Life in the spray zone – overlooked diversity in West African torrent-frogs (Anura, Odontobatrachidae, Odontobatrachus)

West African torrent-frogs of the genus Odontobatrachus currently belong to a single species: Odontobatrachus natator (Boulenger, 1905). Recently, molecular results and biogeographic separation led to the recognition of five Operational Taxonomic Units (OTUs) thus identifying a species-complex. Based on these insights, morphological analy ses on more than 150 adult specimens, covering the entire distribution of the family and all OTUs, were carried out. Despite strong morphological congruence, combinations of morphological characters made the differentiation of OTUs successful and allowed the recognition of five distinct species: Odontobatrachus natator, and four species new to science: Odontobatrachus arndti sp. n., O. fouta sp. n., O. smithi sp. n. and O. ziama sp. n. All species occur in parapatry: Odontobatrachus natator is known from western Guinea to eastern Liberia, O. ziama sp. n. from eastern Guinea, O. smithi sp. n. and O. fouta sp. n. from western Guinea, O. arndti sp. n. from the border triangle GuineaLiberia-Côte d’Ivoire. In addition, for the first time the advertisement call of a West African torrent-frog (O. arndti sp. n.) is described.


Introduction
For a long time all West African torrent-frogs have been assigned to the genus Petropedetes Reichenow, 1874, until their generic distinctiveness from Central African species West African torrent-frogs are nocturnal, inhabit lotic waters and usually occur close to streams with strong currents, cascades or rapids in forested areas. However, Rödel (2003) also collected specimens in gallery forests surrounded by savannah in Mont Sangbé National Park, Côte d'Ivoire. While females are usually found in close proximity to rapids and waterfalls, males may sit on rocky surfaces further away (Rödel 2003). Tadpoles are well adapted to life in torrents. With a dorso-ventrally depressed body and sucker-like mouthparts with enlarged labials, they live attached to rocks in strongest currents or adhere to rock-surfaces in the spray zone. These special mouthparts are only reduced at the very last stages of metamorphosis, which is typical for rheophilous larvae (Lamotte and Zuber-Vogeli 1954;Guibé and Lamotte 1958;Channing et al. 2012). The adult frogs are characterised by a medium to large body length (females reaching > 60 mm snout-urostyle length), the possession of dilated, heart-shaped toe tips, a rough dorsal skin texture with glandular ridges, mandibular fangs in both sexes and femoral glands in males (Boulenger 1905; Barej et al. 2014a). These frogs have a patchy distribution within the Upper Guinea forest region, roughly ranging from western Guinea through Sierra Leone and Liberia to western Côte d'Ivoire (Boulenger 1905;Guibé and Lamotte 1958;Böhme 1994b;Rödel 2003;Rödel et al. 2004a;Hillers and Rödel 2007;Hillers et al. 2008a).
Since the first description, West African torrent-frogs have been regarded as a single species: Odontobatrachus natator (Boulenger, 1905). Although inter-population differences in colouration and shape of dorsal glands have been reported Rödel et al. 2004a), this has not resulted in taxonomic actions. Based on molecular data, five distinct lineages (therein treated as Operational Taxonomic Units, OTUs) were recognised in this supposedly monospecific family, indicating hitherto overlooked cryptic species (Barej et al. 2015). All samples from Sierra Leone, the type locality of O. natator, are grouped in a single clade which has consequently been assigned to the nominate taxon. Two OTUs occur in the westernmost and two more in the easternmost distribution of the family Odontobatrachidae. While the largest area is occupied by the nominate taxon O. natator, OTUs show a tendency to parapatric distribution with little overlap in their potential distribution areas; exceptionally the two western OTUs possess a similar range according to modelled distribution. The recognition of a potential species-complex in the presumably monospecific frog family Odontobatrachidae demonstrated that the current threat classification as "Near Threatened (NT)" (IUCN 2011) is insufficient as recognised OTUs possess very small distribution ranges, demanding a higher threat classification. However, a reassessment of threat categories and subsequent conservation actions require formal description of new species.
We herein present morphological results gathered from more than 150 specimens, covering the entire geographic distribution of the family Odontobatrachidae. Morphological characteristics were analy sed and interpreted in combination with the published molecular data and biogeographic insights after Barej et al. (2015). Consequently, we re-describe Odontobatrachus natator (Boulenger, 1905), describe four new species, and provide the first call analysis for Odontobatrachus.

Species concept and species delimitation
We herein follow the General Lineage Concept of species (de Queiroz 1998(de Queiroz , 1999) and accept distinctiveness on species level based on both morphological and genetic data. The genetic data have already been presented by Barej et al. (2015), indicating four undescribed candidate species and their relationships (compare Vieites et al. 2009). OTUs defined by Barej et al. (2015) were the basis of our morphological analyses. Consequently these molecular clades were taken as a priori group assignments to ensure understanding of the overall morphological character diversity within and between OTUs. Herein, we accept all five OTUs sensu Barej et al. (2015) and consequently four new species are described in the following. We re-describe the nominate species Odontobatrachus natator (blue colour code) and describe OTU1 as O. ziama sp. n. (red), OTU2 as O. smithi sp. n. (yellow), OTU3 as O. fouta sp. n.
(green) and OTU4 as O. arndti sp. n. (orange). For convenience we will use these names without the suffix sp. n. throughout the manuscript, anticipating their formal description below. Environmental Niche Models (see Barej et al. 2015) confirm the overall distribution pattern of the family. No major range extensions are expected and modelled niches of the individual species are very similar. Interestingly niche similarity identified two groups: O. smithi and O. fouta in the first and the remaining three taxa in the second. The distribution of all five OTUs and herein recognised species is illustrated in Figure 1.

Morphology
Collected frogs were anesthetized either with chlorobuthanol or benzocaine solutions and thereafter fixed in 4% formalin or 70% ethanol. All voucher specimens have been transferred to 75% ethanol for long-term storage. Abbreviations of museum collections hosting the investigated vouchers are as follows: The Natural History Museum (BMNH), London, United Kingdom; Natural History Museum of Geneva (MHNG), Geneva, Switzerland; Zoologisches Forschungsmuseum Alexander Koenig (ZFMK), Bonn and Museum für Naturkunde Berlin (ZMB), Berlin, both Germany.
Measurements follow standard procedures and were taken on preserved material with an electronic dial calliper (± 0.1 mm) by one person (MFB). Webbing formulae are composed as follows: dividing different toes by a '-' and differentiating inner and outer side of toe by a '/', thus the example 0-0.5/0-1/0-1/1-0 translates to: the webbing reaches the disc at toe I, webbing extends halfway between the most proximate tarsal tubercle to the disc at the external side of toe II and the disc at the internal side of this toe, etc. Tarsal tubercles are counted from tip of the toe to toe base. Additional qualitative characters such as skin granulation were recorded, but not always ascertainable in all vouchers, probably due to different preservation procedures.

