Systematics of the Australo-Papuan tree frogs known as Litoria bicolor (Anura : Hylidae) inthe Papuan region

Transcription

1 CSIRO PUBLISHING Australian Journal of Zoology, 2008, 56, Systematics of the Australo-Papuan tree frogs known as Litoria bicolor (Anura : Hylidae) inthe Papuan region J. I. Menzies A,C, S. J. Richards B and M. J. Tyler A A School of Earth and Environmental Sciences, University of Adelaide, Adelaide, SA 5005, Australia. B Vertebrates Department, South Australian Museum, Adelaide, SA 5000, Australia. C Corresponding author. james.menzies@adelaide.edu.au Abstract. We examined differences in morphology and advertisement calls of a large sample of frogs from the Australo- Papuan Region that resemble Litoria bicolor, and compared them with examples of that species from Australia. Consistent differences in body size, body proportions, and advertisement call structure among populations demonstrate that at least seven distinct species occur in the Australo-Papuan region, and that only the population represented by the holotype from the Northern Territory of Australia is Litoria bicolor s. s. Herein we describe four new species from the Papuan Region and comment on the origin and evolution of the Papuan members of the Litoria bicolor complex in the region. Introduction The frogs known as Litoria bicolor (Gray, 1842) are small green, bronze, or green and bronze climbing frogs characteristically breeding in open grassy swamps. They are widespread and common in northern Australia and frogs closely resembling L. bicolor occur widely but patchily throughout the lowlands of the Papuan region and have also been introduced to the Mariana Islands (Rodda et al. 1991). The type locality of Litoria bicolor is Port Essington (200 km north-east of Darwin) in northern Australia. Tyler (1968) included the species in his treatise on New Guinean Litoria (then Hyla) on account of specimens from the Aru Islands that consistently tallied with Australian (Northern Territory) material, but he also reported specimens from near Aitape and Yamur Lake in New Guinea, Kerevat on New Britain and on Halmahera Island, that closely resemble bicolor in all respects except relative head width. In the same monograph, Tyler described a new upland species, Litoria contrastens, that resembles L. bicolor and is widespread in the eastern highlands of New Guinea but differs from L. bicolor in having a wider head. Since the publication of Tyler s treatise, we have collected widely in the Papuan region and have found bicolor-like frogs in Seram in eastern Indonesia, throughout the southern and northern lowlands of New Guinea, the d Entrecasteaux Islands of Normanby and Goodenough, and on New Britain and Manus Islands. Apart from slight differences in morphology, we have noted differences in male advertisement calls, which led us to believe that several different species may comprise a bicolor-like complex. Menzies (2006) reported on the distribution and biology of the bicolor complex in the Papuan region but did not allocate specific names to any population, other than what he regarded as true L. bicolor in the TransFly plains of southern New Guinea. Our current investigation attempts to unravel the systematics of this widespread complex of small green or green and bronze tree frogs. Since our investigation commenced, the Normanby (d Entrecasteaux) Island population has been described as a new species, Litoria bibonius, by Kraus and Allison (2004), which we include in our analyses. Kraus and Allison (2004) distinguish L. bibonius from Australian L. bicolor by its shorter head, longer legs and differences in colouration. An investigation into the systematic status of L. bicolor in Australia is not part of this project. However, definition of topotypic Litoria bicolor is necessary before it can be compared with any New Guinean population. To this end, we provide a brief summary of the systematic work that has been carried out on this species in Australia. Copland (1957) was the first to recognise that two taxa were included within Australian bicolor and named them Hyla (now Litoria) bicolor bicolor and H. bicolor glauerti. The differences between the two were not great and concerned head width (wider in glauerti) and distribution (more southerly, but overlapping, in glauerti). Moore (1961) examined small samples from within the ranges of bicolor and glauerti but found no consistent differences and did not recognise glauerti as a distinct taxon. Straughan (1966), with much larger samples, reinstated the two forms but found considerable overlap in distribution in north-eastern Queensland. H. glauerti is now known as Litoria fallax (Peters, 1881), the senior name. Straughan (1969) again turned his attention to the Litoria bicolor complex in Australia, this time calling attention to the differences in male advertisement calls between L. bicolor and L. glauerti (=fallax) as well as confirming differences in head width, but commented that after preservation (the two species) are most difficult to distinguish. Failure of in vitro crosses confirmed that the two species were distinct. For nearly forty years, the status of Australian L. bicolor remained unchallenged until James (1997) systematic molecular investigation of the Litoria bicolor Species Group from Eastern Australia. Her analysis also included a few Ó CSIRO /ZO X/08/040257

2 258 Australian Journal of Zoology J. I. Menzies et al. specimens from New Guinea. Although the principal objective of her thesis was to... establish a molecular phylogeny of the Australian members of the L. bicolor species complex,her results are relevant to the New Guinea problem in that there appeared to be at least two, and possibly three, genetically distinct species in Australia, currently going under the name Litoria bicolor. Unfortunately, James thesis does not go as far as allocating scientific names to these genetic populations. If any one of the forms of Litoria bicolor occurring in New Guinea is identical with any of those in Australia, it would most likely be the one occurring in south New Guinea but her analysis did not include any specimens from there. Her analyses do not indicate close relationships between the few New Guinean examples that she analysed and her Queensland L. bicolor, or between them and her Queensland and Northern Territory L. bicolor (James 1997, figs 2.4, 2.5, 2.10). Since 1968 the status of New Guinean populations of L. bicolor received scant attention until the description of L. bibonius by Kraus and Allison (2004). However, Menzies (1976, 2006) called attention to the widespread occurrence of, and probable existence of, several species within this complex. The present work uses a combination of morphometrics and call analysis to produce a systematic analysis of Litoria bicolor in the Papuan Region and to allocate names to populations that represent undescribed species. Methods Material examined Material used in our analyses is listed in the species accounts that follow and Appendix 1 lists geographical coordinates for all places in the Papuan Region from where bicolor-like frogs have been collected. Figure 1 maps all of these locations, except for those in Australia, where only the type locality of L. bicolor at Port Essington is shown. Institutional abbreviations are as follows: AM, Australian Museum, Sydney; BPBM, Bernice P. Bishop Museum, Honolulu; MCZ, Museum of Comparative Zoology, Harvard; MV, Museum Victoria, Melbourne; NMB, Naturhistorisches Museum, Basel; SAMA, South Australian Museum, Adelaide; SMF, Senckenberg Museum, Frankfurt; UP, University of Papua New Guinea, Port Moresby; ZMB, Zoological Museum, Bogor. Morphometric data In total, 374 specimens from different localities in the Australo- Papuan Region were examined. Specimens were grouped according to their place of origin and 10 variables (listed in Table 1) were measured on each specimen. Statistical methods were taken from the JMP package (Sall et al. 1996) and SPSS ver. 15 (SPSS Inc. 2006). In the text, suffixes m and f after abbreviations refer to data for males or females. Male advertisement call data Advertisement calls were recorded with several different tape recorders and microphones during the 30 years that data were collected for this study. These included, inter alia, Sony Pro- Walkman WMD-6C and Sony TCM-5000EV analogue tape recorders, and a Sony SMZ-200 and Sennheiser ME-66 microphone. Calls were analysed using the Cecil sound analysis program (Hunt 1993). For several populations, only limited and poor-quality call data were available, but all data have been included in the results and have provided evidence for our taxonomic conclusions. Terminology for the description Fig. 1. Distribution of Litoria bicolor, s. l., in the Australo-Papuan Region showing the location of specimens examined. In Australia, only the type locality at Port Essington is shown. Note coincident points at Madang and Aitape. Land over 1000 m is shown shaded.

