ArticlePDF Available

Abstract and Figures

Over the past decade, drones have become a popular tool for wildlife management and research. Drones have shown significant value for animals that were often difficult or dangerous to study using traditional survey methods. In the past five years drone technology has become commonplace for shark research with their use above, and more recently, below the water helping to minimise knowledge gaps about these cryptic species. Drones have enhanced our understanding of shark behaviour and are critically important tools, not only due to the importance and conservation of the animals in the ecosystem, but to also help minimise dangerous encounters with humans. To provide some guidance for their future use in relation to sharks, this review provides an overview of how drones are currently used with critical context for shark monitoring. We show how drones have been used to fill knowledge gaps around fundamental shark behaviours or movements, social interactions, and predation across multiple species and scenarios. We further detail the advancement in technology across sensors, automation, and artificial intelligence that are improving our abilities in data collection and analysis and opening opportunities for shark-related beach safety. An investigation of the shark-based research potential for underwater drones (ROV/AUV) is also provided. Finally, this review provides baseline observations that have been pioneered for shark research and recommendations for how drones might be used to enhance our knowledge in the future.
Content may be subject to copyright.
drones
Review
The Drone Revolution of Shark Science: A Review
Paul A. Butcher 1, 2, * , Andrew P. Colefax 3, Robert A. Gorkin III 4, Stephen M. Kajiura 5, Naima A. López 6,
Johann Mourier 7, Cormac R. Purcell 8,9, Gregory B. Skomal 10, James P. Tucker 2, Andrew J. Walsh 3,8,
Jane E. Williamson 11 and Vincent Raoult 12


Citation: Butcher, P.A.; Colefax, A.P.;
Gorkin, R.A., III; Kajiura, S.M.; López,
N.A.; Mourier, J.; Purcell, C.R.;
Skomal, G.B.; Tucker, J.P.; Walsh, A.J.;
et al. The Drone Revolution of Shark
Science: A Review. Drones 2021,5, 8.
https://doi.org/10.3390/
drones5010008
Received: 26 November 2020
Accepted: 14 January 2021
Published: 21 January 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1NSW Department of Primary Industries, National Marine Science Centre, P.O. Box 4321, Coffs Harbour,
NSW 2450, Australia
2
School of Environment, Science and Engineering, Southern Cross University, National Marine Science Centre,
Coffs Harbour, NSW 2450, Australia; j.tucker.22@student.scu.edu.au
3Sci-eye, P.O. Box 4202, Goonellabah, NSW 2480, Australia; acolefax@scieye.com.au (A.P.C.);
awalsh@scieye.com.au (A.J.W.)
4SMART Infrastructure Facility, University of Wollongong, Wollongong, NSW 2522, Australia;
rgorkin@uow.edu.au
5Department of Biological Sciences, Florida Atlantic University, Boca Raton, FL 33431, USA; kajiura@fau.edu
6Marine Futures Lab, School of Biological Sciences, University of Western Australia, Crawley,
WA 6009, Australia; naima.lopez@research.uwa.edu.au
7UMS 3514 Plateforme Marine Stella Mare, Universitéde Corse Pasquale Paoli, 20620 Biguglia, France;
MOURIER_J@univ-corse.fr
8Department of Physics and Astronomy, Faculty of Science and Engineering, Macquarie University, Sydney,
NSW 2109, Australia; cormac.purcell@mq.edu.au
9Sydney Institute for Astronomy (SIfA), School of Physics, The University of Sydney, Sydney,
NSW 2006, Australia
10
Massachusetts Division of Marine Fisheries, 836 South Rodney French Blvd., New Bedford, MA 02744, USA;
gregory.skomal@state.ma.us
11 Marine Ecology Group, Department of Biological Sciences, Macquarie University, Sydney,
NSW 2109, Australia; jane.williamson@mq.edu.au
12 School of Environmental and Life Sciences, University of Newcastle, Ourimbah, NSW 2258, Australia;
vincent.raoult@newcastle.edu.au
*Correspondence: paul.butcher@dpi.nsw.gov.au
Abstract:
Over the past decade, drones have become a popular tool for wildlife management and
research. Drones have shown significant value for animals that were often difficult or dangerous
to study using traditional survey methods. In the past five years drone technology has become
commonplace for shark research with their use above, and more recently, below the water helping to
minimise knowledge gaps about these cryptic species. Drones have enhanced our understanding of
shark behaviour and are critically important tools, not only due to the importance and conservation
of the animals in the ecosystem, but to also help minimise dangerous encounters with humans. To
provide some guidance for their future use in relation to sharks, this review provides an overview
of how drones are currently used with critical context for shark monitoring. We show how drones
have been used to fill knowledge gaps around fundamental shark behaviours or movements, social
interactions, and predation across multiple species and scenarios. We further detail the advancement
in technology across sensors, automation, and artificial intelligence that are improving our abilities
in data collection and analysis and opening opportunities for shark-related beach safety. An investi-
gation of the shark-based research potential for underwater drones (ROV/AUV) is also provided.
Finally, this review provides baseline observations that have been pioneered for shark research and
recommendations for how drones might be used to enhance our knowledge in the future.
Keywords: artificial intelligence; AUV; drones; protocols; ROV; sharks; UAV
Drones 2021,5, 8. https://doi.org/10.3390/drones5010008 https://www.mdpi.com/journal/drones
Drones 2021,5, 8 2 of 28
1. Overview
Drones, the common term for unmanned aerial vehicles (UAVs) [
1
], have become a
fundamental tool for the shark researcher. The rapid proliferation of the technology as
well as the advancement in visualization capabilities, coupled with increasing cost effec-
tiveness, have enabled new studies for all types of marine-based observations globally [
2
],
particularly in the field of shark research (Figure 1).
Drones 2021, 5, x FOR PEER REVIEW 2 of 28
well as the advancement in visualization capabilities, coupled with increasing cost effec-
tiveness, have enabled new studies for all types of marine-based observations globally [2],
particularly in the field of shark research (Figure 1).
Using drones as a shark research tool is a natural extension of aerial monitoring from
planes and helicopters, which has been performed for decades. Besides certain known
ecosystems (i.e., aggregation sites), it is often difficult to see sharks in the wild and gather
data, particularly in the vast expanse of the ocean. Drones offer on-demand, localised pi-
loting and aerial visualization as an effective way to locate, track and study sharks [3].
