Taxonomic notes on the Asian frogs of the tribe Paini (Ranidae, Dicroglossinae): 1. Morphology and synonymy of Chaparana aenea (Smith, 1922), with proposal of a new statistical method for testing homogeneity of small samples

The species from northern Vietnam described as Rana fansipani Bourret, 1939 is shown to be a synonym of the species from northern Thailand described as Rana aenea Smith, 1922. The male secondary sex characters of this species are described in detail for the first time. Combined with a few other characters, they allow this species to be distinguished from the closely related Rana unculuanus Liu, Hu and Yang, 1960 from southern China. Both species should be placed in a subgenus Chaparana (Chaparana), a member of the tribe Paini of the subfamily Dicroglossinae of the Ranidae. In this study, we propose a new statistical method for testing homogeneity of small samples: this consists of considering the values obtained for each specimen and for each regression factor score as morphometric characteristics of this specimen within the frame of this analysis; these values, or some of them, can be used to compare subgroups of specimens by non‐parametric tests such as the Mann–Whitney U test.


Introduction
This is the first of a series of papers dealing with the taxonomy of the frogs of the tribe Paini, a rather poorly-known group of frogs living, mostly in running waters, in the mountains of South and South-East Asia (Himalayas, mountains of northern Indochina and southern China) (Dubois 1976(Dubois , 1992. Bourret (1939) established a new subgenus Chaparana and a new species Rana (Chaparana) fansipani for a single specimen of torrent frog collected on mount Fan Si Pan in northern Vietnam. In subsequent years, no other specimen was referred to this species, whose status remained unclear. Dubois (1977) re-examined the holotype and pointed to several inaccuracies in Bourret's (1939Bourret's ( , 1942 descriptions. He suggested that this species was close to the group for which he had earlier (Dubois 1975) erected the subgenus Paa of the genus Rana, but his conclusion about the taxonomic allocation of this species remained prudent, as the holotype and only known specimen was a juvenile male: as in other cases of frog species named on the basis of single juvenile specimens (see e.g. Ohler and Dubois 1999), taxonomic allocation of the species was difficult due to the absence in this specimen of male secondary sex characters. Later, Dubois (1992) suggested that, by its external morphological characters, the species Rana fansipani appeared closest to Rana aenea Smith, 1922 from Thailand (a species described on the basis of two juvenile female specimens) and Rana unculuanus Liu, Hu and Yang, 1960 from Yunnan (China), and he grouped these three species in a subgenus Chaparana (Chaparana). In doing so, he assumed that these three species had similar male secondary sex characters, i.e. the presence of spines in the anal region, although this character had actually been observed only in the Chinese species.
In 1997, while doing field work on the Fan Si Pan, the Frontier field staff (see Dubois and Ohler 1998, p 2) collected three frog specimens (MNHN 1999(MNHN .5818, 1999) that proved to be adult males of Chaparana fansipani . These specimens are devoid of spines in the vent region, but on the other hand they show black cornified spines on the throat, the chest and the first three fingers, thus closely resembling several species of the subgenus Paa (Paa), as redefined by Dubois (1992). Comparison of these specimens and of the holotype of Rana fansipani with the type-specimens of Rana aenea and additional specimens from Thailand referred to this species further suggests that these two nominal species are nothing but synonyms: the species they belong to should therefore be known as Chaparana aenea (Smith, 1922). The present paper is first devoted to a redescription of the holotype of Rana aenea Smith, 1922 and the description of an adult breeding male of this species from northern Vietnam, topotype of Rana fansipani Bourret, 1939. The species Chaparana aenea (Smith, 1922) is then compared with its close relative Chaparana unculuanus (Liu, Hu and Yang, 1960). Both for the intraspecific and interspecific comparisons, we propose a new method of comparison between two small samples, based on the results of a principal component analysis. This method may prove useful for statistical morphometric comparisons between samples when only small samples are available for measurement, a common situation in amphibian taxonomy.
Specimens were sexed using their external characters (in the case of adult breeding males) or through a slight lateral incision in order to see the gonad: the state of development of the latter and its ducts allowed two ontogenetic stages to be distinguished, juvenile and adult, as explained in detail by Dubois (1976, p 31-33).
