Underestimated species diversity within the Rhacophorus rhodopus and Rhacophorus bipunctatus complexes (Anura, Rhacophoridae), with a description of a new species from Hainan, China

Taxonomy and species boundaries within the Rhacophorus rhodopus and Rhacophorus bipunctatus complexes are very confusing. In this study, we attempt to delimit the species boundaries and test the currently accepted taxonomic assignments within these two complexes based on newly collected samples and previously published data across their distributions. Phylogenetic analyses revealed that the R. rhodopus and R. bipunctatus complexes consisted of six distinct clades (labeled A‒F) that diverged from each other by genetic distances (p-distance) ranging from 5.3% to 9.2% in 16S rRNA sequences, and accordingly analyses of species delimitation placed them into six species, of which three correspond to known species ( R. rhodopus , R. bipunctatus , and R. na-poensis ) and three represent different cryptic species. Rhacophorus rhodopus (Clade C) is distributed in southern Yunnan, China, northern Laos, and northern and central Thailand; R. bipunctatus (Clade F) is distributed in northeastern India and western and northern Myanmar; and R. napoensis (Clade B) is distributed in Guangxi, China and northern Vietnam. Based on both molecular and morphological evidence, we described the clade consisting of samples from Hainan, China and central Vietnam (Clade A) as a new species, Rhacophorus qiongica sp. nov. There are two cryptic species requiring additional morphological studies: one only contains samples from Motuo, Xizang, China (Clade E), and the other is distributed in western Yunnan, China, central Myanmar, central Thailand, and Malaysia (Clade D). Additionally, our results supported the idea that some old GenBank sequences of R. reinwardtii need to be updated with the correct species name.

The disputes over the taxonomy of R. rhodopus and R. bipunctatus have been going on for many years.Inger et al. (1999) compared R. bipunctatus from northern and central Vietnam with R. rhodopus and concluded that the two species are synonymous.However, Bordoloi et al. (2007) considered that this conclusion is not reliable owing to the fact that Inger et al. (1999) did not compare the specimens from Vietnam with topotypes of R. bipunctatus, and suggested that records of R. bipunctatus from Thailand and Vietnam actually refer to R. rhodopus.Nguyen et al. (2008) also suggested that all records of R. bipunctatus in Vietnam should be classified as R. rhodopus.Fei et al. (2009) considered that R. rhodopus from the type locality obviously differs from R. bipunctatus from northern India in body size and color pattern, so they suggested maintaining the validity of R. rhodopus and considered that R. bipunctatus from Vietnam is more similar to R. rhodopus from China.
Analysis of molecular data can more accurately test the taxonomic hypothesis based on morphology (Jablonski and Finarelli 2009).Previous molecular phylogenetic analyses have demonstrated that the taxonomy of R. rhodopus and R. bipunctatus complexes (Chan et al. 2018) is complicated.Yu et al. (2007Yu et al. ( , 2008) ) found that R. rhodopus is not monophyletic, and R. rhodopus from Hainan, China is closer to R. bipunctatus collected from Vietnam.Li et al. (2012) found that samples of R. rhodopus from Hainan, China and Vietnam form a clade that did not cluster together with the clade containing R. rhodopus from the type locality, indicating that R. rhodopus from Hainan, China and Vietnam likely represents a cryptic species.Nguyen et al. (2014) also revealed that R. rhodopus is not monophyletic since R. rhodopus from Vietnam did not cluster together with the clade of samples from Yunnan and Malaysia.Dang et al. (2015) revealed two distinct lineages within R. rhodopus from Yunnan and considered that one of them could be a cryptic species.Moreover, Chan et al. (2018) suggested tentatively moving Malaysian R. bipunctatus to R. rhodopus and revealed that R. rhodopus from Vietnam is composed of two distinct clades, one only containing samples from Vietnam and one containing samples from Vietnam and Hainan, China.Over all, these previous molecular phylogenetic analyses revealed that R. bipunctatus and R. rhodopus represent two complexes (Chan et al. 2018), and at least two cryptic species may exist in the R. rhodopus complex: one occurs in Hainan, China and Vietnam, and one is only known from Vietnam.Recently, Li et al. (2022) described a new species resembling R. rhodopus from Guangxi, China (R. napoensis).This finding raises another question.That is, whether R. napoensis is conspecific with one of the two potential cryptic species within the R. rhodopus complex mentioned above.Additionally, Li et al. (2012) and Che et al. (2020) found that R. bipunctatus, which is likely restricted to northeastern India, Myanmar, and Xizang, China (Fei et al. 2009;Chan et al. 2018;Poyarkov et al. 2021), is not monophyletic since samples from Motuo, Xizang, China formed a clade whereas samples from northern Myanmar formed another clade.Because no samples from India were included in these two studies, it is necessary to investigate which of these two clades represents the true R. bipunctatus by employing R. bipunctatus samples from northeastern India.
In this study, we attempt to delimit the species boundary and test the currently accepted taxonomic assignments within R. rhodopus and R. bipunctatus complexes based on newly collected and previously sequenced samples across their distributions.Our results revealed that there are probably three cryptic species, and one of them was described as a new species herein based on morphological and molecular evidence.The samples from northern Vietnam belong to R. napoensis, and populations from central Vietnam are conspecific with the new species described here.

