Taxonomic revision of the genus Elmomorphus Sharp, 1888 I. Japanese and Korean species (Coleoptera: Dryopidae)

The species of Elmomorphus Sharp, 1888 occurring in Japan and Korea are redescribed and illustrated: E. brevicornis Sharp, 1888 (Japan, Korea) and E. amamiensis Nomura, 1959 (Japan). The standard barcoding fragment of the mitochondrial gene coding for cytochrome c oxidase subunit I (COI) was sequenced and used together with morphological characters to delimit the taxonomic boundaries of the two species. To assess their morphometric variation, eight morphometric characters were measured and statistically evaluated using principal component analysis. The two species of Elmomorphus formed distinct and well-separated clusters in the COI tree. Their interspecific divergence is very high, ranging from 22.7 to 23.9%. On the other hand, morphometric characters, including those previously presumed to be diagnostic, overlap and per se do not allow unambiguous species identification. Reliable morphological distinguishing characters are described for males and females. Molecular data along with the morphological evidence strongly confirm the species status of E. amamiensis. An identification key to the Japanese and Korean species is provided.


Introduction
The genus Elmomorphus Sharp, 1888 was described by the renowned British coleopterist David Sharp (1840-1922 based on E. brevicornis Sharp, 1888 from Japan. Over the next 130 years, another 17 species were added, mainly from the Oriental Region. The species of the genus Elmomorphus are regarded as True Water Beetles (Jäch 1998;Jäch & Balke 2008). They spend most of their life span submerged in various running water habitats. Most species prefer streams and rivulets fl owing through forested areas. Immature adults undertake dispersal fl ights and are often collected in large numbers at light, at least the widely distributed and common species in Southeast Asia. However, there are numerous undescribed species, especially those living in primary rainforests (e.g., in Borneo), which were never collected at light, even if light traps were exposed near streams inhabited by large Elmomorphus populations.
Due to the remarkable morphological similarity, E. amamiensis was initially described as a subspecies of E. brevicornis by Nomura (1959) from Amami Island in the Ryukyu Archipelago. The specifi c status was proposed recently by Jung & Bae (2014), referring to several more or less evident morphological characters.
We examined the type material of E. brevicornis housed in the Sharp collection in the Natural History Museum London and additional material, including freshly collected specimens of both taxa from Japan. Using an integrative taxonomic approach, we redescribe, illustrate, and diagnose E. brevicornis and E. amamiensis, and provide reliable morphological distinguishing characters for both sexes.
This publication is the fi rst part of a comprehensive taxonomic revision of the genus Elmomorphus. The second part (in prep.) will deal with the Chinese species, and it will include a thorough morphological redescription of the genus. The fi nal parts will deal with the species of the Oriental and Australian regions. The revision of Elmomorphus started in the 1990s, but due to the ever-growing number of specimens and new species collected, its publication has been retarded.

Taxonomic methods
The material examined in this study is deposited in the Natural History Museum London, United Kingdom (NHMUK), Naturhistorisches Museum Wien, Austria (NHMW), and the Ján Kodada collection, Comenius University in Bratislava, Slovakia (JKCB).
Dry specimens were fi rstly relaxed for several hours in water with several drops of concentrated acetic acid. Male genitalia were dissected, cleared in lactic acid for one or two days, and temporarily mounted on a microscope slide with a single cavity in Berlese's fl uid (Swan 1936). After examination/drawing, the male genitalia were mounted in a drop of DMHF on the same card as the respective specimen. Female genitalia were observed in glycerol and stored in micro-vials fi lled with glycerol and pinned under the studied specimen. Examination and dissection were made under a Leica MZ16 stereo microscope with magnifi cations up to 120× and a diffuse LED lighting. Drawings were hand-made with a pencil, using a drawing device attached to a Leica DM 1000 microscope and subsequently traced in Adobe Illustrator CC. The specimens were photographed under a Zeiss Axio-Zoom V-16 stereo microscope and a Canon 5D Mark IV camera attached. Each habitus photography was created by stacking 100-120 focal planes using the software ZereneStacker (https://zerenesystems.com/cms/stacker).
For scanning electron microscopy, specimens were dehydrated in graded ethanol series and air-dried from absolute ethanol, coated with gold, and then viewed in a TESCAN microscope.
Label data are cited verbatim, enclosed within quotation marks; individual labels are separated by a vertical bar; square brackets enclose additional information.
The morphological terminology is primarily based on  with some additions.
Measurements were made under a Leica MZ16 stereo microscope with Leica eyepiece cross micrometre (5 : 100). In total, 12 morphometric characters, as listed in the 'Abbreviations' section below, were scored on 37 specimens. The range of measured values and ratios is followed by mean, standard deviation, and the number of individuals measured.
Principal component analysis (PCA) was conducted on eight morphometric characters: PL, PW, EL, EW, PrTL, MsTL, MtTL, and ID. Datasets were compiled separately for both sexes and included 14 males and seven females of E. brevicornis, and nine males and seven females of E. amamiensis. Data were logtransformed before the analysis, and a variance-covariance matrix was used for the subsequent calculations. Several specimens used for DNA isolation were also included in the datasets. The analysis was performed in the PAST 3.12 software (Hammer et al. 2001). Plots were edited in Adobe Illustrator CC.