Recent studies have recommended that drones have the potential to outperform tradi-
tional aerial surveys [2]. Furthermore, a huge limitation of studying sharks up close is that
some species are potentially dangerous, and drones provide the perfect platform with a
controlled aerial viewpoint to enable researchers to study them safely.
Figure 1. Global representation of research groups using drones for shark research with (a) representation of 32 studies
(see Table 1) conducted with different drone systems, and (b) amplified view of Australia, which per continent (and obvi-
ously surrounded by water and many shark species) has the potentially the most extensive use of drones and shark re-
search globally.
With generally declining populations and increasing anthropogenic threats to sharks,
there is a critical need to fill knowledge gaps as they are often a cornerstone of various
ecosystems [4,5]. Additionally, rare but unfortunate shark interactions from certain spe-
cies can have devastating consequences to animals and humans [6]. There is a recognized
need to better understand shark behaviour to preserve the ocean ecosystems, while miti-
gating negative human–shark interactions.
This review provides a comprehensive analysis of how drones have expanded shark
research. In Section 2, we present the usage of drones in context (i.e., a typical deploy-
ment). Section 3 then takes an in-depth look at how drones have been used for shark re-
search in the key areas of shark behaviour of predation, social interactions and bite miti-
gation, as well as critical environments where sharks reside and species-specific studies.
We further detail in Section 4 how new technology in sensors, automation, and artificial
intelligence, as well as the use of underwater drones, have been developed to increase
data quality and enhance our understanding of sharks. Finally, in Section 5, we provide
insights into the future of drone development for shark research.
Figure 1.
Global representation of research groups using drones for shark research with (
a
) representation of 32 studies (see
Table 1) conducted with different drone systems, and (
b
) amplified view of Australia, which per continent (and obviously
surrounded by water and many shark species) has the potentially the most extensive use of drones and shark research globally.
Table 1.
Location, drone (type and model) and research focus for studies working on shark drone projects. ID codes
correspond to numbers in Figure 1.
ID Location Drone Model Focus Reference
1Moorea, French
Polynesia Multirotor DJI Phantom 2 Shark density Kiszka et al. [3]
2Moorea, French
Polynesia Multirotor DJI Phantom 2 Shoaling behaviour Rieucau et al. [4]
3Bahia de la Paz,
Baja, Mexico Multirotor DJI Spark Co-occurrence Frixione et al. [5]
4Guadalupe Island,
Mexico
Underwater drone
(AUV) REMUS-100 Shark behaviour Skomal et al. [6]
5Guadalupe Island,
Mexico
Underwater drone
(AUV) REMUS-100 Fine scale movements Gabriel [7]
6 La Jolla, CA, USA Underwater drone
(AUV) Not-specified Group movements Ho et al. [8]
7SeaPlane Lagoon,
CA, USA
Underwater drone
(AUV) Oceanserver IVER2 Shark movements Clark et al. [9]
8 Bahamas, USA Multirotor DJI Phantom 2+ Detectability Hensel et al. [10]
9Florida SE Coast,
USA Multirotor DJI Phantom 4 Pro Predatory avoidance
behaviour Doan and Kajiura [11]
10
Beaufort, NC, USA
Fixed-wing drone eBee Detectability of shark
analogues Benavides et al. [12]
11 Cape Cod, MA,
USA
Underwater drone
(AUV) REMUS-100 Shark movements Packard et al. [13]
Drones 2021,5, 8 3 of 28
Table 1. Cont.
ID Location Drone Model Focus Reference
12 Faial Island,
Azores, Portugal Fixed-wing drone Skywalker X8 Detectability of
aggregations Fortuna et al. [14]
13 Sea of the
Hebrides, UK
Underwater drone
(AUV) REMUS-100 Sub-surface behaviour Hawkes et al. [15]
14 Sea of the
Hebrides, UK Multirotor DJI Phantom 3 Pro Social behaviour Gore et al. [16]
15 Mossel Bay, South
Africa Multirotor DJI Phantom 3 and 4 Whale hunting
behaviour
Dines and Gennari
[17]
16 D’Arros and St
Joseph, Seychelles Multirotor DJI Phantom 4
Whale
scavenging/hunting
behaviour
Lea et al. [18]
17 Shoalwater, WA,
Australia Multirotor DJI Mavic Pro Shoaling behaviour López et al. [19]
18 Kimberly, WA,
Australia Multirotor DJI Phantom 4
Whale
scavenging/hunting
behaviour
Gallagher et al. [20]
19
Heron Island,
Queensland,
Australia
Multirotor DJI Phantom 3 Pro Shark movement
tracking Raoult et al. [21]
20 NSW Coast,
Australia Multirotor DJI Phantom 4
Whale
scavenging/hunting
behaviour
Tucker et al. [22]
21 NSW Coast,
Australia Multirotor DJI Phantom 4 Swimming behaviour Colefax et al. [23]
22 NSW Coast,
Australia Multirotor DJI Phantom 4 Swimming behaviour Tucker et al. [24]
23 NSW Coast,
Australia Multirotor DJI Phantom 4 Detection probability Colefax et al. [25]
24 NSW Coast,
Australia Multirotor DJI Phantom 4 Detection probability Colefax et al. [26]
25 NSW Coast,
Australia Multirotor DJI Matrice Detection probability Colefax et al. [27]
26 NSW Coast,
Australia Multirotor DJI Phantom 4 Faunal richness Kelaher et al. [28]
27 NSW Coast,
Australia Multirotor DJI Phantom 4 Helicopter v drone for
shark detection Kelaher et al. [29]
28 NSW Coast,
Australia
Artificial
intelligence Not-specified Detection probability Saqib et al. [30]
29 NSW Coast,
Australia
Artificial
intelligence Not-specified Detection probability Sharma et al. [31]
30 NSW Coast,
Australia
Artificial
intelligence Blimp-based system Shark surveillance Gorkin III et al. [32]
31 NSW Coast,
Australia Multirotor DJI Inspire 1 Beach safety Butcher et al. [33]
32
Flinders Island,
Tasmania,
Australia
Underwater drone
(ROV) BlueROV 2
Post-release behaviour
Raoult et al. [34]
Drones 2021,5, 8 4 of 28
Using drones as a shark research tool is a natural extension of aerial monitoring from
planes and helicopters, which has been performed for decades. Besides certain known
ecosystems (i.e., aggregation sites), it is often difficult to see sharks in the wild and gather
data, particularly in the vast expanse of the ocean. Drones offer on-demand, localised
piloting and aerial visualization as an effective way to locate, track and study sharks [
33
].