3 Systematics of Litoria bicolor (Anura : Hylidae) in the Papuan region Australian Journal of Zoology 259 Code HB TL HL HW EN IN EY TY F3 T4 Table 1. Definition Variables used in the multivariate analyses Head + body length: from the tip of the snout to the distal end of the urostyle Tibial length: external distance between the knee and ankle with the joints held at right angles Head length between the tip of the snout and the jaw articulation Head width at mid-tympanic level Distance between the anterior margin of the eye and the centre of the nostril Distance between the nostrils, centre to centre Horizontal eye diameter Horizontal tympanic diameter Horizontal diameter of the disk on the 3rd finger Horizontal diameter of the disk on the 4th toe of calls is as follows. Calls are discrete sequences of sounds that follow one another at (usually) irregular intervals. Notes are the shorter components of a call and are separated by distinct, usually regular intervals. Pulses are separate bursts of sounds, within notes, that cannot be further subdivided. For example, Fig. 7a shows a single call of L. bicolor that is s in length and contains 6 short notes. The first note, of s, contains 3 pulses of sound at a rate of 188 pulses s 1.Ifa call is not divided into separate notes, call and note are synonymous. Throughout this paper we have used, for the sake of brevity, the following abbreviations: Q (Queensland, Australia); NT (Northern Territory, Australia); TF (TransFly); PM (Port Moresby region) and NB (New Britain). TransFly is the region of the lower Fly and Digul rivers, and the Port Moresby region, as defined here, stretches from the lower Kikori River to Sogeri. Results Preliminary morphometric analyses To check for errors Each variable in each geographical group was examined in a univariate plot to test normality of distribution and then was regressed against HB. Outliers were identified, remeasured and corrected if required. In all groups correlation with HB was significant for most variables and the degree of deviation from normality was acceptable (Shapiro Wilks test). Differences between males and females Females in all groups were found to be ~12% larger in body length than males, but only in two groups (PM and TF) were there enough females to test for other differences. In the PM collection, males had longer legs and larger eyes (P 0.05) than females but no other comparisons were significant. In the TF collection, males had smaller eyes and tympani than females (P 0.05) but no other comparisons were significant. Because there were no females in some of the groups, and females were generally deficient in most groups, we restricted the multivariate analyses to adult males. Problem 1. How many species comprise the L. bicolor complex in the Papuan Region? Morphometric analysis Principal Component and Discriminant Function analyses (SPSS Inc. 2006) were computed from the 10 variables listed in Table 1, after logarithmic transformation. To search for distinct morphotypes in the total population, we extracted principal components (PC) from the 10 body measurements takenfromeachspecimenandtheseareplottedinfig.2a while Fig. 2b plots the centre point of each of the individual geographic groups. The first two (and only) components accounted for 74% of the total variance and Fig. 2c illustrates the factor loadings. All factors contribute positively to the first component, suggesting that overall size is dominant, while EY and TY contributed most to the second, with other factors being less important. Examination of Fig. 2a shows that PC1 separates the total sample into two major overlapping clusters. Cluster 1 (to the right) includes frogs from the north coast of New Guinea and two islands, Manus and Seram, while to the left are specimens from Australia, New Guinea, New Britain and the d Entrecasteaux Islands. In view of the loading of overall size to PC1, it is clear that size separates Cluster 1, with larger frogs, from all the others. This primary division of the total sample is confirmed in Fig. 2d, which plots the head+body length (HB) for each of the major geographical groups. The larger frogs have a mean male HB of >26 mm while the smaller ones are <25. There is little overlap between large and small groups and the means of any one of the large groups and any one of the small groups are significantly different (Tukey Kramer test, P 0.05). Both groups overlap geographically on the north coast of New Guinea between Aitape and Madang. PC2 separates three clusters. Cluster 2 includes frogs from Australia and south New Guinea; Cluster 3 includes frogs from north and south coasts of New Guinea and New Britain while Cluster 4 includes only frogs from the d Entrecasteaux Islands. In view of the minimal overlap between clusters of larger and smaller frogs, they were then examined separately by multivariate (discriminant) analysis (SPSS Inc. 2006). Each individual geographical group was introduced stepwise and distances from other groups tested for significance. Groups that were not significantly different were combined before proceeding to the next step. Individual groups were then examined by univariate or bivariate analyses to identify differences between them. Figure 3a (smaller forms) and Fig. 3b (larger forms) show the results of a discriminant function analysis on those 11 different Papuan groups for which we had a reasonable number of well preserved specimens. All groups were significantly different at a probability of 0.05 but the first two axes accounted for only 77% of the between-groups variance, and 77% (Fig. 3a) and 97% (Fig. 3b) of the specimens were deemed to have been correctly classified. Several factors probably contributed to the poor result in Fig. 3a. Onlyfive of the 10 variables entered in this analysis wereidentified as significant; inorder ofimportancethesewere T4, TY, EN, EY, TL. The heavy weight given to the variables T4 and TY was a cause for concern as the first, the diameter of the 4th toe disk, can be measured accurately only if the specimen has been carefully preserved. If the toes are curled, they must be

4 260 Australian Journal of Zoology J. I. Menzies et al. Fig. 2. (a) Scatter plot of principal component scores, all specimens. *, Australia; +, New Guinea south coast; *, New Guinea north coast;, islands. (b) Centre points of the geographical groups comprising 2(a); (c) Principal component loadings. (d) Head + body length of Papuan Litoria bicolor populations. Diamonds show mean and 95% confidence limits of the mean for each population. 1, Yamur Lake; 2, TransFly; 3, Port Moresby region, 4, d Entrecasteaux Island; 5, Popondetta; 6, Aitape-1; 7, Kerevat, New Britain; 8; Alexishafen; 9, Aitape-2; 10, Sentani; 11 Siewa River; 12 Seram; 13, Manus Island. D2 (a) D2 (b) D D1 Fig. 3. Discriminant analyses comparing 11 populations of Litoria bicolor from the Papuan Region. Crosses are group centroids and ellipses are their 95% confidence limits; D1 and D2 are 1st and 2nd discriminant axes. (a) Smaller forms: 1, Port Moresby region; 2, d Entrecasteaux Island; 3, Popondetta; 4, TransFly; 5, Kerevat, New Britain. (b) Larger forms: 6, Sentani; 7, Alexishafen; 8, Siewa River; 9, Manus Island; 10, Seram. flattened before measuring and this can obviously lead to distortion and a degree of error. Tympanic diameter is always difficult to measure accurately because the tympanic rim is often ill-defined, and we had considered omitting this variable altogether, so it was surprising that it contributed so significantly to the results. As a result of these analyses, we recognise four small morphospecies and one large morphospecies in the Papuan Region. The exact limits of distribution cannot be precisely determined as there are extensive gaps between collection localities. The five taxa are: (1) small frogs from the TransFly and Gulf Region of south New Guinea, (2) small frogs from the Gulf to Port Moresby region of south New Guinea,

