Unoccupied aerial video (UAV) surveys as alternatives to BRUV surveys for monitoring elasmobranch species in coastal waters

Abstract

Effective conservation strategies are founded by baseline information on abundance and diversity estimates. Method choice can influence the success of baseline surveys as method performance is variable and needs to be selected based on habitat and taxa. Here, we assess the suitability of unoccupied aerial vehicle (UAV) surveys, specifically multi-rotor “drones”, and baited remote underwater video (BRUV) surveys in shallow-water habitats to quantify elasmobranch abundance and diversity in the Saudi Arabian central Red Sea. Our results show that the number of elasmobranchs h⁻¹ observed using UAV surveys exceeded that of BRUV surveys by two orders of magnitude, indicating that the increased spatial coverage of UAV surveys is beneficial for long-term monitoring projects. BRUV surveys detected a greater number of species within reef habitats, whereas UAV surveys detected a greater number of species within sandflat habitats, indicating the value of multi-method approaches for regional biodiversity studies. Here, we provide the first insight into elasmobranchs associated with sandflat habitats in Saudi Arabia, emphasising the importance of these habitats to stingrays and the need for further information on elasmobranch habitat use to better inform management and conservation efforts in the face of rapid coastal developments across the Red Sea.

Introduction

Many elasmobranchs (i.e. sharks, skates, and rays) have experienced precipitous population declines globally due to overfishing, habitat degradation, and climate change (Dulvy et al., 2021; Pacoureau et al., 2021). Species of conservation concern may be impacted by threats at varying temporal and spatial scales (Chin et al., 2010). Therefore, monitoring elasmobranch assemblages across habitat types is a key component to formulating conservation strategies that may mitigate further population declines. Shallow-water habitats (i.e. mangrove areas, sandflats, lagoonal fringing reef flats, etc.) in particular are known to serve as crucial elasmobranch nursery areas used for pupping, foraging, and refuge (Vaudo and Heithaus, 2009; Davy et al., 2015; Hoff, 2016; Martins et al., 2018, 2020). Despite their significance, historical baselines of elasmobranch abundance in shallow-water habitats are often limited (Lotze et al., 2006; Spaet et al., 2016; Ward-Paige and Worm, 2017; Ferretti et al., 2018). Accordingly, there is a pressing need to further implement survey efforts in these habitats to develop a better understanding of elasmobranch populations. Yet, it is unclear which method is best suited in these difficult to study areas, and novel technologies may be required to increase shallow-water elasmobranch survey effort.

Numerous studies have compared the biases and accuracy of various methods used to quantify elasmobranch abundance and diversity (e.g. Stat et al., 2019; Raoult et al., 2020; Wetz et al., 2020). Invasive methods, both fishery-dependent and fishery-independent, are commonly used to survey sharks and rays (Carrier et al., 2018). For example, in shallow-water surf-zones seine-nets provide comprehensive biodiversity estimates for target taxa, yet they also remove non-target fauna, and are deemed extractive (Shah Esmaeili et al., 2021). The use of non-invasive methods to investigate species of conservation concern, including sharks and rays, has the clear advantage of alleviating issues such as fishing in protected or sensitive habitats, and post-release mortality (White et al., 2013). Survey methods such as diver operated video surveys, Unoccupied Aerial Vehicle (UAV), and Baited Remote Underwater Video (BRUV) surveys can provide information on abundance, diversity, and habitat function without animal extraction. Amidst precipitous elasmobranch declines, non-invasive methodologies suited for shallow-water environments are vital tools for addressing current knowledge gaps in elasmobranch diversity, abundance, and distribution.