Molecular analyses and species delimitation
The total genomic DNA of the specimens was extracted from liver tissue preserved in 99% ethanol.Tissue samples were digested with proteinase K and purified using standard phenol/chloroform separation and ethanol precipitation.A fragment encoding partial 12S rRNA, complete tRNA val , and partial 16S rRNA (16S) was amplified using the protocol of Yu et al. (2019) and the primer pair L1091 (Kocher et al. 1989)/16H1 (Hedges 1994).Sequencing was conducted using the corresponding PCR primers and the internal primer Rh-int (Yu et al. 2019).
All new sequences have been deposited in GenBank under accession Nos.OP740711-OP740721, PP106375-PP106376, and PP115440-PP115442 (Table 1).Sequences were aligned in MEGA v. 7 (Kumar et al. 2016) using the MUSCLE option with the default parameters.The uncorrected pairwise distances (p-distances) between species were calculated in MEGA v. 7. The best substitution model (GTR + I + G) was selected in jMOD-ELTEST v. 2.1.10( Darriba et al. 2012) based on the corrected Akaike Information Criterion (AICc).Bayesian phylogenetic inference and maximum likelihood analysis were performed based on the best model.Bayesian phylogenetic inference was performed using MrBayes v. 3.2.6 (Ronquist et al. 2012).Two runs were performed simultaneously with four Markov chains starting from a random tree.The chain was run for 3,000,000 generations and sampled every 100 generations.When the average standard deviation of the split frequency was less than 0.01, the first 25% of the sampled trees were discarded as burn-in, and the remaining trees were used to create a consensus tree and estimate the Bayesian posterior probabilities (BPPs).The maximum likelihood analysis was conducted using raxmlGUI 2.0 (Edler et al. 2020) with 1000 bootstrap replicates.
We used Assemble Species by Automatic Partitioning (ASAP; Puillandre et al. 2021) and multirate PTP (mPTP; Kapli et al. 2017) to delineate the species boundary within the R. rhodopus and R. bipunctatus complexes based on 16S rRNA sequences.For the ASAP method, the substitution model of p-distances was used to compute the distances under the default values for other parameters.We selected the partition with the lowest ASAP-score as the best partition, according to Puillandre et al. (2021).The mPTP analysis was conducted in mPTP v. 0.2.5 using a maximum likelihood tree generated from 16S sequences by raxmlGUI 2.0.For this analysis, 10 different runs were performed with the following settings: mcmc run of 50 million generations, samples every 1000 generations, and the first 10 million generations were discarded as burn-in.

Morphology
As the molecular phylogenetic and species delimitation analyses revealed that species diversity in the R. rhodopus and R. bipunctatus complexes was underestimated and Hainan populations represent one of the three putative species (see below), we further conducted morphological analyses to confirm its taxonomic status and to officially describe it.The other two putative species were not included in morphological analyses owing to the fact that not enough morphometric data on them is available for the time being.
Measurements were corrected for size (measurements divided by SVL).We used the t-test in SPSS v. 17.0 (SPSS Inc., Chicago, IL, USA) to evaluate the differences in quantitative characters of adult males between Hainan populations and its two relatives (R. rhodopus and R. napoensis) because the Hainan populations were once placed in R. rhodopus and both the clade containing Hain- an populations and the clade of R. napoensis occur in Vietnam (see below).Principal component analyses (PCA) were conducted based on a correlation matrix of size-corrected measurements of males using SPSS v. 17.0.Scatter plots of the first two PCA factors were used to examine the morphological differentiation between specimens from Hainan, R. rhodopus, and R. napoensis.Females were not included as a separate analysis for both the t-test and PCA analysis owing to the small sample size (n = 2; one female from Hainan and one female of R. rhodopus).

