Evaluating the potential of photo‐identification as a monitoring tool for flapper skate (Dipturus intermedius)

Scottish Association for Marine Science, Dunstaffnage, Oban, UK Scottish Natural Heritage (Argyll and the Outer Hebrides section), Oban, UK Scottish Oceans Institute, University of St Andrews, St Andrews, UK North Connel, Oban, UK Bonawe, Oban, UK 6 Institute for Biodiversity, Animal Health and Comparative Medicine, University of Glasgow, Glasgow, UK Correspondence Steven Benjamins, Scottish Association for Marine Science, Dunstaffnage, Oban, PA37 1QA, UK. Email: steven.benjamins@sams.ac.uk


| INTRODUCTION
Understanding of animal ecology has been transformed through the development of methods allowing reliable identification of individuals over extended periods. The ability to identify individual animals also provides significant benefits to conservation research and practice.
Individual identifications aid in answering important questions on population sizes, distributions, and habitat requirements and are particularly beneficial at low population sizes (Parra, Corkeron, & Marsh, 2006). The largest individual-identification studies are based on bird ringing, with hundreds of thousands of birds being individually tagged each year in the UK and Ireland alone (Walker et al., 2016). Individual identification is particularly important when studying animals in small populations as it allows individual behavioural variability to be taken into account (Austin, Bowen, & McMillan, 2004) and also allows an assessment of growth and mortality rates in situations where large statistically valid cohorts are unavailable (Clutton-Brock & Sheldon, 2010). In addition, certain individuals may have particular ecological or societal significance, and regular recording of recognizable individuals may facilitate outreach efforts (e.g. through sponsorship/adoption).
One way in which individual identification can be achieved is by using artificial tags or markers (Silvy, Lopez, & Peterson, 2012). Many different tag designs are available, ranging from physical tags such as bird rings or seal flipper tags, which must be read after recapture, to passive integrated transponder (PIT) tags and active acoustic tags whose presence is recorded when they are near a receiver, to telemetry tags transmitting data via satellite over great distances (e.g. Ehrenberg & Steig, 2003;Gibbons & Andrews, 2004;Matthiopoulos, McConnell, Duck, & Fedak, 2004;Pomeroy, Smout, Moss, Twiss, & King, 2010).
Photo-identification (hereafter referred to as photo-ID) methods are based on the ability to identify individual animals based on photographs of distinctive natural marks (e.g. skin pigmentation patterns, fin shapes, scars), which can then be used to reliably identify each individual over time and across space. The method works best in cases where recapture rates are relatively high (i.e. populations are not enormously large and contain individually recognizable animals) and individuals can be reliably photographed without excessive effort. If animals have insufficiently discrete external marks to be visually distinguishable, conventional methods such as tagging may be more appropriate.
The method has been applied to a wide range of marine species (often, though not exclusively, large, mobile, and long-lived vertebrates) to assess population abundance, residency, migration pathways, life history parameters, and social structures (e.g. Graham & Roberts, 2007;Karczmarski, Würsig, Gailey, Larson, & Vanderlip, 2005;Smith et al., 1999;Würsig & Jefferson, 1990). Photo-ID methods can also allow application of mark-recapture techniques to estimate population size (Hammond, 1986). In this case, more stringent assumptions apply, including the requirement that natural marks remain recognizable over time and have an approximately equal probability of being (re)sighted. If marks change appreciably over extended periods such that the animal is no longer recognizable, or if animals are unlikely to be recognizable in all but the best observational conditions, using such data as the basis for mark-recapture analyses would produce biased estimates of demographic parameters.
To date, most photo-ID studies in the marine environment have focused on marine mammals, which are regularly available at the surface to be recorded by visual observers (e.g. Baird et al., 2009;Langtimm et al., 2004;Würsig & Jefferson, 1990). There is, however, significant potential for applying this approach to other marine megafauna, including elasmobranchs (see Marshall & Pierce, 2012). Historically, most identification projects on elasmobranchs followed the protocols set out by early marine mammal work and used the dorsal fin of some species to identify individuals. This was successful in great white sharks (Carcharodon carcharias; Gubili et al., 2009) and basking sharks (Cetorhinus maximus; Gore, Frey, Ormond, Allan, & Gilkes, 2016). However, this restricts photo-ID to species that exhibit behaviour allowing regular sightings of the dorsal fin from the water surface. Following on from studies on marine mammals showing that markings on the skin or pelage can reliably be used to identify individuals of certain species (e.g. Paterson et al., 2013), the skin markings of clearly marked elasmobranch species such as whale sharks (Rhincodon typus; Arzoumanian et al., 2005;Graham & Roberts, 2007) have shown to reliably allow for the identification of individuals. This approach is now being used to study numerous other shark and ray species bearing visually identifiable markings (e.g. Bansemer & Bennett, 2008;Castro & Rosa, 2005;Dudgeon, Noad, & Lanyon, 2008;Klimley & Anderson, 1996;Marshall et al., 2011;Van Tienhoven, Den Hartog, Reijns, & Peddemors, 2007).
The ability to reliably identify elasmobranchs over time is a potentially significant tool to underpin or validate crucial assumptions about management approaches for these long-lived species, such as the establishment of marine protected areas (MPAs; e.g. Gormley et al., 2012;Schofield, Katselidis, Dimopoulos, & Pantis, 2008;Wilson, Reid, Grellier, Thompson, & Hammond, 2004). As such, understanding mark stability becomes very important. While markings in many species of marine mammals (e.g. seals) have been shown to be stable from pup stages through to adulthood, allowing them to be used for identification throughout the animal's life (Paterson et al., 2013) (Robbins & Fox, 2012).
Photo-ID, on its own, is non-invasive (Pauli, Whiteman, Riley, & Middleton, 2010) and thus avoids injury to animals during and post tagging through tag loss, fouling, and so on (Kohler & Turner, 2001).
