﻿Taxonomic note on the species status of Epiophlebiadiana (Insecta, Odonata, Epiophlebiidae), including remarks on biogeography and possible species distribution

﻿Abstract The species included in the genus Epiophlebia Calvert, 1903 represent an exception within Recent lineages – they do not belong to either dragonflies (Anisoptera) nor damselflies (Zygoptera). Nowadays, the genus is solely known from the Asian continent. Due to their stenoecious lifestyle, representatives of Epiophlebia are found in often very small relict populations in Nepal, Bhutan, India, Vietnam, China, North Korea, and Japan. We here present a taxonomic re-evaluation on the species status of Epiophlebiadiana Carle, 2012, known from the Sichuan province in China, supplemented with a morphological character mapping on a genetic tree to highlight synapomorphies of E.diana and E.laidlawi Tillyard, 1921. We conclude that E.diana is a junior synonym of E.laidlawi. Furthermore, we discuss the Recent distribution of the group, allowing for predictions of new habitats of representatives of this group.

Adult Epiophlebia are very conspicuous (Fig. 1), and in the field they can easily be identified by the black-yellow striped coloration (Asahina 1954) and their characteristic slow and rather uncoordinated appearing undulating flight (Rüppell and Hilfert 1993). Morphologically, the anisopterous body shape, the zygopterous shape of the wings, and the convex frons are some of the main distinguishing characteristics (Asahina 1954;Büsse 2016). The larvae of Epiophlebia also resemble dragonflies, as they use a rectal chamber for respiration, but jet propulsion, which is typical for Anisoptera (Corbet 1999), has never been observed (Tabaru 1984). Their morphological distinction is rather subtle, so they are easily mixed up with, for example, gomphids or petalurids (Asahina 1954) -as happened to Epiophlebia diana Carle, 2012. The type specimens of E. diana were collected by "Dr. David C. Graham in the mountainous regions of western Szechuan" (Needham 1930). Needham however, misidentified the larvae of Epiophlebia as Gomphidae (Carle 2012).
While the ancestors of present Epiophlebia species were at their peak in the Mesozoic era and were possibly distributed over large areas on the pre-Asian continent (Carpenter 1992;Nel et al.1993), recent species have restricted ranges, often confined to small areas in Asia: Epiophlebia superstes Selys, 1889 in Japan; Epiophlebia laidlawi Tillyard, 1921 in Nepal, India, Bhutan, and Vietnam;Epiophlebia sinensis Li & Nel, 2012 in North Korea and China, and Epiophlebia diana also in China (Asahina 1954;1963;Tani and Miyatake 1979;Büsse et al. 2012;Carle 2012;Li et al. 2012;Büsse 2016) showing a characteristic disjunct distribution (Büsse et al. 2012).
Since the habitat requirements of the genus Epiophlebia seem to be very specific, the range of Recent habitats is extremely restricted. Epiophlebia species prefer cold mountain streams with temperatures of about 4 to 5 °C in winter and about 16-17 °C in summer (data published for E. superstes by Tabaru (1984)) and altitudes between 1,300 to approximately 3,000 m (data published for E. laidlawi by Brockhaus and Hartmann (2009)). This stenoecious lifestyle has restricted the genus Epiophlebia to cold habitats, like glacial refuges (De Lattin 1967;Büsse et al. 2012).
For recently diverged species, or for taxa that are described under the assumption of incipient speciation, it can be challenging to develop a morphological character set that reveals the true pattern of evolutionary history for a taxon. Further complicating matters is that there are often separate, not cross-referenced, descriptions of adults and larvae for Odonata. In the case of the genus Epiophlebia, adults and larvae are described for E. superstes and E. laidlawi, while for E. sinensis only the adults and for E. diana only the larva is known. The species status of E. diana has already been critically discussed and is doubtful (cf. Dijkstra et al. 2013;Büsse 2016). We, therefore, present a taxonomic reevaluation of the species status of E. diana. Unfortunately, the type specimen is untraceable and seems lost (F.L. Carle, author of E. diana as well as J.J. Dombroskie of the Cornell University Insect Collection, New York, USA: personal communication, see Büsse 2016). However, a combination of morphology, phylogeny, and biogeography, described here, lays a solid basis for the designation of E. diana as junior synonym of E. laidlawi.

