KU 340897 (CRH 798), an adult male collected at mid-day on February 8^th 2015 by Shea Maddock Lambert, Emile Rajeriarison, and Ralaivao Jean Fulgence in montane rainforest near the former village of Andemaka in Ranomafana National Park (ca. 21.1287°S, 47.5054°E, elevation ca. 1240m a.s.l.; Fig. 2).

Figure 2. 

Map of Ranomafana National Park and the type locality of Rhombophryne nilevina sp. n.. Map is a composite of Landsat 8 satellite imagery and a hillshade layer created from SRTM 1 Arc-Second Global digital elevation data. Data available from the U.S. Geological Survey.

UADBA-A Uncatalogued (CRH 799), an adult male collected the morning of February 7^th 2015 by Shea Maddock Lambert and Ralaivao Jean Fulgence, otherwise with the same collection information as the holotype.

A frog assigned to the cophyline genus Rhombophryne on the basis of its divided vomer, the possession of clavicles and knob-shaped terminal phalanges (see Scherz et al. 2016a). This species is characterized by the following suite of characters: large size (SVL at least up to 57.2 mm), wide, short head (HW 180.7% of HL), tympanum 58.6% of eye, forelimb 51.1% of SVL, tibia 42.2% of SVL, hindlimb 152.5% of SVL, large inner metacarpal and metatarsal tubercles, supratympanic fold distinct and raised, running from the posterior corner of the eye straight over the tympanum, then sharply down behind it, extending to join the front of the arm, distinct vomerine teeth forming curved rows posteromedial to the oblong choanae, separated medially by a small cleft, second finger shorter than fourth finger, fifth toe distinctly shorter than third, without finger or toe reduction, finger and toe tips not enlarged. Additionally, R. nilevina is separated from all nominal species of Rhombophryne by an uncorrected pairwise distance of at least 4.9% in the fragment of the 16S rRNA gene, and by at least 3.8% from all known candidate species in this genus.

Rhombophryne nilevina is the largest species in the genus Rhombophryne, and can be distinguished based on this character alone from all other described species (SVL 57.2 mm vs. maximums of 56.3 mm and 52.9 mm for the next two largest species, R. laevipes and R. vaventy, respectively). This species differs from all of its congeners as follows: from all members of the R. serratopalpebrosa group (R. serratopalpebrosa, R. coronata, R. vaventy, R. ornata, R. tany, and R. guentherpetersi, plus two species under description by Scherz et al. in review) by the absence of superciliary spines (vs. presence); from R. testudo, R. coudreaui, and R. matavy by less wide head (HW 180.7% vs. 187.6–242.4% of HL in R. testudo and R. matavy), longer forelimb (FORL 51.1% vs. 35.4–49.8% of SVL), longer hindlimb (HIL 152.5% vs. 117.4–140.8% of SVL), and the possession of a clavicle (vs. lack thereof); from R. longicrus and R. minima by its wider head (HW 180.7% vs. 122.5–142.8% of HL), shorter forelimb (FORL 51.1% vs. 70.4–74.7% of SVL), and shorter hindlimb (HIL 152.5% vs. 178.5–183.8% of SVL); from R. savaka and R. mangabensis by its longer forelimb (FORL 51.1% vs. 40.9–47.9% of SVL), well ossified clavicles (vs. poorly ossified), and absence of black inguinal spots and a mid-vomerine diastema (vs. presence in R: savaka); and from R. alluaudi, R. laevipes, and R. botabota by its wider head (HW 180.7% vs. 144.2–173.8% of SVL), absence of light dorsolateral stripes (vs. presence in R. alluaudi), absence of a stark color border between the dorsal and lateral parts of the head (vs. presence in R. botabota), absence of inguinal ocellations (vs. presence in R. laevipes and R. alluaudi).

