﻿A new species of scops-owl (Aves, Strigiformes, Strigidae, Otus) from Príncipe Island (Gulf of Guinea, Africa) and novel insights into the systematic affinities within Otus

﻿Abstract A new species of scops-owl (Aves, Strigiformes, Strigidae, Otus) is described from Príncipe Island, São Tomé and Príncipe (Gulf of Guinea, Africa). This species was discovered for science in 2016, although suspicions of its occurrence gained traction from 1998, and testimonies from local people suggesting its existence could be traced back to 1928. Morphometrics, plumage colour and pattern, vocalisations, and molecular evidence all support the species status of the scops-owl from Príncipe, which is described here as Otusbikegilasp. nov. Phylogenetic analyses suggest that this species descended from the first colonisation of the Gulf of Guinea islands, being sister to the clade including the mainland African Scops-Owl O.senegalensis, and the island endemics Sao Tome Scops-Owl O.hartlaubi and Pemba Scops-Owl O.pembaensis. The most diagnostic trait in the field is its unique call which, curiously, is most similar to a distantly related Otus species, the Sokoke Scops-Owl O.ireneae. The new species occurs at low elevations of the old-growth native forest of Príncipe, currently restricted to the south of the island but fully included within Príncipe Obô Natural Park. Otusbikegilasp. nov. takes the number of single-island endemic bird species of Príncipe to eight, further highlighting the unusually high level of bird endemism for an island of only 139 km2.


Introduction
Species are indeed the face of biodiversity with whom everyone relates to. The discovery of new species consistently makes headlines expressing wonder and joy. And yet, it has been estimated that only ca. 14% of extant species have been described, with invertebrates making most of the undescribed species (Mora et al. 2011). In this age of human-driven extinction (Ceballos et al. 2020), a major global effort should be undertaken to document what may soon not be anymore (Dijkstra 2016). Such new wave of exploration, carried out by professional and amateurs alike, would have the additional benefit of helping to revive a global interest in the natural world and the mysteries it holds. Only by rekindling this link can the current biodiversity crisis be reverted.
The discovery of new species tends to have a higher impact when it occurs in familiar groups like mammals or birds. Birds in particular are likely the best studied animal group, making the discovery of new species more challenging and often restricted to remote locations and/or difficult-to-study groups (e.g., Rheindt et al. 2020;Lane et al. 2021;Milá et al. 2021). This paper illustrates how exploration led to the discovery of a new owl species on the forests of Príncipe Island, Gulf of Guinea, Africa.
Owls (Aves, Strigiformes) are a charismatic bird group that made their way into most human cultures, where they are generally either symbols of wisdom or, on the contrary, omens of bad luck (Marcot and Johnson 2003). This is certainly linked to their nocturnal habits and associated elusiveness, their inquisitive look enhanced by their large eyes facing forward, and their calls heard through the night, which together help in creating an aura of mystery surrounding these species. In many aspects, this mystery has also permeated for a long time the scientific knowledge we have of the group. This is strikingly illustrated by the results of the extensive efforts carried out in recent decades, which through exploration in the field (e.g., Lambert and Rasmussen 1998;Warakagoda and Rasmussen 2004;Sangster et al. 2013) and taxonomic revisions (e.g., Fuchs et al. 2008;Flint et al. 2015;Salter et al. 2020) resulted in the remarkable increase of the number of recognised species of owls from 146 in 1975 (Morony et al. 1975) to up to 230 species in 2021 (Gill et al. 2021). This dramatic increase was supported by the widespread adoption of an integrative taxonomic framework (Padial et al. 2010;Sangster 2018), which combines the use of multiple lines of evidence, such as genetics, morphology, acoustics, geography and behaviour to reach informed decisions on the species status of a given taxon (see also Cadena and Zapata 2021).
Still, compared to other groups, the current discrepancy in the number of owl species accepted by different authorities highlights the challenges associated with the taxonomy and systematics of this group. This stems from their nocturnal habits, making them difficult to study, and from being a group where, at the generic level, morphological variation between species can be similar or lower than within-species (Marks et al. 1999). This is because, in owls, the evolution of plumage pattern and colour is driven by the pressure to remain cryptic during daytime as a defence against predators or to avoid being mobbed by other birds, a common occurrence among members of the family (Marks et al. 1999;König et al. 2008). This led to the convergent evolution of similar camouflaged patterns across species, as happens in other nocturnal bird groups like the nightjars (Caprimulgidae: Holyoak 2001). As such, plumage is generally not diagnostic in owls, with distantly related species often being strikingly similar (Marks et al. 1999). This morphological uniformity is especially evident, and taxonomically challenging, in the most speciose genus of the family: Otus Pennant, 1769, which includes over 50 recognised species, occurring across Asia, Europe, and Africa (Marks et al. 1999;König et al. 2008;Winkler et al. 2020). Commonly known as scops-owls, these small to medium-sized predators show two main plumage colour types, rufous or grey (or grey-brown) morphs, which often occur in the same populations (Pons et al. 2013).
In contrast to plumage, vocalisations of members of the Strigidae family are species-specific. As with most non-passerines and suboscine passerines, owl songs are not learned (Gahr 2000) and therefore have a strong genetic basis. Additionally, contrarily to the difficulty in observing owls, their vocalisations are conspicuous and easily detected as they play a major role in territorial defence and mate attraction (Marks et al. 1999;König et al. 2008). Vocalisations thus represent the most important trait to differentiate species of owls (e.g., Marshall 1978;Sangster et al. 2013;Flint et al. 2015), and new species are often first discovered through their calls Dallimer 2008, 2009;Sangster et al. 2013).
The Gulf of Guinea, Central Africa, has three oceanic islands, Príncipe, São Tomé, and Annobón, in a northeast to southwest line, with São Tomé touching the equator. The rainforests of the islands constitute an independent ecoregion (Gascoigne 2004) characterised by high endemism levels across groups (Jones 1994). Endemism is particularly striking in birds, with the islands having been classified as the third most important in the world for the conservation of forest birds (Buchanan et al. 2011). Relatively to their area, the two larger islands (which together make the Democratic Republic of São Tomé and Príncipe) have by far the highest number of endemic bird species in the world (at least 28 endemic species in 996 km 2 ; Melo et al. 2022).
Although birds are the best-studied group of the Gulf of Guinea islands (Jones and Tye 2006;Melo et al. 2022), the presence of a candidate species of owl on Príncipe Island was only confirmed in 2016 (Ryan 2016;Verbelen et al. 2016), following decades of the accumulation of evidence pointing towards it (Suppl. material 1). In this paper we confirm the distinctiveness of the population of scops-owls from Príncipe using morphometrics, plumage colouration and pattern, song, and mitochondrial and nuclear DNA sequence data. We discuss the origin of this species by placing it within a large-scale phylogeny of the genus Otus. Together with the new species, this phylogeny included 14 taxa never previously analysed, providing novel insights on the relationships within the most speciose genus of the Strigidae family.