Recent studies have recommended that drones have the potential to outperform traditional
aerial surveys [
2
]. Furthermore, a huge limitation of studying sharks up close is that
some species are potentially dangerous, and drones provide the perfect platform with a
controlled aerial viewpoint to enable researchers to study them safely.
With generally declining populations and increasing anthropogenic threats to sharks,
there is a critical need to fill knowledge gaps as they are often a cornerstone of various
ecosystems [
35
,
36
]. Additionally, rare but unfortunate shark interactions from certain
species can have devastating consequences to animals and humans [
37
]. There is a recog-
nized need to better understand shark behaviour to preserve the ocean ecosystems, while
mitigating negative human–shark interactions.
This review provides a comprehensive analysis of how drones have expanded shark
research. In Section 2, we present the usage of drones in context (i.e., a typical deployment).
Section 3then takes an in-depth look at how drones have been used for shark research in
the key areas of shark behaviour of predation, social interactions and bite mitigation, as
well as critical environments where sharks reside and species-specific studies. We further
detail in Section 4how new technology in sensors, automation, and artificial intelligence,
as well as the use of underwater drones, have been developed to increase data quality and
enhance our understanding of sharks. Finally, in Section 5, we provide insights into the
future of drone development for shark research.
2. Drones for Studying Sharks
A standard drone operation consists of an unmanned aircraft, ground control station,
and communications link between the two. Typically, the ground control station serves
as the communication gateway with a live-feed visualization screen from the camera. For
shark research, often the drone will be launched from a beach or vessel serving as the
take-off point and operated along a manual or automated flight path over a target area [
25
].
A pilot who is responsible for controlling the flight operates the drone, and there may be
additional personnel who are operating or analysing the camera feed for sharks and at
times other relevant wildlife and/or people and infrastructure of interest. Once a shark is
spotted, data are collected, mostly in the form of pictures or videos, and recorded either
in internal storage on the drone, as part of the ground station, or by recording the feed
via a stream at an auxiliary location. Depending on the task, the data can be analysed
in real time (for instance to track a specific shark seen on the video feed) and/or videos
can be analysed post flight [
25
,
29
]. There are many parameters to consider when setting
up drone-based research for sharks (Figure 2). These can be separated into two main
categories: the requirements for surveying and data collection, and the conditions at the
study site.
The requirements of the study will dictate the appropriate equipment setup and
corresponding analysis techniques. Primarily, the critical requirements for the survey
area and the flight time (duration of visualization) will significantly influence the drone
type. Fixed-wing and multirotor are two main types of small drones currently suitable for
aerial surveys, although hybrids also exist that attempt to combine advantages from both
platforms. Fixed-wing drones are typically used for speed and energy efficiency. They can
survey comparatively longer distances over 100 s of kilometres and have flight durations
from 20 min to several hours; however, they generally require assistance with taking off
(‘throwing’ by hand or catapult) and a clear area for landing [
2
]. Multirotor drones are a
comparatively new technology and have advantages of rapid vertical take-off and landing
capabilities on coastal beaches and vessels. They can also hover and are more dynamic and
responsive in movement positioning, however, are more aerodynamically unstable and
Drones 2021,5, 8 5 of 28
have shorter flight durations of typically 12 to 40 min [
2
]. Increasing the energy density
of the batteries, using alternative fuel systems, or dual or tethered power systems will
provide greater flight times and allow for longer periods of flight, but may add complexity
and cost to the drone platform.
Drones 2021, 5, x FOR PEER REVIEW 4 of 28
based research for sharks (Figure 2). These can be separated into two main categories: the
requirements for surveying and data collection, and the conditions at the study site.
Figure 2. Schematic illustrating the interrelated factors that researchers should consider when planning and performing
shark research with drones. (Top) factors that influence the type of drone and payload required for a research activity,
(Left) factors that influence pre-flight planning for a research monitoring activity, (Right) factors on the day that heavily
influence successful flight and data collection during a research activity. (Inset) image of underwater ROV/UAV and (Bot-
tom) additional factors for underwater drones.
The requirements of the study will dictate the appropriate equipment setup and cor-
responding analysis techniques. Primarily, the critical requirements for the survey area
and the flight time (duration of visualization) will significantly influence the drone type.
Fixed-wing and multirotor are two main types of small drones currently suitable for aerial
surveys, although hybrids also exist that attempt to combine advantages from both plat-
forms. Fixed-wing drones are typically used for speed and energy efficiency. They can
survey comparatively longer distances over 100 s of kilometres and have flight durations
from 20 min to several hours; however, they generally require assistance with taking off
(‘throwing’ by hand or catapult) and a clear area for landing [2]. Multirotor drones are a
comparatively new technology and have advantages of rapid vertical take-off and landing
capabilities on coastal beaches and vessels. They can also hover and are more dynamic
and responsive in movement positioning, however, are more aerodynamically unstable
and have shorter flight durations of typically 12 to 40 min [2]. Increasing the energy den-
sity of the batteries, using alternative fuel systems, or dual or tethered power systems will
provide greater flight times and allow for longer periods of flight, but may add complexity
and cost to the drone platform.
The type of data to be collected will influence the variant of sensor, the type of pay-
load to be attached to the drone, and ultimately the choice of drone for use. Image and
Figure 2.
Schematic illustrating the interrelated factors that researchers should consider when planning and performing
shark research with drones. (Top) factors that influence the type of drone and payload required for a research activity,
(Left) factors that influence pre-flight planning for a research monitoring activity, (Right) factors on the day that heavily
influence successful flight and data collection during a research activity. (Inset) image of underwater ROV/UAV and
(Bottom) additional factors for underwater drones.
The type of data to be collected will influence the variant of sensor, the type of payload
to be attached to the drone, and ultimately the choice of drone for use. Image and video
collection are the primary techniques for most shark research and the related equipment
is a common feature of the typical drone toolkit. Significant variations on the range of
camera types and features for basic visualization are available, and more recently, thermal
imaging, LIDAR, and hyperspectral cameras have shown potential for research but are yet
to be fully tested. Regardless of imaging equipment, sensor type and drone choice are also
influenced by the “data link” requirements of the research. This includes considering the
timing of data analysis needed, i.e., can the images/videos be processed after collection, or
does analysis need to be in real-time or some step in between. Critically for main studies
which require analysis during the flight, the type and transmission of data collection is
a fundamental consideration, i.e., the resolution of video telemetry, data bandwidth and
meta-data captured. The interrelated requirements of automated flight and data collection
are additional worthwhile considerations, if required by the researcher. Additionally, the
payload selection directly impacts the weight allowances and capabilities for the drone.
