each sampling session we searched for fi shes lying on the bottom or hidden in the leaves and debris. Th is sampling method has been described in detail by Mérigoux & Ponton (1999), and has been proved effi cient to collect quantitative samples of fi sh assemblages in small Neotropical streams ( Mérigoux & Ponton 1999). Nevertheless, as the section sampled by rotenone might not be representative of the overall stream habitat variability, rotenone samples were complemented by dip netting at all sites and by visual underwater observations at the sites S6, S8 and S12. In the latter sites we snorkelled diff erent habitats to search for additional species. Finally, in site S11, the river was too large (> 10 m wide) and too deep (> 1 m deep) to use rotenone and we therefore ran day and night snorkelling observations and complemented these samples using dip and seine netting, baited and unbaited traps, and angling. Th e entire sampling protocol was approved by the Direction of Environment of the French ministry of Environment (DEAL) and the French Guiana National Park (Parc amazonien de Guyane). All the specimens collected benefi ted from the Access and benefi t sharing agreement of the "Our Planet Reviewed" program (APA 973-1) (Touroult et al. 2018). We collected 328 fi sh specimens that were identifi ed in the fi eld with Planquette et al. (1996 and Le . Taxonomy was then updated according to Le  and Melki (2016). For specimens of particular taxonomic interest, tissue samples were taken and stored in 96% ethanol for molecular analyses and the specimen was then fi xed in 5% formaldehyde solution for taxonomic confi rmation.
Fish data were compiled for each sampling site, and fi sh abundance was converted to occurrences in order to combine quantitative (rotenone) and qualitative (snorkelling, nets, traps) data and get species lists from the 14 sites. To ordinate sites according to their fi sh fauna, we ran a Principal Component Analysis (PCA) on fi sh species lists. Presence-absence data was here preferred to abundance data because sampling methods diff er between sites. Th e sites S6 and S7 were excluded from the PCA because their fauna was made of a single species that was not encountered in any of the 12 other sites. Family Species Code S4 S5 S12 S13 S14 S3 S8 S9 S10 S11 S1 S2 S6 S7

RESULTS
Th e altitudinal range of the 14 sites varies from 275 to 410 m a.s.l. Th e sites are characterized by a low conductivity that never exceeded 17.3 μS.cm-1 and a turbidity lower than 8.5 NTU. Temperature and pH were also homogenous among sites and ranged from 23.1 to 23.8°C and between 4.52 and 5.73, respectively. Sites nevertheless diff ered in their size and slope (Table 1). Two sites were torrential streams with current velocities higher than 0.30 m.s-1 and coarse bottom substratum made of large boulders (site S6) or pebbles (site S7). All the other sites were characterised by a low current velocity (<0.25 m.s-1) and with the exception of site S12, a bottom substratum dominated by fi ne particles (sand, silt). Site S12 is a side arm of the Alama River, fl owing on a granitic substratum, explaining its coarse substratum size despite its limited slope and current velocity. Finally, two groups of low current velocity sites can be distinguished according to their width and depths. Sites S4, S5, S12, S13 and S14 were small (width = 1.30 ± 0.40 m) and shallow (depth = 9.00 ± 1.22 cm); whereas sites S1, S2, S3, S8, S9 and S10 were larger (2.96 ± 1.84 m) and deeper (27.51 ± 15.10 cm). Finally, S11 is the Alama mainstream and it is therefore markedly larger and deeper than the other sites (Table 1). Considering fi sh fauna, we collected 38 species in 16 families. Four families were represented by more than two species, namely Characidae (six species), Heptapteridae (six species), Loricariidae (four species) and Crenuchidae (three species). Among the 38 species, two are only known from the Mitaraka Mountains (Anablepsoides gaucheri Keith, Nandrin & Le Bail, 2006 and Guyanancistrus megastictus Fisch-Muller, Mol & Covain, 2018), two (Characidium sp. and Copella aff . arnoldi) are still unnamed but known species ) and another one (Jupiaba sp.) is a new record for French Guiana. We were not able to identify this last species and it might be new to science. Th e fi sh assemblages are characterised by a low species richness ranging from one to 23 species according to the sites, and a marked species turnover between sites. Indeed, only two out of the 38 species occurred in more than half of the sites (Gymnotus carapo Linnaeus, 1758 and Hemibrycon surinamensis Gery, 1962), and 12 species occurred in a single site. Moreover, two sites (S6 and S7) were characterised by the presence of a single species (Ituglanis nebulosus de Pinna & Keith, 2003), that only occurs in those sites.