Molecular methods
Eight specimens of E. amamiensis and 16 specimens of E. brevicornis from Japan, fi xed in 96% ethanol were used for molecular analyses. Unfortunately, no specimens from Korea were available. Tissue samples contained either one leg with coxa and attached muscle, or the entire abdomen of an adult specimen. DNA was isolated with the E.Z.N.A. ® Tissue DNA kit (OMEGA Bio-tek Inc., Norcross, GA, USA) according to the manufacturer's protocol. The 5' end of the mitochondrial gene coding for cytochrome c oxidase subunit I (COI), the standard barcode (Hebert et al. 2003), was PCR amplifi ed with the standard primers LCO1490 and HCO2198 (Folmer et al. 1994). Individual PCR reactions were conducted in a total volume of 15 μl and included 6.67 μl of GoTaq ® Green master mix (Promega, Fitchburg, WI, USA), 0.34 μl of each primer (10 pmol/μl), 50 ng of template DNA, and 4.65 μl nucleasefree water. The PCR thermocycler program was as follows: 94°C for 180 s, 40 cycles of 94°C for 40 s, 52°C for 40 s, and 72°C for 60 s, and 72°C for 10 min. PCR products were extracted from the 1% TBE agarose gel, using the Exo-CIP ™ Rapid PCR Cleanup Kit (New England Biolabs ® Inc., Ipswich, MA, USA) according to the manufacturer's protocol. Purifi ed PCR products were sequenced from both sides in Macrogen Europe B.V. (Amsterdam, Netherlands). Sequences were trimmed and assembled into contigs in Geneious ver. 6.1.8 (https://www.geneious.com).