Consumer-grade UAV systems (i.e. multi-rotor “drones”) are a novel, non-invasive means to improve survey efforts in shallow-water habitats. Primary benefits of UAV surveys include accessing areas that may otherwise be difficult to reach and enabling wide spatial coverage (Hodgson et al., 2017; Brack et al., 2018). In coastal waters, UAV surveys have demonstrated their potential to provide unique opportunities to sample densities of elasmobranchs (Kiszka et al., 2016; Digiacomo et al., 2020; Tagliafico et al., 2020; Ayres et al., 2021; Butcher et al., 2021). Furthermore, UAVs can be programmed to fly semi-automated flight paths, which increase reproducibility, reduce human bias, and are low-risk to operators (Linchant et al., 2015; Kiszka and Heithaus, 2018; Schofield et al., 2019). Although the extent to which UAVs can be applied is limited by many factors, such as weather conditions, water visibility, user biases, and animal behaviour (Hodgson et al., 2013; Ferguson et al., 2018), the use of UAVs for conservation science in the marine environment has steadily increased (Schofield et al., 2019; Butcher et al., 2021). Yet, there is little research that has focused on comparing traditional survey methods, such as BRUV surveys, to the use of UAV surveys (Johnston et al., 2017; Colefax et al., 2018; Raoult et al., 2018; Kilfoil et al., 2020).

BRUV surveys are a minimally invasive, fishery-independent approach to collect ecological data. BRUV units consist of an unattended camera system pointed towards a selection of bait within a container and are typically placed on the seafloor or suspended midwater in pelagic zones. BRUV surveys are a scalable technique (i.e. units are adaptable in size) that can be used in shallow-water (<5 m) (Whitmarsh et al., 2014; Gilby et al., 2016; Jones et al., 2018; Kiggins et al., 2018; Grimmel et al., 2020); however, there is limited information on their relative performance in surveying elasmobranch species in shallow habitats (Shah Esmaeili et al., 2021). Specifically, there is no directly comparative information to assess if UAV surveys are an appropriate substitute for, or complement to, BRUV surveys in determining the abundance and diversity of elasmobranchs in shallow-water habitats.

In this study, commercially available drones and BRUV units were used to identify the abundance and diversity of elasmobranchs within shallow-water environments adjacent to coral reefs and mangrove-associated sandflats. Surveys were conducted in the Saudi Arabian central Red Sea, an area of notable conservation concern for elasmobranchs with little information on their abundance and distribution in shallow-water habitats (Spaet et al., 2011; Dulvy et al., 2014; Spaet et al., 2016; Spaet, 2019; MacNeil et al., 2020). The main objectives of this study were to (i) quantify and compare the species composition of elasmobranchs and their relative abundance in shallow reef and sandflat habitats; and (ii) compare the species composition and abundance quantified by BRUV and UAV surveys in shallow-water environments.

Material and methods

Study site

Saudi Arabia has the most extensive coastline of all Red Sea countries, including long fringing reef systems that are typically backed by shallow-water lagoons (Bird, 2010). We surveyed two primary habitat types in the eastern central Red Sea (Thuwal, Saudi Arabia; 22.2760°N 39.1123°E): coral reefs (or reef-associated areas, hereafter referred to as “reefs”) and sandflats (specifically, sandflats adjacent to coastal mangrove stands). Both habitat types surrounded a lagoonal system adjacent to shore in our study area (Figure 1). There are no formal protections in the survey area, however, due to its close proximity to King Abdullah University of Science and Technology and Al Qadimah Military Port, there is no access to the inner lagoon for local fishers, which protects it from fishing activities. In contrast, the outer lagoon reef edge, which extends from ∼160 to 1200 m offshore, is periodically utilised by artisanal fishers and is thus more susceptible to fishing.