Study site
Príncipe 'E) is located in the Gulf of Guinea, ca. 220 km offshore Gabon (Fig. 1). Part of the Cameroon Line of Volcanoes, its oldest sub-aerial rocks date to the origin of the line at ca. 30 Ma (Burke 2001). With a surface area of 139 km 2 (ca. 17 km long and 8 km wide), it has a relatively flat low-lying area in the north, contrasting with the rugged central and southern region characterised by high ridges that rise up to 948 m a.s.l. at Pico do Príncipe (Jones and Tye 2006). It has an oceanic equatorial climate, with an annual precipitation that can reach 5,000 mm. Most of the accessible regions of the island were cleared and converted to plantations (Jones and Tye 2006). Some of those areas were abandoned and regenerated into secondary forest (Atkinson et al. 1991;Castanheira-Diniz and Cardoso-de-Matos 2002;Jones and Tye 2006). The remaining area is covered with two types of native forest stratified by altitude: lowland and montane rainforest, the latter being restricted to Pico do Príncipe and the surrounding summits (Exell 1944).
Fieldwork for specimen and tissue sample collection, measurements, and additional bioacoustics recordings took place in May 2017, July 2018, and January 2019. All samples and vocalisation recordings were collected within Príncipe Obô Natural Park, in the south of Príncipe (Fig. 1). Locality information was recorded using a GPS receiver (Garmin GPS Map 62s; Garmin International Inc., Olathe, Kansas, United States).

Voucher specimen collection
On May 29, 2017, in the Ribeira Porco area (1°33.03'N, 7°22.29'E, Fig. 1), one individual was captured using mist-nets ( Fig. 2A), measured and a blood sample collected. This individual was euthanised by inhaling an Isoflurane 1 mL/5 L solution, dissected, fixed with absolute EtOH, and preserved in 80% EtOH. Afterwards, the specimen was prepared as a study skin and spread wing, and the partial skeleton was prepared following a modified procedure (Cataldo 2017) from that described by Davis and Payne (1992) and Baker et al. (2003). Previously to being captured, recordings of its vocalisations and of those of a second bird were obtained using a recorder (Edirol R-09HR, Roland, Japan) and a microphone (MKE 400, Sennheiser, Germany); these were elicited by playing back previously recorded vocalisations of this taxon to attract it into the nets. The voucher specimen was photographed to document life colouration and appearance ( Fig. 2A). The voucher was deposited in the ornithological collection of the Natural History and Science Museum of the University of Porto (Table 1).

Taxonomy
In this study, species diagnosis was based upon four lines of evidence: morphometrics, plumage colouration and pattern, song, and DNA sequence data.
Species and subspecies limits of scops-owls are challenging to ascertain, leading to numerous taxonomic arrangements. This study follows the taxonomic arrangement and nomenclature of The Clements Checklist of Birds of the World (Clements et al. 2021). Terminology and description scheme follow Sangster et al. (2013), and the features used for the description are depicted in the Suppl. material 3. Description of colour in life is based on the holotype, with some reference to variation as observed in specimens photographed in the field (Fig. 2).

Morphology
Four scops-owl individuals from Príncipe (including the vouchered specimen) were captured in the field. These were measured together with representatives of four of the five species of the Afro-Palearctic clade (sensu Pons et al. 2013) andO. brucei obsoletus (Cabanis, 1875). The latter was included due to the potential affinities of O. brucei (Hume, 1872) with African species (Pons et al. 2013). The Arabian Scops-Owl O. pamelae Bates, 1937 was not included, but we measured individuals of O. senegalensis feae (Salvadori, 1903) from Annobón Island, considered by some authors as a distinct species (Collar and Boesman 2020;Gill et al. 2021). The morphometric dataset includes measurements collected from museum specimens and living individuals measured in the field (Table 1).
Measurements were taken as follows: bill length from the bill tip to where the culmen enter the feathers (Bilen); bill length from the anterior end of the nares to the tip (Binares); bill width (Biwid) and bill depth (Bidepth) at the anterior end of nares; head+bill  (Hebi), from the tip of the bill to the opposite point on the back of the skull; middle toe length (Midt); tarsus length (Tarlen), from the tibiotarsus joint to the distal end of the tarsometatarsus, when the foot is held to the leg; tail length (Tailen), from where the ruler stops at the root of the central pair of rectrices and to the tip of this same pair (by sliding the ruler between the rectrices and the undertail coverts); body length (Bolen) from the top of the head to the tip of the central pair of rectrices; wing length (Wilen), flattened, from the carpal joint to the tip of the longest primary; wing formula, sequence of primary feathers ordered by size; and length of primary feathers (P4-P10, in which P1 is the closest to the body), which were transformed in shortfall of P4-P10 to tip of longest primary. Body, wing, and tail length were measured with standard wing and tail rulers to the nearest 1.0 mm. The length of the primary feathers was measured to the nearest 1.0 mm with a ruler with a pin at the origin; the pin is inserted between two primary feathers until it touches the skin (Jenni and Winkler 1989). All other measurements were made using a digital calliper (Mitutoyo CD-P15K, Mitutoyo Corporation, Kawasaki, Japan) to the nearest 0.1 mm. All measurements were collected by MM, BF, and RL (Table 1). A constant of one was added to each number to make all shortfalls non-zero. All measurements were log-transformed (base-10) to normalise distributions (McDonald 2014). The four individuals from Príncipe were sexed with a molecular protocol (Griffiths et al. 1998), and comprised three females and one male (Table 1). For the statistical analyses males and females were treated together due to the low sampling size, and the fact that most museum specimens were not sexed.
Morphometric differences were explored using a Principal Component Analysis (PCA), performed using the FactoMineR package (Lê et al. 2008) and carried out using R v. 3.6 (R Core Team 2017) in RStudio v. 1.1.447 (RStudio Team 2015). Measurements were size-standardised to prevent the dominance of variables invol ving larger measurement units, thus allowing comparisons between variables. The wing formula was not used in the analyses. Since several individuals had missing data, to maximise the number of analysed specimens of each species only the following variables were used in the PCA: Bilen, Binares, Biwid, Tarlen, Wilen, and Tailen. This dataset included 44 specimens from seven taxa, including three individuals from Príncipe. Welch's ANOVA (recommended for unbalanced designs, different samples sizes, and different standard deviations; McDonald 2014), was used together with Games-Howell post-hoc comparisons to test whether the groups differed from each other.