Phylogeny and species delimitation
The BI and ML analyses yielded similar topologies, and both analyses revealed that there are six distinct clades in the R. rhodopus and R. bipunctatus complexes (Clades A-F; Fig. 2): Clade A contains samples from Hainan, China (sites 1 and 2) and Gia Lai, Vietnam (sites 3 and 4); Clade B contains types of R. napoensis  The sequences of specimens under the name R. reinwardtii in GenBank did not form monophyly.The two specimens that came from Malaysia (NMBE 1056517 and Rao081205) clustered together with R. borneensis and R. norhayatiae, respectively, and the two specimens that came from Indonesia (Java and Sumatra) formed a clade.
The best partition (score = 2.50) obtained by the ASAP species delimitation analysis grouped all samples of R. rhodopus and R. bipunctatus complexes used in this study into six species completely corresponding to the six distinct clades (A-F) mentioned above (Fig. 3a), with a distance threshold of about 4% (Fig. 3b).The clade consisting of R. borneensis and the specimen under the name R. reinwardtii (NMBE 105617), the clade containing R. norhayatiae and the specimen under the name R. reinwardtii (Rao081205), and the clade comprising of R. reinwardtii from Indonesia were identified as three different species (Fig. 3a).These operational taxonomic units were also supported by the mPTP analysis (Fig. 4).

