Sampling of various species of dicroglossid frogs was carried out across the Andaman and Nicobar Archipelago (Tables 1 and 2). Opportunistic searches were carried out in a wide range of habitats such as primary and secondary forests, agricultural fields, parks, beaches and wayside areas with permanent or temporary water bodies, from sea level up to elevations of nearly 700 m asl. During the breeding season, individuals were often located by calls. Live specimens were photographed in the wild or captive conditions and euthanised in Tricaine methanesulfonate (MS-222) solution. Tissue samples were obtained from thigh (adult) or tail muscle (tadpoles) and preserved in absolute ethanol. Specimens were fixed in 4% formalin and rinsed in water before preservation in 70% ethanol. The sampled specimens are available in the amphibian collection of Zoological Survey of India, Andaman and Nicobar Regional Centre, Port Blair (ZSI/ANRC) or the Systematics Lab at University of Delhi (SDBDU). Geographical coordinates and elevation at the sampling localities were recorded using the WGS84 datum system. Maps were prepared using QGIS (http://www.qgis.org).

Genomic DNA was extracted from 15 samples using the Qiagen DNeasy blood and tissue kit (Qiagen, Valencia, CA, USA). From all the extracted samples, a ~540 bp fragment of the mitochondrial 16S rRNA gene was PCR-amplified using primers from Simon et al. (1994). Three additional gene fragments were sequenced for selected samples, using previously published primers: 385 bp of the mitochondrial 12S rRNA (Richards and Moore 1996), 564 bp of the nuclear recombination activating gene 1 (Biju and Bossuyt 2003), and 603 bp of the nuclear tyrosinase (Bossuyt and Milinkovitch 2000). Sequencing was performed on both strands using a BigDye Terminator v3.1 Cycle Sequencing kit on an ABI 3730 automated DNA sequencer (Applied Biosystems). Raw sequences were assembled and checked in ChromasPro v1.4 (Technelysium Pty Ltd.). Sequences from this study are deposited in the National Center for Biotechnology Information (NCBI) GenBank under accession numbers ON009953–ON009969 and ON010541-ON010544. Additional homologous sequences were retrieved from the GenBank for all known members of the Minervarya andamanensis species group and representatives of other Minervarya species. Ten species from other closely related dicroglossid genera were used as the outgroup taxa for phylogenetic analyses. Datasets for each gene were assembled and aligned using MUSCLE (Edgar 2004) in MEGA 7.0 (Kumar et al. 2016). The alignments for coding DNA were checked by comparison with amino acid sequences, whereas those for the non-coding fragments were manually optimised.

Bayesian inference (BI) and Maximum Likelihood (ML) analyses were performed with a concatenated character matrix of 2,101 nucleotides for 69 taxa (Table 1). The data was partitioned by genes for 16S and 12S, and by codons for Rag1 and Tyr, with a total of eight partitions. The following best-fitting models of sequence evolution for each partition were selected through a greedy search in PartitionFinder 2 (Lanfear et al. 2017) using the corrected Akaike information criterion (AICc): GTR+I+G for 16S and 12S, TVM+I+G for the first codon positions of Rag1 and Tyr, K81UF+I for the second codon positions of Rag1 and Tyr, TRN+G for Rag1 third codon position, and TVM+G for Tyr third codon position. Using this partitioning scheme, the Bayesian phylogenetic inference was performed in MrBayes (Ronquist and Huelsenbeck 2003) with four independent Bayesian runs, each running with four Metropolis-Coupled Markov chain Monte Carlo (MCMCMC) chains for 20,000,000 generations using default priors, chain temperature of 0.1, and tree sampling at every 4,000 generations. The convergence of the runs was determined by the nearing of standard deviation of split frequencies < 0.01 and potential scale reduction factors ~1.0. Stationarity of the likelihood scores and effective sample sizes (ESS) for all parameters were viewed in Tracer v. 1.7 (Rambaut et al. 2018). The Bayesian posterior probabilities (BPP) were summarised after discarding the first 25% trees as burn-in (Huelsenbeck et al. 2001). A partitioned maximum likelihood analysis was also performed for 10,000 ultrafast bootstrap (UBS) replicates, executed with the ‘auto’ model selection option, using IQ-TREE (Minh et al. 2013) on the IQ-TREE webserver (Trifinopoulos et al. 2016). Nodes with BPP≥95% and UBS≥90% were considered well supported.