Unmanned aerial vehicle surveys

Between 25 August and 13 November 2019, 16 UAV surveys (5 reef; 11 sandflat) were conducted using a DJI Phantom Pro 4 V1.2 with an iPad Pro (10.5-inch). For sandflat surveys, the drone was operated from shore, while reef surveys were conducted from a 6 m rigid inflatable boat. As depth has been shown to influence detection probability (Benavides, 2020), drone flights were conducted over water <2 m deep to increase detection probability. All drone flights were conducted at an altitude of 12 m to further increase detection probability and ensure identification of target species while retaining a broader spatial coverage. Minimum disturbance flight practices were exercised through optimal altitude testing and avoidance of sporadic flight movements (Hodgson and Koh, 2016). All surveys were conducted in the morning (0600 to 1100) to maintain comparability in results and to decrease sun glare in UAV footage. All flights were pre-programmed using the program Altizure (www.altizure.com) to create and replicate spatially explicit “lawn-mower” transects at a speed of 3 m s⁻¹ with a 50% overlap between transect strips. Each survey consisted of 12 transect strips in a 100 m by 400 m survey grid (4 ha), which required two flights per survey (see Supplementary Table S1). The integrated 20-megapixel camera (4K) was used to collect video data at a 15-degree angle from NADIR with a polarized lens to avoid water glare. The DJI Go 4 application (https://dji.com/goapp) was used for take-off and landing, as well as manual flight control in the event of an unexpected return or aborted flight was required (e.g. bird disturbance or low-flying coastguard helicopters).

Baited remote underwater video surveys

Between 29 April and 13 November 2019, 60 BRUV surveys were conducted (12 reef; 48 sandflat). BRUV units were constructed using PVC supports (see Supplementary Figures S1 and S2) for ease of walking the units to the desired depth from shore and small boat use. GoPro HERO6 cameras were mounted on the BRUV units at a height of 30 cm and used in wide-angle recording mode (1080p). A rectangular frying basket was mounted at a distance of 50 cm from the camera and used as a bait cage (7 × 12 × 22 cm). Approximately 1 kg of chopped Indian mackerel (Rastrelliger kanagurta) was used as standardised bait. Dive weights were used on each of the four corners of the BRUV unit for stability to ensure a consistent field of view. Three replicate BRUV units were constructed for simultaneous deployments.

Surveys were conducted in a spatially stratified random design. Three BRUV units were deployed simultaneously at a minimum distance of 150 m from each other. Reef-associated and shore-based deployments were conducted at 1–2 m depth. BRUV units were dropped from the boat using a rope attached to the surface marker buoy, with the camera and bait cage oriented away from the reef. Ninety minutes was used as standardised deployment time. All surveys were conducted between 0600 and 1100 hours. Deployment depth, time, and bait were identical between boat and shore-based deployments.

Data analysis—UAV surveys

Aerial footage was analysed at 0.25x speed independently by two experienced observers (A.J.M. and C.T.W). Species known to inhabit the study area were determined to have distinct disc and tail features allowing easy species identification. Species identity trials were hence deemed unnecessary (Hensel et al., 2018). To identify individuals, a suite of variables was recorded for each encounter (time, location relative to benthic features, species, body orientation or direction of travel, behaviour, and any distinguishing features). If two individuals of the same species were sighted in parallel transects, extreme care was taken to ensure they were not the same individual by using identifiable features ( e.g. missing tail, relative size, etc.). If there was doubt of distinct individuals, only one individual was counted to maintain conservative abundance estimates. In some cases where species ID was not possible (i.e. due to excavation plumes), individuals were labelled as “unknown”. Discrepancies in total count data between observers were minimal (n = 6/206), indicating consistent and reliable results between readers during UAV footage annotation. Nevertheless, discrepancies between the observers were revisited by both observers to resolve differences in either species or count.

The sum of individuals sighted was converted to individuals per unit time (elasmobranchs h⁻¹) for each survey by dividing the total elasmobranch count by the flight time in hours so that results were more comparable to BRUV surveys. Elasmobranch density was derived for each habitat type by dividing the total number of individuals per unit area (elasmobranchs ha⁻¹). UAV footage was removed post-hoc if surveys showed deterioration in visibility (i.e. sea-state, turbidity) during the deployment and the full survey area could not be appropriately assessed.