Materials and methods
Here, we examined morphological data from Büsse (2016) and several other publications (i.e., Asahina 1954Asahina , 1961Büsse et al. 2012;Carle 2012;Li et al. 2012;Dorji 2015). A matrix composed of all characters used in the past to evaluate Epiophlebia larval and adult characters is shown in Table 1; briefly this includes larval traits and adult characters related to size and colouration, features of the head, abdomen, genitalia, and appendages. The phylogeny that was used here was based on Büsse et al. (2012; fig. 2). The main justification for using Büsse's and colleagues (2012) phylogeny is that presently there are few overlapping sequence fragments across the species of Epiophlebia. Here, we examined all available mitochondrial and nuclear gene fragments for the genus Epiophlebia from GenBank to evaluate what available data was present for the genus (Table 2). Unfortunately, to date E. superstes is the only species for which sequence data are available for a broad sampling of genes. No genetic data is available for E. diana, presumably because the describing author has misplaced the existing specimens (F.L. Carle personal communication). With such a dataset, we decided to use the phylogeny provided by Büsse et al. (2012) which is the most comprehensively sampled phylogeny for the genus currently available. In terms of character mapping, briefly, characters were traced in Mesquite (Maddison and Maddison 2016) using both the ancestral state reconstruction parsimony and likelihood functions. Consistency index values for a matrix including all traits in Table 1 were evaluated in Mesquite against a tree assuming (E. sinensis (E. superstes (E. diana, E. laidlawi))) and found to be 1.0.
For photography, we used specimens of E. superstes (because of availability) to depict the general habitus of the very similar Epiophlebia species. For stacked photography, a custom-made 3D-printed illumination dome system (Bäumler et al. 2020) and an Olympus OMD 10mkII digital camera (Olympus K.K., Tokyo, Japan), equipped with a Leica 45 mm macro lens (Leica Camera AG, Wetzlar, Germany) was used. All images were subsequently processed in Affinity Photo and Affinity Designer (Serif Ltd, Nottingham, United Kingdom).

Results and discussion
Taxonomy A comparison of the morphological characters used in past studies to the currently accepted phylogeny of the genus Epiophlebia suggests that several characters are not useful for reconstruction of the evolutionary history, as they are only known for adults of all species except E. diana, or only known for larvae of all species except E. sinensis. Using parsimony, we found 10 characters supporting a clade comprising E. laidlawi and E. diana (Fig. 2), but as five of those characters are based on adult traits, there are missing data for E. diana. Furthermore, for several distinguishing characters employed by Carle  (Büsse 2016), (Carle 2012), (Dorji 2015) 0 1 1 ? 2 Scape and pedicle: 0 = scape and pedicle same length as flagellomere or shorter 1 = scape and pedicle always longer than first flagellomere ( (2012), the reported characters of E. diana fall within the trait range reported for E. laidlawi, while some characters even seemed to be poorly scored by Carle (2012). For example, he described the abdominal stridulatory files (ASF) in the genus Epiophlebia. He mentioned for E. superstes that the ASF of segment 3 is well developed, and the ASF segment 4 is about as high as long, and the ASF segment 7 is vestigial. In the data of Büsse (2016), specimens of E. superstes can be found with almost no stridulatory file on segment 3, and the ASF of segment 4 all can be seen higher as long, shorter as long, and as long as high. Furthermore, Carle (2012) suggested for the distinction of E. diana differences in the ASF (for E. diana he listed ASF of segment 7 c. 3/4 length of segment and for E. laidlawi the ASF of segment 7 c. 1/2 length of segment), but these are not valid as there are E. laidlawi in the data showing c. 3/4 as well (Büsse 2016). Next, Carle (2012) listed distinctions between the two species based on the degree of sinuity in the premental margins (in E. laidlawi, prementum with lateral margins slightly sinuous, but in E. diana, prementum with lateral margins strongly sinuous). It is difficult to estimate what slightly and strongly means, as such wording is subjective in nature; other points of view may consider the lateral margins of the prementum in E. laidlawi to be not just slightly sinuous, and without a figure showing data from Carle, it is impossible to say whether the sinuous nature of the prementum in E. diana is more pronounced; this character is not diagnostic. Similarly, Carle (2012) listed the fore-femur as being c. 3.0× as long as wide in E. laidlawi and for E. diana 2.5× as long as wide; this character is not valid for distinction between these species as there is variation in this trait and E. laidlawi have been documented with fore-femur that are only 2× as long as wide and there are E. superstes in the data showing a fore-femur c. 3× as long as wide (Büsse 2016). In fact, Asahina (1961) noted as the distinguishing character of E. laidlawi and E. superstes that the fore-femur of E. laidlawi was longer. It seems this character is very variable and impractical for taxonomic use. Indeed, fore-femur length has been shown to be influenced by ontogeny, and it is rarely used to infer evolutionary history.  Table 1 serves as the synapomorphy. Lastly, Carle (2012) listed E. laidlawi with lateral abdominal lobes slightly protruding on segment 9 and E. diana with lateral abdominal lobes protruding on segment 9. Again, these are subjective descriptions, and Büsse's (2016) data show abdominal lobes slightly protruding on segment 9 in E. superstes and distinctively produdent, forming a distinctive overhanging protrusion at the end of the segment 9 compared to the preceding segments, in E. laidlawi, comparable to Carle's (2012) fig. 3D. Only younger larvae of E. laidlawi seem to have only slight protudents on segment 9; here the abdominal segments resemble each other comparable to Carle's (2012) fig. 2D. In short, although no specimens of E. diana are available to examine, the characters used by Carle (2012) to describe the species do not seem to show a bimodal distribution of character values between E. diana and E. laidlawi, and given known and documented phenotypic variation in these traits for E. laidlawi, we consider E. diana a synonym of E. laidlawi.