Statistical analyses
Potential statistical discrimination of OTUs by morphological data was tested in SPSS 20 and R 3.1.0 (default packages; R Core Team 2013). We only included complete data sets in statistical analyses; damaged specimens or specimens with preservation artefacts were excluded. Furthermore, only measurements of adult frogs were taken into consideration. In order to consider sex-specific characters, e.g. femoral glands (present in males only) or size dimorphism (females growing larger than males), sexes were analysed separately.
Natural Log (ln) transformations were applied on measurements before analysis to obtain a homogeneous data distribution. Principal component analyses (PCA) were performed to explore the overall morphological variation between the putative taxa. Subsequently, we tested for significance of differences between OTUs with non-parametric tests (Kruskal-Wallis H test) since morphological datasets often violated the assumptions of standard parametric statistics and non-parametric tests are generally considered to be more conservative, not relying on assumptions such as random sampling, normality and homogeneity of variance (Anderson 2001 Finally, canonical discriminant function analyses (CDA) were performed on ln-transformed mensural variables to test whether our a priori groupings could be confirmed. These analyses maximised separations between groups based on within-group variance and correlation. CDA were again implemented on female and male datasets independently. Both the PCA analyses and the CDA were performed separately on absolute values and morphometric ratios.

Advertisement calls
Odontobatrachus call recordings were collected from specimens in terraria (vouchers collected on Nimba Mts., Guinea). Oscillograms (waveforms) and audiospectrograms (sonograms) as results of the Fast Fourier Transformation (FFT; frequency spectrum) were examined for spectral and temporal characters (analysis settings: 44.1 kHz sample ratio, 16 bits resolution, FFT length = 256). Call recordings were analysed with the software package Soundruler v0.9.6 (Gridi-Papp 2007), spectrograms and oscillograms were prepared with the software package Seewave for R (Sueur et al. 2008). Values of call duration, dominant frequency, fundamental frequency and number of notes are presented as minimum and maximum only, because of low numbers of recorded calls.

Genetics
Phylogeographic analyses included samples from the entire family range and were based on mitochondrial (12S, 16S, CYTB) and nuclear (BDNF, SIA, RAG1) genes (Barej et al. 2015). Herein, we present uncorrected 16S p-distances between species from Barej et al. (2015); a table providing inter and intra-species distances is provided in the Appendix 1: Table A. A list of samples gathered in addition to Barej et al. (2015) and respective GenBank numbers are given in Appendix 2: Table B. For laboratory procedures see Barej et al. (2014a).

Conservation status
Following IUCN Red List criteria, Barej et al. (2015) calculated the Extent of Occurrence (EOO) and the Area of Occupancy (AOO) using GeoCat (2011) for Odontobatrachus natator and four additional OTUs, herein described as new species. While EOO, often measured by a minimum convex polygon, corresponds to the contained area of a species, AOO refers to the area within the species EOO, excluding cases of unoccupied or unsuitable habitat. According to IUCN regulation, the higher of the two classifications is crucial for assessing global extinction risk.

Results and discussion
All specimens have been assigned to five OTUs a priori (OTUs at the putative species level following molecular results in Barej et al. 2015). Phenetic differences of all OTUs were assessed by carrying out a principal component analysis (PCA) on the respective datasets of a total of 65 males and 93 females. Due to the low number of available data points for O. smithi and O. fouta, placement of their centroids has to be regarded with caution.
PCA results of absolute values in male Odontobatrachus ( Fig. 2a) revealed that GL and GW contributed most to axis 1 and axis 1 accounted for 52.84% of the total variance. Axis 2 contributed an additional 23.23%, summing up to a total of more than 75% of the variance explained (Table 3). Axis 2 consisted mostly of contributions by EN, TD and GW. All contributors to axes 1 and 2 in the analysis of absolute values in males are given in Table 3. Axis 3 increased the total explained variance by only 6.60% (contributors not shown). PCA results of morphometric ratios in male Odontobatrachus ( Fig. 2b) revealed that main contributors to axis 1 were GL/FM, OD/EN, TD/OD and TD/SUL and axis 1 accounted for 41.28% of the total variance. Axis 2 contributed an additional 17.06%, summing up to a total of 58.35% of the variance explained (Table  4). The loading of this axis was mostly made up of contributions by GL/FM and TD/SUL. All contributors to axes 1 and 2 in the analysis of morphometric ratios in males are given in Table 4 (Fig. 2b).
PCA results of absolute values in female Odontobatrachus (Fig. 2c) revealed that IT and TD contributed mostly to axis 1 and axis 1 accounted for 59.02% of the total variance. Axis 2 contributed an additional     14.27%, summing up to a total of more than 73% variance explained (Table 3). Axis 2 was mostly made up of contributions by TD and EN. All contributors to axes 1 and 2 in the analysis of ratios in males are given in PCA results of morphometric ratios in female Odontobatrachus ( Fig. 2d) revealed that main contributors to axis 1 were TD/O and TD/SUL and axis 1 accounted for 48.38% of the total variance. Axis 2 contributed additional 17.84%, summing up to a total of 66.22% of the variance explained (Table 4). Axis 2 was mostly made up of IT/FL. All contributors to axes 1 and 2 in the analysis of ratios in females are given in Table 4. Axis 3 explained an additional 12.49% (contributors not shown).  Generally, PCA results on morphology supported the separation of the five molecular OTUs sensu Barej et al. (2015;O. natator, O. ziama, O. smithi, O. fouta, O. arndti), although males and females often showed overlap in the variance in both analyses (absolute values and ratios) between OTUs. Two major morphological groupings O. natator, O. smithi, O. fouta vs. O. ziama, O. arndti were uncovered in all analyses, with the groups being separated from each other on at least one of the two major axes (Fig. 2).
We tested for statistical differences in particular morphological characters and ratios between species (Kruskal-Wallis H test), considering potential sex-dependant characters (Appendix 3: Table C). These non-parametric tests revealed significant differences in males and females of the five species highlighting their morphological distinctness (Table 5).
Due to the overlap in morphological variation of species (see above), the correct assignment of single individuals would be difficult if their geographic origin is unknown. We therefore applied Detrended Correspondence Analyses (DCAs) to assess how reliably individuals can be assigned to a particular species. DCA results showed high levels of correctly assigned males and females, based on absolute values and ratios of mensural data sets, respectively (Table 6). Combined correct assignments of all five species for absolute values in male Odontobatrachus summed up to 89.2% and combined values of ratios referred to 87.7%. DCA results in female Odontobatrachus were 82.9% for absolute values and 74.2% in ratios. The lowest percentages of correctly assigned individuals were recovered in the species with the lowest voucher numbers (Table 6). Despite the high percentage of correct assignments, the persisting mismatches reflected the PCA results and highlight the morphological overlap between some species.
Based on the combination of the molecular data recognising five OTUs and their respective distribution patterns (indicating spatial partitioning) presented in Barej et al. (2015), as well as morphological distinction presented herein, we consider the five Odontobatrachus OTUs as distinct species. Until now, all Upper Guinean populations were assigned to Odontobatrachus natator (Boulenger, 1905) and thus no synonyms are available. We provide diagnostic characters (Table 7) and formally describe four new species.