Plumage description
We used colour standards (Köhler 2012) to describe the plumage of the species of the Afro-Palearctic clade, except for O. pamelae, but including O. senegalensis feae (see Suppl. material 6). The topographic terms of the scops-owl body are detailed in the Suppl. material 3.

Bioacoustic analyses
We compared the calls of the candidate species with the calls of scops-owls from the Afro-Palearctic clade, O. brucei, and the Sokoke Scops-Owl O. ireneae Ripley, 1966 (Tables 2,  3). Otus ireneae is not part of the Afro-Palearctic clade but was included because it has the most similar calls to the ones of the candidate species. Recordings were collected from Xeno-canto (XC; www.xeno-canto.org), Avian Vocalizations Center (AVoCet; https://avocet.integrativebiology.natsci.msu.edu), The Internet Bird Collection (www.hbw.com/ibc), the private collection of PV and the collection of vocalisations obtained during fieldwork on Príncipe by co-authors. Newly collected calls were deposited in Xeno-canto (Suppl. material 7). From each independent source, only one recording was used (unless there were only very limited recordings available). In total, 43 recordings from ten taxa were analysed (Suppl. material 7). In most owls, both sexes produce similar calls for territorial defence, mate attraction, and pair-bonding (Marks et al. 1999;König et al. 2008), therefore our analyses included both male and female recordings. Vocalisations of scops-owls are generally made up of a simple primary call composed by the repetition of the same note (Marks et al. 1999). The call of the Cyprus Scops-Owl O. cyprius (von Madarász, 1901) is composed by the coupling of one long and one short note (Flint et al. 2015), and both note types were included in the analysis. The candidate species primary call is characterised by a repeated note (Suppl. material 4: Fig. S2A, B), but its repertoire also includes a cat-like "kee-a-u" note (Suppl. material 4: Fig. S2C, D). The latter was not included in the analysis.
Recordings were sampled using a 16-bit accuracy and a sampling rate converted to 12 kHz in Avisoft-SASLab pro v. 4.3 (Avisoft Bioacoustics). The following 16 variables were collected for each note: F1, frequency at start (peak frequency at 0s, Hz); F2, frequency at end (peak frequency at last of four call intervals, Hz); F3, frequency at 25% of total duration (peak frequency at the first interval, Hz); F4, frequency at midpoint (peak frequency at the second interval, Hz); F5, frequency at 75% of total duration (peak frequency at the third interval, Hz); F6, frequency at maximum amplitude (frequency at maximum amplitude of note, Hz); F7, maximum frequency (maximum frequency through the note, Hz); F8, minimum frequency (minimum frequency through the note, Hz); DT1, total duration (duration, s); DT2, time to maximum amplitude (time to maximum amplitude of note, s); DT3, time to maximum frequency (time to maximum frequency of note, s); DT4, internote interval (start time -end time of previous note, s); DF1, frequency drop from start to end (F2-F1, Hz); DF2, frequency range (F7-F8, Hz); DFT1, slope from 25% to 75% of total duration ([F5-F3]/∆t, Hz/s); DFT2, slope from midpoint to end ([F2-F4]/∆t, Hz/s). The 16 variables were extracted from the analysis of the spectrograms. We used a Fast Fourier Transformation size of 512 points, a 100% frame size and a temporal resolution overlap of 87.5% (flat top window type), resulting in a frequency resolution of 86 Hz and a temporal resolution of 4.5 ms. Frequencies were analysed between 0.5 Hz (highpass) and 2.25 Hz (lowpass), except for O. brucei with the highpass set at 0.0 Hz; the greyscale was set to 30%. When background noise hampered the measurement of the variables, frequencies were filtered and adjusted by shortening the interval between the highpass and the lowpass.
For each recording, variables were measured on six notes and their means (Suppl. material 8) were used as sample points to calculate the ranges, means and standard deviations for each taxon. A constant of 3000 was added to each computed value to ensure that the dataset only included positive numbers. All measurements were logtransformed (base-10) to normalise distributions (McDonald 2014). PCA was performed using the FactoMineR package (Lê et al. 2008) using R v. 3.6 (R Core Team 2017) in RStudio v. 1.1.447 (R Studio Team 2015). Measurements were size-standardised to make the variables comparable. Welch's ANOVA and Games-Howell post-hoc comparisons were used to test whether the groups differed from each other (McDonald 2014).
Total genomic DNA was extracted from blood and tissue samples using an overnight Proteinase K digestion (10 mg/ml concentration) followed by a standard high-salt extraction method (Bruford et al. 1992). Before the extraction, blood was removed from the ethanol and left to dry in the incubator. Genomic extractions of toe-pads samples were performed using a specific protocol for museum samples (Dabney et al. 2013).
Standard polymerase chain reactions (PCR) were performed in a final volume of 25 µl using 1 µl of each primer (10 pmol), 0.4 µl of total dNTPs (10 mM; Promega), 0.1 µl of 5 U/ml GoTaq Flexi DNA Polymerase (Promega), 5 µl of 5X Green GoTaq Flexi Buffer (Promega), 4 µl of MgCl 2 (25 mM; Promega). The first PCR of the fragments amplified using a nested PCR approach (KIAA1239, SACS, TTN) were performed in half total reaction volume (12,5 µl). Primers and PCR conditions are provided as Suppl. material 10. For the amplification of the toe pads, we amplified shorter fragments of the MYO2, ND2, and ND3 genes. Successfully amplified products were purified and sequenced using dye-labelled dideoxy terminator cycle sequencing on an ABI 3730XL automated sequencer at Macrogen Inc.
Chromatograms of newly generated sequences were checked by eye, edited and aligned using BioEdit v. 7.0.5.3 (Hall 1999). All newly determined sequences were submitted to GenBank (accession numbers provided in Table 4).