Two specimens were described in detail using the same format and methodology as in several of our previous works on Asian anurans (Dubois and Ohler 1998Ohler and Dubois 1999;Ohler et al. 2000Ohler et al. , 2002Bossuyt and Dubois 2001;Delorme and Dubois 2001;Veith et al. 2001). The webbing formulae are given according to Myers and Duellman (1982).
Measurements of specimens were taken with a slide calliper to the nearest 0.1 mm, or, for values below 5 mm, with an ocular micrometer to the nearest 0.01 mm. The list and description of measurements are given in the Appendix.
Univariate morphometric comparisons between samples were made using the nonparametric Mann-Whitney U test and the Kruskal-Wallis test (Zar 1984). Besides, we performed multivariate factor analyses based on these measurements, using principal component analysis (PCA) with varimax rotation as implemented in the SPSS software (SPSS Inc. 1999, p 426). Special problems are posed for the taxonomic allocation of specimens and for morphometric sample comparisons in the case of small or very small samples . Here we propose a new method for such comparisons. Because of small sample sizes and because multivariate normal distribution cannot be assumed, it is not appropriate to use a parametric test for the significance of a difference between cluster centroids obtained by PCA (Sneath and Sokal 1973, p 287). However, the values obtained for each specimen and for each regression factor score can be used as morphometric characteristics of this specimen within the frame of this analysis. Therefore we used some of these values (actually only those of the first three regression factor scores) to compare subgroups of specimens by the Mann-Whitney U test. Subgroups were composed according to four possible criteria: taxonomic allocation (for the interspecific comparisons); and geographical origin, sex and ontogenetic stage (for intraspecific comparisons). For the three kinds of intraspecific subgroups we corrected for multiple comparisons using Bonferroni correction: a95a/k, with a50.05 and k53. However, the use of Bonferroni correction reduces the possibility of detecting differences, and in the case of comparison of populations it does not guarantee a ''prudent'' interpretation of the results (Perneger 1998). So usage of such a correction should not be automatic, but needs to be in agreement with biological plausible results. We did not apply this correction to the comparisons within a series of regression factor scores, as a significant result in one of these comparisons is enough to document the existence of a difference between groups (see Perneger 1998).
Descriptions of specimens referred to the species Chaparana (Chaparana) aenea (Smith, 1922) Holotype of Rana aenea Smith, 1922(Figure 1a-c) Smith (1922 described Rana aenea on the basis of two juvenile female specimens: (1) MAS 5821 (SVL 35 mm), which was clearly designated as ''type'' (we would now write ''holotype'') and shown in his Figure 1 (reproduced in Bourret 1942, p 277), and (2) MAS 5822 (SVL 38 mm), designated as paratype. Both specimens were later deposited in the collection of the British Museum (Natural History), now the Natural History Museum, London. Following is a detailed redescription of the holotype.
Holotype : BMNH 1947.2.2.31 (ex BMNH 1922, juvenile female, collected in May 1920 by a native collector at Doi Chang (19u239N, 98u529E), on the border of Chiang Mai and Mae Hong Son Provinces, Thailand (then ''northern Siam''), about 1500 m altitude. This specimen has been dried up and is in rather bad condition.
Size and general aspect. (1) Specimen of rather small size (SVL 35.2 mm), body rather slender. Head.