Furthermore, drones of increasing weights are divided into different classes. In regard
Drones 2021,5, 8 6 of 28
to regulation, this can impact the licenses required to fly which must also be considered
when preparing for drone research. The interplay of these factors can significantly affect
the ultimate drone choice.
The conditions of the study site are highly influential on a shark research experimental
plan, impacting both pre-flight planning, and the flight and data collection on the day.
Firstly, there are specific conditions that will influence the ability to fly and collect data.
This includes the telecommunication availability, particularly any lack of network coverage
that can impact flight control and safety, as well as limit visualization of data collection
from drones (transmitted video feeds). Furthermore, it is critical to have safe access to
the study site (for shark observations this can range from a populated urban beach to an
isolated marine reserve). The site-specific conditions can influence how researchers should
approach drone use. For instance, operators often are required to avoid flying over people
when using drones at recreational beaches. As well, certain locations have additional
drone restrictions/prohibitions to reduce wildlife impact. In protected areas like certain
parks and sanctuaries, exemptions to no-fly zones are needed, and conditions like altitude
limits are imposed to protect marine life and to minimize bird disturbance. These factors
are impacted by the technological capabilities of the drone, the “local” aviation laws, the
service providers, etc., all of which are critical for planning.
In addition to the general infrastructure and usage at the site, drone-based shark
studies are highly dependent on the weather and sea conditions. This goes well beyond
above water conditions like wind and rain that limit the ability of drones to fly. Wave
conditions are also required to be minimal to avoid distortion on the surface of the water
and thus provide a sufficiently clear view of the underwater environment. It is well known
that while aerial viewing has significant benefits, the underwater environment imposes
constraints to visualization from drones. Primarily the time of day, angle of the sun and
resulting reflection, also influence optimal visualization [25,33].
The combination of these factors showcases the difficulties in studying sharks, espe-
cially when the most basic principle requires them to be barely sub-surface to be visible.
This can occur anywhere in the ocean but most often in shallow water that constrains
the animals near the surface. In addition, inhabiting waters must be sufficiently clear to
enable the animals to be seen from the air. Some coastal or estuarine habitats are often too
turbid (murky, sandy or muddy, etc.) to see anything below the surface. Any of the above
constraints will limit the utility of drones for observing shark behaviour. Recent advances
in fluid lensing have demonstrated the ability to compensate for surface distortion from
drone video and this is a promising area for future development that could provide a
clearer view of the underwater environment at small scales [38].
One final consideration is that civil aviation regulations often require training, certifi-
cations, and impose specific usage (recreational versus commercial use) and other flight
restrictions depending on the jurisdiction. Generally, these rules are evolving and are
varied from country to country. One major consideration is that regardless of drone type,
size and payload configuration, generally authorities limit flight distances to ‘visual line-
of-sight’, which restricts operations to localized spatial scales (i.e., individual beaches).
Beyond line-of-sight operations are currently possible in some areas and countries, how-
ever, operations are typically expensive and arguably no longer cost-effective compared
with manned aircraft [
39
]. Therefore, although regulations are likely to be more flexible
in the near future, drones have the greatest potential on smaller spatial scales. For safety
and effective research execution, the authors strongly advise that any shark researchers
interested in using drones seek expert advice and training to gain understanding of local
issues and regulations.
The authors would also highlight that most of the factors discussed are similarly
important for underwater drones (ROV/UAV) with a few notable additions. For drone
choice, a similar discussion to the aerial flight area and time can be made for underwater
drones in patrol time and depth required for observations. ROVs are currently directly
operated vehicles, meaning they are better at reacting to changing conditions, whereas
Drones 2021,5, 8 7 of 28
AUVs generally rely on following cues (usually requiring acoustic tags) and using remote
acoustic instructions that may not be as flexible as those of an ROV. In unplanned or
unknown conditions, the ROV offers more flexibility, even if the deployment capabilities of
ROVs are often inferior, especially for range and speed. Both those factors can affect drone
choice, particularly in the usage considerations between a submersible drone controlled
and tethered from a ship versus autonomous submarine type drone systems. In terms of
planning, obviously more consideration must be taken for activity/usage in the water. For
conditions on the day, turbidity or the clarity of the environment, water depth, tide and
current will absolutely influence the ability to collect images/video/data from underwater
based perspectives.
3. Drone Research Areas
The following sections provide illustrations of the limited specific research activities
that have been conducted so far on sharks using drones, highlighting the breadth of
potential of this technology for its application in shark research.
3.1. Drones as a Tool for Shark Hazard Reduction
A fundamental issue with the implementation of new methods or ‘tools’ for shark bite
mitigation is the challenge of reliably assessing the direct reduction in shark bite risk due
to the rarity of incidents. Between 2015 and 2020, research trials were completed in New
South Wales, Australia to assess the utility of drones to provide adequate beach safety with
regards to shark bites and provide insight as to whether drones may play a role in future
shark mitigation strategies.
In terms of mitigation, drones are an extension of well-established methods for shark
identification through aerial monitoring which has occurred for decades [
2
]. It is widely ac-
cepted that specific shark species are more dangerous to humans than others and represent
the majority of injuries and deaths recorded (particularly white, tiger, bull sharks) (https:
//www.floridamuseum.ufl.edu/shark-attacks/factors/species-implicated/). Drones can
be used to locate and identify sharks, and combined with alerting responsible beach per-
sonnel and removing swimmers from the water and/or closing the beach, thereby reduces
the risk of attacks through isolating and eliminating the hazard.
Drones have utility for shark surveillance, but are limited in both conditions that
make flying difficult, as well as those conditions that affect the ability to detect fauna from
aerial positions. From a piloting perspective, most drones can only fly effectively in winds
typically up to ~15–18 knots, and during rain-free periods [
25
,
33
] as well as if visibility is not
impaired by fog or low altitude cloud cover. Regarding visibility from the air, the reliability
of detecting fauna declines as sea conditions and water clarity worsens, just like manned
aircraft surveillance [
40
]. Sighting rates from drones are largely comparable to that of
helicopters [
29
], which can have low detection reliability when conditions are not favourable.