Th e following list of the species observed in the Mitaraka Mountains area follows the arrangement in Le  and provides ecological observations on the species. In accordance with Our Planet Reviewed taxonomic agreement, fi sh specimens are stored at MNHN. Th e following list of species is summarized in the Table 2  OCCURENCES. -S8, S11.
ECOLOGY. -Observed during underwater inventories in rapid and rocky areas.
OCCURENCES. -S12. ECOLOGY. -Th is rare species was only observed in a side arm channel of a rapid area of the Alama River. It was found in small rocky pools with low current velocity. Captured using rotenone.
ECOLOGY. -Only observed in the Alama mainstream. Nocturnal species that hides during the day. Captured using rotenone and observed in underwater inventories.
OCCURENCES. -S4, S5, S11, S12, S14.  OCCURENCES. -S4, S5, S9, S10, S11, S12, S13, S14. OCCURENCES. -S8, S9, S10, S11. Order CYPRINODONTIFORMES Berg, 1940Family RIVULIDAE Myers, 1925Genus Anablepsoides Huber, 1992 Anablepsoides gaucheri (Keith, Nandrin & Le Bail, 2006) (  OCCURENCES. -S2, S4, S11, S14. OCCURENCES. -S11, S12. Th is dispersal limitation eff ect is also confi rmed by the marked species turnover between sites, even within a similar environmental context. For instance, the small and shallow sites (depth ≤ 11 cm and width ≤ 2 m; S4, S5, S12, S13, S14) have a dissimilarity of 77.2% ± 8.4% (mean ± SD, Jaccard dissimilarity index). We identifi ed four types of fi sh assemblages in relation to the physical characteristics of the streams. Th e two torren-tial streams, S6 and S7 (type 4), both located on the same tributary, are characterised by an unusual fi sh fauna made of a single Trichomycteridae species (I. nebulosus). Although this species has already been reported from both fl oodplain and mountain headwater streams (De Pinna & Keith 2003;Mol et al. 2007;Brosse et al. 2013), it is unusual to record a stream assemblage made of a single species. Th is is quite surprising since the torrential hydrology of this stream (see Fig. 24 for a picture of the stream) appears as a favourable habitat for a few other species as predicted by the stream fi sh typology of Cilleros et al. (2017). Moreover, Mol et al. (2007) and Brosse et al. (2013) reported the presence of several Loricaridae (e.g. Guyanancistrus and Harttiella), and/or Rivulidae (e.g. Anableposides or Laimosemion) from similar torrential streams in French Guiana and Suriname. In contrast, small and shallow streams with lower current velocities (type 1) were colonised by more diverse assemblages and host 6.80 ± 3.96 species. Th ose species are able to inhabit shallow waters (e.g. Lamosemion geayi; Keith et al. 2000) and to survive to temporary drying by air breathing while sheltering in relictual water pools (e.g. G. carapo, H. malabaricus Liem et al. 1984;Rantin et al. 1992). Larger and deeper streams (type 2 and 3) have a slightly more diversifi ed fauna with 8.00 ± 4.69 species. Th ose sites are dominated by characids that are more sensitive to anoxia. Finally, the Alama main stream has a more diverse fauna with 20 recorded species, including most of the species found in the above discussed large and deep tributaries, revealing an accumulation of fi sh species according to the distance from the source, as already reported by Cilleros et al. (2017).

SPECIMENS
Among the fi sh assemblages, a few rare or unknown species were recorded, such as G. megastictus, a recently described Loricariidae species (Fisch-Muller et al. 2018) which distribution appears limited to the Mitaraka Mountains. We also recorded P. procera, an endemic species from the Maroni Basin. Its previously recorded southern limit of distribution was the Nouvelle France River (Brosse et al. 2018). Th e occurrence of P. procera in the Mitaraka streams therefore extends its known spatial distribution of more than 100 km to the south. We also collected A. gaucheri in two localities (S1 and S2), while the species was only known from a few Alama drainage tributaries located 10 to 15 km to the west of the study sites (Keith et al. 2006). Th is species, only known from the Mitaraka Mountains, might be endemic from the Alama Drainage. Although probably having a restricted distribution, it occurs in several tributaries of the Alama River. Finally, three adult individuals from a large (c. 60 mm SL) unknown species belonging to the genus Jupiaba were collected in site S2. Th e description of this species is currently pending.
To conclude, the fi sh fauna of the Mitaraka Mountains, although not very diverse, includes some micro-endemic species and unusual fi sh assemblages. Complementary samples in the southern part of French Guiana are thus needed to complement the taxonomic knowledge of the fi sh fauna of French Guiana and to document the structure of Guianese mountain streams fi sh assemblages.