A species delimitation analysis was performed using the multi-gene ML phylogram as input by Bayesian implementation of the Poisson Tree Processor (PTP) method (Zhang et al. 2013) on the bPTP webserver (https://species.h-its.org). To further assess the population structure in the Minervarya andamanensis species group, a haplotype network was constructed using the available 16S rRNA sequences. A dataset of 32 sequences comprising 513 characters, excluding sites with missing data but including the alignment gaps, was used to reconstruct haplotypes using the PHASE algorithm (Stephens et al. 2001) in DnaSP version 5 (Librado and Rozas 2009). A median-joining network was then constructed with 64 recovered haplotype sequences using the software Network 4.6.1.0 (www.fluxus-engineering.com). Intra- and interspecific uncorrected pairwise genetic distances for 16S sequences of the M. andamanensis species group were computed in PAUP* 4.0b10 (Swofford 2002).

We morphologically examined our new collections and compared them with the available type specimens, original descriptions, other topotypic specimens or general collections of all the dicroglossid frogs known to occur in the Andaman and Nicobar Archipelago. Sex and maturity were determined by the presence of secondary sexual characters (such as nuptial pads and vocal sacs in males) or examination of gonads through a small lateral or ventral incision. Only adult (sexually mature) individuals were used for morphometric studies. The following measurements were taken to the nearest 0.1 mm using digital slide-calipers with the aid of a stereomicroscope, following measurements and associated terminologies of Garg and Biju (2017, 2021): snout-vent length (SVL), head width (HW, at the angle of the jaws), head length (HL, from rear of mandible to tip of snout), MN (distance from the rear of the mandible to the nostril), MFE (distance from the rear of the mandible to the anterior orbital border), MBE (distance from the rear of the mandible to the posterior orbital border), snout length (SL, from tip of snout to anterior orbital border), eye length (EL, horizontal distance between bony orbital borders), inter upper eyelid width (IUE, the shortest distance between the upper eyelids), maximum upper eyelid width (UEW), internarial distance (IN), internal front of the eyes (IFE, shortest distance between the anterior orbital borders), internal back of the eyes (IBE, shortest distance between the posterior orbital borders), NS (distance from the nostril to the tip of the snout), EN (distance from the front of the eye to the nostril), TYD (greatest tympanum diameter), TYE (distance from the tympanum to the back of the eye), forearm length (FAL, from flexed elbow to base of outer palmar tubercle), hand length (HAL, from base of outer palmar tubercle to tip of third finger), FL_I–IV (finger length), thigh length (TL, from vent to knee), shank length (SHL, from knee to heel), foot length (FOL, from base of inner metatarsal tubercle to tip of fourth toe), total foot length (TFOL, from heel to tip of fourth toe), FD (maximum disc width of finger), width of finger (FW, measured at the base of the disc), TD (maximum disc width of toe), width of toe (TW, measured at the base of the disc). Digit number is represented by Roman numerals I–V in subscript. Measurements and photographs are mostly for the right side of the specimen, unless a character was damaged, in which case the left side was taken. All measurements provided in the taxonomy section are in millimetres. The body size categories discussed in the text for the purpose of convenience and morphological comparisons follow Garg and Biju (2021). The webbing formulae follow Savage and Heyer (1967), as modified by Myers and Duellman (1982). The amount of webbing relative to subarticular tubercles is described by numbering the tubercles 1–3, starting from the base of the digits.

Abbreviations. Museum acronyms and other frequently used abbreviations are as follows: ZSI (Zoological Survey of India); ZSIC (Zoological Survey of India, Kolkata); ZSI/ANRC (Zoological Survey of India, Andaman and Nicobar Regional Centre, Port Blair); ZSI/SRS (Zoological Survey of India, South Regional Station, Chennai); NHM (Natural History Museum, London), formerly BMNH (British Museum [Natural History], London); Systematics Lab, University of Delhi (SDBDU).

ZooBank registration. This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the International Commission on Zoological Nomenclature (ICZN). The ZooBank LSID (Life Science Identifier) for this publication is urn:lsid:zoobank.org:pub:2780E8F9-1ABF-4708-898A-B14447591063. The LSID and associated information can be resolved through any standard web browser by appending the LSID to the prefix http://zoobank.org.