The method has also become progressively more practical in marine environments as digital cameras, underwater housings, and computers have become cheaper, more capable, and more widely available. This also means that photographs can be collected by members of the general public through directed citizen science projects, potentially resulting in significant expansion of sampling effort and increased engagement with project outcomes by local stakeholders (Dickinson et al., 2012;Dickinson, Zuckerberg, & Bonter, 2010;Marshall & Pierce, 2012). However, the basic requirement for a clear view of the subject remains, and so the majority of studies use photographs of parts of the animal visible above the surface or taken in clear water. For animals in deep, turbid or fast-moving water, one potential solution is to attract animals to bait, so that they can be photographed in situ or even captured and briefly brought to the surface for photography (Dala-Corte, Moschetta, & Becker, 2016).
The flapper or common skate (Dipturus intermedius), previously considered part of the Dipturus batis species complex, is the largest member of the family Rajidae found in European shelf and slope waters (Dulvy et al., 2006;Griffiths et al., 2010;Iglésias, Toulhoat, & Sellos, 2010;Last, Weigmann & Yang, 2016). Originally widespread, the species is now classified as Critically Endangered by the International Union for Conservation of Nature, largely due to its low intrinsic population growth rate and high sensitivity to overfishing (Dulvy et al., 2006). Although effectively extirpated across much of its historical range (Brander, 1981;Jennings, Greenstreet, & Reynolds, 1999;Walker & Hislop, 1998), small populations persist in some areas, including along the western and northern coasts of Scotland (Dulvy & Reynolds, 2002). In recent years, increasing efforts have been put into conservation of these populations, including a 2009 landings ban for EU fishing vessels (although the species remains at risk from bycatch in multispecies trawl fisheries; Simpson & Sims, 2016).
One relict population of flapper skates occurs in inshore waters of western Scotland, centred on the Firth of Lorn (Figure 1). This area contains a number of deep basins (>100 m), where skates have long been, and continue to be, caught by recreational sea anglers. By 1975, concern about declining skate numbers led to the development of a tag-recapture programme aimed at the sea angler community , which generated considerable amount of data on skate movements and site fidelity (Little, 1995(Little, , 1997(Little, , 1998Neat et al., 2015). These data were pivotal in clarifying the significance of the area anglers and charter skippers (as described by Neat et al., 2015). Initially, skates were tagged using cattle ear tags on the trailing edge of the wing; later, anglers switched to Floy™ dart tags that were inserted in the skates' dorsal wing muscle. This tagging approach proved effective in generating distribution, behavioural, and movement data for flapper skates. Nonetheless, concerns have increasingly been raised over risks to tagged animals' health and/or behaviour through improper tag application, as well as the potential for tag damage or loss (Jepsen, Thorstad, Havn, & Lucas, 2015;Marshall & Pierce, 2012;Thorstad, Økland, & Heggberget, 2001). More recently, implanted PIT tag technology has been applied to address these concerns (Kohler & Turner, 2001). Presently, only a small number of trained local charter skippers undertake PIT tagging in Argyll, limiting numbers of skates thus tagged. Moreover, as a dedicated PIT scanner is required to record recapture events, recaptures of PIT-tagged skates by individual anglers may go unrecorded. For these reasons, there was a desire to develop alternative methods to monitor distribution, recurrence, and movement of individual skates within, and across the boundaries of, the MPA in conjunction with existing tagging programmes.
Flapper skates represent a potential suitable candidate for photo-ID studies for several reasons. Animals typically possess a dark brown upper (dorsal) side with lighter spots arranged in a variety of patterns (Neal & Pizzolla, 2006;Stehmann & Bürkel, 1984). Spots are generally distributed in a broadly bilaterally symmetrical pattern but vary widely in terms of placement, size, clarity, and overall density, suggesting that they could enable photo-ID of individual skates. In addition, the species is long-lived (~50 years; Du Buit, 1977, although this work describes "Raja batis" [D. batis] caught off France, which may in fact refer to D. flossada) and occurs predictably in particular areas; as a large predator, it is expected to occur at comparatively low densities, potentially allowing for reasonable recapture rates. Flapper skates typically inhabit deep waters (>50 m), where they cannot be readily observed visually by divers (Neat et al., 2015), which would ordinarily limit the utility of photo-ID approaches. During sea angling, however, hooked skates are often briefly brought aboard to allow safe hook removal and collection of size measurements before being released.
While aboard, skates are also often photographed by anglers and/or charter boat skippers. A database of such photographs, assuming sufficient quality, could thus provide a novel means of recording the presence of a greater number of individually identifiable skates across a wider area than can presently be achieved through tagging.
Initial studies suggested that photo-ID methods could provide a viable additional monitoring strategy for this species (Bradley, 2012;Cooper, 2012).
The aims of the present study were to: (i) confirm whether photo-  For the purposes of this study, a 'trip' was defined as a single day on which angling occurred and at least one photograph of one or more captured skates was taken. No data were available on the total number of trips when no skates were caught or where no photographs were taken; use of photograph date metadata as indicators of angling activity was assumed to provide a reasonable approximation of overall effort distribution. Based on logbook data, approximately 80, 50, and 100 dedicated skate angling trips were undertaken for the years 2014, 2015, and 2016 respectively. A total of 486 photographs were available for this study, all taken by the same person (RC). A small number (37) of photographs were taken during 2011-2013 using a digital Olympus™ S1030SW camera. From 2014 onward, the remaining 449 photographs were all collected using a Samsung™ mobile phone camera. All photographs were taken from the deck of the vessel at a distance of ≤2 m from the skate, but the height of the camera above the deck and the camera angle relative to the skates' anteroposterior body axis varied widely, although the skates were facing the camera in almost all cases. Photograph dimensions varied between 850 × 720 pixels and 2592 × 4608 pixels. Date and time were logged as metadata for each photograph.