Phylogenetic methods
The nucleotide COI sequences were aligned based on the predicted amino acid sequences with MEGA-X (Kumar et al. 2018), using the invertebrate mitochondrial genetic code and the MUSCLE codon algorithm. COI sequences of Pomatinus substriatus (Müller, 1806) and Dryops auriculatus (Geoffroy, 1785) were retrieved from the GenBank database (https://www.ncbi.nlm.nih.gov/genbank/) and served as outgroup. Their accession numbers are specifi ed in the respective fi gure.
Three algorithms were used to infer phylogenetic relationships and to confi rm the consistency of the results. A maximum likelihood tree was constructed with the program IQTree (Nguyen et al. 2015) on XSEDE ver. 1.6.10, using the CIPRES science gateway ver. 3.3 (http://www.phylo.org/) (Miller et al. 2010). The best fi tting evolutionary substitution model was determined with the aid of the BIC criterion, using the in-built ModelFinder program. The maximum likelihood analyses included the BioNJ tree option to build the starting tree, correction for ascertainment bias, 1000 ultrafast bootstrap replicates to assess the reliability of internal nodes, and the bnni algorithm to reduce overestimating support (Hoang et al. 2018). Bayesian inference was carried out in the program MrBayes (Ronquist et al. 2012) on XSEDE ver. 3.2.7 under the best evolutionary substitution model, as selected with the BIC criterion in jModelTest ver. 2.1.10 (Darriba et al. 2012). Markov Chain Monte Carlo (MCMC) simulations included two independent runs each with four simultaneous chains, one million generations, a sampling frequency of trees and parameters at one hundred, and a burn-in fraction at 25%. Convergence of the MCMC analyses to stationary distribution and an adequate sample from the posterior distribution were confi rmed using the in-built MrBayes diagnostics. Finally, a distance neighbour-joining (NJ) tree was constructed with the maximum composite likelihood method, gamma distributed rates among sites (α = 0.5), and heterogenous patterns among lineages. The reliability of nodes in the distance tree was assessed with 1000 bootstrap pseudo-replicates. All trees were computed as unrooted and were rooted a posteriori using the outgroup taxa in FigTree ver. 1.2.3 (http://tree.bio.ed.ac.uk/software/fi gtree/).
For the barcoding analyses, pairwise p-distances were calculated in MEGA-X (Kumar et al. 2018).
Histograms showing the intraspecifi c, interspecifi c as well as the intergeneric distances were prepared in Python ver. 3.6.6, using the libraries scikit-learn (Pedregosa et al. 2011) and Matplotlib (Hunter 2007). Multidimensional scaling (MDS) was carried out on the pairwise p-distances with the scikitlearn library. The ordination MDS diagram was constructed with the metric technique, the SMACOF algorithm, and 250 initializations, whereby each initialization had 20 000 iterations, and ε was set to 10 -8 to declare convergence.
A haplotype network was calculated from the COI alignment using the TCS method (Clement et al. 2000) in PopART ver. 1.7 (Leigh & Bryant 2015). Mutations were displayed as numbers along the network edges.
Abbreviations APD = linear distance between anterior angles of pronotum in dorsal aspect EL = elytral length from the middle of the anterior margin of the scutellum (scutellary shield) to the apex of the elytra EW = greatest width of elytra in dorsal aspect ID = minimum linear distance between eyes in dorsal aspect MsTL = maximum length of mesotibia MtTL = maximum length of metatibia PhL = greatest linear distance between base and apex of phallobase in the lateral aspect PL = pronotal length along the midline PrL = greatest linear distance between base and apex of parameres in the lateral aspect PrTL = maximum length of protibia PW = maximum pronotal width TL = linear distance between the anterior margin of the pronotum and the apex of the elytra along the midline

Phylogenetic analyses
Nineteen new COI sequences were obtained from two species of Elmomorphus; seven sequences from E. amamiensis and 12 sequences from E. brevicornis. Their origin and GenBank accession numbers are summarised in Table 1. Three phylogenetic methods (IQTree, MrBayes, and NJ) were utilised to delimit their taxonomic boundaries and reconstruct their interrelationships. All algorithms brought almost identical results and, therefore, only the maximum likelihood IQTree is presented along with nodal supports from all three statistical methods (Fig. 1A). The two species of Elmomorphus formed strongly to fully statistically supported clusters that were separated by long branches in the COI trees. Despite their deep divergence, E. amamiensis and E. brevicornis grouped with very strong statistical support (98% IQTrees, 1.00 MrBayes, 97% NJ). Their monophyletic origin was recognized not only when the trees were rooted by outgroup taxa, but also when the midpoint rooting method was applied.
The distinctness as well as the deep divergence of the two species of Elmomorphus were also corroborated by barcoding analyses and multidimensional scaling. The intraspecifi c distances ranged from 0.0 to 0.5% within E. amamiensis and from 0.0 to 2.1% within E. brevicornis (Table 2). There was no overlap between intraspecifi c and interspecifi c p-distances, but interspecifi c and intergeneric distances signifi cantly overlapped in the histogram (Fig. 1B). This pattern was caused by the very deep interspecifi c divergence of E. amamiensis and E. brevicornis, which spanned a range of 22.7-23.9% (Table 2). Similarly, each species formed a homogenous cluster well separated from all other taxa in the MDS ordination diagram inferred from the pairwise p-distances (Fig. 1C). Distances among individual species were comparable, documenting their deep genetic divergences as well. Haplotype network analyses supported the monophyletic origin of E. amamiensis and E. brevicornis by 55 shared nucleotide positions within the 567 nucleotides long barcoding COI sequence fragment. Specimens of the former species were separated by 1-3 mutation steps, while those of the latter species by 1-12 mutation steps. The two species of Elmomorphus differed by as many as 109 nucleotide positions, documenting their very deep genetic divergence (Fig. 2).