Our phylogenetic study found ‘Ingerana’ charlesdarwini to be deeply nested within the genus Minervarya (Fig. 1), based on 16S rRNA sequences from four exemplars, as well as additional mitochondrial (12S rRNA) and nuclear markers (Rag1 and Tyr) from two selected samples. Both the BI and ML analyses of a concatenated dataset of 2,101 characters recovered the species as a distinct well-supported (BPP 100, UFB 99) lineage in the M. andamanensis species group (Garg and Biju 2021). It also formed a highly supported (BPP 100, UFB 98) sister-group relationship with M. andamanensis, showing relatively shallow divergence of 2.8–3.6% for the 16S rRNA sequences. In comparison, the two previously recognised members of the group, M. andamanensis and M. muangkanensis, showed a higher divergence of 3.8–4.6%. Further, ‘Ingerana’ charlesdarwini was divergent from M. muangkanensis by 4.4–4.7% for the 16S locus. These interspecific divergences fall well within the range of genetic distances usually observed at the species-level in the genus Minervarya and closely related dicroglossid genera (e.g., Kotaki et al. 2010; Köhler et al. 2019; Garg and Biju 2021). Our results, thus, clearly indicate that ‘Ingerana’ charlesdarwini is a member of the genus Minervarya and should therefore be treated as Minervarya charlesdarwinicomb. nov. Within the M. andamanensis species group, both the insular Andaman species, M. charlesdarwini and M. andamanensis, are more closely related to each other than to M. muangkanensis of mainland Thailand and Myanmar region, which formed the basal lineage (Fig. 1).

Within the focal Minervarya andamanensis species group, the multi-loci species delimitation analysis recovered all the three recognised species as distinct (Fig. 1). At the same time, the mitochondrial 16S median-joining network did not find sharing of haplotypes among the three species. A total of 12 haplotypes were recovered with an overall haplotype diversity of 0.8834 within the species group. In line with the results obtained in the phylogenetic analyses, the mainland member of the group, M. muangkanensis, was the most distinct species showing separation from M. andamanensis by minimum 24 mutation steps and from M. charlesdarwini by 26 steps. Eight haplotypes were detected among individuals of M. muangkanensis. Although the populations from Thailand and Myanmar did not share any haplotypes, the absence of clear genetic structuring indicates ongoing gene flow and admixture between these geographically continuous regions. The populations of M. andamanensis from South Andamans and Little Andamans formed two distinct haplotype clusters without sharing of haplotypes, separated by minimum six mutation steps and considerable genetic differentiation (1.1–1.3%), suggestive of limited gene flow, longer geographical isolation, and hence a potential area of population differentiation between these islands. On the other hand, the two widely co-occurring sister species within the Andaman Islands that do not exhibit similar habitat associations—M. charlesdarwini (largely forest dwelling, phytotelm breeding, and possibly a habitat specialist) and M. andamanensis (occupies a broader range of habitats, breeds in open water bodies, and is possibly a habitat generalist)—formed clearly distinct clusters separated from each other by minimum 18 mutation steps, suggesting that sympatric speciation potentially occurred within this radiation (Fig. 1). These species appear to present an insightful case for future investigations on the patterns of gene flow, speciation and diversification processes, ecological niche segregation, and phylogeography within the islands of Andaman and Nicobar.