(4) Canthus rostralis rounded, loreal region slightly concave, obtuse in cross-section. (5) Interorbital space flat, smaller (IUE 2.63 mm) than upper eyelid (UEW 3.19 mm) and internarial distance (IN 4.50 mm); distance between front of eyes (IFE 6.3 mm) more than one-half of distance between back of eyes (IBE 10.5 mm). (6) Nostrils closer to eye (EN 2.38 mm) than to tip of snout (NS 3.25 mm). (7) Pupil horizontal oval, very dilated. (8) Tympanum (TYD 2.63 mm) distinct, rounded, much wider than half of eye; tympanum-eye distance (TYE 1.25 mm) less than half its diameter. (9) Pineal ocellus absent. (10) Vomerine ridges present, bearing teeth, between posterior parts of choanae, with an angle of 45u to body axis, slightly closer to each other than to choanae, slightly longer than distance between them. (11) Tongue not examined. (12) Supratympanic fold distinct, from eye to shoulder. (13) Parotoid glands absent. (14) Cephalic ridges absent. (15) Co-ossified skin absent. Hindlimbs. (23) Shank more than four times longer (TL 21.1 mm) than wide (TW 4.8 mm), longer than thigh (FL 18.9 mm) and than distance from base of internal metatarsal tubercle to tip of toe IV (FOL 18.9 mm). (24) Toes long, thin; toe IV rather long (FTL 14.4 mm), more than one-third of distance from base of tarsus to tip of toe IV (TFOL 29.4 mm). (25)  Coloration in alcohol. (38) Dorsal and lateral parts of head and body: uniform brown with several darker bands and spots (a band between eyes, spots on anterior part of eyelids, a band from eye to snout, a band enclosing tympanum and around it, and a band along anterior third of latero-dorsal fold); clearer spots on upper lip, and clear triangle on tip of snout; flanks darker in their upper than in their lower part, with distinct small darker spots everywhere. (39) Dorsal parts of limbs: brown, with two darker bands on forelimb and four narrow bands on each hindlimb. (40) Ventral parts: throat orange with brown marblings and numerous darker spots on margins; chest and belly yellow orange; thighs orange; back of thighs brown with clearer spots and with darker spots similar to those of flanks; webbing marbled brown and cream. Size and general aspect.
(2) Head of large size, wider (HW 30.1 mm) than long (HL 27.3 mm; MN 23.1 mm; MFE 18.0 mm; MBE 11.4 mm), flat. (3) Snout rounded, very slightly protruding, its length (SL 11.3 mm) longer than horizontal diameter of eye (EL 7.7 mm). (4) Canthus rostralis rounded, loreal region concave, obtuse in cross-section. (5) Interorbital space flat, smaller (IUE 4.7 mm) than upper eyelid (UEW 5.4 mm) and internarial distance (IN 8.4 mm); distance between front of eyes (IFE 11.2 mm) more than one-half of distance between back of eyes (IBE 17.9 mm). (6) Nostrils closer to eye (EN 4.5 mm) than to tip of snout (NS 7.1 mm). (7) Pupil not observed. (8) Tympanum indistinct. (9) Pineal ocellus absent. (10) Vomerine ridges present, bearing few teeth, between posterior parts of choanae, with an angle of 40u to body axis, slightly closer to each other than to choanae, slightly longer than distance between them. (11) Tongue large, rounded, emarginate. (12) Supratympanic fold prominent, from eye to above shoulder. (13)  Hindlimbs. (23) Shank four times longer (TL 48.4 mm) than wide (TW 12.8 mm), longer than thigh (FL 42.0 mm) and than distance from base of internal metatarsal tubercle to tip of toe IV (FOL 42.8 mm). (24) Toes long, rather thin; toe IV rather long (FTL 23.7 mm), more than one-third of distance from base of tarsus to tip of toe IV (TFOL 62. Coloration in alcohol. (38) Dorsal and lateral parts of head and body: greyish brown with several darker bands and spots (a mid-dorsal chevron, a band from eye to snout, a band enclosing tympanum and around it, and a band along anterior third of latero-dorsal fold); clearer spots on upper lip, and clear triangle on tip of snout; flanks darker in their upper than in their lower part, with some darker spots. (39) Dorsal parts of limbs: brown, with indistinct darker bands; posterior part of thigh dark brown with white marbling. (40) Ventral parts: throat whitish with brown spots on margins; chest, belly and thighs whitish; webbing greyish brown.
Male secondary sex characters. (41) Large-sized, blackish nuptial spines present on prepollex, fingers I-III, chest and throat, forming two separate pads on finger I and prepollex and two widely separated plates of tightly connected spines on chest. Table I gives the numbers of spines in these different places. (42) Forearms distinctly enlarged. (43) Vocal sacs and slits absent.