Multivariate morphometric analyses
To assess the morphometric variation of the two species, principal component analysis (PCA), based on eight morphometric characters, was carried out for each sex separately. In males, the fi rst and second principal components (PC1 and PC2) explained 62.58% and 25.16% of the total variance, respectively. Two homogenous groups, each representing one species of Elmomorphus, were rather well separated along the fi rst ordination axis (Fig. 3A). The pronotal length and width strongly correlated with PC1, while the length of meso-and metatibia with PC2 (Table 3). Thus, specimens with smaller bodies and longer tibiae are clustered in the upper left part of the ordination diagram, while individuals with larger bodies and shorter tibiae are placed in the lower right corner. These positions correspond well with raw measurements according to which the average body size of E. amamiensis males is smaller, while the tibiae of the males of E. brevicornis are on average longer. Although the average values are comparatively different, the character ranges overlap in the two species. Individual characters thus do not allow unambiguous identifi cation. However, the combination of eight morphometric features can differentiate males of the two species without any overlap in the phenotypic space.
In females, PC1 and PC2 explained 77.92% and 8.93% of the total variance, respectively. Females of the two species of Elmomorphus were only vaguely separated along the fi rst ordination axis, as two specimens of E. brevicornis were placed within the rather compact cluster of E. amamiensis individuals (Fig. 3B). The pronotal length and the elytral width strongly correlated with PC1, while the length of the protibia and the interocular distance correlated with PC2 ( females showed a size gradient with the smallest specimens located in the left part of the ordination diagram, where they were mixed with E. amamiensis females. The present multivariate analyses show that females of the two species cannot be unambiguously differentiated even when a combination of eight morphometric characters is considered.

Note
The original description was based on two specimens collected by G. Lewis. Both specimens are still preserved in the collection of David Sharp in the NHMUK. The female specimen is glued on a larger rectangular card with the following handwritten text: "Elmomorphus brevicornis Type D.S. Jokio. Japan Lewis.". This style of labelling and label text is typical of David Sharp's type material. The male specimen and the dissected aedeagus are glued on one card, while its detached appendages are glued on a second card, pinned below the specimen. The two syntypes are conspecifi c. However, while the type locality mentioned in the original description ("Kobé") is written on the original label of the male specimen (Fig. 4C), the name "Jokio" [Sharp's spelling of Tokyo] is found on the label of the female specimen (Fig. 4A). Since no other species is known from Honshu and since taxonomic confusion is improbable, a lectotype designation is considered unnecessary (see ICZN 1999: Recommendation 74G, https://www.iczn.org/the-code/declaration-44-amendment-of-article-74-7-3/).
In the abstract of their most remarkable article, Devi et al. (2016: 371) wrote that "A lectotype is designated for this species [Elmomorphus brevicornis]". However, the term "lectotype" is not used anywhere else in their article, and it does not contain suffi cient information to ensure recognition of the specimen designated. Therefore, the requirements of the International Code of Zoological Nomenclature (ICZN 1999: Art. 74.7) are not fulfi lled.