5 Systematics of Litoria bicolor (Anura : Hylidae) in the Papuan region Australian Journal of Zoology 261 (3) small frogs from the d Entrecasteaux Islands (Litoria bibonius), (4) small frogs from the north coast of New Guinea between Popondetta and Aitape and New Britain Island, (5) larger frogs from the north coast of New Guinea, west of Madang, as well as Seram and Manus Islands (these do not subdivide into separate clusters). Call analysis AcallmadebyLitoria bicolor,fromanyregion,normally consists of one to several pulsed notes of varying length that can be conveniently categorised as long notes ( s, mean 0.35 s) or short notes (less than 0.15 s) and a call may be uttered singly or as one of a series. Likewise, long notes may be uttered singly, in pairs, or in small groups, while short notes may be uttered singly, either before, after, or mixed with long notes, or in longer trains. The composition of a call series may be influenced by the proximity of conspecific males or females, as in Fig. 4a where a second male called 0.3 s after the first, but with a pulse rate twice as fast. Short notes may not be uttered at all by solitary males that are not calling within a breeding community. Long notes usually increase in amplitude as they progress (Fig. 4b), but may be flat (1st note of Fig. 4c), are sharply damped at conclusion and the pulse rate may increase within a note (Fig. 4c). In one case, pulse rate became interrupted to such an extent that a long note gradually changed into a succession of short notes. The similarity in call structure among the different populations makes it difficult to find species-specific call characters. Figure 4d illustrates the prominent frequency bands of each call that we recorded. The larger north-western forms commonly show a prominent frequency band between 3.6 and 4.5 khz while, in the smaller forms, such as true bicolor from the Northern Territory, the most prominent band usually lies between 4.6 and 5.5 khz. This follows the general trend whereby smaller frogs make calls with higher dominant frequencies (Menzies and Tyler 1977). Any differences between populations that we noted are examined in the descriptions that follow in the systematic accounts. Problem 2: Does Litoria bicolor sensu stricto occur in New Guinea? Cogger (2000) indicated that the distribution of L. bicolor in Australia extends from the Kimberley in the north-west, around the northern coast, to Proserpine in Queensland. However, Fig. 5a, based on collections in the Northern Territory and Queensland Museums, shows that there is a gap at the southern side of the Gulf of Carpentaria separating western and eastern populations. Several species of hylid frogs have a wide, bidomicilic, distribution in northern Australia and in the seasonally dry woodlands of southern New Guinea (Tyler 1972), so it would be reasonable to assume that L. bicolor has a similar distribution. It was therefore necessary to compare Australian L. bicolor with similar frogs from southern New Guinea. We compared the morphology and calls of samples from several locations in the Northern Territory and Queensland with frogs from the TransFly region of New Guinea, which is geographically closest to Queensland. Morphological similarity of southern New Guinean and Australian bicolor was tested in a discriminant analysis on samples from Queensland and the Northern Territory of Australia and the TF of New Guinea (Fig. 5b). The first two axes accounted for 95% of the between-groups variance but only 65% of the cases were deemed to be correctly classified. Thus, although the centroids of all groups, except 1 and 4, were separated at a probability of <0.05, there was not a great deal of difference between them. However, two clusters are apparent, including Groups 2 and 3 (both NT) and 1 and 4 (Q and TF). In morphology, TF bicolor more closely resemble Qld bicolor than true bicolor from the NT. Univariate analyses revealed only two significant differences between the TF and Q groups. Q frogs have broader heads, mean HL/HW (Q) 1.10 ( ) cf. 1.78(TF) ( ) and larger eyes, mean EY/HB(Q) 0.11 ( ) cf. 0.10(TF) ( ). There were five significant differences between TF and NT. Acoustic analyses support the distinction between Queensland and true L. bicolor from the Northern Territory and these are discussed in the systematic account that follows. However, the acoustic data also demonstrate differences between the advertisement calls of TF and Queensland populations (Fig. 6) despite their morphological similarity. Conclusions Our only uncontroversial conclusion is that there are smaller forms and larger forms within what is currently known as Litoria bicolor in the Papuan Region. Within the small populations and within the large populations there are few morphological differences and it would be difficult, if not impossible, to assign any preserved specimen lacking locality or advertisement call data to any particular population. That said, our conservative conclusion based on a combination of morphological and acoustic data is that there are four smaller species within the total mainland New Guinea sample that we studied, two of which extend to outer islands, and one larger species that extends from the north coast to outer islands. A sixth species, Litoria contrastens, occurs in parts of the eastern cordillera of New Guinea, between 1000 and 1500 m. It is not known to be sympatric with any of the lowlands species discussed here. Northern Territory and Queensland populations are morphologically distinct and call data support the specific distinction. The type locality of Litoria bicolor is in the Northern Territory, so the Queensland population cannot represent that species and therefore Litoria bicolor s. s. does not occur in New Guinea. Although the TF population is morphologically distinguishable from the Q population only in average head width, there are differences in call structure, discussed later. We do not regard these two populations as conspecific. As there is no name in the literature that could be applied to the TransFly population, we describe it as a new species. We make no comment on the status of the Queensland population as it is the subject of current work elsewhere.

6 262 Australian Journal of Zoology J. I. Menzies et al. Fig. 4. (a) Wave form (upper), spectrogram and frequency spectrum (lower) of two long calls of Litoria viranula probably made by different frogs. Note difference in pulse rates. (b) Wave form, spectrogram and frequency spectrum of a long call of Litoria chloristona recorded at Kopi. Note rise in amplitude as the call progresses. (c) Wave forms and frequency spectra of two coincident calls of Litoria bicolor recorded at Jabiru. The first call includes two long notes; the second, three short and one long. (d) Distribution of prominent frequency bands (%) in calls of Litoria bicolor s. l. Light shade, smaller species; dark shade, larger species. Arrows in (a c) indicate points at which spectra were taken. The Papuan Gulf Port Moresby population differs from the TF population (to the west) and from L. bibonius (to the east) in three morphological characters. Its call includes, inter alia, long trains of up to 20 short notes, and it utters much longer long notes, than we have observed in any other population. This population is also described herein as a new species. Available call data for L. bibonius are not good, even in the original description, and the d Entrecasteaux Islands population differs from the Popondetta population and small Madang population in only one or two characters, but L. bibonius is distinct in Fig. 2 (Cluster 4) so we do not consider it to be identical to the small northcoast species. The Madang and Popondetta populations are

7 Systematics of Litoria bicolor (Anura : Hylidae) in the Papuan region Australian Journal of Zoology 263 Fig. 5. (a) Distribution of Litoria bicolor s. l. (shaded area) in northern Australia, compiled from records in the Queensland and Northern Territory Museums. Arrow indicates position of type locality and asterisk marks a locality noted by Straughan (1969, fig. 1) that we have not been able to confirm. WA, Western Australia; NT, Northern Territory; Q, Queensland. (b) Discriminant analysis comparing Litoria bicolor from Australia and the TransFly region of New Guinea. Crosses are group centroids and ellipses are their 95% confidence limits; D1 and D2 are 1st and 2nd discriminant axes. 1, Queensland, Australia; 2, Northern Territory (Groote Eylande); 3, Northern Territory (mainland); 4, TransFly. Distance between Groups 1 and 4 is not significant (P = 0.098), other distances are significant (P < 0.01). herein described as a new species. We have no call data at all for the population from New Britain but, in morphology, it differs significantly from the Popondetta or Madang populations in only one or two characters. Pending better data, this too is included with the north-coast species. Sample sizes (5 12) and call data for the larger-sized north-coast, Seram and Manus populations are insufficient for absolute determination so we conservatively include them within one species, described herein as new. Extension of a mainland New Guinean species to Seram is not unique as two congeners (L. infrafrenata and L. amboinensis) also occur there. It is convenient to regard all populations as belonging to a Litoria bicolor group without implying that they form a single clade, although future research may or may not show that they do. The large north-western forms, particularly, may be more distantly related to the others as a discriminant analysis (not illustrated) including all species shows a large and highly significant distance between Group 1, including the larger northwestern forms, and all the others. In fact, all between-groups distances in this plot were highly significant (P < 0.05) except between Queensland and TransFly (P = 0.1). Better elucidation of the relationships of these taxa will require genetic studies but clustering of the smaller species (Fig. 3a) suggests two separate invasions into New Guinea. Fig. 6. Comparison of calls made by Queensland and TransFly bicolor.(a) Queensland, wave form of a three-note call, each note with 5 6 pulses. (b) TransFly, wave form of a single call with 11 pulses. (c) Queensland, wave form of a single two-note call. Note difference in pulse rates between 1st and 2nd notes. Taxonomic account Smaller species from Australia and from south and east New Guinea Litoria bicolor sensu lato including L. bibonius and L. contrastens General description Small climbing frogs (HBm 19 26, HBf mm); head as long as or longer than wide, occasionally wider than long, HL/HW mean 1.08 (range ); snout projecting, less so in the larger, north-coast and island forms, rounded in profile; obtusely pointed or rounded from above; canthus rostralis rounded, straight to slightly concave; lores slightly oblique, flat; eyes prominent, large, mean EY/HB 0.12 ( ); pupil horizontal; tympanum small to large, mean TY/HB 0.06 ( ), distinct, except upper margin concealed by a shallow supratympanic fold; nostrils prominent, either closer together than, or wider apart than, their distance from the eyes, mean EN/ IN 1.02 ( ); vomerine teeth present or absent. Legs moderately long to very long, mean TL/HB 0.55 ( ); toes from less than fully webbed, leaving penultimate digits on all toes free, to fully webbed on all except 4th; fingers partially webbed; all digits with terminal disks, those on fingers larger than those on the toes, mean F3/HB 0.045, mean T4/HB 0.041; subarticular tubercles small; inner metatarsal tubercle prominent; male with brown nuptial pad on medial surface of the first finger.

