The moths were collected at night using LepiLED lamps emitting ultraviolet radiation (Brehm 2017) (Fig. 22). Specimens were collected singly in vials and killed with ammonia dripped on rolled paper sticks. The moths were initially mounted with wings half-spread for transportation and later spread to the standard position. The specimens were photographed using a Canon EOS 7D camera attached to a Canon EF 100mm f/2.8L Macro lens. Seven specimens of the new species had their genitalia dissected. Dissections followed Robinson (1976) and were made using Olympus and Leica stereomicroscopes. The genitalia preparations were photographed using the following setup: StackShot Macro rail from Cognisys, Nikon Z8 camera, Nikon AF Micro-Nikkor 200mm f/4D lens with a Nikon LU Plan Fluor 5× WD 23.5 microscope objective mounted in front. 100–150 images of each slide were stacked using Zerene Stacker Version 1.04. The final composite images were edited using Adobe Photoshop 26.3.0, Adobe Lightroom Classic 15.2 and Topaz Sharpen AI 4.1.0. A map was constructed in QGIS Desktop 3.28.5 (Fig. 24).

The genitalia preparations were compared with others made from the closely related species Rhodophaea formosa from different parts of Europe. Comparisons were also made with illustrations of R. exotica (Inoue, 1959) given by Inoue (1959), Paek and Bae (2001), Streltzov (2012) and Bae et al. (2017).

Tissue samples (a single hindleg) from two of the specimens were prepared according to the prescribed standards. These were processed successfully at the Canadian Centre for DNA Barcoding (CCDB, Biodiversity Institute of Ontario, University of Guelph) targeting the 658 base pair segment of the CO1 gene, which is used for DNA barcoding of animals (Hebert et al. 2003).

We used the Distance Summary tool provided in the BOLD Systems Workbench to calculate distances using the Kimura 2-parameter model (Kimura 1980) for substitution and MUSCLE for alignment. Only sequences ≥600 base-pairs in length were included and searches were filtered by Barcode Index Numbers (BINs) (Ratnasingham and Hebert 2013) to avoid possible misidentifications and contaminations.

To assess the taxonomic status of Rhodophaea madaricola sp. nov. and its close relatives in other genera from a molecular perspective, we conducted molecular analyses of a larger number of the publicly available barcodes in the BOLD systems database (Ratnasingham and Hebert 2007), as of April 2026. Suspected allied genera were selected by comparison of one full-length sequence (658 base-pairs) from the BINBOLD:ACC8900 of R. formosa (Process ID: CGUKB170-09) and one full-length sequence of R. madaricola sp. nov. (Process ID: PYRCR038-11) with available sequences in GenBank (Sayers et al. 2025), using Nucleotide BLAST searches (Altschul et al. 1990). All genera represented within ≥94.22 % identity to either of the two queried sequences as of April 2026 were included (this exact distance was chosen as it included all the 100 sequences in the BLAST search of the R. madaricola sp. nov. sequence). All sequences of ≥500 base-pairs in length representing these genera were downloaded from BOLD systems (none directly from GenBank), excluding records flagged as suspected misidentifications, contaminants or containing stop codons. The resulting sequence list included the following genera, in addition to Rhodophaea Guenée, 1845, with the number of BINs and number of assigned species names, respectively, given in parentheses: Ambesa Grote, 1880 (5, 3), Asalebria Amsel, 1953 (6, 4), Canarsia Hulst, 1890 (4, 1), Meroptera Grote, 1882 (4, 5), Ortholepis Ragonot, 1887 (5, 6), Psorosa Zeller, 1846 (5, 4), Pyla Grote, 1882 (16, 17), Quasisalebria Heinrich, 1956 (5, 3), Salebriaria Heinrich, 1956 (25, 22), Sciota Hulst, 1888 (34, 28) and Selagia Hübner, [1825] (4, 3). Including the two barcodes obtained from our collected specimens, this amounted to 1404 sequences in total.

A maximum likelihood (ML) tree and a Neighbour-joining (NJ) tree of the sequences were constructed using this dataset (Fig. 21). Multiple sequence alignment (MSA), testing of substitution models and the NJ analysis were performed in MEGA12 (Kumar et al. 2024). The ML analysis was done with RAxML-NG v. 1.2.2 (Kozlov et al. 2019). The sequences were trimmed down to a maximum length of 658 base-pairs and aligned with MUSCLE (Edgar 2004), after which the alignment was subjected to a brief manual inspection. For ML, the alignment was partitioned by all three codon positions, for which models were tested individually. The models selected were as follows: 1^st positions = Tamura and Nei (1993) (TN93) with gamma distributed (+ G) and invariant (+I) sites; 2^nd positions = Hasegawa et al. (1985) (HKY) + G; 3^rd positions = General Time Reversible Model (GTR) (Tavaré 1986), +G, +I. Four discrete gamma categories were set for all three partitions. For NJ, the Kimura 2-parameter model was used for all positions, with gamma distribution; the gamma parameter was set to 0.61 given by a separate model test of the full alignment. The models were selected based on the lowest Bayesian Information Criterion (BIC) score in the MEGA12 model test for ML, except the 1^st positions ML partition, for which the model with lowest BIC score was not available for analysis with RAxML-NG – the model with the lowest corrected Akaike Information Criterion (AICc) score was instead chosen for this partition. Perfect duplicate sequences were removed from the dataset with RAxML, which resulted in a final MSA of 760 sequences. This MSA was used for both trees. For ML, the tree with superior likelihood values among 10 randomly generated Maximum Parsimony trees with RAxML was used as the starting tree for the heuristic search. 1000 bootstrap replications were performed for both trees.