Other specimens referred to the species Chaparana aenea (Smith, 1922) To our knowledge, beside the two specimens described in detail above, only 11 other specimens referable to this species are known, which are briefly discussed below. Table II gives the measurements of the 13 specimens here discussed, and Figure 2 shows the known localities of this species in China, Thailand and Vietnam, as well as areas of known occurrence of the closely related species Chaparana unculuanus (see below). BMNH 1974.1113. Juvenile (SVL 38.6 mm), paratopotype of Rana aenea from Doi Chang. This specimen, slightly larger than the holotype, has similar characters.

Specimens from Thailand
MNHN 1989.0712 (Figure 3a-c). Non-breeding adult male (SVL 77.4 mm) from Doi Inthanon (18u359N, 98u289E), Chiangmai Province, Thailand. Except for its absence of secondary sex characters, this specimen is very similar to the three adult males from Fan Si Pan mentioned above and below. The major differences between them are: the complete absence of nuptial spines on fingers, breast and throat and the absence of enlargement of the forearms, as in the Vietnamese specimen MNHN 1999.5818. The ventral colour pattern is similar in all specimens, but the specimen from Doi Inthanon has a rather clear fawn dorsal colour, whereas the males from Fan Si Pan are dark brown on the back: this difference may be related to nycthemeral colour variation, as specimens collected at night For each item, the table gives the number of spines on the left side of the body followed by that on the right side: e.g. 20/19. The maximum total given in the last column is the total of the maximum value for each item on either the left or the right side of the body (see Dubois 1976, p 59;Dubois and Matsui 1983, p 900-901). Juv., juvenile (as defined in Dubois 1976, p 31-33); nm, measurement not taken on this specimen. Of the 36 measurements, only 32 are available for all 13 specimens and were used for the principal component analysis of Figure 5 and Table III. Taxonomic notes on the tribe Paini are darker than specimens collected during the day. This specimen is clearly adult, and its absence of spines on the fingers and chest indicates that it was not collected during the breeding season. However, its first finger is notably enlarged, and this finger and the prepollex show structures which correspond to basements of spines. 103u479E), Sa Pa District, Lao Cai Province, Vietnam. This specimen was described in detail by Dubois (1977), who wondered whether this frog, clearly a male, was juvenile, subadult or adult. It is very similar in all its characters to the three adult males collected in 1997-1998 on the same mountain, except for its absence of male secondary sex characters: given its much smaller size (SVL 54.5 mm versus 64.7-74.6 mm in the three adult males), it is now clear that it is not adult, nor even subadult, but juvenile, hence its absence of any secondary sex character. This specimen is also very similar in all its characters to the lectotype and paralectotype of Rana aenea mentioned above.
MNHN 1999.5820. Adult male (SVL 64.7 mm), topotype of Rana fansipani Bourret, 1939, collected on 17 November 1997at Cat Cat (128022u199N, 103u499E), on Fan Si Pan mountain, Vietnam. This second specimen is very similar to the specimen MNHN 1999.5821 described in detail above. The major differences between them are the secondary sex characters that are less developed in MNHN 1999.5820: the spines on the breast are large and blackish, but isolated from each other and less numerous (Table I); on the throat, blackish spines are distinct, but small; the forearms are distinctly enlarged.  ; 22u239N, 103u489E), on Fan Si Pan mountain, Vietnam. This specimen is similar to the other two adult males mentioned above, but in a better condition of preservation. The secondary sex characters are similar to those of MNHN 1999.5820, but the blackish horny covering of the nuptial spines is lost on the breast and on the fingers, and the spines are less numerous (Table I) 2050 m; 23u019N, 104u219E), Huang Lian Shan Nature Reserve, Yunnan, China. This specimen, the only one of this species known from China, is very similar in aspect and coloration to the Vietnamese specimens. It has a very conspicuous chevron on the anterior part of the back.