8 264 Australian Journal of Zoology J. I. Menzies et al. Dorsal skin finely granular, ventral skin more coarsely so. In life, dorsum normally immaculate bright green to olive green, colour ceasing at wrist and ankle and absent from upper arm; dorsum occasionally completely bronze, or with broad median bronze band, or with two bronze dorso-lateral bands; canthus rostralis and upper loreal edge often dark; upper lip pale, this colour continuing as a broken lateral band separating dorsal and ventral colours; ventrum whitish to yellowish, semitransparent, unmarked; concealed surfaces of thigh and groin yellow, orange, cherry red to dark blue or brown; iris bronze to reddish; tympanic membrane bronze or green. After preservation in ethanol, green colour becomes greyish blue and ultimately fades to pale brown. The presence or absence of vomerine teeth has often been used as a diagnostic character in frogs and was the basis for distinguishing Hylella from Hyla (Boulenger, 1882). To a substantial extent, the occurrence of vomerine teeth is merely a question of size as larger frogs usually have them while smaller frogs do not. Unless preserved frogs have been fixed with their mouths open, which is rarely the case with museum specimens, it is not possible to examine the palate without some destruction. However, several specimens examined in the present study support the above generalisation. No examples of L. bibonius or L. chloristona sp. nov. (small species) were found to have vomerine teeth whereas among Litoria eurynastes sp. nov. (the larger north-coast species) they were seen to be present in 12, and absent in nine, specimens. In some cases, although present, the vomerine teeth were poorly developed and difficult to see. Vocalisation A general account of the calls made by male members of the bicolor complex has already been given. All produce complex calls in series of varying length. Several species appear to produce only relatively long calls but the sample sizes for recordings of some populations are very small and unsatisfactory, because calling was sporadic at the time of recording. Furthermore, the pattern of long and short calls produced by the same individual may change dramatically depending on social interactions within a chorus. Differences in call structure that may have diagnostic value are described in the species accounts below and Fig. 8 shows the relationship of note length to pulse rate in the species that we recognise. Distinction from other Litoria species No other plain green adult Litoria species are as small as adult L. bicolor s. l. apart from L. fallax in Australia. Multicoloured, mainly green, Litoria species overlapping at the lower margin of their HB size are members of the Litoria nigropunctata complex with truncate snouts and spotted green dorsum) or L. iris with truncate snout, variegated dorsum and coloured ventrum. Members of the L. dorsalis complex are equally small or smaller but are brown and have feet that are only partially webbed. Juveniles of the larger species L. caerulea, L. infrafrenata and members of the L. gracilenta and L. graminea complexes are plain green but have sloping or truncate snouts. Litoria havina is plain green and would overlap at the lower end of its size range but males have a pointed rostral process and females have a more acute snout. L. havina also utters whistling calls consisting of a long note followed by several short notes. Distribution This complex of tiny, green tree frogs is widespread in northern Australia, throughout the lowlands and uplands (below 1500 m) of New Guinea, the islands of eastern Indonesia and New Britain and the Admiralty Islands but has not been recorded from the Solomon Islands or New Ireland. The distribution of the smaller bicolor-like species in the New Guinea lowlands more or less coincides with the maximum extent of dry broadleaf woodland or open forest vegetation at the lowest Pleistocene sea levels, years ago. These forests extended in a broad moreor-less continuous band from north-eastern Australia across the Sahul Shelf, around the south-east flank of New Guinea and then north-west at least as far as Popondetta and Jayapura (Nix and Kalma 1972; fig. 5.10). Today, such open vegetation exists only in the TransFly, around Port Moresby and in isolated patches along the north coast. Litoria bicolor (Gray) Eucnemis bicolor. Gray, J. E. (1842: 57). Material examined Holotype. BMNH , collected at Port Essington, Northern Territory, Australia. Other material examined. SAMA R , Groote Eylande; SAMA R19781, Jabiru; SAMA R24031, Victoria River; SAMA R , Katherine; SAMA R34273, Georgetown; SAMA R , Nutwood Station; SAMA R , Gulunggooi; SAMA R , Booloola. All localities are in the Northern Territory of Australia. Remarks James (1997) has demonstrated, by allozyme and mitochondrial DNA comparisons, that there are genetic differences between L. bicolor from the Northern Territory and Queensland, Australia, and that there are possibly two species included in the Northern Territory populations. The differences between Queensland and Northern Territory populations are reinforced by comparison of the call recordings described below. Because the type locality of L. bicolor is Port Essington in the Northern Territory, the L. bicolor occurring in Queensland cannot represent that species. The original description of Eucnemis bicolor by Gray (1842) is very brief and contains little useful information on morphology: the foretoes scarcely webbed, the hinder ones webbed to the ends; the toe disks small; tympanum distinct, and that is all. There are neither illustrations nor dimensions. The holotype was in poor condition when Günther (1858), compiling a catalogue, wrote: Hyperolius (?) bicolor Bad state...the condition of the single specimen does not enable me to give the characters with certainty. It has not improved in the interim but we have attempted to obtain some data from this specimen though our measurements can only be treated as approximations. The small size (HB 21 mm) suggests a less than adult animal, mean HBm for all NT specimens 25.0 ( , n = 47). The sex could not be determined but body proportions fall within the range shown by our sample of NT specimens. The poor condition means that this holotype cannot be used as a basis for a diagnosis, so the following diagnosis is based on those Northern Territory specimens that we have examined. Which of James postulated

9 Systematics of Litoria bicolor (Anura : Hylidae) in the Papuan region Australian Journal of Zoology 265 two NT species is the true bicolor is unlikely ever to be determined. Diagnosis and distinction from other species A smaller species of the Litoria bicolor complex, maximum HBm 27.4, HBf 27.6 mm (figures from specimens in the South Australian Museum); otherwise as in the general description above. Northern Territory bicolor are larger than Queensland bicolor, mean HBm (NT) cf (Q). Queensland bicolor have larger heads, mean HL/HB (Q) cf (NT) but there are no differences in head proportions. Queensland bicolor have more widely placed nostrils, mean EN/IN (Q) cf (NT), but none of these characters will provide absolute separation as all measurements and ratios show extensive overlap and so are of little practical value. A combination of two metrics, i.e. HB and HL/HB, provides better, but not absolute, separation. Vocalisation This description is based on recordings made by Stephen Richards and Ian Morris at Jabiru, NT. The division between lengths of short notes (<0.05 s) and long notes (>0.15 s) is absolute (Fig. 8a) and long notes always show a slow pulse rate (<150 s 1 ). Short notes have highly variable pulse rates, anywhere between 67 and 300 s 1. Figure 7a shows a call consisting of six short notes, each with two or three pulses and a pulse rate between 188 and 222 s 1. Figure 7b shows two successive calls with different pulse rates, possibly made by different frogs. The first call (0.206 s), divided into six notes in acceleration, has a pulse rate of s 1. The second call (0.278 s) is not divided into separate notes but contains 31 pulses in acceleration, at a mean rate of 110 s 1. The dominant frequency lies around 5 khz, with subsidiary bands between 2 and 3 khz, rising by ~0.6 khz, though amplitude is more or less flat. All these calls sound like short rattles. Figure 4c illustrates two frogs calling simultaneously. The nearest to the microphone utters two long calls with mean pulse rates of 38 and 51 s 1. The second frog utters three short notes with pulse rates s 1 and one long note with a pulse rate of 39 s 1. In all notes, the pulses are in acceleration. Ecology Tyler et al. (1983) have written on the ecology of this species in the Jabiru area and have described the eggs and tadpoles. Distribution The western limits of L. bicolor lie in the Kimberley region of north-western Australia. The eastern limits lie on the south-west coast of the Gulf of Carpentaria (Fig. 5a) and there is no indication of sympatry with the Queensland bicolor. Litoria bibonius Kraus & Allison Litoria bibonius. Kraus, F. & Allison, A. (2004: 72). Holotype. BPBM collected at Sewa Bay, Normanby Island in Material examined AM , AM132165, AM , AM , AM132185, Esa Ala, Normanby Island; UP , Esa Ala Airport, Normanby Island; UP , Ulutuya, Goodenough Island. Fig. 7. Wave forms (upper), spectrograms (lower left) and frequency spectra (lower right) of calls of Litoria bicolor recorded at Jabiru, Northern Territory, Australia. Arrows indicate points at which spectra were taken. Note the rise in frequency between arrows. (a) Single long call of six notes. (b) Two long calls, possibly made by different frogs.