| Validation
An ongoing SNH/Marine Scotland tagging programme provided an opportunity to test the validity of the photo-ID approach. During March 2016, 39 skates were caught, equipped with PIT tags, and photographed by SNH as part of an ongoing study on skate movements. These photographs and PIT tag records were subsequently matched to tag records and photographs collected later in 2016 during routine angling trips. Matching PIT tag codes would provide independent confirmation of recaptured skates' individual identities and allow an assessment of long-term spot pattern stability.

| Analysis
The 486 photographs were sorted into 373 unique capture events based on associated date and time metadata. Although the available photographs spanned a timeframe from August 2011 until October 2016, 92% of photos were taken from 2014 onward. Sixteen photos contained multiple (two or three) skates, each of which was considered to represent a separate capture event.
Photographs were taken using two cameras (one digital, one mobile phone camera) and under varying conditions, causing considerable variability in photograph quality due to different light conditions, parts of the skate being blocked by people, camera height above deck, angle of skate towards camera, and so on. Prior to matching, the best photograph of each capture event was selected and graded according to several basic parameters describing both picture and mark quality (Urian et al., 2015;Wilson, Hammond, & Thompson, 1999). Apart from blocking out peoples' faces for confidentiality purposes, no post-processing of photographs was undertaken. Photograph quality was assessed based on the following binary scale:

Poor quality
i. Photograph is not in focus.
ii. Photograph has insufficient resolution to reliably detect spot patterns.
iii. The skate is being held up or otherwise not lying flat on deck.
iv. Less than 50% of main body surface (excluding tail) is visible due to poor lighting conditions (shade and/or glare), attached sediment, obstruction by people or objects, and/or being photographed from a very low angle.
v. The skate is photographed from the back, preventing a clear view of spots around the head and leading edges of the fins.
vi. The skate is very small (<50 cm width), as it is not known at what age skates' spot patterns stabilize.