These factors complicate shark surveillance, while increasing wind velocities and sea states
usually negatively correlate with water users, there is a large degree of variation among
beaches, and often there are some water users (e.g., surfers) in circumstances where the
reliability of fauna detections may be significantly compromised [25].
Environmental predictors influence the probability of observing the presence of tar-
get sharks such as bull (Carcharhinus leucas), tiger (Galeocerdo cuvier) and white sharks
(
Carcharodon carcharias
) [
25
]. However, water visibility, wind speed and direction seem to
have little influence on the behaviour of white sharks while near the surf break. Along the
east coast of Australia, swim behaviour of white sharks near the surf break was demon-
strated to be largely predictable for this species, as it was shown to be consistently slow at
~3 km h
1
and parallel to the beach line [
23
] (Figure 3a). The slow and predictable track
trajectories of white sharks compliments surveillance strategies that can make frequent
surveillance passes (Figure 3b). However, although white shark behaviour was consistent
across the various study locations, it has been demonstrated to significantly differ near
abundant food sources, such as in proximity to seal colonies or when whale carcasses wash
Drones 2021,5, 8 8 of 28
near shore [
24
]. Such species-specific information on behaviour can enhance our success of
identifying and tracking sharks under different environmental conditions.
Drones 2021, 5, x FOR PEER REVIEW 7 of 28
favourable. These factors complicate shark surveillance, while increasing wind velocities
and sea states usually negatively correlate with water users, there is a large degree of var-
iation among beaches, and often there are some water users (e.g., surfers) in circumstances
where the reliability of fauna detections may be significantly compromised [7].
Environmental predictors influence the probability of observing the presence of tar-
get sharks such as bull (Carcharhinus leucas), tiger (Galeocerdo cuvier) and white sharks (Car-
charodon carcharias) [7]. However, water visibility, wind speed and direction seem to have
little influence on the behaviour of white sharks while near the surf break. Along the east
coast of Australia, swim behaviour of white sharks near the surf break was demonstrated
to be largely predictable for this species, as it was shown to be consistently slow at ~3 km
h
1
and parallel to the beach line [12] (Figure 3a). The slow and predictable track trajecto-
ries of white sharks compliments surveillance strategies that can make frequent surveil-
lance passes (Figure 3b). However, although white shark behaviour was consistent across
the various study locations, it has been demonstrated to significantly differ near abundant
food sources, such as in proximity to seal colonies or when whale carcasses wash near
shore [13]. Such species-specific information on behaviour can enhance our success of
identifying and tracking sharks under different environmental conditions.
Figure 3. Representation of the types of imagery collected by drone: (a,b) tracking white sharks
(Carcharodon carcharias) along coastal beaches of eastern Australia as part of behavioural and
bather protection programs (Image credit—A Colefax), (c) great hammerhead (Sphyrna mokarran)
predation event on blacktip sharks (Carcharhinus limbatus) (Image credit—S Kajiura), (d) tiger
shark at a humpback whale (Megaptera novaeangliae) carcass off eastern Australia (Image credit—M
Dujmovic), (e) hammerhead sharks (Sphyrna sp.) observed during the austral summer in Western
Australia (Image credit—N.A. López), and (f) Epaulette shark (Hermiscyllium ocellatum, ~50 cm TL)
captured feeding in sediments at low tide on Heron Reef flat, Great Barrier Reef, Australia from an
altitude of 5 m (Image credit—V. Raoult).
Figure 3.
Representation of the types of imagery collected by drone: (
a
,
b
) tracking white sharks (Carcharodon carcharias) along
coastal beaches of eastern Australia as part of behavioural and bather protection programs (Image credit—A Colefax), (
c
)
great hammerhead (Sphyrna mokarran) predation event on blacktip sharks (Carcharhinus limbatus) (Image credit—S Kajiura),
(
d
) tiger shark at a humpback whale (Megaptera novaeangliae) carcass off eastern Australia (Image credit—M Dujmovic), (
e
)
hammerhead sharks (Sphyrna sp.) observed during the austral summer in Western Australia (Image credit—N.A. López),
and (
f
) Epaulette shark (Hermiscyllium ocellatum, ~50 cm TL) captured feeding in sediments at low tide on Heron Reef flat,
Great Barrier Reef, Australia from an altitude of 5 m (Image credit—V. Raoult).
Currently, drones are one socially preferred method for assisting shark bite mitigation
through hazard reduction, along with other monitoring techniques, particularly compared
with cull-based strategies [
41
]. This is despite a sentiment of concern around the reliability
of detection during unfavourable weather conditions, and the ability to effectively discrim-
inate between shark species by the pilots. It demonstrates that further development of
drones as a shark bite mitigation tool is warranted, particularly in the areas of improving
detection reliability and efficiency of individual species identification. Such improvements
in utility should further increase the positive community sentiment and further reduce
shark bite risk potential.
While the reliability of drone-based shark surveillance can likely be further improved
with spectral filtering and recognition software ([
30
,
32
], see Section 4.1 and Section 4.2), the
largest improvements to efficiency will occur with system automation. However, as coastal
air spaces often also have high air traffic activity, it is unlikely that civil aviation authorities
would approve beyond line-of-sight flights there. However, in coming years this will
likely change. Therefore, advancing and integrating detection software in drone-based
shark surveillance will have short-term benefits regarding detection reliability, but also
facilitate automation in the longer-term. Even with those advances, the authors would
Drones 2021,5, 8 9 of 28
caution against an overreliance on detection technologies including drones. It is unlikely
that drones will replace spotting from trained professionals (lifeguards and lifesavers) for
the foreseeable future.
Drones have strong potential for providing adequate risk reduction in the potential for
shark bites, appease public needs and perceptions of beach safety, and support conservation.
Therefore, drones can play a pivotal role in transitioning away from shark bite mitigation
using cull-based strategies and help meet the increasing demand for non-destructive beach
safety [
26
]. In addition, drone-based surveillance can potentially provide opportunistically
collected data to conservation agencies as a by-product [
28
,
42
,
43
]. With further research
and development, the utility of drone-based surveillance will further increase, as should
public perceptions, particularly if operations remain incident-free [41].