10 266 Australian Journal of Zoology J. I. Menzies et al. Fig. 8. Note length and pulse rate for Litoria bicolor s. l. (a) Litoria bicolor recorded at Jabiru, Northern Territory. Spots are long calls, crosses are short calls. (b) Litoria cf. bicolor recorded at Biboorah, Queensland (spots) and derived from Straughan (1969) (crosses). (c) Litoria viranula recorded at Morehead and Bensbach. Crosses are long notes, spots are short notes. (d) Litoria chloristona recorded at various locations. (e) Litoria bibonius recorded at various locations. Spots are long notes, crosses are short notes. (f) Litoria eurynastes recorded at various locations. &, Siewa River; *, Alexishafen; +, Frieda River;, Manus Island. Note that the scales are not all identical. Diagnosis and distinction from other species The following is from Kraus and Allison (2004), but expanded to include additional material. A smaller species of the bicolor-complex, HBm 22 26, HBf mm; head about as long as wide (HL/HWmf ); snout projecting, rounded; eyes large (EY/HBm ; EY/HBf ); nostrils narrowly placed (EN/ INmf ); tympanum visible. Fingers partially webbed; toes fully webbed; adult males with a brown nuptial pad on 1st finger. Differs significantly from L. viranula in having longer legs (mean TL/HB 0.58 cf. 0.55), larger eyes (mean EY/HB 0.12 cf. 0.11) and smaller tympanum (mean TY/HB 0.06 cf. 0.07); from L. chloristona in having longer legs (mean TL/HB 0.58 cf. 0.53), broader head (mean HL/HW 1.00 cf. 1.08; HW/HB 0.36 cf. 0.33), more widely placed nostrils (mean EN/IN 0.91 cf. 1.07), and larger toe disks (mean T4d/HBm cf ); from L. lodesdema in having longer legs (mean TL/HB 0.58 cf. 0.53) and broader head (mean HL/HW 1.00 cf. 1.08, HW/HB 0.36 cf. 0.32). As before, differences are slight and single characters cannot be used for reliable identification. A combination of two characters may give better, but not absolute, separation. Colour in life From Normanby Island: dorsum bright leaf green sharply divided from a pure white ventrum; upper lip white or yellowishwhite; green colour ceasing at ankles; concealed surfaces of thighs and groins bright orange-yellow bordered above with blue-black; iris red. Vocalisation The single record illustrated by Kraus and Allison (2004) consisted of two long notes followed by a short note. The few calls recorded by us near the type locality on Normanby Island, and also on Goodenough Island, were, unfortunately, confused by coincident calls of the larger (and louder) L. infrafrenata but appear to be similar to the long notes recorded by Kraus and Allison (2004). Figure 9a shows one call, of a series, consisting of 9 or 10 notes in acceleration, each with 3 5 pulses of sound. The short notes lasted s (0.040 s) and

11 Systematics of Litoria bicolor (Anura : Hylidae) in the Papuan region Australian Journal of Zoology 267 Fig. 9. (a) Wave form and frequency spectrum of a single long call of Litoria bibonius recorded at Esa-ala, Normanby Island. The heavy double note that obscures the initial three notes of the call was made by Litoria infrafrenata.(b) Wave form, spectrogram and frequency spectrum of a single call of Litoria lodesdema recorded near Popondetta, comprising 13 notes, each with 3 5 pulses of sound. Note the increase in amplitude as call progresses. (c) Wave form and spectrogram of a long call of Litoria lodesdema preceded by two short calls, recorded near Popondetta. In (a) (c), arrows indicate points at which frequency spectrums were taken. each included eight pulses. Figures in parenthesis refer to Kraus and Allison (2004). Figure 8e shows the division of the total sample into short (<0.15 s) and long (>0.25 s) notes. The call is generally similar to the call of L. chloristona, but the long trains of short notes, characteristic of that species, were never recorded and the long calls are not so long. Ecology On Normanby Is. L. bibonius was found around reedy and shrubby swamps and sago swamps, calling up to 2.5 m above water level. The tadpole has not been described. Distribution Restricted to the d Entrecasteaux Islands of Goodenough and Normanby and not yet recorded from Fergusson Island, which lies between them (Fig. 1). Litoria contrastens (Tyler) Hyla contrastens. Tyler, M. J. (1968: 72). Holotype. SAMA R5845, adult female collected near Kundiawa, Simbu Province of Papua New Guinea. Material examined SAMA R5845, R5847a e, Kundiawa; R8267, Kainantu. Diagnosis and distinction from other species Close in size to species of the large bicolor group. HBm mm, HBf mm; legs long, mean TL/HB 0.55 ( ); head usually longer than wide, mean HL/HW 1.04 ( ); nostrils broadly spaced, mean EN/IN 0.89 ( ); otherwise as in the general description above. Our figures differ somewhat from, but are generally in accordance with, the figures given by Tyler in the original description. The concealed surfaces of the legs are bright red in life. Tyler distinguished contrastens from bicolor by their nonoverlapping HL/HW ratios, cf With our Australian NT bicolor sample we find the same difference but the magnitude is much less: mean HL/HW for contrastens 1.04, for NT bicolor 1.10 Because L. contrastens is an upland species, found nowhere below 1000 m and not known to be sympatric with any of the other species in the bicolor complex, we have not carried comparisons any further. Vocalisation A slow succession of clicks followed by a two-part buzz. The first part of the buzz note has a slower pulse rate than the second.

12 268 Australian Journal of Zoology J. I. Menzies et al. Ecology The frogs live in clumps of grass or reeds. By night they are found in the grass, by day many are found in the axils of cultivated banana plants in gardens, in or near the swamps (Parker, in Tyler 1968). Menzies observed this species breeding in artificial lily ponds in the Wau Bulolo area; males were calling from near the water surface on lily pads or grass stems. The eggs have a darkly pigmented animal hemisphere and float on the water surface but the tadpole has not been described. Distribution Known from several locations in the eastern part of the central cordillera of New Guinea between Kundiawa and Lake Trist, m. Localities in Fig. 1 are taken from Tyler (1968) and collections at the University of Papua New Guinea and are not fully inclusive. Litoria cf. bicolor from Queensland Material examined SAMA R09474, R09476, Mitchell River; 09725D,F, Edward River; R , R , Jones Lagoon; R , Cooktown; R14710, Tully; R14776, Alligator River; R15286B, Tully; R15474B D, Manango. All those localities are in Queensland, Australia. The differences in morphology between this form and true L. bicolor have already been described. As this species is being described elsewhere, no formal designation is proposed here. Vocalisation Straughan (1969) recorded calls at various sites between Cairns and Townsville and described the call as biphasic. The first phrase (a long call in our terminology) consists of a single note of 351 ms mean duration, with a mean pulse content of 58 s 1 (54 61). The second phrase (our short call ) consists of 3 5 notes, each of 43 ms mean duration and a mean pulse content of 6(4 7). Both calls have a mean starting dominant frequency of 5.1 khz that rises through 500 Hz and a pulse-repetition rate of s 1 (Fig. 10a, ex Straughan). Calls recorded at Biboorah, Queensland, are generally similar although our descriptive terminology is different, a phrase is our call or series of calls. Short calls last s and contain 2 4 notes, each unpulsed or with 2 16 pulses of sound at a rate of s 1 (Fig. 10b). Notes may be uttered singly or in short series (<7) and sound like tzt. Long calls, as in Fig. 10c, last s, include two to several notes, and sound like zeeep. The first, and sometimes second, part of a long call (Fig. 10c) includes 9 17 notes, each with 2 4 pulses at a mean rate of 241 s 1 while the following part contains pulses at a rate of 151 s 1, but here the rate increases through the note and towards the end individual pulses merge into one another. Short and long calls may be uttered singly or in multiples and in any order of succession. These figures are generally in accordance with Straughan s but his plate 1 is not sufficiently clear to allow a more detailed comparison of his long phrase with our long call. Moreover, short calls do not always follow long calls in the same strict succession that he indicated. As with preceding examples, long notes always have a slow pulse rate while short notes may be slow or fast. The principal difference between calls made in Jabiru, NT (Figs 7, 8a) and in Queensland (Figs 8b, 10) is that some of the Queensland notes reach a pulse rate of 600 s 1, not recorded in any other population. A comparison between calls of the Queensland and TransFly populations follows in the account of that species. Ecology In Australia this species is found in a variety of habitats but not in rain forest. Males may call at any time of year but breeding activity is strongly seasonal with the peak being between December and March, the same time as peak rainfall. Breeding sites include natural lagoons and swamps, man-made dams and temporarily flooded grassland. Males call from vegetation up to a metre above the water surface and, where occurring sympatrically with the very similar L. fallax, L. bicolor prefers the shallower water around the swamp margins. Males may also call outside the peak breeding season and at some distance from potential breeding sites, particularly when stimulated by rain showers (Straughan 1969). Distribution On the eastern side of Australia the species occurs as far south as Proserpine and inland to approximately the annual rainfall isohyet (Straughan 1969; Cogger 2000). The confirmed distribution around the Gulf of Carpentaria is shown in Fig. 5. Straughan (1969, fig. 1) indicates a locality on the western side of the Cape York Peninsula, near Burketown, which is outside the shaded area in Fig. 5 but we have not been able to confirm this record as the location of the specimen(s) on which the record is based is unknown to us. Litoria viranula sp. nov. (Fig. 11) Material examined Holotype. SAMA R63487, adult male, collected by S. Richards at Wegamu Camp, Bensbach River, Western Province, Papua New Guinea on 26.iii Paratypes. AM , Balimo, 2m; AM , Balimo, 3m; AM , , Balimo, 5f; AM30644, Balimo, f; MV29855, Balimo, f; MV49826, Balimo, f; MV49853, Balimo, f; MV49866, Balimo, f; MV49992, Lake Murray, f; MV50062, Balimo, f; SAMA R10264a b, Lake Murray, 2f; R10265, Katow, f; R10535a c, Kiunga, 3 m; R11628, Lake Murray, f; R13110f,h,j, 13110m,p, Morehead, 5m; R13259a d, Morehead, 4f; R13666a,b,c, R13667A D, Merauke, 9m; R17526a h, Daru, 6f 2m; R , Wegamu, 8m; UP , Morehead, 2m. Referred material. AM , Balimo, 2m; AM , Balimo, 3m; SAMA R10293, R10536a f, Balimo, 6m, 1f; SAMA R13190a b, Bamu, m. Diagnosis and distinction from other species A smaller species of the Litoria bicolor complex; maximum HBm 23.5, HBf 26.4; otherwise as in general description above; significantly smaller than L. eurynastes sp. nov.; differs from L. chloristona in narrower head (mean HL/HW 1.14 cf. 1.00), smaller eyes (mean EY/HB 0.10 cf. 0.13), and larger