A strict consensus dendrogram of the ML and NJ trees was made for a subset of the taxa using Adobe Illustrator 29.3, including duplicate sequences earlier removed from the MSA (added manually) (Fig. 21). The full ML tree with the duplicate sequences re-added can be found in Suppl. material 1.

KBE Collection of Kai Berggren, Kristiansand, Norway;

NHMO Natural History Museum, University of Oslo, Norway;

RVO Collection of Reidar J.D.I. Voith, Bodø, Norway;

TLMF Tiroler Landesmuseum Ferdinandeum, Innsbruck, Austria;

ZMUC Zoological Museum, Natural History Museum of Denmark, Copenhagen, Denmark;

ZSM Zoologische Staatssammlung München, Germany.

The Eastern Palaearctic species have largely remained in the traditionally larger genus Nephopterix Hübner, [1825] (sometimes spelt Nephopteryx, an unjustified emendation by Zeller (1839)), even though most of the Western Palaearctic and all the North American Nephopterix have been transferred to Sciota, Pempelia, Elegia Ragonot, 1887 and other genera after it became clear that Phycis rhenella Zincken, 1818 (now Sciota rhenella) had been erroneously interpreted as the type for Nephopterix (Roesler 1988; P. Leraut 2001, 2002; Neunzig 2003). Partial revisions and synonymisations (e.g. Yamanaka 1994, 2004; Streltzov 2012; Slamka 2019; Trofimova et al. 2020) have resulted in a heterogenous circumscription for Nephopterix, especially regarding the Eastern Palaearctic species which have little in common with the type species N. angustella (Hübner, 1796) (see e.g. Nephopterix sensu Bae et al. (2017)). Most of the species currently in Nephopterix have a Palaeotropical or Australian distribution (G. Leraut 2021); areas where Microlepidoptera are typically not well studied compared with Europe and North America, thus updated revisions may lag behind. This seems to be the case for N. shantungella, apparently lacking taxonomic treatment in relevant works (e.g. Qi et al. (2014). In the original description of N. shantungella, Roesler (1969) erected the subgenus Clasperopsis, which included N. shantungella and N. nigerrimella (Caradja, 1916). The latter was synonymized with Sciota fumella (Eversmann, 1844) by Yamanaka (1994) (as Nephopterix (Clasperopsis) fumella). P. Leraut (2002) synonymised it with Catastia umbrosella Erschoff, 1877 based on the female genitalia and transferred it to Sciota. Hence, regardless of which synonymisation is correct, N. nigerrimella was transferred to Sciota, and it stands to reason that N. shantungella should also have been transferred along with its congener, hence the following new combination is proposed: Sciota shantungella (Roesler, 1969), comb. nov. This is substantiated by the illustration of the female genitalia given by Roesler (1969) which shows clear affinity with Palaearctic members of Sciota (S. adelphella (Fischer von Röslerstamm, 1836, S. hostilis (Stephens, 1834) etc.) and bears little resemblance to N. angustella. Furthermore, Roesler (1969) mentioned that it is externally similar to S. hostilis. In any case, the genus Nephopterix should no longer be associated with Sciota or related genera, which is corroborated by molecular data. Despite being represented by several trustworthy barcoding sequences, the type species N. angustella did not show up in our BLAST searches, nor did any other Nephopterix species except N. shantungella. The threshold of 94.22 % similarity is admittedly arbitrary and CO1 alone is probably insufficient in determining genera. However, we also calculated the actual distance between N. angustella and the Nearctic type species of Sciota, the Palaearctic representative Sciota fumella (chosen as a probable proxy for Sciota shantungella comb. nov.) and European Rhodophaea (Table 1). The distances, both minimum and mean, between N. angustella and all of these is almost twice their internal distances, the latter being very similar.

