The term ankarafensis is a specific epithet deriving from the species’ terra typica, the Ankarafa Forest. The name is used as an adjective in the nominative singular.

MRSN A6973, adult male (Fig. 2D and Fig. 3D) collected at Ankarafa Forest (Sahamalaza Peninsula, north-western Madagascar), 14°22.85'S, 47°45.52'E; ca 140 m a.s.l., transitional forest, 26 January 2013, leg. G. M. Rosa.

MRSN A6974 adult female (Fig. 3D), same data as holotype. MRSN A6975, adult male (Fig. 2A–B) collected at Ankarafa Forest (Sahamalaza Peninsula, north-western Madagascar), 14°22.83'S, 47°45.47'E; ca 130 m a.s.l.; transitional forest, 21 November 2011, leg. S. G. Penny. MRSN A6976, adult male (Fig. 3A) collected at Ankarafa Forest (Sahamalaza Peninsula, north-western Madagascar), 14°22.85'S, 47°45.51'E; ca 130 m a.s.l., transitional forest, 12 January 2012, leg. S. G. Penny.

A treefrog assigned to the genus Boophis based on absence of femoral glands in males, the presence of an intercalary element between the ultimate and penultimate phalanges of fingers and toes (verified by external examination), presence of nuptial pads in males, general morphological resemblance to other Boophis species, and molecular evidence. Assigned to the Boophis rappiodes group based on small size (adult males 22.9–24.0 mm and one female 28.5 mm SVL), absence of lateral fringes along lower arm and tarsus, greenish and slightly translucent dorsal colouration and translucent venter (inner organs can be clearly seen through the skin in live specimens). Boophis ankarafensis sp. n. is distinguished from Boophis erythrodactylus by lack of reddish colour of fingertips, and from Boophis erythrodactylus and Boophis tasymena by lack of regular pattern of red dorsal spots. Distinguished from Boophis viridis by its smaller size (SVL up to 31 mm in males and 35 mm in females of Boophis viridis), iris colouration (Boophis viridis shows a distinctive brown inner-iris area and blue outer-iris area), and the presence (vs. absence) of yellowish dorsolateral stripes. Compared to Boophis rappiodes, the new species is distinguished by a more extensive darker pattern on the dorsum, which is most evident in living or freshly preserved specimens; while the dorsal pattern in Boophis rappiodes is intensely red, and remains red in preservative before it eventually fades, the pattern in Boophis ankarafensis sp. n. is reddish-brown in life and becomes persistently dark brown in preservative, often covering almost the entire dorsum. Boophis ankarafensis sp. n. is most similar in morphology to Boophis bottae and no apparent morphological feature distinguishes them. However, the new species can be distinguished by molecular analysis and by its advertisement calls (see below).