13 Systematics of Litoria bicolor (Anura : Hylidae) in the Papuan region Australian Journal of Zoology 269 Fig. 10. (a) Complete diphrasic call of Litoria cf. bicolor, from Straughan (1966), recorded in Queensland. (b) Wave form (upper), spectrogram (lower left) and frequency spectrum (lower right) of a short call of Litoria cf. bicolor with three notes, recorded at Biboorah, Queensland. Arrow indicates point at which spectrum was taken. (c) Wave form, with 10 notes, of a long call of Litoria cf. bicolor recorded at Biboorah, Queensland. Note difference in structure in first and second parts of the call. tympanum (mean TY/HB 0.07 cf. 0.06); from L. lodesdema in larger body size (mean HBm 23.8 cf. 22.0), narrower head (mean HL/HW 1.11 cf. 1.07, mean HW/HB 0.30 cf. 0.32), more widely placed nostrils (mean EN/IN 1.01 cf. 1.07), smaller eyes (mean EY/HB 0.10 cf. 0.13) and larger tympanum (mean TY/HB 0.07 cf. 0.06). Comparison with L. bibonius has been made in the account of that species. As before, differences are slight and single characters cannot be used for reliable identification. A combination of two characters may give better, but not absolute, separation. Description of the holotype Head longer than wide (HL/HW 1.09); snout obtusely pointed from above, rounded, projecting in profile; lores flat, oblique; canthus rostralis rounded, straight;nostrils lateral, hardly visible from above; eye large (EY/HB 0.10); pupil horizontal; tympanum small (TY/HB 0.07), visible except upper margin obscured by skin fold terminating above axilla; fingers virtually unwebbed; disks large (F3/HB 0.47); brown nuptial pad on medial surface of digit 1; legs long (TL/HB 0.51); toes almost completely webbed leaving only terminal phalanges free on digits 1, 2, 4; dorsal skin granular; ventral more coarsely so, except smooth on throat. Intraspecific variation Summarised metric variables and ratios are set out in Table 2 and the range in body size is expressed in Fig. 2d. Legs long to very long (TL/HB ); head as wide as long (occasionally) to longer than wide (HL/HW ); nostrils equidistant from eyes or further apart (EN/IN ); finger disks larger than toe disks (mean F3/HB 0.44, mean T4/HB 0.39); finger webbing variable, usually unwebbed except between 3rd and 4th, where there is sometimes a basal web, as in the holotype, or occasionally up to one-third webbed between 3 and 4; toes sometimes completely webbed, except the 4th where the terminal phalanx is always free, but webbing usually as holotype. Colour in life Dorsum green with mid-dorsal bronze band; narrow dark canthal stripe continuing through eye and tympanum but becoming indistinct between green dorsum and white ventrum; upper lip white, colour continuing below eye and tympanum before merging with ventral colour; male throat pale yellow, ventrum elsewhere white; groin and concealed surfaces of thigh dark brown. This species was illustrated in life by Menzies (2006; plate 44) as Litoria bicolor.

14 270 Australian Journal of Zoology J. I. Menzies et al. Fig. 11. Holotype of Litoria viranula sp. nov., SAMA R Left foot in dorsal view. Scale line = 5 mm. Table 2. Summary of biometric data for Litoria viranula sp. nov. Variable Female (n = 35) Male (n = 33) Range Mean s.d. s.e. Range Mean s.d. s.e. HB TL/HB HL/HW HW/HB EN/IN EY/HB TY/HB F3d/HB T4d/HB Vocalisation Calls recorded near Morehead and Bensbach consisted of short/fast and long/slow notes uttered singly or in various combinations. Figures 4a and 12a show that the dominant frequencies lay either between 2.5 3, around 4 or 5 6 khz. However, these differences may not be significant because the recordings were made with different microphones that may have had a different sensitivity. Once again, combination of length and pulse rate clearly distinguishes the two types of notes (Fig. 8c). Short notes with a mean length of s ( s) contain 1 20 pulses of sound at a mean rate of 180 s 1 ( s 1 ). Long notes with a mean length of s ( ) contain up to 20 pulses at a mean rate of 39 s 1 (22 146).