Morphological study
Morphological data are summarized in Table 3.The t-tests revealed that the male specimens from Hainan differ significantly (p < 0.05) from male topotypes of R. rhodopus in tympanum diameter (TD), upper eyelid width (UEW), and distance between nostril and eye (DNE; Table 4), and differ from R. napoensis in body size (SVL), head length (HL), internarial distance (IND), and distance between nostril and eye (Table 5).PCA analysis on Hainan populations and R. rhodopus revealed that the first two principal components accounted for 66.13% of the total variance (Table 6), and loadings for PC2 were heavily loaded on TD and UEW, which separated Hainan samples from R. rhodopus along the PC2 axis (Fig. 5a).PCA analysis on Hainan populations and R. napoensis showed that loadings for PC2 were heavily loaded on HL, which can effectively separate Hainan populations from R. napoensis along the PC2 axis (Fig. 5b).Additionally, morphological comparison indicated that specimens from Hainan can be distinguished from known members of R. rhodopus and R. bipunctatus complexes and other related species of Rhacophorus by a series of characters (see below).
Based on the above molecular and morphological evidence, we considered that misidentifications were involved in the R. rhodopus and R. bipunctatus complexes, and herein, the clade consisting of specimens from Hainan and central Vietnam (Clade A) is described as a new species.Type material.Holotype.GXNU YU000691, adult male, collected on 14 July 2023 by Lingyun Du from Diaoluo Mt., Lingshui, Hainan, China (18°43'28"N, 109°52'12"E, ca 914 m a.s.l.).Paratypes.GXNU YU000690, an adult male, collected at the same time as the holotype from the type locality by Lingyun Du and Jiaqi Luo; GXNU HN110501-HN110503, three adult males, collected on 20 July 2021 by Fanrong Xiao from the type locality; and three adult males (GXNU YU000693, GXNU YU000696, GXNU  Etymology.The specific name qiongica is derived from Qiong (琼), referring to Hainan, China, and meaning good and beautiful in Chinese.The specific name means that this species is very beautiful, and in China, it is distributed in Hainan.We suggest the English common name "Hainan flying frog" and the Chinese common name "琼树蛙 (Qióng Shù Wā)".
Diagnosis.The new species is assigned to Rhacophorus by the presence of intercalary cartilage between terminal and penultimate phalanges of digits, terminal phalanges of fingers and toes Y-shaped, the tip of the digits expanded into disks with circummarginal grooves, fingers webbed, tarsal projections present, dermal folds along the forearm and tarsus present, and a horizontal pupil (Jiang et al. 2019).Rhacophorus qiongica sp.nov.differs from its congeners by a combination of the following characters: 1) medium body size (adult males SVL 35.1-38.2mm); 2) dorsal surface red brown; 3) entire web between fingers and toes; 4) webbing between toes purely scarlet; 5) small black blotches on flank; 6) bands on limbs distinct; 7) throat smooth; 8) absence of dermal appendage on snout tip; 9) absence of small black spots on head sides; 10) palm smooth without small tubercles; and 11) tibiotarsal articulation reaching beyond eye.
Description of holotype.Adult male, body size medium (SVL 37.8 mm); head width (HW 13.2 mm) longer than head length (HL 12.0 mm); snout blunt pointed, sloping in profile, protruding beyond the margin of lower jaw in ventral view; snout length (SL 5.5 mm) longer than diameter of eye (ED 4.4 mm); canthus rostralis distinct, curved; loreal region oblique, concave; nostril oval, lateral, slightly protuberant, slightly closer to tip of snout than to eye; internarial space (IND 3.8 mm slightly smaller than interorbital distance (IOD 4.2 mm), nearly equal to the width of the upper eyelid (UEW 3.6 mm); pupil horizontal; pineal ocellus absent; tympanum distinct (TD 2.3 mm), rounded, about half eye diameter (ED 4.4 mm); supratympanic fold narrow, flat; tongue cordiform, attached anteriorly, notably notched posteriorly; choanae oval; vomerine teeth present in two series, touching the inner front edges of the choanae; an internal single subgular vocal sac; a vocal sac opening on the floor of the mouth at each corner.
Forelimbs thin, length of forearm and hand (FHL 18.2 mm) is about half snout-vent length; relative length of fingers I < II < IV < III; tips of all fingers expand into discs with circummarginal and transverse ventral grooves, disc of finger I smaller than discs of other fingers; entire web between fingers, webbing formula: I2-2II1-1.5III1-1IV;subarticular tubercles rounded and prominent, formula 1, 1, 2, 2, proximal one smaller than distal one on the third and fourth fingers; supernumerary tubercles below the base of finger absent; metacarpal tubercle single, inner, oval and prominent (Fig. 7).
Dorsal skin smooth with very fine granules; throat and ventral surface of forelimbs smooth; chest, belly, and ventral surface of thighs granular (Figs 6, 7); dermal folds on forearm, tarsus, heels, and vent present.
Coloration in life.Iris light brown; dorsal surface red brown, mottled with two discontinued rows of dark patches and scattered with small black spots on dorsum; dark brown bands and small black spots on dorsal surface of limbs; upper part of flank orange red and lower part of flank orange yellow, scattered with a few small black blotches; skin of ventral surface semi-transparent, mottled with orange yellow on throat and belly; ventral, anterior, and posterior surfaces of limbs orange yellow; discs of fingers and dorsal surface of fingers I-III orange yellow; discs of toes and toes I-IV red; web between fingers yellow, mottled with red; web between toes completely red.Color of holotype in preservative.The color faded, but the pattern remained the same as in life.Dorsal surface brown, with dark patches and spots; dorsal side of limbs barred with dark brown; throat, chest, belly, webbing, ventral surface of limbs, and anterior and posterior parts of thighs faded to yellowish; a few small black blotches on flank.
Sexual dimorphism.The body size of males is smaller than that of female; adult males have an internal single subgular vocal sac with a pair of slit-like openings on the floor of the mouth at each corner.Additionally, adult males have a milk-white nuptial pad on the inner side of the base of finger I.
Morphological variation.The number of small black spots on the flank varied among specimens.The holotype GXNU YU000691 and two paratypes (GXNU YU000698 and GXNU HN110502) have multiple small black spots on flank; paratypes GXNU YU000690 and GXNU YU000697 have no black spots on flank; and paratypes GXNU YU000693 and GXNU YU000696 have two small black spots on flank (Fig. 8).Additionally, dorsal color pattern also varied among specimens, as the two paratypes GXNU YU000696 and GXNU YU000698 have yellowish-brown blotches on dorsal surfaces of the body and limbs (Fig. 9).

Distribution and ecology.
The species is distributed in Hainan, China and Gia Lai, Vietnam.In Hainan, the species was found usually in shrubs and small arbors at elevations ranging from 600 to 850 m (Shi 2011; as R. rhodopus) and called from 19:00 to 03:00 every night during the breeding season (from May to July), with a peak at about 22:00 (Sun et al. 2017;as R. rhodopus).The types in this study were found in roadside bushes ca.1-2 m above the ground (Fig. 10).There were temporary puddles under the bushes, and there is a lake (Tianchi) and a stream nearby the road in the type locality.Chirixalus doriae Boulenger, 1893, Kurixalus hainanus (Zhao, Wang & Shi, 2005), and Polypedates megacephalus Hallowell, 1861 were also found in sympatry at the type locality.