MRSN A6973, adult male in a good state of preservation. SVL 24.0 mm (see Table 1 for more detailed morphometric measurements). The body is slender with the head much wider than the body. Snout rounded in dorsal view, slightly truncate in lateral view, nostrils directed laterally, slightly nearer to tip of snout than to eye; canthus rostralis and loreal region both slightly concave; tympanum distinct, rounded, 40% of eye diameter; supratympanic fold not recognizable; tongue ovoid, distinctly bifid, posteriorly half free; vomerine odontophores distinct; positioned posteromedian to choanae; choanae small, rounded. Arms slender, subarticular tubercles single, round; metacarpal tubercles unrecognizable; fingers with weak webbing; webbing formula 1(1), 2i(1.75), 2e(0.75), 3i(2.5), 3e(1.75), 4(1); relative length of fingers 1 < 2 < 4 < 3 (finger 2 distinctly shorter than finger 4); finger disks moderately enlarged; small unpigmented nuptial pads faintly recognizable on inner side of first finger. Hindlimbs slender; tibiotarsal articulation reaches nostril when hindlimb is adpressed along body; lateral metatarsalia separated by webbing; inner metatarsal tubercle distinct, no outer metatarsal tubercle; webbing formula between toes 1(0), 2i(0.75), 2e(0), 3i(0.75), 3e(0), 4i(1), 4e(1.5), 5(0.25); relative length of toes 1 < 2 < 5 = 3 < 4; toe disks slightly enlarged. Skin on dorsal surfaces and ventrally on throat smooth; ventral skin on belly and around cloacal opening glandular. The ground dorsal colour, including limbs, is light green. The webbing, finger and toe disks are green-yellow in colour. Speckles of reddish-brown and yellow pigment can be seen covering the dorsum and limbs. Thin yellow dorsolateral stripes run from behind the eye to the forelimb and then fade towards the mid-body. On the head a reddish-brown pigment forms a band between the eyes and covers the supra-ocular area, interspersed with yellow speckling. This reddish-brown pigment also forms a faint rostral stripe between the eye and nose tip. The pupil is horizontal, with a beige iris containing darker brown patches and reticulations; the iris periphery is blue. The specimen has a white venter with some translucence exposing the inner organs, and a bluish throat. At day-time the green on the dorsum became pale and the red markings fade out to a pale brownish red colour (Fig. 2D). In preservative it is in good condition and comparatively more hydrated. The colouration, at about 4 months from the capture is vivid: the dorsum is pigmented and the dark bar between the eyes is contrasting and evident, as well as the pigmentation around the nostrils. The specimen has the last phalanx of the 4th toe of the left foot missing for genetic analysis.

The paratypes (MRSN A6974-6976) closely match the holotype but with slightly different patterning of pigment patches. Finer regular black spots were observed on the dorsum and limbs (possibly single melanophores); a feature similarly observed in the related Boophis bottae (Vences and Glaw 2002). Based on other individuals photographed in nature, specimens matched the holotype and paratypes. However, pigment patterns vary in their distribution and intensity. The males MRSN A6975 and MRSN A6976 are in mediocre state of conservation and dehydrated due to fixation and preservation in ethanol. The colouration, after about 18 months from the capture, faded from green to whitish-yellowish except for pigmentation on the upper eyelids and for the bar between the eyes. The dorsum and upper parts of legs are finely pigmented. Eyes are blackish, with whitish pupillae. The 5th toe of the left foot was removed and fixed in 96% ethanol for molecular analyses. The female specimen (SVL 28.5 mm) collected more recently (MRSN A6974) is in better condition, similar to the holotype. This specimen also has the last phalanx of the 4th toe of the left foot missing for molecular identification. See Table 1 for detailed morphometric measurements.

The molecular data confirms the attribution of Boophis ankarafensis sp. n. to the Boophis rappiodes group (Glaw and Vences 2007). The four analysed specimens of Boophis ankarafensis are genetically uniform and did not show any intraspecific divergence, while the analysed specimens of Boophis bottae, Boophis rappiodes, Boophis tasymena and Boophis viridis are more heterogeneous and show intraspecific uncorrected divergence ranging from 1.2% (Boophis viridis) to 2.2% (Boophis rappiodes), in the 16S rRNA gene analysed fragment (Table 2). The species showing the highest values of intraspecific divergences is Boophis rappiodes (2.2%). This value of intraspecific divergence is computed based on the analyses of sequences coming from Andasibe and Ranomafana, two sites that are about 300 km far apart. The population of Boophis ankarafensis is about 400 km apart from the closest population of Boophis bottae (at Betampona) and their genetic distance is of 4.9%. As shown in Table 2, available values of intraspecific genetic divergence values never reach the minimum value of genetic divergence observed between different species of the Boophis rappiodes group (comparison Boophis ankarafensis/Boophis bottae).

The genetic distance between Boophis ankarafensis and the five other species of the Boophis rappiodes group ranges between 4.9% (comparison between Boophis ankarafensis and Boophis bottae) and 11.4% (comparison between Boophis ankarafensis and Boophis erythrodactylus). Among the species of the analysed species group the smallest genetic distance is observed between the newly described species and Boophis bottae (4.9%) and the highest value between Boophis erythrodactylus and Boophis viridis (13.0%). More details on genetic distances between species of the Boophis rappiodes group are provided in Table 2.