3.2. Drone Studies of Shark Predation Events
Natural predation is notoriously difficult to document in the wild. Predation events
occur infrequently, and the events are often brief. Therefore, the probability of being in
the right place at the right time to witness a natural predation event is low and it is easier
to see the results after the fact. For example, scientists have used the presence of a large
blood slick on the surface of the water to determine that a white shark had predated
upon a pinniped, despite not seeing the actual predation event itself [
44
]. In addition, the
presence of the observer can alert prey or impact the behaviour of the predator, potentially
resulting in an aborted predation attempt. These complications are amplified when dealing
with underwater animals. In the underwater realm, all the same constraints apply with
some additional difficulties. The time available to wait and observe a predation event
underwater is limited by scuba or rebreather capabilities, with available duration inversely
proportional to increasing depth. Underwater visibility is greatly limited compared to in
air. This requires the observer to be much closer to the animals which can, in turn, impact
their natural behaviour. However, under some circumstances it is possible to bypass some
of these constraints to remotely observe natural predation.
Drones provide an opportunity to unobtrusively observe underwater animals, if they
are sufficiently near to the surface. At altitudes greater than 5–10 m, the sound produced by
most small drones is undetectable above background levels underwater [
45
]. A small drone
flying overhead would also be nearly impossible to distinguish from underwater. Thus,
unless the drone is close to the water’s surface where it could be detected by the animals,
the behaviours documented are likely to be natural and not impacted by the observer.
Another advantage of using drones to record natural predation events is the stability that
they confer. Unmanned aerial vehicles can hover in place using GPS positioning. Coupled
with gimbal-stabilized cameras, this provides a highly stable video recording platform that
facilitates quantification of predator and prey movement with respect to each other. This
enables observations to incorporate quantitative data, such as distance between individuals,
or swimming velocity in body lengths per second, rather than being merely descriptive [
46
].
An ideal location that satisfies the requirements to visualize shark predation using
drones is the southeast coast of Florida, USA. The nearshore environment is characterized
by a uniform, light sandy seafloor that allows the dark shape of sharks to be clearly seen. In
addition, the continental shelf narrows in southeast Florida, which allows the Gulf Stream
current to transport clear water close to shore providing good visibility. Each winter,
thousands of blacktip sharks (Carcharhinus limbatus) aggregate in the shallow water along
the coast where they are observed from the air [
47
]. These sharks are prey to larger sharks,
such as the great hammerhead (Sphyrna mokarran) (Figure 3c) [
48
]. Drones have been used to
document great hammerhead sharks attempting to prey upon blacktips in the shallows [
11
].
These natural predation events reveal hammerhead sharks cruising slowly in the nearshore
environment with numerous blacktips nearby. The hammerhead will suddenly accelerate
rapidly and chase down a blacktip. The featureless sandy seafloor provides no structure or
shelter and the blacktips often flee to the shallow water adjacent to the beach. The much
larger hammerhead shark is unable to follow into the shallows and turns back to deeper
Drones 2021,5, 8 10 of 28
water, allowing the blacktips to escape. Documentation of the use of shallow water as
a refuge was only possible because of the aerial view provided by the drone. Although
other studies have examined social behaviour of sharks with drones [
4
,
8
,
16
,
18
], Doan and
Kajiura’s [
11
] study is the first to document predator avoidance behaviour by large adult
sharks. The regular predictable occurrence of large numbers of sharks in a nearshore
environment with clear water provides a rare opportunity to use drones to observe and
study natural predation in the wild.
3.3. Drone Studies of Shark Behaviour and Social Interactions
Drones have recently been used to record and study collective shark behaviour. Drone
footage can be analysed in depth to quantify swimming alignment, nearest-neighbour
distances, velocity (based on static fly and landmarks) and tail beat frequency (Figure 4).
For example, Rieucau et al. [4] used an image analysis-based technique applied to drones
and showed that blacktip reef sharks (Carcharhinus melanopterus) displayed increased
alignment with shoal companions when distributed over a sandflat where they are regularly
fed for ecotourism purposes as compared with when they shoaled in a deeper adjacent
channel. Using similar methods, it could be possible to study the collective response
of shoaling sharks to predation risk using drone-based methods [
11
] for example, by
measuring the transfer of information between individuals following the approach of the
predator. Drones can also be used to reveal the fine-scale interaction rules of mass migrating
elasmobranchs [16,47,49], as has been done for other terrestrial mammal species [50].
Drones 2021, 5, x FOR PEER REVIEW 10 of 28
Figure 4. Drone footage can be analysed to quantify swimming alignment, nearest-neighbour distances, velocity and tail
beat frequency (Image—modified by J Mourier from Rieucau et al., 2018).
Quantifying social interactions and building social networks may be limited using
drones because it requires identifying individuals within observed groups over multiple
sampling periods to infer the association indices calculated from the repeatability of in-
teractions—not currently an easy task with drones. However, drones can be used to rec-
ord group size in elasmobranchs [31] or to document social behaviour and frequencies of
such behaviours [27]. For example, Gore et al. [27] used a combination of boat surveys,
snorkelling and drones to record the frequency of close-following, parallel or echelon
swimming and breaching of basking sharks (Cetorhinus maximus). Post-processing tech-
niques can also be used to determine the track and path of multiple individual sharks
within a frame [33], which can be used to measure interactions and encounters between
multiple individuals. Although such analysis may be limited due to the short sampling
increments of the drone, some social measurements can still be recorded such as the fre-
quency of associating with another shark versus swimming alone.
Drones are still constrained by many practicalities that would limit the study of col-
lective behaviour and social interactions in elasmobranchs. The first one is that aerial
drone surveys are limited to record behaviour at the surface in shallow environments.
While some shark and ray species can display collective and social behaviour at the sur-
face that can be easily captured by drones, most species spend considerable time at rela-
tively greater depths beyond the vision of the drones. Moreover, drone operations can be
significantly impacted by environmental conditions as flights require good weather con-
ditions and optimal sunlight, and are restricted to daylight surveys [34]. Another current
important issue is the short battery life and flight duration. While battery technology and
subsequent capacity delivery for a given weight continuously improves, current multi-
rotor flights rarely last more than 20–30 min, which can affect the robustness of sampling
social interactions.