Two species of the genus Limnonectes Fitzinger, 1843 are purported to occur in the Andaman Islands. The reports of both L. doriae (Boulenger 1887) by Annandale (1917) and L. hascheanus by Boulenger (1920) are based on three out of the four reported type specimens of Rana gracilis var. andamanensis Stoliczka, 1870 (current name combination: Minervarya andamanensis). While describing M. andamanensis, Stoliczka (1870) mentioned examination of “four specimens from Port Blair”, of which three types available in the collection of ZSI, Kolkata were stated as “2732, 3538–9” “Types of R. gracilis, var. andamanensis, Stol.” by Sclater (1892) under the name Rana limnocharis (on page 6), and not as Rana doriae (on page 4, reported only from Burma) as later suggested by Chanda et al. (2001 “2000”). Of these, Annandale (1917) found two to be labelled as types and selected 3539 as the type; therefore by implication designating it as the lectotype, which he found distinct enough to be recognised as a subspecies or a local race of R. limnocharis (current name combination: Fejervarya limnocharis Gravenhorst, 1829). At the same time, however, it was Annandale (1917) who stated that the larger and better preserved of the two labelled types undoubtedly belongs to R. doriae Boulenger, 1887 (current name combination: Limnonectes doriae) and the same was followed by Boulenger (1920), who additionally discussed that “one of the types received from the Indian museum in 1893” belonged to Rana hascheana Stoliczka, 1870 (current name combination: Limnonectes hascheanus). Following these works, Smith (1941) included L. doriae in the herpetofauna of Andamans, but did not mention L. hascheanus. Sarkar (1990) reported the distribution of L. doriae in both the Andaman and Nicobar group of islands based on examination of “16 frogs”—two from Stoliczka’s Andaman collection (possibly referring to two syntypes of M. andamanensis), another of Stoliczka’s Nicobar collection, and several other collections made by subsequent workers. He also included L. hascheanus in the faunal list of Andamans, following Boulenger (1920), although clearly stating “I have got no specimen in my disposal” (Sarkar 1990). In addition, Sarkar (1990) provided a vouchered record of Rana macrodon var. blythii Boulenger, 1920 (currently, a composite of Limnonectes blythii, L. leporinus, and possibly L. malesianus) based on material from “Tribeni Nullah, Campbell Bay, Great Nicobar” collected in 1977. Dutta (1997) stated the number of the type of Rana gracilis var. andamanensis housed in the NHM, London collection as BMNH 1947.2.1.23. Later, Chanda et al. (2001 “2000”) corrected the catalogue numbers for two types deposited at ZSI, including the lectotype, as ZSI 8538 and ZSI 8539, while also stating that two of the three types “cannot be located at present” (possibly referring to ZSI 2732 and ZSI 3538 / 8538 that were regarded as belonging to L. doriae). The available lectotype of Minervarya andamanensis, however in our observation, carries the number “3539” on the original label found inside the specimen jar and “ZSI 8539” on the outside label.

Over the years, both Limnonectes doriae and L. hascheanus have been included in the regional faunal lists of the Andaman Islands (e.g., Das 1999; Harikrishnan et al. 2010, 2012; Chandramouli et al. 2015; Rangasamy et al. 2018), however without any new vouchered records. Neither has any subsequent study attempted to provide morphological diagnoses or clear explanations in support of what became the first reports of these species and the genus Limnonectes from the region. Harikrishnan and Vasudevan (2018) emphasised on the need for detailed studies to confirm the occurrence of L. doriae and L. hascheanus, and two unnamed Limnonectes mentioned by Das (1999), in the Andaman group of islands. Inger and Stuart (2010) have previously discussed that L. hascheanus is restricted to high elevations of about 1000 feet above sea level in southern parts of the Malay Peninsula, and expressed doubts on its occurrence in the Andamans. Following this, Harikrishnan and Vasudevan (2018) suggested the record of L. hascheanus from Andamans to be considered tentative.

The present study failed to locate the original syntypes (now paralectotypes) of Rana gracilis var. andamanensis that were identified as belonging to Limnonectes hascheanus and L. doriae (SDB personal observation at NHM in 2010; SDB and SG personal observation at ZSI in 2018). Neither did we locate any other specimens referable to these species from this region in potential museums such as ZSIC (Kolkata, India), ZSI/ANRC (Andaman and Nicobar, India), and NHM (London). The additional material reportedly studied by Sarkar (1990) for L. doriae originated from various surveys and lacks accompanying voucher or museum information, making their traceability difficult. Hence, during our study, we made an extensive effort to locate frogs possibly referable to L. doriae and L. hascheanus in the Andaman Islands, particularly at the type locality of Minervarya andamanensis in Port Blair and surroundings. Instead of locating these species, we found the populations of M. andamanensis collected from Andaman Islands to be extremely variable in size, skin texture, dorsal colouration and markings (Figs 5, 6), including the absence of the distinctive chestnut-brown dorsal colouration with dark brown lateral surfaces that are considered typical of this species (Annandale 1917; Chandramouli et al. 2015, 2021). Several morphologically variable individuals were also included in our molecular analyses and found to be conspecific or shallowly divergent, providing evidence that even though these populations exhibit morphological variations they represent a single widely distributed species. Hence, the variation among M. andamanensis individuals (such as variable skin colouration, markings, texture, and their overall robust and stout appearance) could have been a source of confusion leading to the presumed occurrence of Limnonectes doriae and L. hascheanus in India.