Molecular analyses
Six different datasets were compiled for different purposes, detailed below.

Dataset 1
ND2 sequences of the species belonging to the Afro-Palearctic clade and of O. brucei, whose taxonomic affinities are not resolved, although it appears to be closer to African and Indian Ocean islands species than to the Asian species (Pons et al. 2013). This alignment contained 16 sequences belonging to the candidate species from Príncipe, the five species of the Afro-Palearctic clade, including the mainland and Annobón subspecies of O. senegalensis, and O. brucei. This dataset was used to compute the mean genetic distances matrix (uncorrected p-distance in percentage, using the pairwise deletion option) within and between taxa. Distances were computed using MEGA, v. 7.0.21 (Kumar et al. 2016).

Dataset 2
Molecular dataset used for the phylogenetic and divergence time analyses. This dataset comprised 51 individuals from 39 taxa and a final concatenated sequence alignment of 12,925 bp. This dataset was built with the sequences produced in this study together with previously available sequences. We used homologous sequences of the Tawny Owl Strix aluco Linnaeus, 1758 and of the Eurasian Eagle-Owl Bubo bubo (Linnaeus, 1758) for outgroup rooting. The input files for phylogenetic inference were prepared in "Pipelogeny" (Muñoz-Pajares et al. 2019). Sequences were automatically aligned using the mafft algorithm (Katoh and Standley 2013). The best model of molecular evolution and the best partition scheme to analyse the molecular dataset was identified using PartitionFinder2 v. 2.1.1 (Lanfear et al. 2012(Lanfear et al. , 2016 using the greedy algorithm and applying the Bayesian Information Criterion (BIC).
Bayesian inference (BI) analyses were computed in MrBayes v. 3.2.6 (Ronquist et al. 2012). Two runs of 100 million generations (starting with random trees) and four incrementally heated Markov chains were performed, using default heating values and sampling the Markov chains at intervals of 1000 generations. The first 40% of trees were discarded as burn-in, and the remaining trees were retained and summed up to gene rate a 50%-majority rule consensus tree. Chain mixing, stabilisation and convergence of likelihood values was assessed by examining the standard deviation of split frequencies.
PartitionFinder and the partitioned BI analyses were run on the CIPRES gateway server (Miller et al. 2010) on XSEDE. The purpose of this phylogenetic analysis was: 1) to confirm that the representatives of the candidate species form a monophyletic group; 2) to infer the phylogenetic relationships of the candidate species within Otus.
We estimated divergence times (one partition per locus) using Beast 1.10.4 (Drummond et al. 2012). Substitution models for each marker were selected using MEGA X (Kumar et al. 2018) and used default prior distributions for the substitution models parameters. We specified uncorrelated lognormal clock models (Drummond and Rambaut 2007) for the seven mitochondrial loci and strict clock models for six nuclear loci. As calibrations, we used the substitutions rates and corresponding associated uncertainties from Lerner et al. (2011) and specified the following priors: ND2normal distribution with mean 0.029 and standard deviation 0.0029; ND3 -normal distribution with mean 0.024 and standard deviation 0.003; ATP6 -normal distribution with mean 0.026 and standard deviation 0.0026; CYTB -normal distribution with mean 0.014 and standard deviation 0.0012. We used uniform distributions for each of the nuclear loci (lower bound: 0.0, upper bound: 0.5). We specified a Birth-Death prior for the tree prior. MCMC chains were run for 50 million iterations with trees and parameters sampled every 1000 iterations. Analyses were run on the CIPRES 3.1 gateway server (Miller et al. 2010). We used Tracer v. 1.7 (Rambaut et al. 2018) to help ensure that the effective sample size for all Bayesian analyses of the underlying posterior distribution was adequate (> 200) for meaningful estimation of parameters.

Datasets 3-6
The alignments of each nuclear gene were analysed separately to obtain evidence for genetic differentiation of lineages from unlinked loci and, hence, to provide further support to their status as distinct species following the criterion of genealogical concordance (Avise and Ball 1990;Avise and Wollenberg 1997). These datasets contained the nuclear markers KIAA1239, MYO2, TGFB2 and TTN. In these four datasets we included all species of the Afro-Palearctic clade (sensu Pons et al. 2013) and, when available, se quences of O. brucei. We trimmed all sequences to equal length (KIAA1239: 653 bp, MYO2: 131 bp, TGFB2: 403 bp, and TTN: 706 bp). Sequences were phased using the PHASE algorithm (v. 2.1.1) with default settings (Stephens et al. 2001) as implemented in the software DnaSP (v. 6.12.01; Rozas et al. 2017). PHASE parameters were 1000 iterations, one thinning interval and 100 burn-in iterations and a posterior probability threshold of 0.95 to determine the most probable inferred haplotypes for each nuclear sequence. Analyses were repeated three times with different seed values. Haplotype network reconstruction of phased sequences was performed using the software TCS, v. 1.21 (Clement et al. 2000). This software applies the method of Templeton et al. (1992) to calculate the minimum number of mutational steps between haplotypes, computing the probability of parsimony for pairwise differences until the probability exceeds 0.95. This analysis was used to detect the occurrence and extent of haplotype sharing in the studied Otus species. The resulting networks were plotted using the online tool tcsBU (Múrias dos Santos et al. 2016). We interpreted the lack of haplotype sharing among individuals of different phylogenetic lineages as independent evidence of their evolutionary distinctiveness.