Diagnosis
Elmomorphus brevicornis is a medium-sized (TL 2.59-3.51 mm), elongated oval, dorsally moderately convex species, with the highest point at the anterior third of the elytral length (lateral view). Body outline discontinuous between pronotum and elytra; yellowish plastron microscales cover the entire surface of the cranium, lateral portions of the pronotum and the entire dorsal surface of the elytra, so most of the dorsal surface appears to be matt (Fig. 5A). Each elytron has nine rows of large round punctures arranged in shallow striae. Legs conspicuously long and robust, only moderately shorter than the combined length of pronotum and elytra.
Elmomorphus brevicornis closely resembles E. amamiensis in external morphology, and it can be distinguished by the following characteristics: (1) elytral sides in E. brevicornis subparallel in the anterior half, while in E. amamiensis they are arcuate and more convergent (Fig. 5B), thus the outline appears to be generally broader and less oval in E. brevicornis; (2) sides of pronotum usually more rounded in E. brevicornis, while they are almost straight in E. amamiensis; (3) males of E. brevicornis with several longer setae in a transverse row on the labrum, and with similar setae in two pairs of clusters on each prosternal process and on the admedian portions of metaventrite, row and clusters of setae rather inconspicuous (Fig. 6A, C, E); similar, longer setae also near the apex of the fi fth ventrite (Fig. 7D); in the male of E. amamiensis the setal row and all setal clusters are well developed, conspicuous (Figs 6B, D, F, 7E); (4) parameres weakly curved ventrad in E. brevicornis (Fig. 8B), but strongly curved in E. amamiensis (Fig. 8D); (5) bursa copulatrix of the two species distally with one cluster of microspines on each side (Fig. 9D), while in E. amamiensis, besides these microspines there are several additional larger sclerites (Fig. 9E). Except for differences in the morphological characters, the two species differ by 22.7-23.9% in the partial mtDNA encoding for cytochrome c oxidase subunit I.

Redescription
BODY. Elongated oval, widest behind mid-length of elytra, moderately convex dorsally, with the highest point at anterior third of elytral length (lateral view). Colouration black, except reddish-brown mouthparts, antennal clubs, tarsi, claws, trochanters, and ventral portions of femora. Tibiae and remainder of femoral surface dark brown to black. Dorsal pubescence consists mainly of short, thin decumbent yellowish setae arising from small punctures. Plastron microscales (Fig. 7B) on the entire cranium, on pronotum in two lateral bands, covering ca 0.2 of pronotal width on each side, and on the entire surface of elytra. Ventral surface with dense, thin plastron hair-like setae (Fig. 7C) except for prosternal process and median part of metaventrite.
THORAX. Pronotum transverse, widest at base, strongly convex, PW/PL: ♂♂ 1.60-1.85 (1.72 ± 0.07, n = 14), ♀♀ 1.60-1.76 (1.68 ± 0.05, n = 7); plastron in lateral band on each side along entire pronotal length, mesally reaching level of third elytral row; disc smooth, shiny, with large round setiferous punctures, punctures nearly as large as facets, separated by 0.50-1.00× of a facet diameter; punctures on plastron area smaller, similar to those on head. Anterior angles strongly defl exed, protruding and acute, third as long as interocular distance; pronotal sides convergent anteriad, distinctly arcuate near middle, moderately explanate along entire length. Hypomeron widest behind mid-length. Prosternal process wide and short, lateral margins divergent and moderately arcuate; posterior margin widely rounded; lateral portion raised, wide and in male with group of longer setae (Fig. 6C) forming a hardly discernible setal cluster (these setae only moderately longer than other hair-like ventral setae and often abraded); median keel moderately convex, ca ⅓ as wide as width of prosternal process. Scutellum longer than wide, smooth, with small, sparse punctures. Metaventral process with lateral sides strongly raised, anterior margin not raised. Metaventral disc fl at in female, fi nely depressed in male; longitudinal step-like elevation along sides delimits lateral area with plastron from smooth central one, surface irregularly and moderately sparsely punctate and in some specimens fi nely irregularly wrinkled; discrimen well impressed, distinct; a small, hardly discernible setal cluster present in front of metakatepisternal suture in males on each side of metaventral disc (Fig. 6E). Elytra oblong, widest behind middle; moderately convex dorsally, in dorsal view with the highest point at anterior third; sides subparallel in anterior half, more strongly arcuate in posterior half; EL/EW: ♂♂ 1.53-1.69 (1.57 ± 0.04, n = 14), ♀♀ 1.53-1.66 (1.61 ± 0.04, n = 7); surface entirely covered by plastron scales and with numerous minute punctures distinctly smaller than a facet diameter, punctures separated by about 0.50-1.00× a facet diameter. Strial punctures large and deep on elytral disc and subequal in size with facet size, laterad and posteriad progressively smaller. All tibiae straight, to slightly bent near mid-length; protibia ca 1.25× as long as protarsus; PrTL/PL: ♂♂ 1.01-1.20 (1.10 ± 0.05, n = 14), ♀♀ 1.00-1.08 (1.05 ± 0.03, n = 7). Terminal tarsomere in male foreleg ca 1.20× as long as all preceding tarsomeres combined; male foreclaws strongly curved, not widened or thickened, about half of terminal tarsomere length, both similar to each other (Fig. 7A) and similar to female foreclaws.