Vocalisations of Boophis ankarafensis sp. n. were recorded on thirteen occasions between October 2011 and January 2012. No males were found in calling activity between January and February 2013 (later in the wet season). Calls were detected from just 8 of the 37 locations monitored on the peninsula (Fig. 4). All eight of these locations were along the banks of a single stream in Ankarafa Forest. Vocalisations were detected within 3978 minutes of acoustic recordings and averaged 306 ± 181 minutes per night (n = 13). A total of 59.8% of these minutes contained vocalisations emitted by a single male caller, with the remainder emitted as a chorus of two or more callers. Total activity equalled 5576 minutes and average per night activity amounted to 429 ± 273 minutes (n = 13). Levels of calling fluctuated throughout the night (Fig. 5); there was a peak in activity just after dusk, followed by a slow decline until dawn, when the fewest vocalisations were detected (y = 0.019e^-0.015x; R² = 0.582).

The acoustic repertoire consists of two note types: a multi-pulsed trill (type 1; Fig. 6A) and a 1-3 pulsed click (type 2; Fig. 6B). Only 11 of the 13 occasions produced recordings of a sufficient audio quality to perform spectral analysis. Measurements were taken from five representative notes of each call type per night, thus a total of 55 trill notes and 55 click notes were measured (Table 3). From a subset of five trill notes, acoustic parameters were measured for ten sequential broadband pulses and ten sequential narrowband pulses per note. The environmental temperature of the recorded trill notes ranged between 25.2 and 30.3 °C (27.4 ± 1.08 °C, n = 55) and of the click notes between 22.9 and 29.7 °C (26.7 ± 1.41 °C, n = 55).

The trill note measured 4.29 ± 1.29 s (1.614–8.091 s, n = 55) and is composed of alternating broadband and narrowband pulses. These notes had a broadband pulse rate of 15.0 ± 1.19 pulses/s (11.7–17.3 pulses/s, n = 25) and peak frequency of 4.14 ± 0.131 kHz (3.69–4.48 kHz, n = 55). Broadband pulses measured 3.47 ± 0.485 ms (2.50–5.06 ms, n = 50) whereas narrowband pulses were slightly longer at 17.9 ± 2.56 ms (11.2–25.1 m; n = 50). The broadband pulse had a peak frequency of 4.19 ± 0.122 kHz (3.91–4.34 kHz; n = 50), whereas the narrowband pulse measured 4.15 ± 0.108 kHz (4.00–4.31 kHz, n = 50). Bandwidths were 0.288 ± 0.169 kHz (0.120–0.780 kHz, n = 50) for the broadband pulse compared to just 0.155 ± 0.036 kHz (0.129–0.280 kHz, n = 50) for the narrowband pulse when measured at a -10 dB threshold. The majority of click notes consisted of two pulses, although singular and triple pulses were also observed. Click notes had a total duration of 52.9 ± 5.20 ms (40.0–68.7 ms; n = 55) and peak frequency of 3.95 ± 0.162 kHz (3.57–4.30 kHz; n = 50). Bandwidth of the click note measured 0.538 ± 0.234 kHz (0.120–0.143 kHz) at a -10 dB threshold.

The advertisement call of Boophis ankarafensis sp. n. sounds similar to that of the morphologically similar Boophis bottae; with both species possessing a trill note and a click note in their acoustic repertoires (Vences et al. 2006; Rosa et al. 2011). However, a number of differences can be used to distinguish them. In our comparison with Boophis bottae we selected only those notes recorded at a similar ambient temperature, and we have also distinguished the two pulse types of the trill by their relative differences in bandwidth, rather than duration as used by Vences and Glaw (2002); the broadband pulse and narrowband pulse of Boophis ankarafensis sp. n. are equivalent to the shorter and longer duration pulses of Boophis bottae respectively.