Data analysis—BRUV surveys

BRUV footage was analysed in 1.0x speed. Two independent observers (A.J.M. and C.T.W.) annotated the arrival time, species, sex (where possible), unique identifying features (e.g. spot patterns, injuries), and body to tail length ratios of each elasmobranch that came into the field of view. Numbers of elasmobranch sightings were expected to be low in this region (Spaet et al., 2016); therefore, we used the maximum number of individuals (MaxIND) per BRUV deployment and between simultaneous deployments, to avoid double-counting individuals. MaxIND reflects the total number of uniquely identifiable individuals throughout the deployment (Sherman et al., 2018). Individuals of common stingray species (blue-spotted lagoon ray Taeniura lymma and whipray species Himantura spp.) were identified by unique spot patterns alongside other key features such as sex (i.e. presence of claspers), damage to disc/tail, and body to tail length ratios using ImageJ, version 1.8.0 (Schneider et al., 2012). During footage annotation of simultaneous BRUV deployments, there were no instances in which individual rays were documented in multiple deployments using MaxIND, thus confirming independence. Footage was excluded from the analysis if the visibility deteriorated over the course of a deployment (Lowry et al., 2012). BRUV sightings per unit time (elasmobranchs h⁻¹) was calculated using MaxIND per soak hour.

Statistical analysis

All statistical analyses were conducted using R version 4.0.3 (R Core Team, 2020). An alpha significance value p-value of ≥0.05 was utilised across all analyses. Shapiro–Wilk and Filgner–Killeen tests were used to assess the assumptions required for parametric tests through significance testing. Kruskal–Wallis testing was conducted to determine if there were significant differences in abundance between habitats per method. Additionally, the relationship between stingray abundance and tide height were explored using the Kruskal–Wallis test to account for the possibility of tidally influenced movement patterns (i.e. Brinton and Curran, 2017; Kanno et al., 2019; Ruiz-García et al., 2020). Tide height information was extracted from tidal charts developed for Rabigh, Saudi Arabia (Saudi Aramco, 2019). No other environmental variables were investigated.

Relative abundance indices (i.e. sighting frequency) were calculated for each species per method and habitat as the number of sightings divided by deployment/flight time in hours. The mean and standard error of relative abundance indices was calculated by method and habitat, including deployments/flights with zero sightings (Carbone et al., 2001; Lauretta et al., 2013). Rarefaction curves were generated to compare species richness for each method using the vegan package (100 iterations in a randomised order) (Oksanen et al., 2019). Multivariate analysis was used to further compare the structure of assemblages characterised by UAV and BRUV surveys. Given the lack of shark observations, this analysis was conducted for myliobatiformes only. Elasmobranch sightings per unit time (elasmobranchs h⁻¹) of the respective methods were log-transformed [log(x + 1)] for each species per deployment. The assumption of homogeneous multivariate dispersion was evaluated between survey methods with the permutational distance-based test: PERMDISP (Anderson, 2006). Comparisons of assemblage structure derived from UAV and BRUV surveys were made using the permutational multivariate analysis of variance (PERMANOVA) (Anderson, 2017). Data were visualized with non-metric multidimensional scaling (NMDS) using Bray–Curtis distances (Bray and Curtis, 1957).

Results

Using both UAV and BRUV surveys, we observed a total of seven elasmobranch species (five rays, two sharks) belonging to five families (Figure 2). Due to cryptic speciation within the Himantura complex and some unresolved range distributions, we grouped all observations of Himantura spp. to the genus level, however, they are most likely reticulate whiprays Himantura uarnak (Borsa et al., 2021). UAV surveys identified four species: two species over reef habitats (T. lymma, spotted eagle ray Aetobatus narinari) and four over the sandflat habitats (T. lymma, Himantura spp., cowtail ray Pastinachus sephen, A. narinari). BRUV surveys identified five species: four species over reef habitats (T. lymma, Himantura spp., mangrove whipray Urogymnus granulatus, sicklefin lemon shark Negaprion acutidens) and three over sandflat habitats (T. lymma, H. uarnak, tawny nurse shark Nebrius ferrugineus). Only two species (T. lymma, Himantura spp.) were recorded using both methods. Two species (P. sephen, A. narinari) were only observed by UAV surveys, and three species (N. ferrugineus, N. acutidens, U. granulatus) were only observed by BRUV surveys.