Morphometric comparisons
Intraspecific comparisons within Chaparana aenea (Smith, 1922) Despite the low number of specimens referable to the species Chaparana aenea available, we tried to carry out comparisons between them. As mentioned above, in their overall aspects and colours these 13 specimens are very similar. Part of the variation observed is related to age (size) and sex (especially in breeding adult males). However, the shapes and proportions of these specimens, as studied through 32 measurements (Table II), do not show any clear geographical trend, as shown in Figure 5 which presents the distribution of the 13 specimens according to the first two regression factor scores provided by the PCA. Before and after application of Bonferroni correction, Mann-Whitney U tests based on the first three regression factor scores of PCA (Table III) do not show significant differences between subgroups of these 13 specimens, based either on sex (males versus females) or on main geographic origin (mountains of northern Vietnam and southern Yunnan versus mountains of northern Thailand). However, prior to Bonferroni correction, when ontogenetic stages (juveniles versus adults) are considered, the first regression factor scores of juveniles are statistically different from those of adults. After Bonferroni correction the value is slightly too high for significance. The existence of significant differences between different ontogenetic stages is meaningful as it points to allometric growth in this species, as is usual in frogs, and especially in this group of torrent frogs (Dubois 1976). From these limited available data, we conclude that, for the time being, no significant difference is documented between populations from Thailand, Vietnam and China referred to this species, and we consider the name Rana fansipani Bourret, 1939 as a junior subjective synonym of Rana aenea Smith, 1922. Further data, based on more specimens and using additional information (molecular data, bioacoustics, tadpole morphology) will be necessary to confirm or refute this conclusion.
Comparisons between Chaparana aenea (Smith, 1922) and Chaparana unculuanus (Liu, Hu and Yang, 1960) Besides Chaparana aenea (Smith, 1922) and its synonym Chaparana fansipani (Bourret, 1939), Dubois (1992) included a third nominotypical species in his subgenus Chaparana (Chaparana), namely Chaparana unculuanus (Liu, Hu and Yang, 1960). The latter (Figure 4d-f) is indeed very similar to Chaparana aenea in general morphology, and both species occur in southern Yunnan, although in different areas ( Figure 2). However, these two species differ markedly in their male secondary sex characters: as discussed above, breeding males of Chaparana aenea have enlarged forelimbs and black cornified spines on the fingers and chest, but no spines around the vent. In contrast, breeding males of Chaparana unculuanus are devoid of spines on fingers and chest and their forelimbs are not enlarged, but the skin around their vent is distended and bears spines. We examined and measured two adult males and four adult females of Chaparana unculuanus from Jingdong Xian (Yunnan, China) (Table IV). Univariate comparisons by Mann-Whitney U tests of measurements (expressed as ratios of SVL, except for SVL itself) between the four adult specimens of Chaparana aenea listed in Table II and the six adult specimens of Chaparana unculuanus listed in Table IV show that the two samples differ statistically in seven measurements out of 27 (HL, FLL, TFL, TL, IN, TYE, IMT). If these 10 specimens are included in a PCA, the two samples appear as two well-separated groups ( Figure 6) and the second regression factor scores for the two samples are significantly different (Table V). These data confirm the morphological distinctness of the two samples, independently from their male secondary sex characters, and they support the interpretation that the two samples represent different biological species. This is consistent with the result of the analysis of partial sequences of mitochondrial 12S and 16S rRNA genes provided elsewhere (Jiang et al. forthcoming), which confirms that the two species have distinct, although closely similar, sequences.  (Smith, 1922) from China, Thailand and Vietnam (see Table II).