Biogeography
The described stenoecious lifestyle has restricted the genus of Epiophlebia to cold habitats, indicated by the recent distribution in glacial refuges (Fig. 3B;De Lattin 1967;Büsse et al. 2012) -E. superstes from the Japanese refuge, E. laidlawi from the Nepalese and the Yunnanian refuges, E. sinensis from the Manchurian refuge (more precisely Ussurian secondary centre), and 'E. diana' from the Sino-Tibetan refuge -thus, clearly showing a separation in a western and eastern habitat zone (Fig. 3C). Due to the distribution in the mentioned glacial refuges, we predict that other Epiophlebia habitats may exist in the Sino-Pacific refuge, the Sindhisian refuge, the Mongolian refuge, and further populations in the Manchurian refuge because it is divided into secondary refuges, as well as the Kamtchatian refuge. Whether one can expect new species of Epiophlebia or new populations of a known species in these possible habitats is to be answered.
Indeed, the connection between Japan and the Asian mainland, as well as regions of the Himalayas and other parts of Asia, has been well documented by Ikeda and Ohba (1998) and is known as the Sino-Japanese floristic region during the last ice ages (Ikeda and Ohba 1998;Büsse et al. 2012). The question remains as to when the extant Epiophlebia species diverged, as two contradicting hypotheses are plausible: i) Epiophlebia dates back to the Jurassic when Pangaea broke apart (Brockhaus and Hartmann 2009), or ii) Epiophlebia diverged during to the last or second last ice age period (Büsse et al. 2012;Büsse 2016). To substantiate one of these biogeographic scenarios, a re-analysis is absolutely necessary.
Nowadays, the habitats of Epiophlebia species are widely separated. Japan is separated from the mainland by sea-straits with depths of approximately 55 m north of Hokkaido and 130 m between the southern island of Kyushu and Korea (Millien-Parra and Jaeger 1999). In addition to the ocean, there are approximately 3000 km (respectively more than 3600 km) of temperate lowlands separating Japan, inhabited by E. superstes and the known ranges of E. laidlawi and 'E. diana'. The same is true for the habitat of E. sinensis in Heilongjiang province, China , as it is more than 3000 km away from the cold mountain habitat of E. laidlawi in the Himalayas and separated by temperate lowlands. The eastern and western habitat zones are, thus, separated by unsuitable, temperate lowlands (Fig. 3C). The location where the synonymized 'Epiophlebia diana' was found in Sichuan province, China (Carle 2012), is also part of the western habitat zone, as the known range of E. laidlawi. The known distributions of 'E. diana' and E. laidlawi are around 1000 km apart but are connected by the mountain range of the Himalayas, which contains ample suitable habitats for an Epiophlebia species.