Systematics of the Odontobatrachus natator-complex
Frogs belonging to the genus Odontobatrachus are all characterised by the following external morphological characters: tusk-like odontoids on the lower mandible in both sexes; posteriorly curved teeth on premaxillaries and anterior maxillaries; presence of vomerine teeth; eye diameter distinctly larger than tympanum diameter; pupil horizontally elliptical; tympanum rather indistinct; skin

Loadings (ratios)
Tibiofibula length to snouturostyle length ( Table 5. Morphological comparison of absolute measurements and ratios in males (lower left corner) and females (upper right corner). Number of samples per sex in each species and level of significance are given; lack of significant differences in absolute values or ratios is marked with "X". See material and methods section for abbreviations.  Table 6. Results from statistical discrimination of species (DCA) using morphological data and pooling individuals according to sex. Percentage of correct assignments, number of cases (N) and overall correct classification rate for absolute values and ratios in males and females are given. texture granular and heterogeneous; males with femoral glands, external vocal sacs, velvety nuptial excrescences on finger I. These characters apply to all species treated herein and are not repeated in the specific diagnoses below. For further osteological characters see Barej et al. (2014a, b).
During his service in western Africa, Captain (Local Major) F. Smith researched tropical diseases, prepared species lists of pests and elaborated respective preventive measures (Smith 1902(Smith , 1905. A part of his contribution contains the local fauna around barracks (Smith 1905) and Major F. Smith mentioned "a local frog (a new species named Petropedetes natator)…". He was based in Freetown predominantly surveying the area of Mt. Aureol, Tower Hill and Kortright but likewise carried out short travels to Port Lokkoh (today: Port Loko) and Rotifunk in the close hinterland (Smith 1902). However, Smith (1902) searched in the latter region for swampy areas as potential breeding habitats of the mosquito genus Anopheles, a habitat type inappropriate for torrent-frogs. Consequently, we herein restrict the type locality of Petropedetes natator Boulenger, 1905 to the Freetown area, Sierra Leone. A more detailed restriction appears unreasonable.
We refrain from designating a single lectotype as subsequent species descriptions are possible with comparison to the whole syntype series.
Genetics. Odontobatrachus natator is genetically well differentiated from all congeners and known populations form a well-supported and monophyletic clade (Barej et al. 2015). Uncorrected 16S p-distances between O. natator and other Odontobatrachus species range from 3.40-5.40% (Appendix 1: Table A), while maximum intrataxon differences of O. natator reach 1.98% (one-to-one pairwise comparisons N = 703), maximum intra-subclade difference values for the two subclades of O. natator are 0% (N = 1) and 0.72% (N = 630) respectively (Appendix 1: Table A). These two subclades correspond to the disjunct distribution of I) the Freetown area and II) all remaining localities further inland; divided by unsuitable habitat in-between (Barej et al. 2015). In case taxonomic changes are made in the future, the Freetown clade should retain the nominate form following the restriction of the type locality.
Description of male syntype. The male syntype (BMNH 1947.2.30.68) has been assigned to this taxon in both DCA analyses (absolute values and ratios). The male syntype has a robust body shape: snout-urostyle length of 46.1 mm; head width 17.0 mm; head slightly longer than broad; snout in lateral view short, slightly rounded at the snout tip (Fig. 3); snout in dorsal view fairly rounded; lower jaw with sharp tusk-like prolongations and single small knob at lower jaw symphysis with corresponding socket in-between premaxillae; upper premaxillae and maxillae with numerous teeth, posteriorly curved; vomerine teeth present, arranged in two small odontophores, closer to each other than to choanae; tongue broadly heart shaped; horizontal eye diameter 6.4 mm; interorbital distance 5.3 mm; pupil horizontally elliptical; eye diameter distinctly larger than tympanum diameter (Fig. 3); tympanum indistinct (horizontal diameter 2.7 mm); nares closer to snout than to eye; snout as long as eye diameter; canthus rostralis rounded; loreal region concave; paired lateral vocal sacs (Fig. 3); forelimbs moderately slender, forearms slightly hypertrophied, fingers slender; prepollex absent; relative finger lengths III>IV≥II>I (Fig. 3); velvety nuptial excrescences on finger I weakly developed; subarticular tubercles large, subconical; supernumerary tubercles absent; fingertips dilated, triangular, notched in the middle; femur length 23.2 mm; tibia length 23.8 mm; femoral glands large (length × width: left: 10.3 × 5.4 mm, right: 9.5 × 5.5 mm); femoral glands positioned on the posterior part of the ventral side of femur (Fig. 3); relation femoral gland length to femur length: 0.43; minuscule circular glands running along upper side of tibia; foot length (incl. longest toe) 29.9 mm; relative toe lengths IV>III≥V>II>I; inner metatarsal tubercle elliptical; toe tips broadened forming triangular dilated discs; inner metatarsal tubercle prominent (2.8 mm); number of subconical subarticular tubercles on toes I-V: 1, 1, 2, 3, 2; supernumerary tubercles absent (Fig. 3); prominent skin fold on posterior side of feet; dorsal skin texture heterogeneous; dorsum and flanks covered with slender dorsal ridges of app. 3.0 mm length (partially flattened on the dorsum due to preservation); venter smooth; flank texture as on dorsum; webbing fully developed (0-0.5/0-1/0-1/1-0), running as a skin fold along toes III and IV to the disc, webbing between toes hardly concave, almost straight.

Colouration in preservation.
Specimen overall brownish in colour (Fig. 3); dorsum darker than ventrum; throat darker than belly, ventrum lacking any marbling or patterns. Damage of the male syntype: third toe of left foot (in dorsal view) cut off (Fig. 3); left side (in dorsal view) with cut along flank; transverse cut on throat; discs on toes and fingers partially shrivelled due to drying-out. Variation in webbing formulae of examined specimens in the covered distribution range corresponds to the extent in the type series (Table 7); although skin folds running along toes in the male syntype are more distinct than in many other specimens, showing an almost fully extended webbing state. Dorsal ridges form either slender lines as in the male syntype (Fig. 3, see also Fig. 4b, c) or are short and knob like (Fig. 4a). Number of distinct dorsal ridges (counted from spine to flank) range between two and six, usually three to five ridges per body site. However, both characters were not recognisable due to preservation artefacts in many specimens. Glandular ridges on tibia are usually built of small to large conic glands and form more or less interrupted lines (Fig. 4a-e). Dorsal colouration (in life) varies from uniform brownish, to mottled patterns with greenish or light brownish background and darker spots, usually arranged along dorsum. Male femoral glands are rose-coloured but colouration may be attenuated by the ventral colouration (Fig. 4e). Belly colouration (in alcohol) ranges from completely pale, dirty whitish, dark throat and pale belly, dark with few pale markings, to entirely dark colouration, showing no sex-dependant colour differentiation.