Rhombophryne nilevina is morphologically similar to terrestrial members of the genus Plethodontohyla, but aside from being distinguishable from this genus by the combination of the possession of clavicles with knob-shaped terminal phalanges, this species can be distinguished from P. inguinalis by its smaller size (SVL 57.2 vs. 62.2–99.1 mm), the absence of enlarged fingertips, absence of dark inguinal spots (vs. occasional presence), and absence of a strong dorsolateral color border (vs. occasional presence); from P. notosticta, P. guentheri, P. fonetana, and P. mihanika by the absence of enlarged fingertips, absence a strong dorsolateral color border (vs. presence in all but P. fonetana), and shorter forelimb (FORL 51.1% vs. 57.5–71.9% of SVL); and from P. bipunctata, P. tuberata, P. brevipes, and P. ocellata by the absence of inguinal spots (vs. presence in all but P. tuberata) and larger size (SVL 57.2 vs. 24.6–44.7 mm) and from P. tuberata by the presence of smooth skin (vs. granular skin).

Although the bioacoustic repertoires of cophylines is far from completely known, bioacoustically, this species’ call is strongly distinct from the other known calls by being strongly amplitude modulated (Fig. 4). To the human ear, this call most closely resembles the genetically distant R. testudo (Table 2), but the call of R. testudo differs by having a much longer duration and lacking significant amplitude modulation (Fig. 4). No other known calls can be confused with those of this species.

Morphology of the holotype. An adult male specimen in an excellent state of preservation. The vocal sac is still somewhat loose and malleable. The tongue was removed as a tissue sample.

Body rotund; dorsal and ventral skin smooth, with subtle bumps on the dorsal skin (more rugose in life). Head considerably wider than long (HW 180.7% of HL), snout rounded in dorsal and lateral view; nostrils protuberant, directed laterally, closer to the snout than the eye; canthus rostralis distinct and concave; loreal region concave and oblique; tympanum indistinct, oval, horizontally 58.6% of eye diameter; pupil dilated in preservative but more or less round in life (Fig. 3a, 3d); supratympanic fold distinct and raised, running from the posterior corner of the eye straight over the tympanum, then sharply down behind it, extending to join the front of the arm; tongue removed as a tissue sample, was attached anteriorly and posteriorly free; vomerine teeth distinct, forming curved rows posteromedial to the choanae; choanae relatively large, oblong.

Figure 3. 

Photos in life of Rhombophryne nilevina sp. n. (a) Dorsolateralview of the holotype (KU 340893). (b) Dorsal view of the holotype. (c) Ventral view of the holotype. (d) Dorsolateral view of the paratype (CRH 799, UADBA-A Uncatalouged). (e) Dorsal view of the paratype. (f) Ventral view of the paratype.

Arms strongly built, relatively short; fingers without webbing, short, with distinct, rounded subarticular tubercles, relative lengths 1<2<4<3, the second finger marginally shorter than the fourth (and marginally longer than the first), without enlarged terminal discs; inner metacarpal tubercle strong, oblong, 28.1% of hand length; outer metacarpal tubercle indistinct, round. Legs relatively long and thick (HIL 152.5% of SVL; TIBL 42.2% of SVL), position of the tibiotarsal articulation when adpressed along the body not possible to assess without breaking the hindlimbs; toes long, unwebbed, with indistinct round subarticular tubercles, relative toe lengths 1<2<5<3<4, third toe distinctly longer than fifth; inner metatarsal tubercle present and distinct, 12.7% of foot length; outer metatarsal tubercle absent.

Coloration of the holotype. In preservative, the holotype is chocolate brown dorsally with a loosely reticulated pattern of ebony to burnt umber markings, including an indistinct interocular bar. There are no inguinal spots. The loreal region has a grey marking in it. The forelimb is as the dorsum, with dark patches on the elbow and a crossband on the forearm. A distinct light annulus is present before the terminus of each finger. The hindlimb is dorsally as the back, with three dark crossbands on the thigh and shank. The posterodorsal thigh has weak cream spots, as does the anterior thigh. The dorsal foot is brown speckled with cream. The toes are even more flecked with cream, and also possess a light annulus before the terminal phalanges. The ventral abdomen is brown with numerous small cream flecks. The chin is darker and mostly solid dark brown. The ventral arms are as the trunk. The subarticular and metacarpal tubercles are lighter in color than the rest of the hand. The ventral hindlimbs are as the abdomen. The color in life was as in preservative (Fig. 3).