15 Systematics of Litoria bicolor (Anura : Hylidae) in the Papuan region Australian Journal of Zoology 271 Figure 4a shows two long notes, the second with a faster pulse rate than the first. It is possible that this was made by a second frog in response to the first, as in Fig. 7b. In view of the morphological similarity of L. viranula sp. nov. and Queensland bicolor, itis necessary to compare their calls. The most obvious difference between Queensland and TransFly calls is in the maximum pulse rate, 600 cf. 330 s 1 (Fig. 8b, c). Short calls of TF (Fig. 6b) are not divided into distinct notes but the pulse rate of Queensland, within notes, is similar (Fig. 6a). In L. viranula, slow and fast pulse-rate notes only appear in different calls (Fig. 4a), and are not combined as slow or fast pulse notes in the same call, as they sometimes are in Litoria Queensland (Fig. 6c). We conclude that the Queensland and Transfly populations represent different species though we caution that the Queensland data are based on a single set of recordings. Ecology Field observations in the TransFly area around Bensbach and Morehead during the wet season revealed that this species breeds in temporary flooded areas and in permanent waterways. It was one of the few frog species that were calling from vegetation overhanging, and flooded by, the Bensbach River. Males called from vegetation between 30 cm and 3 m above the water. The climate of the region is highly seasonal, with peak rainfall between December and March and it is likely that breeding ceases during the dry season. The recording in Fig. 4a was made during peak wet season. Neither eggs nor tadpoles have been described. Distribution Throughout the woodlands of the Digul and Fly River plains, from the coast to as far north as Kiunga, where woodlands are replaced by forest. Our sample includes specimens from (west to east) Merauke to Balimo and (north to south) from Kiunga to the coast, sea level to ~50 m. To the east, the distribution of this species may overlap that of L. chloristona but the western limit is currently unknown. Etymology The specific epithet, viranula, is compounded from the Latin viridis (green) and ranula (diminutive of rana, frog). Litoria chloristona sp. nov. (Fig. 13) Material examined Holotype. UP2188, adult male, Iarowari High School grounds, Sogeri, Central Province of Papua New Guinea, collected by J. Menzies in Paratypes. SAMA R4716a c, Epo, 2m f; R9122a, Waigani, m; R13251a c, Brown River, 3m; R50860, Moitaka, f; R , Kopi, 8m; R , 3m f; UP , Sogeri, m f; UP , Brown River, 5m f; UP1365, Brown River, m; UP , Brown River, 6m f; UP , Brown River, 4m 3f; UP , Sogeri, 5 m; UP , Sogeri, 3m 3f; UP2429, Sogeri, m. Diagnosis and distinction from other species A smaller species of the Litoria bicolor complex, HBm , HBf mm; otherwise as in the general description above. Distinguished from L. eurynastes by significantly smaller size. Comparison with L. viranula and with L. bibonius has been made in the accounts of those species. Distinguished from L. lodesdema by smaller tympanum, mean TY/HB 0.05 (cf. 0.06) and more narrowly placed nostrils, mean EN/IN 1.07 (cf. 0.98). As before, differences are slight and single characters cannot be used for reliable identification. Description of holotype Head longer than wide, HL/HW 1.07; snout obtusely pointed from above, rounded and projecting in profile; lores flat, sloping; canthus rostralis rounded, straight; nostrils semilateral, visible from above; eye large, EY/HB 0.12; tympanum small, TY/HB 0.06; visible, except upper margin concealed below narrow postorbital skin fold that fades above the axilla; fingers partially webbed; light brown nuptial pad medially on 1st finger; toes not completely webbed, terminal phalanx free of web on all digits; dorsal skin finely granular; ventral somewhat more coarse; skin fold across throat. Intraspecific variation Summarised metric data and ratios are set out in Table 3 and HB range is expressed in Fig. 2d. Head nearly always longer than broad, mean HL/HW 1.07( ); nostrils nearly always closer together than their distance from the eyes, mean EN/IN 1.07 ( ); finger webbing variable but generally half webbed between fingers 3 and 4, sometimes less, as in Fig. 15; legs long to very long, occasionally short, mean TL/HB 0.53 ( ). Colour in life Dorsum bright green, tending towards bronze on the flanks; white labial stripe continuing below eye separates dorsal green from ventral whitish-transparent; green colour ceasing at wrist and ankle; concealed surfaces of thighs orange-red; dark canthal stripe; iris bronze; tympanum bronze with green centre; ventrum whitish, translucent on male throat. This species was illustrated in life by Richards (2002) as Litoria sp. nov. 3 and by Menzies (2006, plate 45) as Litoria cf. bicolor. Vocalisation Includes long notes and short notes uttered in combination or singly. A long note, as in Fig. 4b, lasts s and is produced singly, repeated several times, or followed immediately by a long, rapid, sequence of short clicking notes with the total acoustic impression:zeeep, click-click-clickclick-click... as in Fig. 12b. Longcallsmayormaynotbe divided into separate notes (Fig. 12c). A single-note long call includes pulses of sound at a mean rate of 211 s 1. Amplitude increases through the note and is sharply damped at the conclusion. In Fig. 12a, b, there is no division of the long call into separate notes but incipient division is noted in some examples (Fig. 12c). Short notes average s ( ) with a mean pulse rate of 207 s 1 ( ).However,Fig.8d does not give a complete picture as many short notes emitted by this species are unpulsed and therefore do not appear in the figure. The principal difference between this species (Fig. 8d)

16 272 Australian Journal of Zoology J. I. Menzies et al. Fig. 12. (a) Wave form (upper), spectrogram (lower left) and frequency spectrum (lower right) of a call of Litoria viranula with three short notes and one long note, recorded at Morehead. Arrow indicates point at which section was taken. (b) Spectrogram of a call of Litoria chloristona recorded at Sogeri. The long note is followed by fifteen short notes. Unfortunately, this picture is confused by the operation of a nearby water pump, responsible for the heavy, low-frequency background. (c) Wave form of the end of long call of Litoria chloristona recorded at Kopi showing incipient separation into distinct notes. Fig. 13. Female paratype of Litoria chloristona sp. nov., UP2422. Left foot in dorsal view. Scale line = 5 mm. and L. viranula (Fig. 8c) is that long notes have the same pulse rate as short notes so the total sample cannot be divided into long-slow and short-fast notes. Also, the highest pulse rate recorded (223 s 1 ) is lower than that of other species and long calls are much longer than those of L. viranula (maximum 0.58 versus maximum 0.5 s). The overall pattern is more similar to

17 Systematics of Litoria bicolor (Anura : Hylidae) in the Papuan region Australian Journal of Zoology 273 Table 3. Summary of biometric data for Litoria chloristona sp. nov. Variable Female (n = 14) Male (n = 41) Range Mean s.d. s.e. Range Mean s.d. s.e. HB TL/HB HL/HW HW/HB EN/IN EY/HB TY/HB F3d/HB T4d/HB that of L. bibonius but the mean pulse rate is higher and the long notes are longer. None of the other populations examined in this study produces such long trains of short calls. Ecology Common and characteristic of open, grassy or reedy swamps, roadside ditches through forest, ponds, and water-filled ditches in town suburbs and similar situations, including temporarily flooded grassland. Males call while clinging to grass stems a short distance above the water. The type locality is a shallow pond created as a water reservoir for Iarowari School, and vegetation consisted only of tall grass. The site near Brown River is a small lake that varies in extent due to a highly seasonal rainfall. Surrounding vegetation is Eucalyptus woodland in a transition zone to nearby monsoon forest where the species was less common. Near Kopi, males called from flooded roadside ditches in primary lowland rainforest. Data for other localities are unrecorded. Eggs are tiny, black, and float on the water surface in a single layer. The tadpole, figured in Menzies (2006; fig. 7a, b) is morphologically typical of Litoria species breeding in still water. Tail ~1 1.5 times body length, tapering rather abruptly to a point and boldly marked with irregular light and dark vertical bands; dorsal fin broad, arched; ventral fin narrower. Distribution From Kopi on the lower Kikori River to Port Moresby and Sogeri, sea level to 500 m. The western limit is not confirmed beyond the Kikori River but computer identification of seven Balimo specimens allocated four to viranula and three to chloristona and one from the Bamu River to viranula. This could be a transition zone where both species occur but morphological limits of the species overlap and some specimens could be allocated equally to either species. The eastern limit is not known either because the country between Port Moresby and Milne Bay has not been thoroughly surveyed. Etymology The specific epithet, chloristona, is compounded from the Greek, chloros (green), elaschistos (diminutive) and ontos (living thing) and Latinised to give the feminine suffix. Litoria lodesdema sp. nov. (Fig. 14) Material examined Holotype. UP3601, adult male, collected by J. Menzies by the Popondetta Cape Killerton Road, Oro Province, Papua New Guinea, June Paratypes. SAMAR , 2; R7049, m; R , 2m; R , 2m; R8439a,b, 2f; all from Kerevat, New Britain; UP3593, m, UP , m, collected with the holotype; UP , m, collected by J. Menzies, April 1987, Nagada Harbour, Madang Province, Papua New Guinea. Referred material. MCZ , m, Aitape, East Sepik Province, Papua New Guinea. Diagnosis and distinction from other species A small species of the Litoria bicolor complex (mean HBm 22 mm ( ); females unknown), differing from L. bibonius, L. viranula and L. chloristona only in size, tympanic ratio and nostril placing. Comparison with these species has been made in the accounts of those species. It is morphologically most similar to L. chloristona but the colour and the vocalisation are different. Description of the holotype HB 22.7 mm; snout obtusely pointed, rounded and projecting in profile; lores flat, sloping, canthus rostralis slightly concave, rounded; nostrils dorso-lateral, but visible from above; head about as long as wide (HL/HW 0.98); internarial and eye nostril distance about equal (EN/IN 0.97); eye large (EY/HB 0.12); tympanum moderate (TY/HB 0.06); visible except upper margin; postorbital skin fold slight, ending at forelimb; outer fingers half webbed, others with only basal web; brown, simple nuptial pad; legs long (TL/HB 0.52); toes fully webbed except 4th; dorsal skin granular, ventral more so; vomerine teeth absent. Intraspecific variation A summary of biometric data is given in Table 4; finger web varies from scarcely any to almost half webbed. Colour in life At Popondetta: dorsum immaculate yellowish green of different shades with vague bronze median band; some specimens with