Distribution.
Odontobatrachus natator has the widest distribution of all congeners (Fig. 1). The species is known   (Barej et al. 2015), one of them being restricted to the Freetown Peninsula in coastal Sierra Leone (FP sensu Barej et al. 2015) and the other covering all remaining localities (IL sensu Barej et al. 2015) of this taxon.
Conservation status. The EOO, combining both subclades of O. natator (Barej et al. 2015;therein natator) sums up to 180,231 km 2 , resulting in the IUCN Red List category "Least Concern (LC)". However, due to the habitat requirements of this family the AOO is restricted to 224 km 2 and thus classifies the species as "Endangered (EN)". When considering the genetic subdivision of O. natator (see Barej et al. 2015), the distribution areas further diminish dramatically, especially for the Freetown Peninsula subclade. While IUCN categories remain constant for the widely distributed subclade, the Freetown Peninsula subclade possesses an AOO of only 20 km 2 classifying it as EN and an EOO of 34 km 2 placing it as "Critically Endangered (CR)" if treated as its own taxonomic unit. Diagnosis. Medium sized frogs, robust body shape; head narrow, smallest tympanum diameter/eye diameter ratio in the family, webbing fully developed, leaving up to 0.5 of the distal phalange free at the inner side of toe II, leaving up to 0.5-0.75 of the distal phalange free at toe IV; male femoral glands dark orange; glandular lines on tibia contain minuscule to small conic glands forming a pretty continuous line, belly pattern highly variable. Genetically O. ziama differs by a minimum of 2.89% in the mitochondrial 16S gene from its congeners.

Differential diagnosis.
Odontobatrachus ziama can be distinguished from its congeners by a combination of characters (for all significant differences see  (Figs 4, 6, 8, 10). Glandular lines on tibia contain minuscule to small conic glands forming almost continuous lines (Fig. 6a-d), while small to large glands form more or less interrupted lines in O. natator (Fig. 4a-e), small to mean conic glands form predominantly interrupted lines in O. smithi (Fig. 8a, b), small to large glandular conic glands, rather interrupted lines in O. fouta (Fig. 10b, c). and similar to O. ziama small to mean glandular conic glands form hardly interrupted lines in O. arndti (Fig. 12b, c). Morphologically the species is most similar in size and colour pattern to O. arndti (Table 7); however, they differ in several mensural characters: male O. ziama have larger SUL, but smaller HW, TD, O and extremities (FM, TI, Genetics. The species is genetically well differentiated from all congeners and known populations form a well-supported and monophyletic clade (Barej et al. 2015). Uncorrected 16S p-distances between O. ziama and other Odontobatrachus species range from 2.89-5.41%, while maximum intrataxon differences of O. ziama add up to 0.38% (mean value 0.18%; N = 496; Appendix 1: Table A).
Holotype description. The male holotype has been correctly assigned to this taxon in both DCA analyses (absolute values and ratios). The holotype is an adult male with a moderately robust body shape (Fig. 5): snout-urostyle length of 46.1 mm; head width 15.5 mm; head slightly longer than broad; snout in lateral view short, flattened and slightly pointed at the snout tip; snout in dorsal view pointed; lower jaw with sharp tusk-like prolongations and single small knob at lower jaw symphysis with corresponding socket in between premaxillae; upper premaxillae and maxillae with numerous teeth, posteriorly curved; vomerine teeth present, arranged in two small odontophores, closer to each other than to choanae; tongue broadly heart shaped; horizontal eye diameter 6.5 mm; interorbital distance 4.8 mm; pupil horizontally elliptical; eye diameter distinctly larger than tympanum diameter; tympanum indistinct (horizontal diameter 3.5 mm); nares closer to snout than to eye; snout as long as eye diameter; canthus rostralis rounded; loreal region concave; paired lateral vocal sacs; forelimbs moderately slender, forearms slightly hypertrophied, fingers slender; prepollex absent; relative finger lengths III>IV>II>I; velvety nuptial excrescences on finger I weakly developed; subarticular tubercles large, subconical; supernumerary tubercles absent; fingertips dilated, triangular, notched in the middle; femur length 24.1 mm; tibia length 24.4 mm; femoral glands large (length × width: left: 12.7 × 7.6 mm, right: 12.8 × 7.5 mm); femoral glands positioned on the posterior part of the ventral side of femur; relation femoral gland length to femur length: 0.53; minuscule circular glands running along upper side of tibia; foot length (incl. longest toe) 32.0 mm; relative toe lengths IV>III≥V>II>I; shortest toe 5.5 mm; inner metatarsal tubercle elliptical; toe tips broadened forming triangular dilated discs; inner metatarsal tubercle prominent (3.7 mm); number of subconical subarticular tubercles on toes I-V: 1, 1, 2, 3, 2; supernumerary tubercles absent; prominent skin fold on posterior side of feet; dorsal skin texture heterogeneous; dorsum and flanks covered with slender dorsal ridges of app. 2.6 mm length (partially flattened, but recognisable as darker spots); venter smooth; flank texture as on dorsum; webbing fully developed (0-0.25/0-0.75/0-1/1.25-0); webbing between toes hardly concave. Damage to the male holotype: left femur (in dorsal view) with short cut; third toe of left foot (in dorsal view) clipped for tissue sample; glandular dorsal ridges partially not recognisable due to preservation. (Fig. 5). Dorsum brownish, marbled with small dark spots (partially indicating presence of former dorsal gland ridges), a pale marking between shoulders; hind limbs on upper side with large dark blotches, surrounded with blurred pale lines; throat dark showing pale markings of scratches (scars); venter dark; femoral glands pale, clearly contrasted against femora, with few minuscule dark dots; femora and tibia dark as belly.  (Tables 1 and 2). However, ratios are predominantly similar between the two sexes, although males show higher values in FL/ SUL (Z = -2.119, p < 0.05), FM/SUL (Z = -1.883, p = 0.06), HW/SUL (Z = -3.119, p < 0.01) and TD/SUL (Z = -1.942, p = 0.52). Both sexes possess enlarged tusk-like prolongations in the lower jaw as well as the name-bearing 'teeth' on the upper jaw. Male secondary sexual characters are femoral glands, velvety nuptial excrescences on finger I and presence of vocal sacs. Male femoral glands are dark orange (Fig. 6f). Several specimens marked as males in the field lacked obvious secondary sexual characters (femoral glands) and flanks were opened to assess primary sexual characters. Probably due to preservation, femoral glands were not contrasted against the femora, and skin of vocal sacs shrivelled and retracted; however, one male showed no trace of skin modification on femora even if the typical gland position was cut open (ZMB 78262). Webbing formulae show very extensive webbing (Table 7). However, few specimens show a little reduced webbing on toe IV leaving almost the whole distal phalange free (1/1). Dorsal ridges are either long and slender ( Fig. 6a-c) or are roundish and knob-like (Fig. 6d). Number of distinct dorsal ridges (counted from spine to flank) ranges between three and seven, usually four to five ridges per body site. However, both characters were not recognisable due to preservation artefacts in many specimens. Glandular ridges on tibia are usually built of small to large conic glands and form more or less interrupted lines (Fig. 6a-d). Dorsal colouration (in life) varies from uniform dark brown or olive, to dark brownish with pale irregular markings, to ochre with brownish spots and dorsal ridges are set off in terms of colour by usually being darker than the remaining dorsum ( Fig. 6; . Male femoral glands are dark orange (Fig. 6f). Belly colouration (in alcohol) is very variable, ranging from completely pale, to dirty smeared pale-dark, to pale reticulation on dark belly, to dark throat and pale belly, to dark throat and belly with pale longitudinal lines to dark with few pale markings, to completely dark, showing no sex-dependant differentiation.

