Regional movements of satellite‐tagged whale sharks Rhincodon typus in the Gulf of Aden

Abstract To gain insight into whale shark (Rhincodon typus) movement patterns in the Western Indian Ocean, we deployed eight pop‐up satellite tags at an aggregation site in the Arta Bay region of the Gulf of Tadjoura, Djibouti in the winter months of 2012, 2016, and 2017. Tags revealed movements ranging from local‐scale around the Djibouti aggregation site, regional movements along the coastline of Somaliland, movements north into the Red Sea, and a large‐scale (>1,000 km) movement to the east coast of Somalia, outside of the Gulf of Aden. Vertical movement data revealed high occupation of the top ten meters of the water column, diel vertical movement patterns, and deep diving behavior. Long‐distance movements recorded both here and in previous studies suggest that connectivity between the whale sharks tagged at the Djibouti aggregation and other documented aggregations in the region are likely within annual timeframes. In addition, wide‐ranging movements through multiple nations, as well as the high use of surface waters recorded, likely exposes whale sharks in this region to several anthropogenic threats, including targeted and bycatch fisheries and ship‐strikes. Area‐based management approaches focusing on seasonal hotspots offer a way forward in the conservation of whale sharks in the Western Indian Ocean.

. In addition to undertaking large-scale movements (e.g., >7,000 km; Hueter et al., 2013) and crossing international boundaries, these filter-feeding sharks aggregate seasonally at numerous locations around the world (Sequeira et al., 2013). The reliable aggregation of whale sharks at coastal localities has facilitated the development of an increasingly valuable tourism industry in the past three decades, where divers and snorkelers are able to swim with and observe individuals (Anderson et al., 2014;Zimmerhackel et al., 2019). Despite the successes of these ecotourism ventures (Zimmerhackel et al., 2019), anthropogenic impacts, such as targeted fisheries catches, bycatch in nets, and vessel strikes, continue to jeopardize global whale shark populations (Pierce & Norman, 2016) and, as a consequence, the persistence of this tourism industry. Recent population declines have resulted in the upgrading of whale sharks to globally Endangered on the IUCN Red List of Threatened Species in 2016 (Pierce & Norman, 2016). Global efforts over the past decade to conserve this species have aimed to better understand its movement ecology (Andrzejaczek et al., 2016;McKinney et al., 2017); however, knowledge gaps still remain in understanding the movement patterns of whale sharks after departing their seasonal aggregations.
Photo-identification (photo-ID) and satellite tracking are two common approaches to investigating whale shark movements at local to cross-ocean scales (Andrzejaczek et al., 2016;Berumen et al., 2014;Pierce et al., 2018;Sequeira et al., 2013). Photo-ID uses unique natural markings to recognize individuals and explore residency and regional movement patterns .
While this may be an effective technique in regions such as the Western Central Atlantic Ocean where high tourism and directed research efforts exist , it may not be viable in areas where direct access to sharks is limited, particularly on remote coastlines and in offshore regions. In addition, these methods are largely limited to when sharks are in surface waters and can lead to misleading conclusions about seasonal habitat use if sharks remain present yet move into deeper, less accessible waters on a seasonal basis . Alternatively, satellite tracking is an effective approach that can enable identification of previously undetected habitat hotspots (Diamant et al., 2018;Robinson et al., 2017) and has the added benefit of recording detail about the vertical habitats and thermal environment encountered by a tagged individual.
Numerous aggregations of whale sharks have been identified in the Western Indian Ocean (WIO) including the Maldives, the Gulf of Oman, the Arabian Gulf, Djibouti, the Red Sea, the Seychelles, Tanzania, Mozambique, Madagascar and South Africa (Berumen et al., 2014;Brooks et al., 2010;Diamant et al., 2018;Perry et al., 2018;Pierce & Norman, 2016;Riley et al., 2010;Robinson et al., 2016;Rohner et al., 2015). Despite the remote nature of many of these sites, knowledge of regional movement patterns has been gained through satellite tracking. A total of 131 tracks have so far been collected and published from seven general localities spanning several thousand kilometers (Table 1). These tracks revealed that sharks predominantly remained within the region in which they were tagged; however, many whale sharks were also recorded to cross international boundaries (Berumen et al., 2014;Robinson et al., 2017;Rohner et al., 2018). Increased tagging efforts throughout the WIO are required to continue to build a regional picture of whale shark connectivity within the Indian Ocean and develop management strategies to ensure the long-term viability of those populations.
In the Gulf of Aden, reports of whale shark movement and habitat use have been limited to a seasonal aggregation in the Arta Bay region of the Gulf of Tadjoura, Djibouti (Figure 1).
This area is thought to serve as a feeding ground for juvenile whale sharks in the winter months (October to February), where individuals aggregate to forage upon the dense zooplankton and baitfish patches that result from upwelling following the south-west monsoon (Boldrocchi & Bettinetti, 2019;Boldrocchi et al., 2018;Rowat et al., , 2011. This aggregation was first formally described in 2004 and a photo-ID study identified upward of 290 individual whale sharks visiting the aggregation between 2003 and 2010, with individuals re-sighted up to six years after being first recorded there (Rowat et al., , 2011. Individuals were predominately male, and relatively smaller (mean size 3.7 m) than those observed at other aggregations in the Indian Ocean (Rowat et al., 2011). To date, the movement of whale sharks away from the Arta Bay aggregation site has been poorly documented, with records limited to a short (nine day) satellite track from a juvenile individual . As this aggregation site is situated close to one of the world's busiest shipping lanes, as well as countries, such as Somalia, with high estimated shark catch through artisanal fisheries (Cashion et al., 2018), these Endangered sharks may be exposed to a myriad of threats while moving throughout the Gulf of Aden. There is therefore a need to further understand the movements and habitat use of whale sharks in Djibouti, and the greater Gulf of Aden region. Given the remote and inaccessible nature of the region surrounding Arta Bay, further electronic tracking offers the best approach to uncovering movement patterns of whale sharks in the Gulf of Aden.
In this study, we describe regional movements and patterns of vertical habitat use of whale sharks on departure from the seasonal aggregation in the Arta Bay region. We discuss the likely drivers of these patterns, the potential for overlap with anthropogenic activities, and the conservation and management implications of our results.