At the same time, we observed that Minervarya charlesdarwini and M. andamanensis occur sympatrically in most of the reported and surveyed localities. With both the species being extremely variable in dorsal colour and markings, the possibility of L. doriae and L. hascheanus being misidentifications of M. charlesdarwini cannot be ruled out. While describing L. hascheanus from Peninsular Malaysia, Stoliczka (1870) discussed a “W mark” (page 147 and pl. IX, fig. 3), which was also discussed to be present in the NHM specimen of Rana gracilis var. andamanensis (BMNH 1947.2.1.23) by Boulenger (1920). We have found several similar-sized specimens of M. charlesdarwini possessing a W-shaped mark (Figs 3, 4), providing support for possible misidentifications between the two taxa. As for L. doriae, based on the description of specimens Sarkar (1990) regarded as belonging to this species, most of the discussed characters could be confused with Limnonectes species, as well as their ecology “collected from marshy area in deep forests” appear to be comparable with M. charlesdarwini. Hence, we believe that the suggested occurrence of both L. doriae and L. hascheanus in Andamans is likely to have been based on misidentifications of either M. charlesdarwini or M. andamanensis, or possibly even a mix of both these morphologically variable and highly confusing species (Figs 3–6). Chandramouli (2017) also reported on some overlooked museum specimens collected by Annandale and deposited under the name “Rana doriae andamanensis” as belonging to M. charlesdarwini.

In light of the above and the fact that no recent surveys, especially since the description of Minervarya charlesdarwini, have reported new specimens referable to the two Limnonectes species, except for their mostly unverified inclusion in regional checklists, the occurrence of both L. doriae and L. hascheanus in Andamans should not only be considered erroneous but the two should be excluded from the list of Andaman amphibians to avoid further confusions. It is also interesting that Stoliczka could have collected the enigmatic M. charlesdarwini over a century ago, in 1869. However, we may not know with certainty, unless the discovery of the ‘lost’ specimens from Stoliczka’s collection, or the verification of at least some specimens examined by Sarkar (1990), whose judgement was based on both Stoliczka’s and subsequent additional collections.

The only minervaryan species to be reported from the Nicobar group of the Andaman and Nicobar Archipelago is Minervarya nicobariensis (Stoliczka, 1870). This taxon was originally described as a new variety “var. nicobariensis” of Rana gracilis Wiegmann sensu Günther, 1864 (= Fejervarya limnocharis Gravenhorst, 1829) from “the Nicobars, in the neighbourhood of the Nancowri harbour”. The description was based on “one peculiar young specimen” measuring “1¼^th inch” (= 31.75 mm), which was later stated to be ZSIC 2679 by Sclater (1892). The type was reported as lost (Dubois 1984) or unlocatable (Chanda et al. 2001 “2000”), and later, as “lost or destroyed” by Chandramouli and Prasad (2020) who also designated ZSI/ANRC/T/12326 from “Munak, Camorta Island [in the vicinity of the holotype locality fideStoliczka 1870]” as a neotype. While trying to locate the original name-bearing type at ZSI Kolkata we found two young specimens ZSIC 3567 (SVL 13.6 mm) and ZSIC 3570 (SVL 13.4 mm) both labelled as “syntype”. However, these numbers have not been reported in any of the previous works, hence a further investigation will be necessary to ascertain their type status.

Since the original description, this taxon has been moved to three currently recognised dicroglossid genera, Limnonectes, Fejervarya, and Minervarya, chiefly owing to the different genus-level reorganisations proposed within the family (e.g., Dubois 1987; Dubois and Ohler 2000; Sanchez et al. 2018). Chandramouli and Prasad (2020) redescribed the species based on fresh adult and larval collections and provided a revised diagnosis as M. nicobariensis. They also briefly reported on the natural history, call characteristics, and distribution of the species in the Nicobar Islands. Garg and Biju (2021) assigned this species to the M. andamanensis species group based on morphological affinities. In the present study, our detailed morphological study of several additional new and museum specimens of this species, including the neotype, reveals a close relationship between M. nicobariensis and M. charlesdarwini of the Andaman Islands (Fig. 7).