Colouration of holotype in alcohol
Distribution. Distribution of Odontobatrachus ziama is restricted to isolated mountains north of the Nimba Mts. in south-eastern Guinea (Fig. 1). Its range apparently overlaps with O. natator as the latter is found in proximity to the Simandou Mountain Range, Massif du Ziama or Mt. Going. However, no syntopic populations are known so far. At present no differing habitat requirements or ecological adaptations are known (Barej et al. 2015), which could explain their spatial separation. Presence of O. natator in lower altitudes (e.g. Liberia, Grand Gedeh 250-500 m a.s.l.) could be a factor but both species co-occur in altitudes of app. 500-1300 m a.s.l. in the distribution range of O. ziama. Natural history remark. Odontobatrachus ziama is known as a host of the endoparasitic mite Endotrombicula pillersi, otherwise known from members of the family Phrynobatrachidae (Wohltmann et al. 2007).
Etymology. The species epithet ziama is a noun in apposition, therefore invariable, referring to the species' type locality, the Ziama Forest, in eastern Guinea.
Common name. We advise to use the term ''Ziama torrent-frog'' in English and ''grenouilles des torrents de Ziama'' in French.

Conservation status.
The EOO of O. ziama is 7797 km 2 , placing the species in the category "Vulnerable (VU)" while the AOO of 104 km 2 classifies the species as "Endangered (EN)" (Barej et al. 2015).   (Figs 4,6,8,10). Glandular lines on tibia contain small to mean sized conic glands forming predominantly interrupted lines (Fig. 8a, b), while small to large glands form more or less interrupted lines in O. natator (Fig. 4a-e), small to large glandular conic glands, rather interrupted lines in O. fouta (Fig. 10b, c) and small to mean glandular conic glands form hardly interrupted lines in O. arndti (Fig.  12b, c). Morphologically, the species is most similar in size and colour pattern to O. fouta. However, they differ in a few characters, namely colouration in male femoral glands and belly pattern with both taxa possessing a dark belly colouration, but only O. fouta specimens show a smeared pattern.
Genetics. The species is genetically well differentiated from all congeners and known populations form a well-supported and monophyletic clade (Barej et al. 2015). Uncorrected 16S p-distances between O. smithi and other Odontobatrachus species range from 3.79-5.55%, while maximum intrataxon differences of O. smithi add up to 0.54% (mean value 0.20%; N = 45; Appendix 1: Table A).
Holotype description. The male holotype has been assigned to this taxon in the DCA analysis of ratios. The holotype is an adult male with a robust body (Fig.  7): snout-urostyle length of 60.4 mm; head width 21.9 mm; head slightly longer than broad; snout in lateral view short, flattened and slightly pointed at the snout tip; snout in dorsal view triangular, pointed; lower jaw with sharp tusk-like prolongations protruding the skin and single triangular knob at lower jaw symphysis, corresponding socket in between premaxillae weakly developed; upper premaxillae and maxillae with numerous teeth, posteriorly curved; vomerine teeth present, single prolongations; odontophores arranged in short lines, closer to each other than to choanae; tongue broadly heart shaped; horizontal eye diameter 7.9 mm; interorbital distance 5.6 mm; pupil horizontally elliptical; eye diameter distinctly larger than tympanum diameter; tympanum distinct (horizontal diameter 4.0 mm); nares closer to snout than to eye; snout as long as eye diameter; canthus rostralis rounded; loreal region concave; paired lateral vocal sacs; forelimbs robust, forearms hypertrophied, fingers slender; prepollex absent; relative finger lengths III>IV>II>I; velvety nuptial excrescences covering finger I; subarticular tubercles large, subconical; supernumerary tubercles absent; fingertips dilated, triangular, notched in the middle; femur length 29.9 mm; tibia length 30.4 mm; femoral glands large (length × width: left: 15.7 × 7.0 mm, right: 15.4 × 7.3 mm); femoral glands positioned on the posterior part of the ventral side of femur; relation femoral gland length to femur length: 0.52; minuscule circular glands running along upper side of tibia; foot length (incl. longest toe) 40.0 mm; relative toe lengths IV>III≥V>II>I; shortest toe (7.1 mm); inner metatarsal tubercle elliptical; toe tips broadened forming triangular dilated discs; inner metatarsal tubercle prominent (4.7 mm); number of subconical subarticular tubercles on toes I-V: 1, 1, 2, 3, 2; supernumerary tubercles absent; prominent skin fold on posterior side of feet; dorsal skin texture rough; dorsum and flanks covered with broad dorsal ridges of app. 3.0-4.0 mm (partially flattened, but recognisable as darker spots); venter somewhat rough and slightly granular; flank texture rough and granular as dorsum; webbing ful-ly developed (0-1/0-1/0-1/1-0) ), skin fringe running along toe III, webbing between toes hardly concave. Damage to the male holotype: transverse cut at pectoral region (liver tissue sampled) and skin cut on throat; glandular dorsal ridges partially not recognisable due to preservation. (Fig. 7). Dorsum dark brownish, few pale marblings recognisable; hind limbs with dark blotches on upper side; entire throat dark showing minuscule pale dots; venter dark; pale colouration between axillaries and elbows, femoral glands pale, clearly silhouetted from femora, with blurred minuscule dark dots; femora and tibia dark as belly.