18 274 Australian Journal of Zoology J. I. Menzies et al. Fig. 14. Male paratype of Litoria lodesdema sp. nov., UP7353. Right hand in dorsal and right foot in ventral views. Scale lines = 5 mm. faint yellow postorbital stripes, fading before groin; concealed surfaces of thighs faintly orange, heavily speckled black on dorsal edge; male throat pale yellow, venter elsewhere white; iris bronze. At Madang: dorsum immaculate yellow green to bright green, one duller and brownish; indistinct yellowish postorbital stripe; bronze lateral stripe, tip of snout to mid body, separates dorsal and ventral colours; upper lip cream and colour continues to groin; concealed surfaces of thighs sparsely to moderately sprinkled with blue-black; tympanum bronze, iris gold; male throat yellow, otherwise white below. Vocalisation A recording made near Popondetta (Fig. 9b) is similar to that of L. bibonius (Fig. 9a) and shows a long call, lasting s, Table 4. Summary of biometric data for Litoria lodesdema sp. nov. Variable Male (n = 22) Range Mean s.d. s.e. HB TL/HB HL/HW HW/HB EN/IN EY/HB TY/HB F3d/HB T4d/HB divided into 12 distinct notes in acceleration (150!222 s 1 ) each with 3 5 pulses of sound. Figure 9c shows part of a longer series of calls, including two short notes followed by a long call that is not subdivided. The pulse rate of the long call (180 s 1 ) does not change through the note. This example shows that a series of short calls is merely a long call with the individual notes well separated. The calls heard near Madang sounded similar but no satisfactory recordings were made, due to the acoustic dominance of calls made by L. infrafrenata and Rana papua at the same site. Ecology At the type locality, L. lodesdema was found in a breeding aggregation at a shallow, grassy roadside swamp and near Madang males called from bushes around the margin of a heavily shaded swamp in much-degraded forest adjacent to banana plantations, together with L. infrafrenata. There are no data on the specimens from Aitape and the tadpole has not been described. Distribution Lowlands of northern New Guinea from Popondetta to as far west as Nagada Harbour, near Madang, and possibly further west to Aitape, but the Aitape specimens are too poorly preserved for meaningful comparison with other material. Thus, distribution west of Madang has yet to be confirmed.

19 Systematics of Litoria bicolor (Anura : Hylidae) in the Papuan region Australian Journal of Zoology 275 Etymology The trivial epithet is a contraction of the expression loca demissa septentrionalis domicilium habemus : in the lowlands of the north we have our home. Larger species from northern and western New Guinea and islands Although we have samples from seven different geographical populations, samples sizes are small (maximum 12) and voice recordings are all of less than ideal quality. There are morphological differences between populations, which are to be expected in view of their allopatry, but small sample sizes urge caution when giving weight to these apparent differences. Therefore our treatment of this group is conservative and we recognise just one species, ranging from Madang westwards to Seram Island and northwards to Manus Island (Fig. 1). It is appreciated that acquisition of better data may alter the taxonomic resolution in the future. The general description given under the smaller species heading is also applicable here, except that mean HBm is 29.8 ( ) compared with Litoria eurynastes sp. nov. (Figs 15, 16) Material examined Holotype. UP2321, adult male, collected from a pond, near Alexishafen, Madang Province of Papua New Guinea, sea level, by J. Menzies, July Paratypes. MZB , Siewa River, imm. 4m; RMNL5262, Sentani Lake, 7m 6f; RMNL5307, Sentani Lake, f; SAMA R , Siewa River, 4m; R , Lorengau, Manus Island, 5m; UP , Alexishafen, 3m; UP , Alexishafen, 9m; UP2325, Alexishafen, f; UP , Yala River, Seram, 4m; UP5276, Yala River, Seram, m; UP , Frieda River, m f. Referred material. MCZ , Aitape, sex not determined. Diagnosis and distinction from other species A larger species of the Litoria bicolor complex; HBm , HBf , vomerine teeth usually present, otherwise as in the general description above. Description of the holotype Snout rounded and projecting in profile, somewhat angular from above; lores slightly sloping, flat; canthus rostralis rounded, straight; nostrils lateral, slightly visible from above; eyes large, EY/HB 0.11; tympanum large, TY/HB 0.07, distinct, except extreme upper margin; shallow postorbital fold fades before axilla; outer fingers about half webbed, others webbed only at base; simple, pale brown nuptial pad; all fingers with disks, larger than on the toes (mean F3/HB 0.046, T4/HB, 0.041); toes fully webbed, except 4th with terminal phalanx free, all with disks; dorsal skin finely granular; ventral more coarsely; throat wrinkled; vomerine teeth in two small patches between choanae. Intraspecific variation A summary of biometric data is presented in Table 5. Differences in proportions between the five populations analysed are recounted in the multivariate analyses, but in view of the small size of the male samples (5 12), we attach no significance to these. Fig. 15. Male paratype of Litoria eurynastes sp. nov., MZB14.653, Siewa River. Left foot in dorsal view. Scale line = 5 mm.

20 276 Australian Journal of Zoology J. I. Menzies et al. Fig. 16. Male paratype of Litoria eurynastes sp. nov., UP5276, Seram. Left foot in dorsal view. Scale line = 5 mm. Colour in life Dorsum immaculate yellow green to bright green (Alexishafen) or dull brownish green (Seram), one specimen duller and brownish; indistinct yellowish stripe eye groin; bronze dorsolateral stripe (Alexishafen) or vague pale dorso-lateral stripe (Seram) from snout tip to mid body separates dorsal and ventral colours; upper lip cream below the bronze band but continues to groin; concealed surfaces of thighs with sparse to moderate sprinkling of blue-black; tympanum bronze, iris gold; ventrum white except throat yellow. UP (Frieda River) are similar to those from Alexishafen. This species was illustrated in life in Menzies (2006, plate 46) as Litoria cf. bicolor. Vocalisation Although we have a good sample of calls from Alexishafen, recording was made during rain and call structure is confused by heavy background noise. Calls ranged from to s in length with a pulse rate of s 1 (Fig. 17a, b). Recording was repeated with captive frogs, which made shorter calls ( s) with a slower pulse rate ( s 1 ). The acoustic impression is zeeep. The shortest calls, which were often produced as couplets or triplets before the next long call, sound more like clicks but relatively few short calls (<0.1 s) were recorded (Fig. 8f). The dominant frequency commences at ~3 khz, rising to 3.8 khz, with secondary bands around 2 and khz. Frequencies of Alexishafen calls are not comparable in detail with the Manus and Siewa recordings, described below, as they were made using different microphones. Only a few recordings of this species were obtained (by Richards) at Siewa because not many males were calling, and only sporadically. Calls consisted of a series of long, buzzing notes (Fig. 17d). However, he also heard shorter calls that were probably attributable to this species. Note length ranged from to s and, in the longer calls, amplitude rose initially and then remained flat until a final decay. Dominant frequency rose from ~3.5 to 4 khz during the call (Fig. 17c, d). Table 5. Summary of biometric data for Litoria eurynastes sp. nov. Variable Female (n = 11) Male (n = 37) Range Mean s.d. s.e. Range Mean s.d. s.e. HB TL/HB HL/HW HW/HB EN/IN EY/HB TY/HB F3d/HB T4d/HB

21 Systematics of Litoria bicolor (Anura : Hylidae) in the Papuan region Australian Journal of Zoology 277 Fig. 17. (a) Wave form (upper), spectrogram and frequency spectrogram (lower) of a long call of Litoria eurynastes recorded near Alexishafen, unfortunately obscured by rain responsible for the heavy low-frequency background. (b) Wave form of three calls of Litoria eurynastes recorded near Alexishafen. (c) Wave form (upper), spectrogram (lower left) and frequency spectrum (lower right) of a single long call, with 7 pulses, of Litoria eurynastes, recorded at Siewa River. (d) Wave form (upper), spectrogram and frequency spectrum (lower) of two overlapping long calls of Litoria eurynastes recorded at Siewa River. Arrows indicate points at which spectra were taken. On Manus Island, conditions were dry and frogs called only sporadically. In a sample of seven calls, only long, buzzing notes ( s), with a pulse rate from 108 to 223 s 1 were heard (Fig. 18a). Dominant frequency was slightly above or slightly below 4 khz and frequency rose by ~200 Hz as the call progressed. A secondary band at 2 khz is prominent and higher frequencies are less emphasised. The amplitude during the first half of the call is very low but then increases sharply through the call and pulse rate increases so that in the final part pulse rate is difficult to determine (Fig. 18a). The acoustic impression of the Manus and Siewa calls was similar. Only a few calls were recorded on Seram (by Menzies) and these were obscured by coincident calls of L. infrafrenata so that detailed analysis was impossible. Only nine long calls (0.445 s) were recorded, and eight of these were uttered in pairs. As in other calls of the large bicolor complex, amplitude increased,