UAV surveys recorded a total of 206 rays (belonging to four species and two families) and zero sharks. During each deployment, UAV surveys recorded at least one ray, with an average of 13 ± 12 rays per deployment (minimum = 1, maximum = 41). Significantly more rays were recorded in UAV surveys over sandflat habitats compared to coral reef habitats (total survey n = 193 and 13, respectively; Kruskal–Wallis: χ² = 6.82; df = 1; p = 0.009).

From 60 BRUV surveys, 17 recorded elasmobranchs. Two sharks (N. ferrugineus, N. acutidens) and 23 rays (U. granulatus, Himantura spp., T. lymma) were identified using MaxIND. There were no discrepancies in MaxIND values between observers. There were no significant differences in the number of elasmobranchs observed using BRUV surveys between habitats (Kruskal–Wallis: χ² = 2.38; df = 1; p = 0.12).

Species accumulation curves suggest that UAV and BRUV surveys yield similar levels of diversity (Figure 3). Species accumulation curves for BRUV surveys show a higher species diversity in reef habitats, whereas the opposite was true for UAV surveys in sandflat habitats. Species accumulation curves from sandflat habitats (both UAV and BRUV) appear asymptotic, suggesting that it is unlikely that more species from this community would have been be detected without a significantly greater sampling effort using either of these methods. Species accumulation curves from reef habitats appear linear; however, this may be a bias from single observations of distinct species, an un-equal sampling across habitat types, and overall small sample sizes (Heck et al., 1975).

UAV and BRUV surveys characterised different myliobatiform assemblages (PERMANOVA: F_{(1, 29)} = 7.73, p < 0.001). Between survey methods, the magnitude of variation within assemblages was similar (PERMDISP: F_{(1, 29)} = 3.10, p = 0.09) enabling valid multivariate comparisons. NMDS revealed distinct, largely non-overlapping clustering between assemblages derived from UAV and BRUV (Figure 4).

Elasmobranchs h⁻¹ values were significantly higher in UAV surveys compared to BRUV surveys with total abundances of 22.07 ± 5.26 and 0.28 ± 0.07 elasmobranchs h⁻¹, respectively (Kruskal–Wallis: χ² = 44.91; df = 1; p < 0.001). Within UAV observations there was a higher number of individuals in sandflat habitats (sandflat: 30.08 ± 6.31 elasmobranchs h⁻¹; reef: 4.46 ± 0.87 elasmobranchs h⁻¹), whereas within BRUV observations more individuals were detected in reef habitats (reef: 0.53 ± 0.02 elasmobranchs h⁻¹; sandflat: 0.004 ± 0.001 elasmobranchs h⁻¹). Similar trends in the relative abundance indices for T. lymma and Himantura spp. were observed between methods in the sandflat environment (Figure 5). There were no significant differences in stingray abundance with tide height in UAV (reef: p = 0.41; sandflat: p = 0.44) or BRUV (reef: p = 0.91; sandflat: p = 0.47) surveys.