Nomenclature review
This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The LSID (Life Science Identifier) for this publication is: urn:lsid:zoobank.org:pub:0731A37D-B363-43C9-A1AC-69F5E10F6810. The electronic edition of this work was published in a journal with an ISSN, and has been archived and is available from the following digital repositories: https://zookeys.pensoft.net/.

Data availability
Molecular data are deposited in GenBank. Photographs and audio recordings are deposited in Macaulay Library and Xeno-canto, respectively. All other datasets underpinning this article are available as supplementary files (Suppl. materials 1-16).

Justification for species delimitation
Following the integration by congruence approach (Padial et al. 2010), we consider independent evolutionary lineages as separate species if two or more independent lines of evidence support their distinctiveness. The candidate species: i) differs in morphology from other Otus species (Fig. 3); ii) has a very distinct call (Tables 2, 3; Figs 4, 5); iii) forms a distinct monophyletic group (Fig. 6) that iv) differs from the nearest taxa by a mitochondrial uncorrected pairwise sequence divergence (p-distance) similar or larger than the divergence observed between other currently accepted sister species pairs of the genus Otus (Table 5); v) has no haplotype sharing at some of the analysed nuclear markers (Fig. 7). We interpret the concordance between these independent lines of evidence as strong support for its distinctiveness and species status (Avise and Ball 1990;Padial et al. 2010;Sangster 2018). Therefore, we conclude that the candidate species from Príncipe is a distinct species that we describe herein as Otus bikegila sp. nov., and for convenience we will use this name throughout the manuscript, anticipating its formal description below.

Morphological differentiation
In the PCA, the first two components presented eigenvalues higher than one (Suppl. material 11), representing 81.7% of the variation. PC1 (60.9% of the variance) was negatively correlated with all variables related with bill, tarsus length and tail length, whereas PC2 (20.8% of the variance) was positively correlated with wing length (Suppl. material 11). Otus bikegila sp. nov. is within the variation of O. pembaensis and separated from the other species (Fig. 3). All other taxa are separated from each other, except O. senegalensis feae that is within the variation of O. senegalensis senegalensis (Swainson, 1837).
For the Welch's ANOVAs, O. brucei obsoletus was not included due to the small sample size. All variables differed significantly between species (P < 0.05), except Bidepth (Suppl. material 12). Games-Howell post-hoc test provided significant values for all the species only for Bilen and Biwid because of missing data from some specimens. This test was used to identify the morphological diagnostic characters detailed in the Diagnosis section.