We consider the BINBOLD:AAG9106 to represent the type species of Sciota: S. rubrisparsella (Ragonot, 1887). The BIN only contained one sequence under that name at the time of writing; most of the sequences are named S. celtidella (Hulst, 1890), a species closely related to S. rubrisparsella. However, all the photographs provided in BOLD of S. celtidella imagines in BINAAG9106 show forewing characteristics that better align with S. rubrisparsella as described by Neunzig (2003) and Heinrich (1956) as well as the photographs of S. rubrisparsella (without sequences) in BOLD. Furthermore, they appear distinct from all the photographs of S. celtidella in BINBOLD:AAA7237 (which better align with S. celtidella) in BOLD. These two BINs formed distinct, relatively distant, clusters in the ML analysis (Suppl. material 1).

A full revision of any Phycitini genus is not possible with the taxon selection and the genetic data (CO1) used in the ML analysis, nor is it within the scope of this study or intent of the analysis. The resolving power of CO1 for deeper relationships is low even for Phycitini genera, which is evident from the very low bootstrap values of the older nodes in our analysis. The genetic signal, however, was strong enough to unite the four or five clusters (the Finnish specimen with BINBOLD:AFS7949 is very close to the main clade of Rhodophaea formosa) of the Rhodophaea group, with relatively high bootstrap support. Regarding the true origin of the Chinese sequences, it is doubtful that their identifications are trustworthy. These sequences were mined from Genbank to BOLD systems and were not intended for expanding the library of reference sequences for Lepidoptera in BOLD. Most of them originate from the large dataset of Hao et al. (2020), with no information as to how specific entries were identified. The dataset evidently contains some contaminations, misidentifications or a mixture of both, as outlined in the Results section. The Phycitini species across multiple genera are notoriously similar; Sciota fumella being a prime example with very subtle colouration and markings on the forewings, which could easily be confused with worn representatives of other Phycitini species, hence an unsurprising candidate of misidentification. The two Chinese clusters within the Rhodophaea group might represent undescribed species of Rhodophaea or described species such as R. exotica (or even species of other genera). We find it unlikely that any of them represent the true Sciota shantungella comb. nov., as we would expect this species to cluster closer to (the true) S. fumella and similar Palaearctic species of Sciota, such as S. adelphella and S. hostilis, given the morphological similarities within these species, particularly in the female genitalia. However, the three conspicuous cornuti of the phallus (Roesler 1969, Fig. 5) would be unique within Sciota and could conceivably be evidence of affinity with Rhodophaea. Regardless of the true identity of these sequences, our analysis strongly suggests that all of them should be included in Rhodophaea. Given that none of them represents R. formosa sensu Streltzov (2012), there should be more than two species of Rhodophaea in the Eastern Palaearctic fauna. All in all, the reinstatement of the genus Rhodophaea (P. Leraut 2001) without consideration of the related and morphologically diverse genus Sciota appears inconsistent. We believe Sciota and Rhodophaea should probably be united into one genus. Due to priority this would make Rhodophaea the valid name for the genus and cause the transfer of numerous Sciota species to Rhodophaea. However, we believe that such an action needs more comprehensive data, especially from the Eastern Palaearctic.

Rhodophaea has been represented in the Western Palaearctic by a single, widespread species, R. formosa. Its distribution spans from Iberia and the British Isles in the west to Scandinavia in the north, extending eastward to the Ural Mountains and in the southeast to Armenia. Notably, R. formosa is absent from mainland Greece, with the nearest records from Italy, Bulgaria and Turkey. All specimens of R. madaricola sp. nov. were collected in the vicinity of the Omalos Plateau and the species has not been recorded elsewhere. This isolated occurrence from the continental R. formosa indicates that R. madaricola sp. nov. is an island endemic. Despite the differences revealed by DNA barcoding and dissection of the genitalia, R. madaricola sp. nov. is superficially similar to R. formosa. R. formosa could easily be regarded as a “supposedly unmistakable” species (Huemer 2025). This highlights the limitations of identifying Lepidoptera based solely on external characters. It was the molecular data that provided the primary impetus for a detailed morphological investigation. As highlighted by Huemer et al. (2025) nearly 10% of the Cretan Lepidoptera fauna currently lacks a clear taxonomic assignment, proving gaps in the knowledge of the island’s fauna. Much of Crete’s rare flora and fauna is, according to current knowledge, restricted to specific, inaccessible gorges and alpine areas. This includes several Lepidoptera species from the same primary habitat: Cryphia omalosi Fibiger & Svendsen, 1998 (Noctuidae), Herrichia vilhelmseniella Karsholt & Lvovsky, 2018 (Oecophoridae), Peribatodes omalensis Rajaei, Skou & Hausmann, 2019 (Geometridae), Teleiodes albiluculella Huemer & Karsholt, 2001 (Gelechiidae) and Pyralis papaleonei Huemer, Kaila & Segerer, 2026 (Pyralidae). It is likely that additional undescribed species exhibit similar micro-endemic distributions. Additional surveys are needed to further document the unique entomological diversity of the island.