The trill note (type 1) of Boophis ankarafensis has a faster broadband pulse rate than the trill of Boophis bottae (13.4–13.5 pulses/s, 25.2 °C versus 8.12–11.6 pulses/s, 23.0 °C). The click note (type 2) of Boophis ankarafensis sp. n. usually contains just two pulses, and only rarely consists of three, in contrast the click notes of Boophis bottae are usually three-pulsed. Although only slightly divergent, it is discernable that the spectral frequency of the click note is lower in Boophis ankarafensis sp. n than in Boophis bottae (3.75–3.84 kHz, 22.9 °C versus 4.23–4.42 kHz, 23 °C). The trill note of Boophis ankarafensis shows a slightly lower broadband pulse peak frequency (4.10–4.24 kHz, 25.2 °C versus 4.30–4.64 kHz, 23 °C) but a similar narrowband pulse peak frequency (4.18–4.31 kHz, 25.2 °C versus 4.23–4.42 kHz, 23 °C) in comparison to Boophis bottae.

A total of 54 individuals of Boophis ankarafensis sp. n. were encountered between 29 October 2011 and 05 January 2012. An additional amplexing couple was found on 26 January 2013. Individuals were found at night in Ankarafa Forest along the banks of two streams; these streams became fast flowing after heavy rains during the wet season. No individuals were detected from sections of stream in open habitat where vegetation was absent, and so it should be considered a forest species.

Of the 56 encounters, 48 frogs were male and 8 female. Males of this species were found calling from vegetation approximately 0.5 to 2 m high. Vocalising males were often within close proximity to one another, positioned on different leaves of the same plant. All but one of the eight females were found in axillary amplexus with males (Fig. 3D–E); mating pairs were positioned on leaves overhanging the stream bank, and in one instance on a rock. Calling and breeding activity were most intense during the first few months of the rainy season (Oct–Dec), decreasing later in the season (Jan–Feb). A lone female was found in a tree during the day approximately 3 m high and 30 m from a stream.

The suggested conservation status of this species was assessed using the criteria and guidelines of the IUCN Red List (IUCN Standards and Petitions Subcommittee 2011). This species is endemic to Madagascar and so far only recorded in a small forest fragment (Ankarafa Forest). Individuals were found along the vegetated banks of two streams, with forested sections of these streams extending for approximately 2 km of the more northerly stream (Fig. 7B), and just 1 km for the stream south of this (Fig. 7A). No individuals were found upstream of these two locations, where the streams are replaced by Raphia-dominated swampland or downstream where the stream-banks are cleared of vegetation. An average occurrence of 3 specimens per 200 m of river was determined from those transect searches where the specimens were detected (Table 4). This species was not detected from Anabohazo Forest, which aside from Ankarafa represents the only other large area of intact forest remaining on the peninsula. A survey of the Vavan’aneno River between Antafiabe and Ambinda villages, and a survey of Betsimpoaka village also failed to detect it.

If suitable habitat is considered to be all areas of Ankarafa Forest (likely an over-estimate) then this area totals less than 5 km², giving an EOO (extent of occurrence) of less than 100 km². If plots with a scale of 2 km² are used to estimate AOO (area of occupancy), then this species occurs within 4 km² of habitat, resulting in an AOO of less than 10 km². Therefore the Critically Endangered thresholds for extent of occurrence and area of occupancy are both met (EOO < 100 km² and AOO < 10 km²) (CR B1+2). The most serious threat to the species is habitat destruction through tavy practice (slash and burn agriculture), small-scale logging and the uncontrolled burning of neighbouring grasslands; a large out of control fire could easily affect the two subpopulations as they are separated by a distance of less than 2 km. Therefore, all individuals can be considered to occur within a single location only (CR B1a+2a). Given this on-going destruction of suitable habitat, population declines can be expected to continue unless some remedial action is taken (CR B1b(i, ii, iii, iv, v + B2b(i, ii, iii, iv, v)). Thus the species should qualify as Critically Endangered under criterion B (CR B1ab (i, ii, iii, iv, v)+2ab(i, ii, iii, iv, v) of the IUCN Red List (IUCN Standards and Petitions Subcommittee 2011).