Colouration of holotype in alcohol
Variation. Females (N females = 6) grow larger than males (N males = 3), mean SUL in females 54.1 mm and 48.9 mm in males and accordingly absolute values for extremities are larger too. However, ratios between the two sexes overlap in their range, and are very similar in their mean values showing only minor differences (Tables 1 and  2). Both sexes possess enlarged tusk-like prolongations in the lower jaw as well as the name-bearing 'teeth' on the upper jaw. Male secondary sexual characters are femoral glands, velvety nuptial excrescences on finger I and presence of vocal sacs. Male femoral glands are pale orange (Fig. 8d). Webbing formulae showed little variance in this character (Table 7). Dorsal ridges are usually elongated and slender (Fig. 8a, b). Number of distinct dorsal ridges (counted from spine to flank) ranges between four and seven, usually five to six ridges per body site. However, this character was not recognisable in all vouchers due to preservation artefacts. Glandular ridges on tibia usually are built of small to large conic glands and form more or less interrupted lines (Fig. 8a,  b). Dorsal colouration (in life) ochre coloured with dark brown markings along dorsal glandular ridges or greenish with dark grey markings. Male femoral glands are pale orange (Fig. 8c). Belly colouration (in alcohol) is mainly uniform dark, only few specimens possess paler markings or show a dirty smeared colouration, showing no sex-dependant differentiation.
Distribution. Distribution of Odontobatrachus smithi is restricted to localities in western Guinea on the western and southern edge of the Fouta Djallon Highlands and its western extensions to the Kindia region (Fig. 1). Its easternmost localities are in proximity of O. fouta. However, O. smithi seems to occupy lowland to mid-altitudes (app. 100-650 m a.s.l.) while O. fouta occurs in mid-altitudes (app. 650-900 m a.s.l.).
Etymology. The species epithet smithi refers to Major F. Smith of the Royal Army Medical Corps (R.A.M.C.). In addition to his studies on blackwater fever he contributed to our knowledge on West African amphibians and collected the first specimens of Petropedetes natator Boulenger, 1905 in Sierra Leone during his military service in West Africa. Common name. We advise to use the term ''Smith's torrent-frog'' in English and ''grenouilles des torrents de Smith" in French.
Conservation status. The EOO of Odontobatrachus smithi is 12673 km 2 , placing the species in the category "Vulnerable (VU)" while the AOO of 40 km 2 even classifies the species as "Endangered (EN)" (Barej et al. 2015). Diagnosis. Medium to large sized frogs, robust body shape; head narrow, low mean eye diameter/eye-naris distance ratio, highest tympanum diameter orbita diameter ratio in the family, webbing fully developed, leaving 0.75 of the distal phalange free at the inner side of toe II, leaving the distal phalange at toe IV free; belly colouration typically dark, male femoral glands bright orange; glandular lines on tibia contain mean conic glands forming frequently interrupted lines. Genetically O. fouta differs by a minimum of 3.79% in the mitochondrial 16S gene from its congeners.  (Figs 4, 6, 8, 10). Glandular lines on tibia contain small to large glandular conic glands, rather interrupted lines (Fig. 10b, c), while similar to O. fouta small to large glands form more or less interrupted lines in O. natator (Fig. 4a-e), and small to mean glandular conic glands form hardly interrupted lines in O. arndti (Fig. 12b, c).
Holotype description. The male holotype has been assigned to this taxon in both DCA analyses (absolute values and ratios). The holotype is an adult male with a robust body (Fig. 9): snout-urostyle length of 55.6 mm; head width 21.6 mm; head slightly longer than broad; snout in lateral view short, flattened and slightly rounded; snout in dorsal view triangular, tip fairly rounded; lower jaw with sharp tusk-like prolongations protruding the skin and single triangular knob at lower jaw symphysis, corresponding socket in between premaxillae weakly developed; upper premaxillae and maxillae with numerous teeth, posteriorly curved; vomerine teeth present, single prolongations; odontophores arranged in short lines, closer to each other than to choanae, skin around vomerine teeth dark; tongue broadly heart shaped; horizontal eye diameter 7.7 mm; interorbital distance 5.9 mm; pupil horizontally elliptical; eye diameter distinctly larger than tympanum diameter; tympanum distinct (horizontal diameter 3.1 mm); nares closer to snout than to eye; snout as long as eye diameter; canthus rostralis rounded; loreal region concave; paired lateral vocal sacs; forelimbs robust, forearms hypertrophied, fingers slender; prepollex absent; relative finger lengths III>IV>II>I (Fig. 9); velvety nuptial excrescences covering finger I; subarticular tubercles large, subconical; supernumerary tubercles absent; fingertips dilated, slightly triangular; femur length 27.8 mm; tibia length 28.9 mm; femoral glands large (length × width: left: 14.2 × 8.0 mm, right 14.3 × 8.7 mm); femoral glands positioned on the posterior part of the ventral side of femur; relation femoral gland length to femur length: 0.51; minuscule circular glands running along upper side of tibia; foot length (incl. longest toe) 38.0 mm; relative toe lengths IV>III≥V>II>I (Fig. 9); shortest toe 7.2 mm; inner metatarsal tubercle elliptical; toe tips broadened forming triangular dilated discs; inner metatarsal tubercle prominent (4.5 mm); number of subconical subarticular tubercles on toes I-V: 1, 1, 2, 3, 2; supernumerary tubercles absent; prominent skin fold on posterior side of feet; dorsal skin texture rough; dorsum and flanks covered with slender dorsal ridges of app. 2.0-5.0 mm, mainly positioned dorsolaterally (partially flattened); venter somewhat rough and slightly granular; flank texture rough and granular as dorsum; webbing fully developed (0-0.75/0-1/0-1/1-0), skin fringe running along toe III, webbing between toes hardly concave. Damage of the male holotype: transverse cut at pectoral region (liver tissue sampled); glandular dorsal ridges partially not recognisable due to preservation. (Fig. 9). Dorsum dark brownish; hind limbs with dark blotches on upper side, few pale lines recognisable; entire dirty blurred dark and pale, with several scratches (scars); venter as throat on the anterior part, more reticulated pattern on the belly; colouration between axillaries and elbows brighter; femoral glands pale, clearly silhouetted from femora, with blurred minuscule dark dots, posterior part darker; femora and tibia dark as belly.