Taxonomic consequences of these observations at the generic level
Beside the synonymization of Rana aenea and Rana fansipani, the observations presented above have taxonomic consequences at the generic level. The nominotypical species Rana fansipani being the type-species of Chaparana Bourret, 1939, the status of the latter name needs to be re-evaluated. Dubois (1992) recognized two genera in the tribe Paini Dubois, 1992 of the subfamily Raninae Rafinesque-Schmaltz, 1814: Chaparana Bourret, 1939 (with four subgenera) and Paa Dubois, 1975 (with four subgenera). Based on unpublished molecular data, this tribe was transferred to the subfamily Dicroglossinae of the Ranidae Anderson, 1871 by , a taxonomy followed by Dubois (2003) and Roelants et al. (2004). Furthermore, following the results of the work of Jiang and Zhou (2003), the genus Nanorana Gü nther, 1896, placed by Dubois (1992) in the Ranini, must be included in the tribe Paini. In Dubois's (1992) taxonomic arrangement, breeding males of all species of the genus Chaparana were stated to be devoid of breeding spines both on the forelimbs and on the breast, whereas most (but not all) species of Paa had such spines, at least on the first finger. Besides, breeding males of several species referred to Chaparana were known to have a specialized zone of skin, either swollen or bearing spines, around the vent. It is now known that the type-species of Chaparana is devoid of this latter character but has spines on the fingers and chest, thus fitting with the genus Paa as understood by Dubois (1992). This Table III. Comparison by Mann-Whitney U test of the first three regression factor scores (RFS) generated by principal component analysis based on 32 measurements between subsamples of a sample of 13 specimens of Chaparana aenea (Smith, 1922) (see Table II Criteria of construction of subsamples: (G) geographic origin: Thailand (G1) versus Vietnam + China (G2); (S) sex: males (S1) versus females (S2); (D) ontogenetic stage: juveniles (D1) versus adults (D2). The sex of two juveniles is unknown, hence the total sample of 11 instead of 13 for the criterion ''sex''. For each criterion of comparison, minimum and maximum values of RFS given for subsamples are those obtained by PCA for the first (RFS 1), second (RFS 2) and third (RFS 3) regression factor scores. Significance levels before Bonferroni correction (BBC): ns, not significant, P.0.05; *P(0.05; significance levels after Bonferroni correction (ABC): ns, not significant, P.0.0167; *P(0.0167. species, but this species alone, must be removed from the group called Chaparana by Dubois (1992), and placed in the group called Paa by this author. The name Chaparana has priority to designate this genus. However, two significant differences remain between Chaparana aenea and all species referred by Dubois (1992) to the subgenus Paa (Paa): (1) whereas, in breeding males of the former, both sides of the chest are covered by a well-delimited patch of densely packed spines, whose general shape recalls that of a bean or kidney (Figure 7), in breeding males of those species of Paa that have breast spines (i.e. only a part of the species of this group), the two patches are composed of unequally spaced spines and have variable limits, so that these Ad., adult (as defined in Dubois 1976, p 31-33); nm, measurement not taken on this specimen. Of the 36 measurements, only 27 are available for all six specimens and were used for the principal component analysis of Figure 6 and Table V. patches do not have a common general shape (see e.g. Dubois 1976, p 52, 89, 105, 121); (2) in Chaparana, the first finger is usually longer than the second (both fingers may be subequal in some juveniles), whereas in Paa the first finger is usually shorter than the second (in some large adults, both fingers may be subequal, or the first one may be exceptionally longer in the species Paa liebigii (Gü nther, 1860): see Dubois 1976, p 65): in this character, Chaparana aenea is similar to the frogs placed by Dubois (1992) in the subgenus Paa (Quasipaa). This suggests that the relationships between Chaparana and Paa Figure 6. Plots of factors 1 and 2 of principal component analysis based on varimax rotated coefficients for logtransposed characters (25 measurements) for four adult specimens of Chaparana aenea (Smith, 1922) (see Table II) and six adult specimens of Chaparana unculuanus (Liu, Hu and Yang, 1960) (see Table IV). Table V. Comparison by Mann-Whitney U test of first three regression factor scores (RFS) generated by principal component analysis based on 27 measurements between a sample of four adult specimens of Chaparana aenea (Smith, 1922) (see Table II) and six adult specimens of Chaparana unculuanus (Liu, Hu and Yang, 1960) (see Table IV).

Chaparana unculuanus (n56)
Mann-Whitney U test should be explored further. As for the other groups referred by Dubois (1992) to several other subgenera of Chaparana and Paa, their taxonomy and nomenclature will also have to be modified as a result of a re-evaluation of the significance of some characters following the results presented above. To start clarifying this matter, we recently proposed a preliminary cladistic analysis of the Paini based on partial sequences of mitochondrial 12S and 16S rRNA genes of 17 species of the clade (Jiang et al. forthcoming), but this analysis only provides partial information. Therefore we explored the phylogenetic relationships in this clade on the basis of external morphological characters of adults and of a few other characters as available. These results and their implications will be presented elsewhere.