39
Both the present and previous phylogenetic analyses revealed that R. norhayatiae, R. reinwardtii, R. kio, R. borneensis, and R. helenae are imbedded in the R. rhodopus and R. bipunctatus complexes.The new species can be easily distinguished from these five species by the dorsal surface being red brown (vs.green) and the web between toes being red with no black pigmentation (vs.black pigmentation present).

Discussion
The taxonomy within the R. rhodopus and R. bipunctatus complexes is complicated owing to the similar external morphology among members of these two complexes, which has heavily hampered the identification of species and understanding of the species boundary in these two complexes.For example, the Xizang population was once recorded as R. rhodopus (e.g., Fei et al. 2010;Li et al. 2012) or R. bipunctatus (Che et al. 2020); Nguyen et al. (2008) and Poyarkov et al. (2021) suggested that existing records of these two complexes from Vietnam are actually of R. rhodopus, but it has been revealed that Vietnamese R. rhodopus may represent a cryptic species (Li et al. 2012), suggesting that a species delimitation within these two complexes based on broad sampling is necessary.In this study, based on newly collected samples and previously published data, our phylogenetic analyses revealed that the R. rhodopus and R. bipunctatus complexes contain six distinct clades (A-F), and they were assigned to six species by the analysis of species delimitation, indicating the species diversity of these two complexes was underestimated and the distribution range of members of these two complexes needs to be re-defined.
According to previous phylogenetic analyses (Yu et al. 2007;Li et al. 2012), there are two distinct clades (C and D) in Yunnan, China.Clade C contains samples from southern Yunnan, China, northern Laos (Phongsali and Louangphrabang), northern (Phetchabun, Phitsanulok, and Chiang Mai) and central (Chachoengsao) Thailand.Undoubtedly, this clade represents the true R. rhodopus because the topotypes of this species were grouped in it.Clade D contains samples from western and southwestern Yunnan, central Myanmar (Mandalay), central Thailand (Prachuap Khiri Khan), and Malaysia.Although it was recovered as the sister taxon to the clade C that represents the true R. rhodopus, genetic divergence in 16S sequences between them reaches 5.3%, which is greater than the distance between R. helenae and R. kio (4.5%), the distance between R. helenae and R. borneensis (4.9%),Table 7. Morphological comparison between the new species and members of Rhacophorus rhodopus and R. bipunctatus complexes.Characters are: ① dorsal color: 0 = brown, 1 = green; ② black spots on flank: 0 = small, 1 = large; ③ bands on limbs: 0 = distinct, 1 = indistinct; ④ throat: 0 = smooth, 1 = granular; ⑤ snout: 0 = pointed without appendage on tip, 1 = pointed with appendage on tip; ⑥ black spots on head side: 0 = absent, 1 = present; ⑦ palm: 0 = smooth without tubercles, 1 = rough with tubercles; ⑧ tibiotarsal articulation: 0 = reaching beyond eye, 1 = reaching eye."?" means unknown.and the distance between R. norhayatiae, R. borneensis, and R. reinwardtii (4.0%-4.8%;Table 2), and they were assigned into two different species by the analyses of species delimitation.Therefore, we consider that clade D probably represents a cryptic species pending additional morphological studies.Rhacophorus bipunctatus was originally described from Khasi Hills, Northeast India.Previous phylogenetic analyses (Li et al. 2012;Che et al. 2020) revealed that samples of R. bipunctatus from Myanmar and southern Xizang, China form two distinct clades.Owing to the absence of R. bipunctatus samples from Northeast India in these two previous studies, it is confusing which clade represents the true R. bipunctatus.In this study, we found that R. bipunctatus from Northeast India (Mizoram), north Myanmar (Kachin), and west Myanmar (Chin) form a clade (Clade F), whereas samples from southeastern Xizang form an independent clade (Clade E).Considering that Mizoram is close to Khasi Hills, the type locality of R. bipunctatus, and records from there have been confirmed based on morphology (Decemson et al. 2020), we consider that clade F is true R. bipunctatus and the Xizang population (Clade E) probably represents a cryptic species.Morphologically, the Xizang population differs from R. rhodopus and R. qiongica sp.nov.by throat granular (Che et al. 2020) since the latter two are smooth (Liu and Hu 1960; this study), and from R. napoensis by smaller body size (male SVL 31.6-38.7 mm, mean = 34.9mm, n = 10), head slightly longer than wide, and tibiotarsal articulation reaching eyes (Che et al. 2020) (vs. SVL 39.7-44.2 mm [mean = 41.3, n = 5] in males, head wider than long, and tibiotarsal articulation reaching snout; Li et al. 2022).Additionally, bands on limbs in the Xizang population are distinct (Che et al. 2020), but bands on limbs in specimens of R. bipunctatus from northeastern India and Myanmar are indistinct (Bordoloi et al. 2007;Fig. 11).