Discussion

This study presents the first direct comparison of UAV and BRUV surveys in monitoring elasmobranch abundance and diversity in shallow, nearshore habitats. Further, it provides novel insights on elasmobranch populations in the historically overexploited and understudied Red Sea (Spaet et al., 2011; 2016; Berumen et al., 2013; Spaet and Berumen, 2015). UAV and BRUV surveys varied in their detection rates between habitats in terms of abundance and species richness. UAV surveys showed higher species richness and abundance in sandflat compared to reef habitats, whereas the opposite was true for BRUV surveys. Overall, multivariate analysis revealed that UAV and BRUV surveys characterised dissimilar myliobatiform assemblages when incorporating indices of both diversity and abundance. We were able to observe 11% of the 64 elasmobranch species known to inhabit the area, showing that the combined performance of UAV and BRUV was better than individual method performance (6% UAV c.f. 8% BRUV) (Bonfil and Abdallah, 2004; Spaet et al., 2016; Golani and Fricke, 2018). Given the lack of overlap in the observed community composition, elasmobranch monitoring and biodiversity studies could benefit from the complementary nature of UAV and BRUV integration.

Employing more than one survey type is not always possible; therefore, it is important to understand the strengths and limitations of each survey method. UAV surveys can cover more ground in a short amount of time, which can provide rapid information (often in near real time) of a larger area. This provides highly valuable data that may fill spatial and temporal data gaps which could complement the in-water observations that BRUV provide. BRUV surveys require greater field-sampling durations regarding deployment, retrieval, as well as the time set to soak, which can range from 60–120 min (Currey-Randall et al., 2020). Although multiple cameras can be deployed at the same time to provide a high number of BRUV replicates (Cappo et al., 2004), a much greater sampling effort would be required to achieve a similar number of sightings to that made by UAV surveys. UAV surveys present a cheaper alternative to sampling larger areas in terms of cost per unit area covered. The initial cost of purchasing BRUV structures may be lower than the single upfront cost of a drone, however, the cost associated with BRUV deployments may be more expensive over time as bait and boat time is required for each deployment. Despite their individual shortcomings, both methods gather large amounts of data (e.g. species abundances, diversity, habitat quality, etc.) that can provide long-term, if not permanent, records for future research needs.

Despite their financial and logistical differences, the suitability of the respective method also varies depending on the research question, regional logistics, habitat, and target species (Raoult et al., 2020). For example, UAV surveys are particularly useful for baseline studies in habitats that are remote or considered fragile, or where there is a lack of resources to get to study sites (e.g. boats). UAV surveys may also be beneficial for some regions of the world in which it is difficult to obtain boat permits, such as Saudi Arabia, yet in other locations obtaining drone permits may be even more challenging. Estimating detection probability using UAV surveys may be affected by habitat complexity or extreme animal movements (Edwards et al., 2021). In the present study, we assumed that UAV detection was 100% given that surveys were conducted in <2 m depth, yet there are many factors that could influence detection estimates outside of depth, such as wind speed and time of deployment (Benavides, 2020). Specifically, UAV surveys are more likely than BRUV surveys to miss individual animals if they are buried or under/within structure (e.g. rays), as they spend less time in one space to allow for detection and do not use bait to attract animals. Moreover, as depth increases individuals will be overlooked entirely by UAV surveys. Detectability trails (i.e. deployment of objects that mimic the study species with known locations) may be required to fully understand what UAV surveys can measure in specific habitats and would facilitate a more comprehensive comparison of UAV and BRUV surveys in terms of detection probabilities (Benavides et al., 2020). Furthermore, UAV surveys have the capacity to increase spatial coverage, yet the data derived from BRUV surveys often have higher spatial resolution (i.e. sampling distance to target species), which facilitates greater confidence in species identification. The lower spatial resolution of UAV surveys could introduce misclassification errors, where species are either misidentified or double-counted (i.e. single individual counted twice or multiple individuals counted as one) (Brack et al., 2018; Edwards et al., 2021). The ability to use “MaxIND” in BRUV surveys may aid in reducing misclassification errors owing to the greater spatial resolution to distinguish between conspecifics, but does not fully remove the potential for double-counting individuals (Sherman et al., 2018).