Colouration of holotype in alcohol
Variation. Females (N females = 4) grow larger than males (N males = 3), maximum SUL in females 62.5 mm and 57.0 mm in males, and absolute values for extremities are accordingly larger, too (Tables 1 and 2). However, males and females have similar ratios and mean values. Both sexes possess enlarged tusk-like prolongations in the lower jaw as well as the name-bearing 'teeth' on the upper jaw. Male secondary sexual characters are femoral glands, velvety nuptial excrescences on finger I and presence of vocal sacs. Webbing formulae showed little variance (Table 7). Dorsal ridges are short and knobbed (Fig. 10a) or elongated and slender (Fig. 10b). Number of distinct dorsal ridges (counted from spine to flank) ranges between three and six ridges per body site, usually four to five ridges per body site. However, this character was not recognisable due to preservation artefacts in all specimens. Glandular ridges on tibia usually are built of small to large conic glands and form rather interrupted lines (Fig. 10b, c). Dorsal colouration (in life) ochre coloured with dark brown markings along dorsal glandular ridges or almost uniform dark with few whitish markings along flanks and on dorsum. Male femoral glands are bright orange (Fig. 10d). Belly colouration (in alcohol) is mainly uniform dark, only few specimens possess paler markings or show a dirty smeared colouration, showing no sex-dependant differentiation.
Distribution. The distribution of Odontobatrachus fouta is restricted to isolated peaks in the central Fouta Djallon Highlands in western Guinea (Fig. 1)  Etymology. The species epithet fouta is a noun in appo-Diagnosis. Medium to large sized frogs, robust body shape; head narrow, highest eye diameter/eye-narisdistance ratio in the family, low mean tympanum diameter orbita diameter ratio, webbing almost fully developed, leaving 0.25-0.5 of the distal phalange free at the inner side of toe II, leaving 0.75-1 of the distal phalange free at toe IV, belly pattern very variable, glandular lines on tibia contain mean conic glands forming frequently interrupted lines. Genetically O. arndti differs by a minimum of 2.89% in the mitochondrial 16S gene from its congeners.  (Table 7). Glandular lines on tibia contain small to mean conic glands forming hardly interrupted lines in O. arndti (Fig. 12b, c), while similar to O. fouta small to large glands form more or less interrupted lines in O. natator (Fig. 4a-e).
Genetics. The species is genetically well differentiated from all congeners and known populations form a well-supported and monophyletic clade (Barej et al. 2015). Uncorrected 16S p-distances between O. arndti and other Odontobatrachus species range from 2.89-5.55%, while maximum intrataxon differences of O. arndti add up to 0.58% (mean value 0.05%; N = 861; Appendix 1: Table A).
Holotype description. The male holotype has been assigned to this taxon in both DCA analyses (absolute values and ratios). The holotype is an adult male with a slightly robust body (Fig. 11): snout-urostyle length of 48.8 mm; head width 17.1 mm; head slightly longer than broad; snout in lateral view short, flattened and rounded at the snout tip; snout in dorsal view triangular, rounded; lower jaw with sharp tusk-like prolonga-tions protruding the skin and single triangular knob at lower jaw symphysis, corresponding socket in between premaxillae; upper premaxillae and maxillae with numerous teeth, posteriorly curved; vomerine teeth present, arranged in two small odontophores, closer to each other than to choanae; tongue broadly heart shaped; horizontal eye diameter 7.6 mm; interorbital distance 5.2 mm; pupil horizontally elliptical; eye diameter distinctly larger than tympanum diameter; tympanum distinct (horizontal diameter 2.7 mm); nares closer to snout than to eye; snout shorter than eye diameter; canthus rostralis rounded; loreal region concave; paired lateral vocal sacs; forelimbs robust, forearms hypertrophied, fingers slender; prepollex absent; relative finger lengths III>IV>II>I; velvety nuptial excrescences weakly developed on finger I; subarticular tubercles large, subconical; supernumerary tubercles absent; fingertips dilated, triangular, notched in the middle; femur length 24.6 mm; tibia length 26.3 mm; femoral glands large (length × width: left: 12.9 × 8.0 mm, right: 13.1 × 7.8 mm); femoral glands positioned on the posterior part of the ventral side of femur; relation femoral gland length to femur length: 0.53; minuscule circular glands running along upper side of tibia; foot length (incl. longest toe) 34.9 mm; relative toe lengths IV>III>V>II>I; shortest toe 6.4 mm; inner metatarsal tubercle elliptical; toe tips broadened forming triangular dilated discs; inner metatarsal tubercle prominent (3.8 mm); number of subconical subarticular tubercles on toes I-V: 1, 1, 2, 3, 2; supernumerary tubercles absent; prominent skin fold on posterior side of feet; dorsal skin texture heterogeneous; dorsum and flanks covered with slender dorsal ridges of app. 2.5-3.5 mm (partially flattened); venter with fine granulation; flank texture rough and granular as dorsum; webbing fully developed (0-0/0-1/0-0.75/0.75-0); webbing between toes hardly concave. Damage of the male holotype: cut at pectoral region (liver tissue sampled) and skin cut on right lumbar region (in ventral view); glandular dorsal ridges partially not recognisable due to preservation.
Acoustics. Three calls of Odontobatrachus arndti were recorded from specimens in terraria. Calls sound like a repeat of "chucks", consisting of several tonal notes. Two harmonics were visible (Fig. 13), the second harmonic being the dominant frequency (2842.4-3359.2 Hz), the first being the fundamental frequency (1421.2-1679.6 Hz). The call duration ranged from 1.2 to 3.0 s. One call comprised 22 notes (Fig. 13a) and the other two comprised five notes each (Fig. 13b). Each note had a duration of 34.7 ± 0.01 ms (N = 32). The notes were separated from each other by pauses of 238.0 ± 0.01 ms (N = 12) with the two calls comprising five notes (Fig. 13b). Pause duration in the third call (22 notes) was decreasing from 160.0 to 67.0 ms from the beginning towards the end of the call (Fig. 13a). Rödel's (2003) anecdotal report of a torrent frog call from Mt. Sangbé (Côte d'Ivoire) comprising a series of click sounds with ever-shorter intervals corresponds to the second call type (22 notes).

Distribution.
Odontobatrachus arndti is known to occur on the Nimba Mts. in Guinea and Liberia, the adjacent areas at Mt. Gangra (Liberia) and Déré (Guinea), as well as the Mt. Sangbé in western Côte d'Ivoire (Fig. 1). This taxon represents the easternmost representative of the family. Localities at the southern end of the Nimba Mts. and along Mt. Gangra are in very close proximity to O. natator. Both species inhabit similar altitudes at the foot of the Nimba Mts. However, at present no differing habitat requirements or ecological adaptations are known (Barej et al. 2015), which could explain their spatial separation.
Etymology. The species epithet arndti was chosen in order to honour Prof. emerit. Dr. Rudolf G. Arndt, New Jersey USA, for his trust in young academics and his invaluable support of this study.
Common name. We advise to use the term ''Arndt's torrent-frog'' in English and ''grenouilles des torrents d'Arndt" in French.