Good quality
i. Photograph is in focus.
ii. Photograph has sufficient resolution to reliably detect spot patterns.
iii. The skate is approximately flat on deck and photographed from the front or side (≤90°of anteroposterior body axis).
iv. Most (>50%) of main body surface (excluding tail) is visible, allowing spots to be observed clearly. Poor lighting conditions (shade and/or glare), attached sediment, and/or obstruction by people or objects may locally affect spot visibility.
Photographs were graded as "poor quality" on this scale if one or more of the listed criteria were noted. Poor-quality photographs were not used in the present analysis to minimize the risk of incorrect reidentification (Urian et al., 2015). Good-quality photographs were only identified as such if all the listed characteristics were observed.
Given that most photographs were taken with the skate facing the camera, matching efforts focused on spot patterns on the front of the body. Spot patterns around the head (inclucing the rostrum), the proximal part of the vertebral column, and the leading edge of the pectoral fins proved particularly useful for photo-ID. The presence and location of individual spots, linear aggregations, and spot clusters relative to skates' eyes, spiracles, tip of rostrum, anteroposterior body axis, and/or pectoral fin edge were used to confirm individual identities. Occasionally, spots on other parts of the body, notably the pelvic fins and the tail, could also be used, but these were considered of secondary importance. Examples of locations of spot patterns used for photo-ID in this study are provided in Figure 2a-d.
The vast majority of skates possessed clearly visible spot patterns spread across their entire dorsal surface. This meant that mark quality was typically sufficiently good to allow individual identification even if only a portion of a skate's dorsal surfaces was visible. Nevertheless, some skates displayed only very few spots, which might affect identification probability of such poorly marked animals. For the present study, a skate's mark quality was therefore also classified using the following binary scale:

Poorly marked
The skate possesses no or very few, small spots; the dorsal surface appears almost monochrome. McEachran & Konstantinou, 1996) and any permanent injuries (fin nicks, scars, etc., including those resulting from historical tagging efforts). Confirmed individuals were given an individual alphanumeric code (e.g. Di000001) for inclusion in a master database. Inter-sex differences in (re-) capture rates were assessed through χ 2 tests (Zar, 1999). Graphs underpinning results were created using the R package ggplot2 (Wickham, 2009 January. However, 57% of trips in these years occurred during the March-May period, which is the peak season for dedicated skate angling ( Figure 3a). Average catch rates (number of skates per trip) remained broadly consistent between years; there was, however, considerable variability within years (Figure 3b). Gender could be determined for 96% of skates photographed during 2014-2016; on average, 81% of these skates were females. A small subset of individuals was designated "gender unknown," pending better photographs of these individuals becoming available.

| Identification
Twenty-four of the 373 original capture events (approximately 6%) were excluded from further analysis based on poor-quality photographs, mainly due to poor lighting conditions, skates being lifted off the deck, and the photograph having been taken from behind the animal (Table 1). It is worth noting that several pictures of skates photographed from behind were of sufficient quality to allow retrospective successful matching with other good-quality photographs, but this could not be achieved consistently in all cases. Six capture events involved very small (<50 cm wingspan) skates, likely juveniles.
These were also excluded from further analysis because they were typically being lifted completely off the deck by anglers while being