Bioacoustic differentiation
In the PCA, four components presented eigenvalues higher than one (Suppl. material 13), representing 87.6% of the variation. PC1 (56.8% of the variance) mostly represented frequency characteristics of the note, being positively correlated with the  Means of all bioacoustic variables differed significantly (P < 0.05) between species (Suppl. material 14). Low sample size prevented performing the Games-Howell posthoc test between all taxa pairs. This test identified the bioacoustic diagnostic characters detailed in the Diagnosis section.
The topology of the majority rule consensus tree (Dataset 2; Fig. 6) is largely in agreement with previously published phylogenies of the genus Otus (Fuchs et al. 2008;  Bioacoustic parameters -F1: frequency at start; F2: frequency at end; F3: frequency at 25% of total duration; F4: frequency at midpoint; F5: frequency at 75% of total duration; F6: frequency at maximum amplitude; F7: maximum frequency; F8: minimum frequency; DT1: total duration; DT2: time to maximum amplitude; DT3: time to maximum frequency; DT4: internote interval; DF1: frequency drop from start to end; DF2: frequency range; DFT1: slope from 25% to 75% of total duration; DFT2: slope from midpoint to end. Pons et al. 2013), and increased the resolution of these by resolving some po lytomies. An important improvement relatively to Fuchs et al. (2008) and Pons et al. (2013) was the inclusion of 13 additional taxa besides O. bikegila sp. nov. Otus icterorhynchus was found to represent a relatively basal lineage, sister to the clades containing the Afro-Palearctic and the Indo-Malayan/Indian Ocean species (PP = 1), rather than being sister to O. ireneae, as often hypothesised. The two subspecies of O. icterorhynchus were recovered as sister taxa, albeit with a very high genetic divergence (only one sample per taxon). Otus brucei (from the Arabian Peninsula to Asia) was the sister lineage to the Afro-Palearctic clade (PP = 1). Other novel insights are detailed in the discussion.
Otus bikegila sp. nov. samples were recovered as monophyletic, and formed a clearly distinct lineage belonging to the Afro-Palearctic clade. It was recovered as the sister lineage (PP = 1) of the clade formed by O. senegalensis senegalensis, O. senegalensis feae, O. hartlaubi, and O. pembaensis (Fig. 6).  The best model of sequence evolution for each marker used for the divergence times analyses are listed in the Suppl. material 16. The genus Otus started to diversify ca. 7.8 mya (95% high posterior density [HPD]: 6.2-9.6). The two primary cla des diverged at ca. 6.3 mya (95% HPD: 5.1-7.7), and went on to diversify at similar times: Nuclear markers independently supported the evolutionary independence of the taxa of the Afro-Palearctic clade. The taxa included in the analysis shared no haplotypes for markers KIAA1239 and TGFB2; the latter was the most variable of the analysed nuclear markers with a total of 19 haplotypes (Fig. 7). For MYO2, O. hartlaubi and O. brucei had no shared haplotypes, whereas O. bikegila sp. nov., O. scops and O. senegalensis had some unique haplotypes but shared the most common one (Fig. 7). Otus hartlaubi and O. scops did not share any TTN haplotype, with O. bikegila sp. nov. having both one unique haplotype and one shared with O. senegalensis feae (Fig. 7).
We provide here a diagnosis relatively to the closely related species belonging to the Afro-Palearctic clade and also to O. ireneae due to the similarity in their calls. The diagnosis is based on the following analysed morphological characters: 1) Biwid; 2) Binares; 3) Tarlen; 4) Wilen; 5) Tailen); 6) SP10; 7) SP9; 8) SP4; on the following analysed bioacoustics characters: 9) F1; 10) F2; 11) F3; 12)   Otus bikegila sp. nov. differs from O. scops in seven morphological and 11 bioacoustic characters: higher Biwid (9.0 to 11.8 mm vs. 5.5 to 7.8 mm), larger Binares (11.3 to 12.6 mm vs. 9.0 to 11.5 mm), longer Tarlen    Underparts: Breast overall Pale Buff 1 with Sepia 286 irregular markings and Robin Rufous 29 shading. Breast feathers Pale Pinkish Buff 3 with Sepia 286 dots and markings proximally and Pale Pinkish Buff 3 shaft, distally Sepia 286 with Pale Pinkish Buff 3 dots and markings forming irregular bands, middle Sepia 286 stripes along shaft. Belly overall similar to breast but with colours more defined and with Light Buff 2 shadings. Belly feathers Pale Pinkish Buff 3 in proximal section followed by a Sepia 286 V stripe. This is followed by a Pale Buff 1 broad band delimitated distally with a thin Pale Pinkish Buff 3 line followed by a Sepia 286 line. Distally, these feathers are Pale Buff 1 with irregular spots Sepia 286 and Pale Pinkish Buff 3. Some feathers on the belly and the vent have a marked middle and broad Sepia 286 line along shaft. Vent is similar to breast and belly but with feathers Cinnamon 21 in proximal section followed by a Sepia 286 stripe. This is followed by a Pale Buff 1 broad band delimitated distally with a thin Cinnamon 21 followed by a Sepia 286 line. The feather then becomes Pale Buff 1 with a Cinnamon 21 thin band followed by a Sepia 286 line, ending with a Pale Buff 1 colouration with Sepia 286 irregular dots and markings. Flank feathers Pale Buff 1 with Cinnamon 21 shading followed by a broad Cinnamon 21 V stripe followed by a Sepia 286 line, ending with a broad Pale Buff 1 section with Sepia 286 markings only in the very distal portion. Undertail coverts are similar to flanks but with more defined bands. Feathers are Pale Buff 1 followed by a broad Cinnamon 21 band defined distally by a thinner warm Sepia 40 line. This colouration is repeated twice. Feathers end with a broad Pale Buff 1 band fol-lowed by a Cinnamon 21 band with irregular Warn Sepia 40 markings. Tarsus covered with feathers to base of toes. Feathers overall similar to flank but with less Pale Buff 1 and more Cinnamon 21 shading and one or two Sepia 286 dots in distal section. Tarsus feathers have a larger proportion of Pale Buff 1 close to toes. Tarsus feathers are Pale Buff 1 proximally, followed by Cinnamon 21 shading and Sepia 286 markings distally (approximately 1/4 of the feather distally), no middle stripe along shaft. Toes feathers are Pale Buff 1 with Cinnamon 21 shadings and Sepia 286 markings only in the distal section.
Wing Vocalisations: Call recordings collected at the moment of specimen collection included the call of the holotype and a second individual (XC audios: XC619445, XC619447): one emitted the main call type (the single repetitive note used in the bioacoustic analyses), and the other the cat-like call. We believe that the holotype individual was the one giving the main call, but this was uncertain. Thus, it is not possible to provide bioacoustics parameters specific to the holotype.
Variation: Morphometric variation in O. bikegila sp. nov. is based on the analysis of three additional individuals, of which one is a male (Table 1; Fig. 2). The male (P9-038) had shorter tarsus and wing length than the female holotype and the other two females. This result is consistent with the reversed sexual dimorphism in size described for all species of scops-owls (Marks et al. 1999;König et al. 2008). Two colour morphs (rufous and grey-brown) have been documented in the field (Figs 2,9). Molecular sexing of the four captured individuals has shown that colour morph is not associated with sex. Examples of the grey-brown morph include the holotype (Figs 2A, 8), individuals P9-037 and P9-038 (Fig. 2B), and individuals photographed in the field (Fig. 2D, E); examples of the rufous morph include the first photographed individual of this species (Fig. 2C) and individual P8-001 (Fig. 2F). Plumage pattern and colour of the latter is similar to the holotype, although in the rufous morph the eyebrows are less marked, the underparts are more similar in colour to the upperparts and have more prominent sepia marks and stripes along the feather shafts. In the field, we observed no differences in the rate of occurrence of the two morphs.
Vocalisations were recorded at the type locality by MM in 2002MM in , 2007MM in , 2011MM in , 2018 and 2019 and at Boca do Inferno in 2019, and by PV at the type locality in 2016. The call of O. bikegila was described in Melo and Dallimer (2009). Among vocalisations of Otus species, the primary call of O. bikegila sp. nov. is unique in consisting in a short, undulated note emitted at a fast repetition rate, reminiscent of insect calls, of ca. one note per second (Tables 2, 3; Fig. 4; Suppl. material 4: Fig. S2A). Vocalisations were often performed in duet (Suppl. material 4: Fig. S2B), with intercalated or overlapping notes. Otus bikegila sp. nov. is able to produce a cat-like "kee-a-u" note, which is emitted both in duets (Suppl. material 4: Fig. S2D) and by single birds (Suppl. material 4: Fig. S2C). We confirmed in the field that the same individual can produce both calls. Bioacoustic parameters (mean ± standard deviation) of the primary and of the cat-like notes are available in Tables 2, 3.
Etymology. The species name is a patronym honouring Ceciliano do Bom Jesus, known as 'Bikegila' (Suppl. material 5). The species epithet name is intentionally defined as an invariable noun in apposition (not a noun in the genitive case) for better pronunciation; no confusion with the species authority is possible because the noun is an oral nickname.
Bikegila, a native of Príncipe Island, began the 'Príncipe Scops-Owl saga' in 1998, when he shared with MM reports of two sightings of birds that looked like owls in parrot nests. Since then, Bikegila took part in every field effort that led to the bird's discovery for science; he also led the capture of all sampled individuals, including the holotype, which required ingenious ways to erect canopy nets. For almost 25 years, Bikegila has put all his resources, including bottomless fieldwork skills and a vast knowledge of Príncipe, towards the successful completion of innumerable research projects in a terrain that the collector José Correia considered to be the "bad among the bad or the worse among the worse" [sic] (Diary, 2 September 1928, Archives AMNH, New York). Besides his skills, Bikegila's "cheerful temperament, possibly the first requirement for an undertaking in inhospitable regions" (von Humboldt 1841), coupled with an unbeatable gift for story-telling and an underlying quiet wisdom, contributes as much to making the expeditions he leads memorable and successful. A former parrot harvester, Bikegila became a warden of Prín cipe Obô Natural Park soon after its creation; he is now a much sought-after nature guide.
We believe that most field researchers are grateful to the 'Bikegilas' with whom they are/were honoured to work with. As such, the name is also in recognition of all the people, around the world, who through their deep relationship with and knowledge of the regions they inhabit, play key roles in the description of new species and of new sites to science.
Common name. We propose the English common name Principe Scops-Owl, the name for São Tomé and Príncipe as Kitóli-do-príncipe, and the name for the Portuguese list of the birds of the world as Mocho-do-príncipe. All common names refer to Príncipe Island, from where it is endemic.
Distribution and natural history. All records from O. bikegila sp. nov. come from old-growth native lowland rainforest with mid-height (14-20 m) trees (Fig. 10), with the species apparently preferring lower elevations (Melo and Dallimer 2009;Freitas et al. 2022). Its area of occurrence is fully within the limits of Príncipe Obô Natural Park. Detailed surveys have been carried out to determine the area of occupancy of this species, to estimate its population size, ecological requirements, and to propose an IUCN Red List category (Freitas et al. 2022).
The holotype (Figs 2A, 8; female MHNC-UP-AVE7000), collected on 29 May 2017, was undergoing a well-advanced moult, a process that takes place after the breeding season. The female captured close to Ribeira Porco, in January 2019 (P8-001) had a fully developed brood patch (Fig. 2F), whereas the female captured at Boca do Inferno on the same month (Fig. 2B, left) was growing back the belly feathers, suggesting that she had a recent brood patch. This indicates that breeding takes place in December-January, as with most bird species of the islands of São Tomé and Príncipe (Jones and Tye 2006;Madeira 2018).
Otus bikegila sp. nov. starts calling at dusk and continues throughout the night. Contrarily to the Sao Tome Scops-Owl O. hartlaubi that regularly vocalises during the day, O. bikegila sp. nov. seems to require darkness to sing, although on a single occasion one individual was heard during the day (Melo and Dallimer 2009). Response to the playback of its call was fast and intense at all times of the year we were able to test it, with birds of either sex approaching the speaker. This indicates that O. bikegila sp. nov. is territorial all-year round as it is known from most sedentary cavity-nesting owls (Marks et al. 1999;König et al. 2008). During the day it may roost outside of tree cavities, as suggested when we accidentally flushed one bird when taking habitat measurements. In this situation the bird raised its ear tufts, which are otherwise seldom observed (Fig. 2D).