Records in Vietnam were once placed in R. bipunctatus or R. rhodopus, and recently, Poyarkov et al. (2021) conjectured that all existing records of R. rhodopus and R. bipunctatus complexes in Vietnam are actually of R. rhodopus.In this study, we found that there are two distinct clades in Vietnam (Clades A and B).The samples from northern Vietnam formed Clade B with R. napoensis, indicating that they actually refer to R. napoensis, while the samples from central Vietnam (Gia Lai) and Hainan, China formed Clade A, which is morphologically distinguishable from other known species of R. rhodopus and R. bipunctatus complexes and is described as R. qiongica sp.nov.herein.The amphibian and reptile flora of Hainan Island is clearly dominated by Oriental species (Shi 2002;Wang et al. 2004Wang et al. , 2009)), and the species composition seems to be closely related to the southern part of the Chinese mainland and Vietnam.Although Hainan is separated from the Asian mainland by the Beibu Gulf and Qiongzhou Strait, they have been joined frequently over the past million years (Ali 2018).Rhacophorus qiongica sp.nov.likely colonized Hainan via landbridge dispersal during Pleistocene climatic oscillations, which has been suggested for another tree frog in Hainan, K. hainanus, a species widely distributed in Vietnam and southern China (Yu et al. 2020).
Like Li et al. (2012) and Chan et al. (2018), phylogenetic relationships within R. rhodopus and R. bipunctatus complexes were also not resolved well in the present study.Our results strongly supported that the clade containing topotypes of R. rhodopus (Clades C) is sister to the clade containing samples from western Yunnan (Clade D), and the two together are sister to the clade composed of samples from Xizang, China (Clade E), which is consistent with Li et al. (2012).Chan et al. (2018) also recovered the clade containing samples from Laos (labeled as Clade 2 in Fig. 4 of Chan et al. (2018) and corresponds to Clade C of this study) as sister to the clade containing samples from Malaysia (labeled as Clade 3 in Fig. 4 of Chan et al. (2018) and corresponds to Clade D of this study) with strong support, but they recovered these two clades together as sister to the clade containing samples from Hainan, China and Gia Lai, Vietnam (labeled as Clade 1 in Fig. 4 of Chan et al. (2018) and corresponds to Clade A of this study).This difference may be caused by the absence of samples from Xizang, China in Chan et al. (2018).In addition, Chan et al. (2018) did not resolve the phylogenetic placement of the clade of the R. reinwardtii group, but both the present study and Li et al. (2012) recovered that the clade consisted of the R. reinwardtii species group as sister to the group formed by clades C-E with strong support.Neither this nor the two previous studies (Li et al. 2012;Chan et al. 2018) resolved the phylogenetic placement of the three clades corresponding to R. bipunctatus, R. napoensis, and R. qiongica sp.nov., respectively.Therefore, more studies are needed to resolve the phylogenetic relationship among the R. rhodopus and R. bipunctatus complexes.
Additionally, it is worth noting that some old Gen-Bank sequences of R. reinwardtii need to be updated with the correct species name.Rhacophorus norhayatiae and R. borneensis were described as distinct species by Chan and Grismer (2010) and Matsui et al. (2013), respectively, and both of them were confused with R. reinwardtii prior to their naming (Chan and Grismer 2010;Matsui et al. 2013).The type locality of R. reinwardtii is Mount Pangerango, Java, Indonesia.In this study, we found that the two samples under the name of R. reinwardtii from Malaysia (Rao081205 and NMBE 1056517) were not grouped together with the lineage of R. reinwardtii from Indonesia (Java and Sumatra), but clustered together with R. norhayatiae and R. borneensis, respectively.The sequence from the specimen Rao081205 (JX219443) was submitted to GenBank in June 2012, posterior to the taxonomic revision of the R. reinwardtii/R.norhayatiae group (Chan and Grismer 2010), and the sequence from the specimen NMBE 1056517 (JN377366) was submitted to GenBank in July 2011, prior to the taxonomic revision of the R. reinwardtii/R.borneensis group (Matsui et al. 2013).These evidences suggest that the specimen Rao081205 is a misidentification of R. norhayatiae, and the name of the specimen NMBE 1056517 is not yet updated in GenBank.This result is partially consistent with the viewpoint of Chan and Grismer (2010) that records of R. reinwardtii from Malaya are referable to R. norhayatiae.
In summary, based on newly collected samples and previously published data, we obtained a clearer delineation of species boundaries within the R. bipunctatus and R. rhodopus complexes.We recovered six distinct clades (A-F) in these two complexes.