Osteology of the holotype (Fig. 5, Suppl. material 1). The skeleton of the holotype is typical of Rhombophryne. It is well ossified and robust. The right femur shows signs of an old break toward its distal end that has healed.

Anterior braincase laterally closed by the sphenethmoid. Interior braincase containing calcified material. Nasal in medial contact with contralalteral and posterior contact with frontoparietal. Frontoparietal broadening anteriorly from narrow waist anterior to lateral flanges, possessing a strong, posteriorly elongated dorsal process. Prechoanal vomer simple, triradiate. Neopalatine and postchoanal vomer distinguishable. Vomerine teeth not medially fused, without diastemata, oriented oblique to antero-posterior body axis, curved. Maxillary teeth minute. Otic capsule dorsally poorly ossified.

Sternum not ossified. Clavicle robust, curved. Humerus proximally broad, distally rather narrow; possessing a well-developed crista ventralis along roughly 50% of its length; crista lateralis weak. Terminal phalanges of fingers and toes with small distal knobs. Phalangeal formula of fingers 2-2-3-3; of toes 2-2-3-4-3. Femur without cristae. Prepollex strong, blade-like, half length of first metacarpal. Prehallux strong, approximately half length of first metatarsal.

Neural spines decrease in size posteriorly, the sixth and seventh lacking spines altogether. Neural arches of atlas fused. Dorsal crest of urostyle running roughly 80% along its shaft. Iliosacral articulation type IIA sensu Emerson 1979. Iliac shafts with well developed dorsal tubercles and deep oblique grooves; dorsal crests running most of their length. Pubis partially ossified.

The paratype UADBA-A Uncatalogued (CRH 799) strongly resembles the holotype, but has a slightly more distinct color border between the lateral and dorsal head (see Fig. 3 for comparison).

We analysed a total of seven calls from R. nilevina, and compared these to the call of R. testudo (Fig. 4; Table 3). We presume that the calls we recorded come from one individual, the holotype (see Materials and methods). We further assume that the recorded call is an advertisement call, as no other call types (except distress calls) are known from cophylines. This call sounds like a slow groan to the human ear.

Each call is rapidly pulsed, with 3–5 (3.5 ± 0.534) amplitude modulated peaks occurring throughout the call, and peak amplitude occuring in the last 50% of the call. The call duration is 505–544 (536 ± 1.7) ms with an inter-call interval duration of 42.5–99.5 (68.8 ± 24.0) s. The fundamental frequency is 236.9–279.9 (261.5 ± 22.9) Hz. The mean dominant frequency throughout the call was 528.3–555.9 (537.9 ± 9.2) Hz and the first harmonic frequency is 775.2–818.3 (796.8 ± 17.6) Hz (Fig. 4).

Figure 4. 

The osteology of Rhombophryne nilevina sp. n. Skull in (a) lateral, (b) dorsal, and (c) ventral view; and full skeleton in (d) dorsal and (e) ventral view. Abbreviations: angspl, angulosplenial; angspl.cp, angulosplenial coronoid process; col, columella; exoc, exoccipital; fpar, frontoparietal; fpar.dop, frontoparietal dorsal process; max, maxilla; max.pf, maxillary pars fascialis; mmk, mentomeckelian bone; npl, neopalatine; pmx, premaxilla; povom, postschoanal vomer; proot, prootic; prvom, prechoanal vomer; prsph.ap, parasphenoid alary process; prsph.cp, parasphenoid cultriform process; pter.ar, pterygoid anterior ramus; pter.vr, pterygoid ventral ramus; pter.mr, pterygoid medial ramus; qj, quadratojugal; qj.pvp, quadratojugal posteroventral process; smx, septomaxilla; spheth, sphenethmoid; sq, squamosal; sq.or, squamosal otic ramus; sq.zr, squamosal zygotic ramus.

The specific epithet “nilevina” is a Malagasy word meaning “buried.” This name was chosen to recognize the fossorial habits of this species. It is to be treated as an invariable noun in apposition.