Otus bikegila: a new bird species, endemic to Príncipe Island
Multiple lines of evidence were brought together to demonstrate, unambiguously, that the recently discovered population of scops-owls on Príncipe Island makes a well-differentiated species, Otus bikegila. Genetic distances, and associated divergence times, to its closest relatives were in the range of those separating currently accepted species (Table 5). Morphological differences, although present, did not stand out (for the human eye at least), whereas vocalisations were unique and clearly distinctive (even for the human ear) and, in fact, it was bioacoustics that led to the discovery of the population of the Principe Scops-Owl. Its unique vocalisations were closest to those of O. ireneae, an Otus species from which it is distantly related, underscoring the value of song in scops-owls to assess taxonomic status but not for inferring taxonomic affinities (Fuchs et al. 2007). Phylogenetic data placed O. bikegila as the sister lineage of the clade containing all African scops-owl species of the Afro-Palearctic clade (sensu Pons et al. 2013 and Fig. 6), rather than as the sister species of O. hartlaubi endemic to the neighbouring island of São Tomé. This leads to the curious conclusion that Príncipe was likely the first island in the Gulf of Guinea to be colonised by a species of scops-owl, albeit the last species to be discovered and described for science. It also begs the question if an undescribed scops-owl waits to be discovered in the extensive rainforests of Bioko Island, the only island of the Gulf of Guinea without records of a scops-owl. This apparent absence is puzzling as Bioko is a land-bridge island, which has been connected to the mainland in multiple instances in the past (Rohling et al. 1998;Lambert and Chappel 2001), and currently lying at ca. 30 km from Cameroon where O. icterorhynchus, a rainforest specialist, is present. Although it may seem odd for a bird species to remain undiscovered for science for so long on such a small island, this is by no means an isolated case when it comes to owls. For example, the recently described Rinjani Scops-Owl O. jolandae Sangster, King, Verbelen & Trainor, 2013 was found to be a previously undescribed species from Lombok Island, Indonesia (Sangster et al. 2013). Similarly, the Anjouan Scops-Owl O. capnodes (Gurney, JH, 1889) was rediscovered in 1992 (Safford 1993), 106 years after its last observation, in an area of primary forest that is smaller and more regularly visited than that of Príncipe, and the Flores Scops-Owl O. alfredi (Hartert, E, 1897), rediscovered in 1994, 98 years after the previous report (Widodo et al. 1999).