Conclusive summary
Only recently, biogeographic separation of molecular lineages identified the monospecific West African torrent-frog family Odontobatrachidae as a complex of cryptic species (Barej et al. 2015). In contrast to studies that predominantly rely on genetics in diagnoses of new species (Jörger and Schrödl 2013;Satler et al. 2013;Petzold et al. 2014), no taxonomic actions were conducted in the case of the Odontobatrachus natator-complex. Phylogeographic insights formed the basis for our assessment of morphological characters which led to the formal description of four new species with distinguishing morphological characters.
Odontobatrachus species are phenetically very similar and show an overlap in their morphometrics. Nonetheless, males and females of all species are statistically distinguishable in their metrics and following McLeod et al. (2012), we could confirm that DCAs provided a reasonable method to assign individuals to single species when morphometrics and ratios show no obvious differentiation.
The application of qualitative characters for species differentiation was difficult and previously used diagnostic characters to distinguish Odontobatrachus populations (see Rödel et al. 2004a), which are appropriate and important in non-related genera, like shape arrangement of dorsal glandular ridges in Ptychadena (Guibé and Lamotte 1957;Perret 1979;Rödel 2000) or ventral colouration in Phrynobatrachus (Rödel et al. 2012b;Zimkus and Gvoždík 2013), are not applicable in this family. Specimens belonging to one species from a single locality for example, already show high character variability (see e.g. Fig. 6c, d for glandular ridges in O. ziama). Yet, despite problems due to preservation artefacts, a few qualitative diagnostic characters (e.g. shape of the glandular line on tibia, differences in webbing formulae) could be successfully used (Table 7). All subtle differences between Odontobatrachus species are supported genetically, with interspecies differences of 2.89-5.55% in 16S rRNA uncorrected p-distance (Barej et al. 2015; Appendix 1: Table A) corresponding to species-level in non-related taxa and additionally, Odontobatrachus species are geographically isolated.
Thus, knowledge of the origin of vouchers can narrow down the potential species assignment, because only O. natator is widely distributed, from western Guinea to eastern Liberia and southeastern Guinea, while O. fouta and O. smithi occur only in the westernmost range of that distribution and O. arndti and O. ziama occur only in the easternmost range. Still, it would be somewhat unsatisfactory, if solely genetics provided a warranted identification of single specimens in areas of distributional overlap between morphologically rather indistinguishable species (Real et al. 2005). Fortunately, following an integrative approach, the consideration of molecular data, distribution patterns, and morphology, rendered recognition of different Odontobatrachus species comprehensible despite their superficial similarity. The similarity in morphology of Odontobatrachus species likely results from speciation lacking distinct external changes (Bickford et al. 2007), probably because the most conspicuous characters are all adaptations to the habitat of fast flowing streams. The Odontobatrachus species all exhibit a torrent-frog'-habitus as likewise independently developed in various non-related taxa (e.g. Petropedetidae: Petropedetes, Arthroleptides, Barej et al. 2010Barej et al. , 2014aHylodidae: Hylodes, Haddad and Giaretta 1999;Heleophrynidae: Heleophryne, Hadromophryne, Minter et al. 2004;Ranidae: Staurois, Matsui et al. 2007). Adults usually possess a rather flattened body shape and head, allowing them to hide between crevices and under rocks. Their extremities are long, terminal phalanges enlarged and digits on hands and feet spatulated offering a larger contact area with the slippery substrate (Minter et al. 2004;Scott 2005;Kamermans and Vences 2009). Torrent-frog tadpoles have a streamlined habitus with distinct tail musculature and a sucker-like mouth which are used to cling or climb on rocks (Barej et al. 2010;Minter et al. 2004). It seems likely that any radical deviation from that morphotype could have negative effects on species survival.
Recognition and description of species is just a first step which provides the baseline for subsequent studies to gather further data on the ecology or behaviour -or   simply: naming does not mean knowing a species. Our knowledge on the family is still incomplete, as calls of four species remain unknown and data on tadpole morphology is lacking. Lamotte and Zuber-Vogeli (1954) published a detailed description on tadpoles collected in Liberia and the Nimba Mts., thus we cannot rule out that their description is based on material from two different species (O. natator and O. arndti). Solely Guibé and Lamotte (1958) described a series of tadpoles from the River Zougue on the Nimba Mts. (described as O. arndti in the present work) and provided a short and superficial description of the habitus. Our insights on the ecology of adult Odontobatrachus are scarce, too. While streams with rapids in primary forests are the preferred habitat of torrent-frogs, detailed ecological studies are missing. Differences in habitat requirements and ecology have been recognised as further characters of importance to distinguish species in other genera (e.g. Amietophrynus regularis vs. A. maculatus Amiet 1976; Böhme 1994a; Phrynobatrachus guineensis and P. phyllophilus Rödel and Ernst 2002). 'West African Forests' are recognised as one of the world's biodiversity hotspots (Myers et al. 2000;Bakarr et al. 2001) and biogeographically demarcated from the Central African forest block (Penner et al. 2011).
Although West Africa, defined as ranging from Senegal to Nigeria, is regarded as one of the better known regions on the continent, more than ten new amphibian species have been described in the last decade (e.g. Blackburn et al. 2008;Ernst et al. 2008;Hillers et al. 2008b;Rödel 2007;Rödel et al. 2003Rödel et al. , 2009aRödel et al. , b, c, 2010Rödel et al. , 2011Rödel et al. , 2012a and more await formal description (Rödel et al. unpubl. data.). We agree with McLeod et al. (2012) that it is crucial to identify the "true" biodiversity and although species that are morphologically difficult to tell apart become condemned, their scientific recognition is an indispensable tool for conservation management.
In the past, only a single torrent-frog species, O. natator, was known to occur in West Africa and it has been listed as "Near Threatened (NT)" according to the IUCN Red List (IUCN 2011). However, we herein demonstrated the presence of four new Odontobatrachus species with dramatically constrained distribution ranges. According to Barej et al. (2015) all five species (therein treated as OTUs) require the IUCN category "Endangered (EN)" resulting from analyses of the range criteria Extent of Occurrence and Area of Occupancy. Environmental Niche Modelling (Barej et al. 2015) confirmed large distributional gaps and thus justify the use of AOO as criterion. Both sub-clades of O. natator should be treated distinctly with regard to conservation concern as evolutionary significant units (Moritz 1996(Moritz , 2002Ennos et al. 2005) and if recognised as distinct species, the Freetown Peninsula population demands the category "Critically Endangered (CR)". As habitat loss is ongoing due to forest fragmentation and conversion (Chatelain et al. 1996;FAO 2006;Norris et al. 2010), conservation efforts need to be made soon. In the case of West African torrent-frogs, only O. natator, O. ziama and O. arndti occur in protected areas like National Parks and Biosphere Reserves (Barej et al. 2015) and even if lower priority areas are considered, they fail to protect all five species. Distribution patterns in Odontobatrachus cover ranges of various endemic species in the Fouta Djallon and the Nimba Mts.-Massif du Ziama-Simandou Mountain Range (Angel 1943;Porembski et al. 1994Porembski et al. , 1995. Furthermore, these areas are assumed to have played an important role as forest refugia in Upper Guinea (Maley 1987;Porembski et al. 1994;Sosef 1994). Barej et al. (2015) suggested that the Loma Mts. and Tingi Hills in Sierra Leone, as the highest elevation occurrences in the Guinea Highlands, could represent a refugium in central Upper Guinea for O. natator and any conservation effort within these areas could consequently be beneficial in many nonrelated taxa and assure their long-term survival.
In summary, the diversity in the family Odontobatrachidae has been raised to five species. While our knowledge on this West African endemics is far from complete, nomination of OTUs recognised by Barej et al. (2015) is of importance for promoting immediate conservation actions as all species require the IUCN category "Endangered (EN)". West African torrent-frog species are at risk of becoming extinct because of habitat loss in the Upper Guinean biodiversity hotspot, whose "true" biodiversity is still far from being completely known. Table C. Kruskal-Wallis test statistics for Odontobatrachus male and female absolute values (A) and ratios (B). χ 2 -value, degree of freedom (DF) and the asymptotic significance (Asymp. Sig.) are given. See material and methods section for abbreviations.