Due to morphological and size similarities, as well as geographic distribution, two existing names must be considered for this species: Phrynocara laeve Boettger, 1883, and Plethodontohyla laevis tsianovohensis Angel, 1936. Both of these names are currently considered to be junior synonyms of Rhombophryne alluaudi. We examined the morphology and osteology of the holotypes of both of these taxa (P. laeve: SMF 4286; P. laevis tsianovohensis: MNHN 1936.47), and our new species differs critically from both in the possession of a well-developed clavicle (vs. absence/strong reduction; Scherz unpubl. data). Their taxonomy, as well as that of Rhombophryne alluaudi, will be discussed in a future article, and we here simply rule out the possibility that they are conspecific with R. nilevina sp. n. based on the presence vs. absence of a clavicle. The type specimen of P. laevis tsianovohensis was collected from Tsianovoha, which is around 60 km south of Ranomafana, suggesting the possibility of sympatry or parapatry with R. nilevina.

Natural history. Both known specimens of R. nilevina were obtained from a relatively flat, poorly drained section of moist montane forest adjacent to a stream, with the holotype found along the bank of this stream. Nearby habitats include a swamp with many large Pandanus and steep forested slopes with relatively smaller trees. However, the calls of R. nilevina seemed to emanate mostly from the flatter, forested area. Males were heard calling during the day, particularly during overcast conditions and after rainfall. Advertisement calls were not heard at night, however, the night-time chorus of other frogs, including Boophis, Spinomantis, Gephyromantis, and Anodonthyla, may have interfered with detection. When heard from a distance, the call is reminiscent of that of an owl. When heard from close proximity, the call sounds like a groan, and is far less melodic. Both specimens were both located by auditory tracking, and found calling from underground: one from a cavity under the roots of a large tree, and the other from a burrow in soft, moist soil alongside the stream. In order to collect the holotype from its burrow, excavation was required. Based on these observations and suggestive morphology, we presume that R. nilevina spend much of their lives underground, possibly coming to the surface for short periods during rainfall, similar to other fossorial Rhombophryne species (Glaw and Vences 2007, D’Cruze et al. 2010). We also note that R. nilevina was discovered in the middle of the wet season, and after a week-long period of particularly heavy, sustained rain.

Figure 5. 

Comparative spectrograms (top), oscillograms (center) and power spectra (bottom) between the calls of (A) Rhombophryne nilevina sp. n. and (B) R. testudo (from Vences et al. 2006). Spectrogram was created using a Hanning window size of 1024.

Rhombophryne nilevina has thus far been detected at a single site, near the former village of Andemaka, in the north-west of Ranomafana National Park (Fig. 2). This locality is relatively high-elevation for Ranomafana National Park (ca. 1240 m). To our knowledge, R. nilevina has not been detected by any previous survey, including several conducted by CRH and SML at similarly high-elevation sites in the northern (Miaranony), central (Vohiparara), and southern (Maharira) regions of Ranomafana. Nevertheless, we do not rule out here the possibility that R. nilevina occurs elsewhere in the park. This is in large part due to the secretive habits and potentially ephemeral activity periods of this species (see Natural history). In addition, much of the high-elevation forest of Ranomafana is difficult to access and thus remains sparsely or completely unsurveyed for herpetofauna. Although it is possible that R. nilevina has been overlooked in other eastern rainforest patches, current information suggests that this species is endemic to Ranomafana National Park, and potentially to a much smaller area within the park.

Although the type locality of R. nilevina is within Ranomafana National Park, its occupancy within the park is potentially highly restricted, elevationally and geographically, as it has not been detected in any other herpetological surveys of the park. However, its secretive lifestyle means that it icould be easily overlooked. Given this large uncertainty in area of occupancy, we suggest an initial IUCN categorization of Data Deficient. If R. nilevina is for instance, restricted to the type locality, then habitat destruction, chytrid fungus (recently detected in Madagascar, Bletz et al. 2015), and/or climate change could easily place the only population of R. nilevina sp. n. at risk of extinction.