Although direct methodological comparisons are often difficult owing to their respective and fundamental biases, preliminary evaluations between them can be made using reasonably available means. For example, comparisons of UAV and boat-based surveys to assess turtle breeding dynamics show that UAV surveys more than doubled the number of sightings per minute effort, yet had a lower success rate of identifying individual sexes (Yaney-Keller et al., 2021). Comparisons of reef fish communities made between BRUV and diver-operated video (DOV) surveys show that BRUV surveys can be more cost-effective (Langlois et al., 2010), and provide greater estimates of relative biomass for carnivores and general species richness (Langlois et al., 2010; Andradi-Brown et al., 2016). Alternatively, a similar comparison of BRUV surveys and underwater visual census (UVC) showed that although UVC provides greater species richness, BRUV surveys record proportionally more mobile predators (Colton and Swearer, 2010; Lowry et al., 2012). The direct comparison of UAV and BRUV surveys, however, is inherently challenging owing to the differences in spatial and temporal coverage of the two methods. As such, UAV surveys have the capability of covering large areas in a relatively short amount of time, while BRUV surveys cover smaller areas over longer periods of time. Within this context, comparative assessments should be based on the temporal and spatial resolution of the gear types. Previous, exploratory comparisons of UAV and BRUV surveys in relation to black-tip reef shark (Carcharhinus melanopterus) abundance did not find significant differences between the two technologies when space and time were standardised (Cushenan, 2019). In the present study, UAV surveys were used over a greater spatial area than BRUV surveys, resulting in over two orders of magnitude greater “sightings per unit time” (i.e. counts over time). The increased number of sightings was to be expected when sampling larger areas, yet this result further supports UAV surveys as a time-efficient tool in quantifying elasmobranch abundance in shallow-water habitats. As the flying speed of the drones may also influence the number of encountered individuals over time, survey standardisation requires additional quantification of spatial differences between UAV and BRUV surveys. Although the area covered by UAV surveys is known, both the visual area covered by each BRUV unit and the “gear affected area” (i.e. the size of the bait plume used to attract elasmobranchs to the BRUV) is unknown. Future spatial comparisons between UAV and BRUV surveys should consider integrating correction factors that could be derived from characterising the specific hydrodynamics of the study site (i.e. current velocity and direction; Taylor et al., 2013) and the resulting bait-plume predictions. Quantifying bait-plume coverage was not logistically feasible in the current study; however, this may be an imperative component to making direct comparisons of UAV and BRUV surveys in the future.

In line with previous research, UAV surveys in the present study observed higher stingray abundances over sandflat habitats compared to reef habitats (Kiszka et al., 2016). The most common species recorded in this study was T. lymma, a species that is elsewhere rarely documented in such abundances far from coral reefs (e.g. Australia) (Chin et al., 2010; O'Shea et al., 2012; Last et al., 2016). The higher abundances of T. lymma in sandflat habitats may be driven, at least in part, by reduced predation risk provided by inhabiting shallow-water (Cartamil et al., 2003; Vaudo and Heithaus, 2013; Davy et al., 2015; Kanno et al., 2019). Stingrays are known to exhibit tidally influenced movements (Brinton and Curran, 2017; Kanno et al., 2019; Ruiz-García et al., 2020). However, the lack of variance in stingray abundance across tidal heights suggests that tide may have little influence on the observed patterns. Alternatively, the observed abundances could be related to mesopredator release, where low predator abundance in the surrounding waters allows for their population increase (Spaet et al., 2016; Sherman et al., 2020). Although stingrays were observed in BRUV surveys, sightings were rare compared to UAV surveys. Given that UAV surveys cover much larger areas than BRUV surveys, there is a greater opportunity for UAV to observe stingrays in the sandflats as they use this open habitat to feed. Overall, it is the dorso-ventral compression of myliobatiformes (i.e. larger surface area) that facilitates the conspicuous detection of individuals when using UAV surveys. In the current study, stingray species encountered were distinguishable through disc shape alone. Prior knowledge of the species inhabiting the study site is imperative as some species complexes (e.g. Himantura spp.; Borsa et al., 2021) or morphologically similar groups of species (e.g. Aetobatus spp.) may not be as distinguishable though UAV surveys. In instances where foraging stingrays are excavating, the resultant sediment plume can mask individuals and further decrease identification certainty to the species level. Despite their morphological similarities, myliobatiformes are suitable candidates for drone research (Kiszka et al., 2016; Hensel et al., 2018; Colefax et al., 2019; Frixione et al., 2020; Oleksyn et al., 2020; Tagliafico et al., 2020), particularly in sandflat habitats (see Supplementary Figure S3).