| Study site and tag deployments
Satellite tags were deployed on whale sharks in the Arta Bay region of the Gulf of Tadjoura (11.57°N, 42.77°E; Figure 1) in January and/ or December in 2012 (n = 2), 2016 (n = 3) and 2017 (n = 3; Table 2). Sharks were visually located by boat-based searches from a 6 m long skiff with a single outboard engine and then approached slowly.
Sharks larger than 3.5 m were targeted (a) to satisfy a minimum size for tagging and (b) due to the assumption that only sharks of a certain size migrated from the Gulf of Tadjoura study site. Free-divers entered the water from the vessel to tag and measure sharks, as well as take photo-ID images. Tags were deployed by a pole-spear with a welded plate and rubber buffer to prevent insertion greater than 8-10 cm. 2012 tags were leadered with a ~15 cm length of 45 kg nylon filament covered with several layers of heat shrink tubing and attached via a titanium flat anchor M dart (Wildlife Computers) and placed at the base of the first dorsal fin, on the left side. 2016 and 2017 tags were connected to a large titanium anchor (Wildlife Computers) via a 50 cm stainless steel tether. Tether lengths were selected to allow the anchor to be placed 8-10 cm below the skin and leave space to let the tag lie flat against the body surface in the case of the former, and to facilitate breaking the air-surface barrier for transmission during deployment for the latter. Individual sharks were measured (total length, TL) by one of two methods; (1) by using visual observation and comparing the shark to an object of known size, and/or (2) by an intense photogrammetric laser measurement campaign using the methods as described in Jeffreys et al. (2013).
Photo-ID images were also taken of the left and right flanks of tagged individuals and matched with the existing Djibouti database using the public domain pattern-recognition software I 3 S (Interactive Individual Identification System; Van Tienhoven et al., 2007). All fieldwork was approved by and conducted with the knowledge of the Ministry of Environment, Djibouti and local authorities in Arta.
All procedures followed standard international guidelines for tagging whale sharks and staff were trained by experts in the field (D. Rowat and M. Meekan;Robinson et al., 2017;Wilson et al., 2006).
Location data and/or processed archived data were transmitted and retrieved through the Argos satellite system when the tags detached. Detachment of the tag from the shark was identified by a combination of near-continuous high quality Argos transmissions for the first few hours of each day and depth summaries from histograms consistent with surface records (Hearn et al., 2013). The tags deployed in 2016 and 2017 also transmitted data when sharks swam at the surface, and, in addition, housed a Fastloc global positioning system (GPS) for acquiring location information.