Novel insights in the phylogenetics of the genus Otus
Our phylogenetic analyses confirmed the supported nodes from previous phylogenies (Fuchs et al. 2008;Pons et al. 2013), resolved previously unsupported nodes, and provided novel insights in the affinities of species not previously included.
In relation to African taxa, the most interesting result came from the inclusion of samples from the two subspecies of the only African Otus species never sequenced before: the Sandy Scops-Owl O. icterorhynchus. Together with O. ireneae, this is the only species on the African continent that is a lowland forest specialist (albeit each species occupies very distinct forest types), and the two species were widely hypothesised as being closely related (Marks et al. 1999;König et al. 2008;Holt et al. 2020). Perhaps more surprisingly, O. icterorhynchus has been considered to form a superspecies ('yellowbilled scops-owls') with two Asian taxa, the Andaman Scops-Owl O. balli (Hume, 1873) and the Sumatran Otus spilocephalus stresemanni (Robinson, 1927) (Marshall 1978), although 'stresemanni' could be an anomalous form of another species and/or a hybrid (Pamela Rasmussen in Holt et al. 2020). Our phylogenetic analyses clarify the affinities of O. icterorhynchus, which was found to be sister to the clade containing the Afro-Palearctic and the Indo-Malayan/Indian Ocean clades. Each of its subspecies was available to us by a single individual, but their genetic divergence levels overlap with the levels found between many currently accepted scops-owls sister species pairs. This study better resolved the branching sequence within the Afro-Palearctic clade, except for the position of O. pamelae that could not be determined, contra Pons et al. (2013) who recovered it as the sister lineage of this clade. Instead, O. brucei, with populations extending from the Arabian Peninsula into Asia, was recovered as the sister lineage of the clade which then branches into African and Eurasian subclades (assuming that O. pamelae is sister to O. scops).
The internodes separating O. senegalensis (mainland and Annobón Island), O. hartlaubi (São Tomé Island), and O. pembaensis (Pemba Island) are very short, indicating that the divergence between these three species (i.e., the colonisation of both islands from their mainland ancestor) occurred almost simultaneously, creating a hard polytomy. Our analyses failed to identify solid lines of evidence for the distinctiveness of O. senegalensis feae from O. senegalensis senegalensis, although we did identify a diagnostic morphological character (bill length from tip to nares, Suppl. material 12) and three molecular diagnostic characters at the TGFB2 gene (C vs. T in site 285, T vs. G in site 344, C vs. G in site 392). In a dataset with a wider taxonomic sampling of O. senegalensis but fewer sequencing data, the Annobón sample nested within the mainland samples (unpublished data). A better sampling of O. senegalensis senegalensis from across its range will help to resolve this taxonomic issue.
Our sampling increased considerably the taxon coverage for the centre of the diversity of the Otus genus, the Indo-Malayan region (Marks et al. 1999), but has failed to solve most of the many taxonomic pending issues. The Ryukyu Scops-Owl O. elegans (Cassin, 1852), a species restricted to small oceanic islands from the northern Philippines to Japan, and included for the first time in a phylogenetic study, was recovered as the fourth Asian representative of the Indo-Malayan/Indian Ocean clade ( Fig. 6; sensu Pons et al. 2013), albeit with no statistical support. At this stage, our results mainly highlight the need for using a phylogeographic-level sampling scheme across the region (i.e., multiple samples per site covering all areas of occurrence) to enable a thorough systematic revision of the Indo-Malayan taxa, a crucial step towards reconstructing the diversification history of the genus Otus.
Our molecular dataset confirmed the low levels of divergence (well within intraspecific variation) of three taxa pairs that are currently treated either as separate species or subspecies. These pairs are: i) O. senegalensis senegalensis (mainland Africa) and O. s. feae (Annobón Island), treated as separate species by del Hoyo (2020) and Gill et al. (2021) based on Collar and Boesman (2020); ii) O. scops and O. cyprius, treated as a distinct species by Gill et al. (2021) and Clements et al. 2021, based on Flint et al. (2015; and iii) the two species from Madagascar, the Malagasy Scops-Owl O. rutilus (Pucheran, 1849) and the Torotoroka Scops-Owl O. madagascariensis Grandidier, A, 1867, whose specific status was proposed by Rasmussen et al. (2000) and adopted by most authorities (e.g., Clements et al. 2021, del Hoyo 2020, Gill et al. 2021), but contested by Fuchs et al. (2007) using a representative geographic sampling.

Conclusions
The discovery of a new bird species inhabiting the forests of Príncipe Island in 2016 (here formally described as Otus bikegila) underscores both the actuality of field-based explorations aiming at describing biodiversity (Dijkstra 2016), and how such curiosity-driven endeavour is more likely to succeed when coupled with local ecological knowledge, the participation of keen amateur naturalists, and persistence. . Superscript values indicate significance levels (* P < 0.05; ** P < 0.01; *** P < 0.005; **** P < 0.001) and NA (Not Available, significance level not possible to obtain due to low sample size). Some individuals had missing data, not allowing for the test performance (indicated by - . Superscript values indicate significance levels that are indicated with asterisks (* P < 0.05; ** P < 0.01; *** P < 0.005; **** P < 0.001) and NA (Not Available, significance level not possible to obtain due to low sample size). Bioacoustic parameters -F1: frequency at start; F2: frequency at end; F3: frequency at 25% of total duration; F4: frequency at midpoint; F5: frequency at 75% of total duration; F6: frequency at maximum amplitude; F7: maximum frequency; F8: minimum frequency; DT1: total duration; DT2: time to maximum amplitude; DT3: time to maximum frequency; DT4: internote interval; DF1: frequency drop from start to end; DF2: frequency range; DFT1: slope from 25% to 75% of total duration; DFT2: slope from midpoint to end. Copyright notice: This dataset is made available under the Open Database License (http://opendatacommons.org/licenses/odbl/1.0/). The Open Database License (ODbL) is a license agreement intended to allow users to freely share, modify, and use this Dataset while maintaining this same freedom for others, provided that the original source and author(s) are credited. Link: https://doi.org/10.3897/zookeys.1126.87635.suppl14