The present study provides the first characterisation of shallow-water (<2 m) elasmobranch assemblages in the historically overexploited and understudied Red Sea (Spaet et al., 2011, 2016; Berumen et al., 2013; Spaet and Berumen, 2015). Despite their importance in artisanal fisheries and concerning declines (Spaet, 2019; Dulvy et al., 2021), there is an apparent lack of data on the abundance, distribution, and habitat use of elasmobranchs in the Red Sea. Without this information, changes to elasmobranch dynamics in shallow-water habitats in the presence of growing anthropogenic pressures could go unnoticed. While we only provide information from a single location, the Saudi Arabian Red Sea hosts vast areas of shallow-water habitats with unique fish communities, such as the Farasan Islands in the southern Red Sea (Roberts et al., 1992; Coker et al., 2018; Berumen et al., 2019) and the Al Wajh banks in the northern Red Sea (Atta et al., 2019; Berumen et al., 2019). Notably, Saudi Arabia is undertaking major coastal developments on the Red Sea (Saudi Arabian Public Investment Fund, 2021), including in the Al Wajh banks. Although, these coastal developments intend to proceed with the highest environmental standards (Cziesielski et al., 2021), they could benefit from a combination of UAV and BRUV technology to more adequately understand shark and ray populations utilising the sensitive coastal habitats in their project areas and develop a more robust understanding of this system. Even though our study was limited in the number of deployments, we believe that our results reflect the low abundance of sharks regionally (Spaet et al., 2016; Spaet, 2019).

This study provides a preliminary comparison of UAV and BRUV surveys while assessing elasmobranch abundance and diversity in shallow-water habitats. Our results indicate that method selection is fundamentally dependent on the research question. Those that wish to study specific taxa should select their method based on the size of the survey area and habitat, while studies that aim to maximise species check-lists or survey entire elasmobranch assemblages should consider employing a multi-method approach where possible. For example, studies that wish to explore reef-based diversity should continue to employ BRUV surveys to allow more cryptic elasmobranch species to be detected. On the other hand, studies that wish to quantify stingray abundance in sandflat habitats should primarily consider UAV surveys as they can cover larger areas in less time and are less intrusive than BRUV surveys. In conclusion, our results emphasise the importance of Saudi Arabia's sandflat habitats to stingrays and the need for further information on elasmobranch habitat use to better inform management and conservation efforts in the face of rapid coastal developments across the Red Sea.

Funding

This research was supported by King Abdullah University of Science and Technology (baseline research funds to M.L.B).

Authors’ contributions

AJM: conceptualisation, data curation, formal analysis, investigation, methodology, visualisation, and writing—original draft, review, and editing. JLYS: methodology, supervision, and writing—review and editing. CTW: data curation, investigation, and writing—review and editing. MLB: funding acquisition, supervision, and writing—review and editing.

Conflict of interest statement

The authors declare no competing interests.

Data availability statement

The data underlying this article will be shared on reasonable request to the corresponding author.

Acknowledgements

The authors are grateful to the members of Reef Ecology Lab for their assistance, notably Alexander Kattan, Walter A. Rich, and Charlotte Young for their assistance in the field; Royale Hardenstine for initial discussions; and to Ute Langer, GIS Specialist, for preparing the map. Three anonymous reviewers are thanked for their valuable comments.

References