Sexual dimorphism
Females are, on average longer and broader than males. For an unambiguous distinction between males and females, the apex of the fi fth ventrite should be examined: it is distinctly excised in males (Fig. 7D), while it is rounded, with a short, smooth keel in females. The presence of several longer semi-erect setae on the labrum, prosternal process, metaventral disc and on the apex of the fi fth ventrite in males is hard to observe even in well-preserved specimens, and even under a high-quality microscope at a magnifi cation of about 100×.

Distribution
Japan: Honshu, Izu-Ôshima, Oki, Shikoku, Kyushu, Tsushima, and Tanegashima (Nakajima et al. 2020); South Korea (Fig. 10A). Devi et al. (2016) erroneously recorded E. brevicornis from India. Their record was published in the unreviewed "Journal of Entomology and Zoology Studies", a so-called predatory journal (see https://predatoryjournals.com/journals/). The "redescription" provided in this article does not allow any conclusions to be drawn on the identity of the specimens. The "Aedagues" [sic] depicted in their fi g. 1c is, in fact, an ovipositor! Judging from their habitus photograph (fi g. 1a), these specimens, beyond any doubt, do not belong to E. brevicornis or any other species similar to E. brevicornis.

Diagnosis
Elmomorphus amamiensis is medium-sized (TL 2.80-3.22 mm), elongated oval, dorsally moderately convex, with the highest point slightly before the middle of the elytral length (lateral view); body outline slightly discontinuous between pronotum and elytra; plastron microscales covering the entire surface of the cranium, lateral portions of the pronotum and the whole dorsal surface of the elytra (Fig. 5B). Elytron with nine rows of moderately large, round punctures in shallow striae. Legs long and robust, somewhat shorter than the combined length of pronotum and elytra, claws moderately curved. Differences to the morphologically most similar E. brevicornis are discussed above.

Redescription
BODY. Elongated oval, widest behind mid-length of elytra, moderately convex dorsally. Colouration black; mouthparts, antennae, tibiae and tarsi reddish-brown; femora dark brown. Main pubescence consisting of very short and thin, decumbent yellowish setae, which arise from small punctures. Dorsal plastron present on the entire surface except for the pronotal disc. Ventral surface except for prosternal process and disc of metaventrite with hair-like plastron setae.

Sexual dimorphism
Females are on average larger than males, and the apex of their fi fth ventrite is rounded, while in males it is triangularly excised. Moreover, males have long conspicuous setae arranged in a transverse row on the labrum and in clusters situated on lateral sides of the prosternal process, lateral sides of the metaventrite and on the apex of the fi fth ventrite; females lack these groups of long setae.