3.4. Shark Behaviour around Whale Carcasses
Natural shark behaviours are often difficult to observe due to their habitats and rel-
atively low abundance. Larger shark species, such as bull, tiger and white sharks, are often
transitory in nature making it more difficult for researchers to observe natural behaviours
[35]. Feeding events are situations where the natural behaviour of low-density organisms
can be observed. Many species of shark feed on whale carcasses and research suggests
that sharks can locate these carcasses over great distances [36]. This makes whale carcasses
excellent opportunities for observing shark behaviour (Figure 3d).
Figure 4.
Drone footage can be analysed to quantify swimming alignment, nearest-neighbour distances, velocity and tail
beat frequency (Image—modified by J Mourier from Rieucau et al., 2018).
Quantifying social interactions and building social networks may be limited using
drones because it requires identifying individuals within observed groups over multi-
ple sampling periods to infer the association indices calculated from the repeatability of
interactions—not currently an easy task with drones. However, drones can be used to
record group size in elasmobranchs [
49
] or to document social behaviour and frequencies
of such behaviours [
16
]. For example, Gore et al. [
16
] used a combination of boat surveys,
snorkelling and drones to record the frequency of close-following, parallel or echelon swim-
ming and breaching of basking sharks (Cetorhinus maximus). Post-processing techniques
can also be used to determine the track and path of multiple individual sharks within a
frame [
21
], which can be used to measure interactions and encounters between multiple
individuals. Although such analysis may be limited due to the short sampling increments
Drones 2021,5, 8 11 of 28
of the drone, some social measurements can still be recorded such as the frequency of
associating with another shark versus swimming alone.
Drones are still constrained by many practicalities that would limit the study of collec-
tive behaviour and social interactions in elasmobranchs. The first one is that aerial drone
surveys are limited to record behaviour at the surface in shallow environments. While
some shark and ray species can display collective and social behaviour at the surface that
can be easily captured by drones, most species spend considerable time at relatively greater
depths beyond the vision of the drones. Moreover, drone operations can be significantly
impacted by environmental conditions as flights require good weather conditions and opti-
mal sunlight, and are restricted to daylight surveys [
51
]. Another current important issue is
the short battery life and flight duration. While battery technology and subsequent capacity
delivery for a given weight continuously improves, current multirotor flights rarely last
more than 20–30 min, which can affect the robustness of sampling social interactions.
3.4. Shark Behaviour around Whale Carcasses
Natural shark behaviours are often difficult to observe due to their habitats and
relatively low abundance. Larger shark species, such as bull, tiger and white sharks,
are often transitory in nature making it more difficult for researchers to observe natural
behaviours [
52
]. Feeding events are situations where the natural behaviour of low-density
organisms can be observed. Many species of shark feed on whale carcasses and research
suggests that sharks can locate these carcasses over great distances [
53
]. This makes whale
carcasses excellent opportunities for observing shark behaviour (Figure 3d).
Drones can be used to observe sharks hunting and scavenging from whales in their
natural habitat with minimal disturbance to individuals [
17
,
20
,
22
] (Figure 3d). Bull, tiger,
tawny nurse (Nebrius ferrugineus), and white sharks have been observed scavenging whale
carcasses using drones [
22
]. Small and cost-effective drones have the resolution to capture
shark behaviours and interactions at whale carcass scavenging events from heights that do
not influence natural behaviour (~30 m). Behaviours observed include test biting, head
shaking, palatoquadrate protrusion, ocular rotation, nictitating membrane use and intra
and interspecific interactions [
22
]. The behaviour of white sharks hunting a live whale
has been recorded including approach direction, approach angle, head shaking, and bite
location and frequency [17].
Drones are an excellent tool for observing shark behaviour in these events as the
top-down view and high resolution allows researchers to observe from a height that
includes all sides of a whale simultaneously and the area directly adjacent. Therefore,
behaviours such as approach angle/direction and intra and interspecific interactions can be
observed. Sharks feeding underneath a whale can also be observed (depending on water
clarity) which has not been possible in the past without the use of a shark cage that may
influence behaviour. This gives researchers an additional view of an event and behaviours
displayed, including some subtle communication behaviours sharks are known for [
54
].
Researchers are also able to pilot a drone from a distance that does not influence shark
behaviour unlike traditional, handheld cameras which require an operator be close to a
carcass. Dicken [
55
] recounts white sharks biting the propellers and pontoons of their boat
while filming a scavenging event with traditional cameras, suggesting that these methods
do affect behaviour and that drones can be considered a more effective tool for collecting
data without altering shark behaviour.
Stranded whale carcass management is a controversial topic due to perceived shark
attraction [
56
,
57
] and initial results suggest that shark behaviour is altered by the presence
of a stranded carcass [
24
]. Behaviours can be compared to that of sharks not near stranded
carcasses [
23
] to better understand shark behaviour around certain stimuli. Shark speed,
total length, and track straightness and sinuosity can be effectively recorded using drones
by remaining directly over a shark while matching its movement speed and heading with
the camera set in nadir [
23
]. Results from these studies are relevant to beach management
as significant food sources may alter the behaviour of large sharks and increase risk to
Drones 2021,5, 8 12 of 28
water users. Drones have been shown to be effective tools for bather protection in normal
conditions [
33
] and can be utilised in a similar manner, around whale carcass standing
events, to reduce risk of shark interactions with water users.
Operating drones around cetaceans and sharks can be more complex than operating
around sharks alone. Some countries have cetacean exclusion zones for drones. In Australia,
for example drones are prohibited to come within 100 m of live whales and dolphins in
New South Wales and within 300 m in South Australia. While some cost-effective drones
have the zoom and/or resolution to collect detailed observations from 100 m altitude, 300 m
would require specialty cameras, and therefore larger, more expensive and more complex
drones, to collect data in the resolution required to effectively analyse. While this is not an
issue around cetacean carcasses, observing interactions between sharks and live cetaceans,
including potential predation, may be difficult in some countries. This, coupled with the rarity
of witnessing such an event, may limit the observational data drones can collect. Advances
in drone and camera technologies will likely increase their usefulness in these situations as
cameras with higher resolution and better zoom are becoming increasingly smaller, lighter
and more cost effective, thereby allowing for their inclusion on smaller drones.
3.5. Drone Research of Pelagic Shark Aggregations
Pelagic sharks are amongst the most threatened of vertebrates globally, with at least
three-quarters of all species assessed as Threatened or Near Threatened by the IUCN [
35
,
58
].