| Track reconstruction
A combination of techniques was used to estimate the most probable track for a given individual based on the type and quality of the data transmitted (Table 2).

| MiniPAT track reconstruction
The two individuals tagged in 2012 were fitted with MiniPAT tags which did not have the capability to record locations during deployment. Track locations were thus estimated by light-levels based on data received via Argos transmission after the tags had detached from the host shark; consequently, these were not contiguous data streams. The transmitted data were first processed through the Data Analysis Program software suite (WC-DAP, Wildlife Computers, Redmond, WA) to extract the dawn/dusk light level as well as the temperature and depth data for each 24-hr period. These data  However, rather than using unlikely speeds, the tracks were re-run in reverse, starting at the detachment location, and then the two tracks were combined and the most parsimonious daily locations were retained for track output.

| MK10 track reconstruction
The six satellite tags deployed on individuals in 2016 and 2017 were programmed to acquire a location estimate from ARGOS and GPS satellites while at the surface. Position estimates acquired from ARGOS satellites were provided with an associated error (Location Class 3: <250 m, 2:250-500 m, 1:500-1,500 m, 0: >1,500 m, A and B: not specified, www.argos -system.org), and Fastloc GPS positions were expected to have an error of <100 m (Bryant, 2007). For the deployment period, all locations reported from above sea level were removed, as well as a small number of locations with A and B error classes that were obviously erroneous (<1%), that is, they were well beyond the bounds of possible distances the shark could have travelled based on both earlier and later location estimates of higher accuracy for the track. More advanced filtering methods, similar to the ones used for the 2012 tracks, were attempted. However, none of the models converged when using the WC-GPE3 processor, and the Iknos Walker routine described above could not be applied to the data. These convergence issues were due to large gaps in light, SST, and location data (location and light data available for 9%-68% of tracking days for tracks more than six days in length), as well as the highly coastal nature of the tagged individuals, with locations being classified as "on land" in several cases in the 0.25° grids of GPE3.
For the longer tracks, all but one tag displayed transmission gaps >20 days (up to 86 days), preventing unbiased interpolation of tracks between consecutive locations (Queiroz et al., 2016). Attempts were also made to thin known locations from tracks in order to reduce clustering and facilitate model convergence as per Lipscombe et al. (2020); however, resulting track paths diverged significantly from known locations and were deemed unreliable.

| RE SULTS
Eight satellite tags were deployed on eight unique juvenile whale sharks in the Arta Bay region of the Gulf of Tadjoura, Djibouti, in January 2012, January and December 2016 and December 2017 (Table 2). Six males and two females were tagged, and individuals ranged in length 3.5-6.3 m (  Figure 1). For the tags deployed in 2012 which did not have the capacity to transmit locations during deployment, there were discrepancies between the first transmission time after release and depth records, which suggested that the tags may have floated at the surface for some time after release before transmitting location data. As a result, no Argos locations were assigned to the final estimated track end times for these two deployments ( Figure 2).

| Horizontal movements
Shark movements ranged from local-scale movements around the Djibouti aggregation site (<100 km), regional movements along the north-west coastline of Somaliland, a return offshore loop, movements north into the Red Sea, and a large-scale (>1,000 km) movement to the east coast of Somalia, outside of the Gulf of Aden week later ( Figure S1).

| Diving behavior and temperature
Time 10 m during the day and 64% at night, with this shifting to 58% and 57% for day and night, respectively, during its offshore loop.