Figure 1 .
Figure 1.Map showing the collection sites of samples of the R. rhodopus and R. bipunctatus complexes used in this study.Sites are labeled as inTable 1, and the star represents the type locality of R. qiongica sp.nov. in Hainan, China.

Figure 2 .
Figure 2. Bayesian phylogenetic tree of R. rhodopus and R. bipunctatus complexes and related species inferred from 12S rRNA, tRNA Val , and 16S rRNA genes.The numbers above and below the branches are Bayesian posterior probabilities (BPP) and maximum likelihood (ML) bootstrap values, respectively (only values greater than 50% are shown).

Figure 3 .
Figure 3. ASAP species delimitation based on 16S rRNA sequences used in this study.The best partition with the lowest score is highlighted with a red frame.

Figure 4 .Figs 6- 9
Figure 4. mPTP species delimitation based on ML tree generated from 16S rRNA sequences.The support values above the branches indicate the fraction of sampled delimitations in which a node was part of the speciation process.

Figure 5 .
Figure 5.The scatter plot of the principal component analysis based on size-adjusted morphological data from males of the new species plus R. rhodopus (a) and data from males of the new species plus R. napoensis (b).

Figure 7 .
Figure 7. Dorsal, ventral, and lateral views of the holotype of R. qiongica sp.nov.(a-c) in preservative and ventral views of its hand (d) and foot (e).

Figure 8 .
Figure 8. Variation of black spots on flank among paratypes of R. qiongica sp.nov.from Hainan, China.

Figure 10 .
Figure 10.Habitat of R. qiongica sp.nov. at the type locality.
Rhacophorus rhodopus (Clade C) is distributed in southern Yunnan, China, northern Laos, and northern and central Thailand; R. bipunctatus (Clade F) is only distributed in northeast India and western and northern Myanmar; R. napoensis (Clade B) is distributed in Guangxi, China and northern Vietnam; and R. qiongica sp.nov.(Clade A) is distributed in Hainan, China and central Vietnam.There are two cryptic species requiring additional morphological studies: one only contains samples from Motuo, Xizang, China (Clade E), and the other is distributed in western Yunnan, China, central Myanmar, central Thailand, and Malaysia (Clade D).More studies are needed to resolve the phylogenetic relationship among the R. rhodopus and R. bipunctatus complexes.

Table 1 .
Species used in phylogenetic analyses of this study.

Table 2 .
Mean uncorrected pairwise distances (%) between clades of Rhacophorus rhodopus and R. bipunctatus complexes and related species based on 16S rRNA sequences.
ic analyses supported that clades C, D, and E form a monophyly, and clade C is sister to clade D. Clade F was recovered as sister to the clade of R. helenae and R. kio with moderate support, and clade A was recovered as sister to clade B with weak support.

Table 4 .
Results of the t-test between male specimens of Rhacophorus qiongica sp.nov.and R. rhodopus based on the size-adjusted data except SVL.

Table 5 .
Results of the t-test between male specimens of Rhacophorus qiongica sp.nov.and R. napoensis based on the size-adjusted data except SVL.

Table 6 .
Factor loadings of first two principal components of 13 size-adjusted morphometric characteristics of male specimens of Rhacophorus qiongica sp.nov., R. rhodopus, and R. napoensis. CharacterR.