Key to the adults of the species of Elmomorphus from Japan and Korea
1. Male labrum with several long, rather widely spaced setae arranged in a more or less distinct transverse row (Fig. 6A); similar, moderately longer setae arranged in small clusters on prosternal process and metaventrite (Fig. 6C, E), and on apex of ventrite 5 (Fig. 7D); long setae on labrum maximally half as long as interocular distance, most elongate ventral setae moderately shorter than those on labrum. Parameres in apical half slightly curved ventrad, nearly straight (Fig. 8B). Female: bursa copulatrix with a cluster of minute spines on each side in distal portions (Fig. 9D) 6B); similar setae in extensive clusters on prosternal process, metaventrite (Fig. 6D, F), and on apex of ventrite 5 (Fig. 7E); most elongate setae on labrum about ⅔ of interocular distance, longest ventral setae moderately longer than interocular distance. Parameres in apical half strongly curved (Fig. 8D). Female: bursa copulatrix with cluster of minute spines and with several larger sclerites scattered over its proximal portion (Fig. 9E) Nomura, 1959 Discussion Elmomorphus brevicornis has been recorded on six Japanese islands and has been quite recently, discovered in South Korea (Jung & Bae 2014;Nakajima et al. 2020). Interestingly, an extensive collection of Elmomorphus specimens sampled during the China Water Beetle Survey from several hundred Chinese localities, including Taiwan, did not contain any specimens of E. brevicornis. Similarly, there have been no records from the Russian Far East, Vietnam, Laos, or Thailand. Therefore, E. brevicornis was for a long time considered to be an endemic species of Japan (Kodada & Jäch 2006).  Nomura, 1959.
European Journal of Taxonomy 758: 97-121 (2021) 118 The subspecies E. brevicornis amamiensis was established based on a female holotype collected on Amami-Ôshima, one of the largest islands of the Ryukyu Archipelago (Nomura 1959).
Jung & Bae (2014) examined specimens of E. brevicornis from Korea and E. brevicornis amamiensis from Amami Island. Using differences in the structure of the male genitalia, they raised the subspecies amamiensis to species level. In their identifi cation key, E. amamiensis was characterized by the long mesofemur, tarsi and claws small, and parameres strongly curved apically; while E. brevicornis possesses a shorter mesofemur, large tarsi and claws, and parameres, which are not strongly curved apically (Jung & Bae 2014: 7).
The present principal component analysis classifi ed males of the two species of Elmomorphus in two homogenous groups that were separated along the fi rst ordination axis given the eight morphometric characters studied. The pronotal length and width are strongly correlated with PC1, while the length of meso-and metatibia are correlated with PC2. This corresponds well to raw measurements according to which the average body size of E. amamiensis males is smaller, while the average tibial length of E. brevicornis males is higher than in E. amamiensis. Although the average values are comparatively different, the ranges of individual characters overlap in the two species. Individual characters thus do not allow unambiguous species identifi cation. However, the combination of eight morphometric features can differentiate males of the two species without any overlap in the phenotypic space. Females of the two species were only vaguely separated along the fi rst ordination axis, as two specimens of E. brevicornis were placed within the relatively compact cluster of E. amamiensis individuals. The present multivariate analyses show that females of the two species cannot be unambiguously differentiated even when a combination of eight morphometric characters is considered. Nevertheless, we cannot exclude that inclusion of a signifi cantly larger number of individuals from a wider distributional area into multivariate analyses might bring moderately different results. So far, the best characters for species identifi cation of males are the conspicuous long hair-like setae on the labrum, prosternal process, metaventrite, and on the fi fth ventrite in combination with the specifi c form of the parameres. Females can be distinguished by the different patterns of various sclerites on the bursa copulatrix.
Contrary to the morphological similarity between the two species, the present barcoding analyses show a profound interspecifi c divergence of E. amamiensis and E. brevicornis and confi rm their species status as proposed by Jung & Bae (2014). In addition, the haplotype network analyses supported the monophyletic origin of E. amamiensis and E. brevicornis by 55 shared nucleotide positions within the barcoding COI sequence fragment, although the two species differ in 109 nucleotide positions. This deep interspecifi c divergence is in congruence with the different geographic distribution of the two species. Elmomorphus amamiensis is confi ned to the Ryukyu Archipelago and has so far not been recorded north of the Watase Line (Tokara gap), a biogeographical barrier between the Holarctic and the Oriental regions, nor has it been found south of the Miyako Strait (Kerama gap) south of Okinawa. Thus, E. amamiensis represents one of the numerous endemic species of the Ryukyu Archipelago. On the other hand, E. brevicornis occurs north of the Tokara gap on several Japanese islands and in South Korea. The deep divergences within the COI sequences of the two Elmomorphus species suggest their different evolutionary origin rather than their sister-group relationship.