| D ISCUSS I ON
Satellite telemetry revealed regional (<500 km) and large-scale

| Drivers of movement patterns
Following tag deployment in December and January, individual whale sharks remained in the vicinity of the Djibouti aggregation site for up to one month. Previous observations have indicated that these filterfeeders primarily use this site to forage on the dense aggregations of zooplankton that occur in shallow waters just off the shoreline from October to February (Rezzolla & Storai, 2010;. Sampling of the surface zooplankton community here has revealed an increasing trend in biomass from November to December, and a decrease from January to February (Boldrocchi et al., 2018) increased during offshore movements, with several dives to more than 1,000 m being recorded. Given available evidence both here and from previous studies, we hypothesize that such vertical behaviors may be driven by foraging at depth, thermoregulation and/or navigation. Shark 165699 showed a two-week return loop into the Gulf of Aden, which we hypothesize was motivated by foraging. Previous tracking studies have linked deep dives to foraging on meso-and bathypelagic layers (Brunnschweiler et al., 2009;Tyminski et al., 2015), and signature fatty acid analysis has suggested that whale sharks from both Ningaloo Reef, Western Australia and Mozambique, attain a significant component of their diet from waters greater than 200 m deep (Marcus et al., 2016;Rohner et al., 2013). As tropical waters tend to be oligotrophic with patchy distributions of prey, declines in prey abundance on the coast may have driven this shark offshore to forage at depth. High use of surface waters (<2 m) may have been a strategy to maintain a preferred body temperature for this ectothermic shark (Thums et al., 2013), following deep dives where ambient temperatures reached a minimum of 4°C at depth. Similarly, the whale shark moving out of the Gulf of Aden also made several dives >1,000 m with high use of the surface ten meters. Alternating between deep and shallow waters may alternatively be a navigational strategy in this case, attaining light and/or celestial cues at the surface (Carey et al., 1990;Lohmann et al., 2008), and magnetic cues by detecting gradients in local field intensity with depth (Klimley, 1993).

Such cues have been linked with navigation in migrating white sharks
Carcharodon carcharias, where tracked sharks primarily occupied the surface two meters of the water column, interspersed with deep dives to >300 m (Bonfil et al., 2010;Weng et al., 2007). Future deployments of high resolution multi-sensor biologging tags with triaxial sensors would provide further insight into the drivers of these vertical movement patterns (Andrzejaczek et al., 2019;Gleiss et al., 2011Gleiss et al., , 2013.
Inter-individual variation in DVM patterns may also be a function of the local distribution and behavior of zooplanktonic prey.
Of the six individuals with depth data, four recorded patterns of reverse DVM, one of normal DVM, and one with no apparent diel differences in depth use. Notably, opposite patterns were displayed by the two individuals tagged in 2012, with one displaying normal DVM and the other reverse DVM, despite both heading north and into the Red Sea at the same time of year. Additionally, one shark displayed reverse DVM patterns while on the coast, and no diel difference while offshore for two weeks. Reverse DVM patterns have frequently been reported for whale sharks at coastal aggregation sites (Brunnschweiler et al., 2009;Gleiss et al., 2013;Robinson et al., 2017;Rowat & Gore, 2007), with some individuals switching to normal DVM (Brunnschweiler et al., 2009) or no diel pattern  on departure from these locations. As DVM patterns have traditionally been explained by the movement of zooplankton to deeper, cooler, and darker waters during the day to reduce detection by visual predators (Hays, 2003), other processes must be driving these reverse behaviors. Local-scale oceanographic phenomena, for example, could be triggering high productivity and therefore high zooplankton concentrations at the surface during certain diel periods, as is the case for whale sharks foraging in surface waters during the day off the Yucatan Peninsula in Mexico (Motta et al., 2010). Similar to the contrasting patterns recorded in our study, basking sharks (Cetorhinus maximus)-another large, filter-feeding elasmobranch-tracked in the English Channel, displayed patterns of normal DVM in deep, well-stratified waters and reverse DVM in shallow, inner-shelf areas near thermal fronts, with such changes in movement related to those of the local zooplankton community (Sims et al., 2005). Alternatively, diel changes in vertical movement patterns of whale sharks may be related to thermoregulatory behaviors (Thums et al., 2013); however, given the lack of temperature change recorded in the coastal habitats occupied here, this seems unlikely.