| Spot pattern variability
Although most skates possessed many spots, the observed degree of variability between individual skate's patterns was considerable ( Figure 4). The size and clarity of individual spots also varied greatly within and between individuals. Although spot patterns were broadly bilaterally symmetrical in terms of locations of distinctive aggregations of spots, considerable variation between left and right sides was observed. Some skates' spots were clearly delineated with a sharp outer edge, while other spots were less distinct or presented a "doughnut" appearance with a dark centre surrounded by a lighter ring. Some skates also possessed highly distinctive dark spots. The often large number of unique spot patterns observable across skates' Skates were observed to gain and lose a variety of non-permanent marks over time. Such marks were superimposed over their permanent pigmentation patterns and were sometimes visually prominent.
Some skates had extensive longitudinal scars across one or both pectoral fins, potentially derived from predation attempts, intraspecific aggression or interactions with fisheries (Figure 7a,b). A few animals possessed extensive white or grey patches of unknown origin, which appeared to change appearance over time (Figure 7c,d). Many photographed skates suffered from ectoparasites, particularly skate immediate surrounding area were observed, their presence would be unlikely to obscure overall spot patterns, and so the potential for incorrect matches is minimal.
A discovery curve was plotted to explore the relationship between the detection rate of "new" individuals against an everincreasing number of capture events (Figure 8; Williams, Dawson, & Slooten, 1993). As shown in Figure 8, the detection rate of new individuals had slowed down over time but had not yet stabilized.
Since little is known about skate recruitment rates and movement patterns, these preliminary results should be treated with some  of life (e.g. Compagno, 1984;Wilson & Martin, 2003). Limited data from captive-bred barndoor skate (Dipturus laevis, a closely related species) suggest that hatchling skates might already possess some form of the adult spot pattern at birth (Parent, Pépin, Genet, Misserey, & Rojas, 2008), but further work is required to confirm the age at which flapper skates' spot patterns become fixed. Once established, the stability of these spot patterns over the life of the individual also needs to be assessed. The permanence of markings on elasmobranch skin is still poorly understood, and there are conflicting studies in the literature both supporting long-term stability (Anderson et al., 2011;Holmberg et al., 2009;Meekan et al., 2006) and short-term variation (Robbins & Fox, 2012). Our study suggested that spot patterns in flapper skates were stable over periods of at least 4 years; nonetheless, this apparent long-term stability of markings needs to be better understood before photo-ID can be fully relied upon as a long-term (i.e. decadal) monitoring tool for this species. The continuing combination of photo-ID work and identification tagging Photographs used in this study were taken for the benefit of customers and as circumstances allowed. As a result, they varied widely in terms of camera angle relative to the deck, ambient light levels, extent of light reflection or shadow across the skate, and how much of skates' dorsal surfaces was visible. This complicated attempts to match poor-quality photographs, which were therefore excluded from the present analysis. To improve the probability of a reliable match, multiple spot patterns were used across the dorsal surface area, such that a match could still be made even if part of the skate was obscured or poorly lit. However, further work is required to understand which areas are crucial for successful photo-ID.
While flapper skates possess many markings that can be used for identification, this is facilitated if pictures are taken in a particular manner: • Photographs should be in focus and taken at as high a resolution as possible.
• The skate should be photographed while lying flat on deck (i.e. not lifted or held up, to prevent spot pattern distortion through bending of fins or body).
• The photograph should be taken from as close to vertical (i.e. looking straight down onto the skate) as possible, to ensure that all spots are clearly visible and foreshortening of distant spots is avoided.
• The entire skate's dorsal surface should be clearly visible to maximize the number of spots that can be used for identification.
• Care should be taken that spot patterns are not obscured by people, obstacles, shadows, and reflections, where possible and practical.
• Efforts should be made to include the pelvic region in the photo-  collect good-quality photographs. There is, therefore, the potential to engage the wider Scottish sea-angling community in a dedicated collaborative citizen science project to photographically record captures of skates across and outside the Loch Sunart to the Sound of Jura MPA and compile these observations in a central catalogue.
The present photographic database is intended to form the basis of such a photo-ID catalogue of flapper skates in Argyll waters, which will also seek to incorporate current and historical images from other sources (e.g. scientific surveys). Such a catalogue will increase understanding of common skate abundance, distribution, and movement patterns in and around the MPA. It will also improve understanding of factors affecting individual skates' health, including prevalence and healing rates of dermal infections, injuries and scars. In future, collection of photographs for the photo-ID catalogue could also be accompanied by sampling of genetic tissue to increase understanding of population substructure (e.g. Barker, Nosal, Lewallen, & Burton, 2015;Griffiths et al., 2010;Wright & Bentzen, 1995). In this manner, photo-ID data will generate novel insights into flapper skate biology and abundance. Such data will complement existing monitoring approaches to assist in the conservation of this critically endangered species in Scottish waters and beyond. More broadly, this approach should also be considered for studying other elasmobranch and teleost species of conservation concern that are targeted by catch-and-release recreational angling programmes.

ACKNOWLEDGEMENTS
Financial support for this study was received from SNH through the

CONFLICT OF INTEREST
The authors are unaware of any existing conflicts of interest that would preclude or bias publication of the results contained in this manuscript.