Not only do these two species share several unique morphological traits (such as scattered dorsal and lateral tubercles with black spots, presence of discontinuous skin folds on dorsum, upper ⅔rd of tympanum and inner margin of tympanic fold dark brown, absence of prominent markings on groin, and ventral surfaces of hand and foot light grey to blackish-brown) compared to other members of the Minervarya andamanensis group, but also exhibit similarities in being primarily associated with forest habitats and their phytotelm breeding preferences. In view of the surprising phylogenetic position of M. charlesdarwini revealed in the present study, which had alluded researchers for several years, a sister relationship of M. nicobariensis with M. charlesdarwini is likely to be expected, for which a future molecular assessment can provide a conclusive evidence.

Distribution.Minervarya nicobariensis is endemic to the Nicobar Archipelago of India, where we find it to be widely distributed in the central and southern group of islands: Central Nicobar (Bompoka Is., Kamorta Is., Nancowry Is., and Katchal Is.) and South Nicobar (Little Nicobar and Great Nicobar). This species has been observed between elevations of nearly sea level up to 170 m asl (Fig. 2; Table 2).

Minervarya agricola is one of the most widely distributed species of minervaryan frogs having a distribution across the Indian mainland (from Tamil Nadu, Kerala, Karnataka, Maharashtra, Gujarat, Rajasthan, Punjab, Haryana, Delhi, Uttarakhand, Uttar Pradesh, Chhattisgarh, Andhra Pradesh, Odisha, West Bengal, Bihar, up to Assam), Nepal, Bangladesh, and Sri Lanka (Garg and Biju 2021). Our samples of a smaller-sized minervaryan species from North, Middle, South, and Little Andamans were phylogenetically (Fig. 1; Table 1) and morphologically (Fig. 8) conspecific with M. agricola, providing the first record of this species from these islands. The 16S gene sequences from the Andamanese M. agricola are identical to those from the typical mainland populations of the species. Morphologically also, the individuals from Andamans exhibit only minor variations in skin colouration and markings (Fig. 8) that are usually observed in this species across its entire known range. This species could have been previously misidentified either as Fejervarya ‘limnocharis’ or Minervarya andamanensis, both of which are frequently reported to occur in the Andaman Islands (e.g., Sclater 1892; Sarkar 1990; Pillai 1991; Dutta 1997; Das 1999; Harikrishnan and Vasudevan 2018; Rangasamy et al. 2018). However, the taxon with which M. agricola may have been confused with remains unclear as the name F. ‘limnocharis’ was also applied to F. moodiei populations from the Andamans by studies in the past (Chandramouli et al. 2020b), while M. andamanensis, even though widely reported and frequently included in regional checklists, is only known with certainty from a handful of available museum specimens (Annandale 1917; Chandramouli et al. 2021). A record of Rana keralensis (= Minervarya keralensis) from Andamans based on a specimen “measuring 30 mm” (Pillai 1991) could also be a misidentification of M. agricola; although none of the subsequent studies seem to have included this taxon in the regional checklists. During our examination of the ZSI/ANRC collection, we did though locate some specimens of M. agricola labelled as Fejervarya limnocharis, a name that has been used extensively for several misidentified minervaryan and fejervaryan species across South and South-east Asia for nearly two centuries. Hence, in addition to providing a new distribution record of M. agricola from the Andaman Archipelago, our study provides further support for the absence of F. limnocharis from this region.

Distribution.Minervarya agricola is a widely distributed species of South and Southeast Asia, being found in India, Sri Lanka, Bhutan, Nepal, Bangladesh, Myanmar, Thailand, and southern China (Garg and Biju 2021). In the Andaman Archipelago of India, we provide new reports of this species from all the major groups of islands: North and Middle Andamans (North Andaman Is., Baratang Is., and Middle Andaman Is.), South Andamans (South Andaman Is., Neil Is., and Havelock Is.), up to the Little Andaman Island. This species has been observed between elevations of sea level up to elevations of nearly 130 m asl (Fig. 2; Table 2).