| Connectivity to other known aggregations
Long-distance movements recorded both here and in previous studies suggest that connections between the whale sharks tagged at the Djibouti aggregation and other documented aggregations are likely.
Roughly 10%  ity are thought to be the major driver of movements to and from aggregation sites; however, it is also possible that such long-distance movements might reflect other aspects of the whale shark's life history Sequeira et al., 2013). For the latter, we are yet to attain satisfactory sample sizes among a representative cross-section of the population to assess this hypothesis in the Western Indian Ocean. To date, available data from satellite tags deployed in this region have enabled long-distance movements to be recorded, demonstrating that whale sharks are capable of moving between aggregation sites (Figure 4). Direct evidence of such links, however, has been limited by satellite tag attachment durations. To enhance the probability of recording such links, satellite tags should be deployed toward the end of the peak aggregation season (i.e., February at the Djibouti site) in order to increase attachment time throughout migratory periods (Sequeira et al., 2013). currents (Zajonz et al., 2016), however, we cannot rule out the possibility that the tags coincidentally drifted into these urban areas, or alternatively that just the tag itself was landed. In either scenario, to improve conservation practices, this whale shark population needs to be managed as a single unit, irrespective of jurisdictional boundaries, throughout its migration cycle (Lascelles et al., 2014). However, as protecting the entire area of whale shark habitat use in the Gulf of Aden region is politically unrealistic and impractical, area-based conservation approaches focusing on seasonal hotspots and/or regular migratory behavior are likely to be a more constructive way forward (Germanov & Marshall, 2014). Future research should endeavor to refine the current understanding of the patterns of movement of whale sharks throughout their annual migratory cycle in the Gulf of Aden, in order to identify key areas of habitat use and to strengthen the design and implementation of management strategies.

| Management and conservation implications
In addition to the threat of capture by fisheries, whale sharks in the Gulf of Aden are also exposed to the threat of ship strike in one of the world's busiest shipping lanes, as well as boat strike by smaller vessels in coastal areas. Similar to results reported elsewhere, whale sharks tracked from Djibouti spent a high proportion of time in surface waters (<2 m) (Motta et al., 2010;Robinson et al., 2017;Thums et al., 2013;Tyminski et al., 2015). At least two individuals tagged in this study made offshore excursions and spent a high proportion of time in surface waters during these ventures. One individual (shark 165699) spent 46% of the time shallower than two meters while in the central Gulf of Aden, increasing its susceptibility to ship strike.
Within the Arta Bay region of Djibouti, 15 of 23 individuals observed over a five-day period had scarring attributable to boat or propeller strikes . Management at both domestic and international levels will be required to reduce the impact of these anthropogenic threats and could involve initiatives such as establishing no-go areas and/or reduced speed limits in important migratory corridors and foraging hotspots and limiting the expansion of marine roads (Pirotta et al., 2019).

| CON CLUS IONS
Our study used satellite tags to reveal patterns of habitat use of whale sharks in the Gulf of Aden, as well as potential connectivity between an aggregation in Djibouti and previously described aggregations in the Western Indian Ocean. The broad horizontal distribution and vertical niche of these sharks in the Gulf of Aden expose them to fishing and shipping activities, threatening the viability of this population. The information collected both here and in previous studies in the region, in combination with continued research efforts, should be used to inform conservation and management strategies at both domestic and international levels.

ACK N OWLED G M ENTS
Whale shark tagging in Djibouti was undertaken by staff and volunteers of Megaptera and the Marine Conservation Society Seychelles. Activities were supported and approved by the Ministry of Environment, Djibouti and local authorities in Arta. We thank EXAGONE TERIA for its financial support, Dolphin Excursions, and Sovereign Global Solutions for their logistical support and in particular the crew that made each trip a success. We also thank our partners in Djibouti: NGO DECAN, Dr Bertrand Lafrance, Husain Al Qallaf from Senyar Dive Team in Kuwait, the Djiboutian authorities and the Arta disctrict Préfet and local communities. Thanks to Heather Baer from Wildlife Computers for support with track processing; and David Obura from CORDIO and Ziad Samaha from the IUCN for information regarding marine protected areas in Djibouti and Yemen respectively. We are also grateful for the critical comments from Jesse Cochran and one anonymous reviewer that improved the draft manuscript. SA thanks Lauren Peel for valuable feedback on an earlier version of this manuscript.

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest. Funding acquisition (equal); Investigation (supporting); Methodology (supporting); Project administration (lead); Supervision (lead).

E TH I C S S TATEM ENT
Fieldwork for this study was conducted in Djibouti with the support and approval of the Ministry of Environment, Djibouti, and local authorities in Arta. All tagging procedures followed standard international guidelines for tagging whale sharks and research staff were trained by experts in the field (Dr Rowat and Dr Mark Meekan).

DATA AVA I L A B I L I T Y S TAT E M E N T
Raw tagging data are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.fqz61 2js1. Photo-IDs have been submitted to whaleshark.org.