They are considered extremely valuable in commercial and recreational
fisheries [5961]
and, consequently, this group has been heavily exploited during the last decades deci-
mating their populations across all oceans [
36
,
62
,
63
]. Their conservative life-history traits,
including longevity, late sexual maturity and few offspring, renders pelagic sharks highly
vulnerable to anthropogenic threats [64,65].
A number of pelagic sharks show aggregation behaviour close to the surface or in
shallow coastal areas during different life stages. These aggregations can represent feeding
or breeding locations when adults [
16
,
66
], and nurseries or growing grounds during early
life stages [
67
69
]. Their aggregation behaviour also increases their vulnerability to ex-
ploitation [
70
,
71
], hence monitoring their distribution and behaviour during these periods
is vital for appropriate management and conservation [
70
,
72
74
]. For example, knowing
where and when they aggregate can inform spatial protection management strategies,
such as marine protected areas and fisheries management. Technological advances in
new non-invasive and cost-efficient methods such as drones provide researchers with the
opportunity to study the fine-scale movements and behaviour of pelagic shark species in
shallow coastal environments, complementing the use of traditional methods.
Hammerhead sharks (Sphyrnidae) are an excellent candidate for drone studies and
provide a clear example of the potential of drones to study other pelagic shark species that
aggregate in shallow coastal waters. Their laterally elongated head shape, or cephalofoil,
makes them unmistakable from other sharks in the Carcharhiniformes [
75
]. The great,
scalloped and smooth hammerheads are considered large species within the group [
76
]
that translates into easier aerial detection. These species have global distributions and
are known to travel hundreds to thousands of kilometres [
77
,
78
] between shallow coastal
habitats [
79
,
80
]. Their behaviour differs noticeably: while great hammerheads tend to
inhabit coastal waters and are solitary [
76
], scalloped and smooth hammerheads occur in the
open ocean and are known to form large aggregations near oceanic islands and seamounts
when adults [70,81] and coastal aggregations while in neonate and juvenile stages [8284].
The first aerial study of the spatial ecology of hammerhead sharks (Sphyrna spp.) using
manned aircraft dates from the 1980s [
85
]. This study quantified seasonal patterns in the
abundance of hammerheads in relation to sea surface temperature and the Gulf Stream in
Florida, USA. Since then, records of hammerheads have been reported in multiple marine
megafauna aerial studies [
86
88
] but have yet to be the focus of drone-based studies in
terms of ecology or behaviour. To date, three studies explored hammerhead detectability
from manned and unmanned aerial vehicles [
12
,
14
,
40
]. In 2014, Robbins et al. [
40
] studied
Drones 2021,5, 8 13 of 28
shark detectability from fixed-wing and helicopter aircrafts using shark analogues with
different shapes including, white, tiger and hammerhead sharks (
Sphyrna spp.
). Although
they observed overall low detectability rates, the only mock “sharks” successfully identified
by the spotters where those with the hammerhead shape. Two further studies have
evaluated hammerhead shark detectability using fixed-wing drones: Fortuna et al. [
14
]
tested the detectability of juvenile hammerhead aggregations (Sphyrna spp.) off the coast
of Faial Island, Azores, demonstrating the potential of drones for identifying hammerhead
aggregations. More recently, Benavides et al. [
12
] demonstrated the effect of environmental
variables on the detection probability of mock bonnethead sharks (Sphyrna tiburo), a small
coastal species, in a temperate estuarine area, concluding that depth had the strongest
effect on detectability rates. As of writing, there was only one study that documented
hammerhead shark behaviour using drones [
11
]. The authors documented predatory
avoidance behaviour of great hammerhead sharks and blacktip sharks in shallow coastal
waters off Florida, USA, demonstrating that drones can be successful in elucidating the
behaviour of hammerhead sharks [
11
]. In addition, drones have been used to study the
southernmost aggregation of hammerhead sharks (Sphyrna sp.) in Western Australia,
following reports of a consistent seasonal but unstudied aggregation [López, unpublished
work; Figure 3e].
The potential of drones as a non-invasive and cost-effective method to study ham-
merhead sharks in shallow coastal areas is clear, prompting future work and should be
extended to other endangered sharks that school. Application of demonstrated methodolo-
gies using drones to study shark shoaling behaviour [
4
], movement trajectories, and habitat
use [
3
,
21
,
23
] will be extremely beneficial to better understand how these endangered species
are using coastal areas to inform better spatial management and conservation strategies.
3.6. Drone Studies of Reef Sharks
Shallow water sharks found in coral reefs are the focus of scientific research for most
elasmobranchs [
89
]. Among species of reef shark, there is a dearth of research on many
smaller or cryptic genera in this category [
90
]. Low numbers of studies may be driven
by the difficulty accessing the shallow reefs where many reef-associated species such as
epaulette sharks (Hermiscyllium ocellatum) occur (Figure 3f). Shallow reefs are often complex
habitats that are generally only accessible by foot at low tide or by boat at high tide: this
makes using common abundance or behaviour survey techniques difficult to use effectively.
Most studies to date examining the movement and behaviour of reef sharks in shallow
lagoons rely on acoustic tagging [
91
,
92
], but these have limits in lagoon environments due
to unreliable detection distances [
93
]. Baited remote underwater video systems (BRUVS)
that are often used to assess reef shark abundance and diversity rely on bait to attract
species of interest [
94
], meaning that they are not able to assess natural behaviour and
habitat use. Researchers can use RUVS to remove the effect of bait on behaviour, but they
are limited in their visibility and probability of observation making them less efficient for
determining abundance and assessing behaviours of sharks. When being in the water
is possible, snorkelers or divers also impact natural shark behaviour [
95
,
96
] making the
separation of natural from human-induced behaviours difficult. Drones are less limited by
tides in shallow reef environments and can circumvent some of these issues by providing a
platform to assess abundance, diversity, movement and behaviour of these animals with
low impact and at relatively low cost. In addition, the barrier reefs that often surround
these environments limit wave clutter that can make drone research more difficult [
97
]. In
shallow reef environments, drones can obtain information that would otherwise require
tagging or visual censuses. For example, by using the drone onboard GPS it is possible to
assess links between movement and habitat use and behaviour of sharks at very high (<1 m)
spatial resolutions, albeit over relatively shorter timeframes (<20 min) than with tagging
approaches with current drone battery technology [
21
]. One way to extend these tracks is
to use at least two drones to relay flights as batteries run out as in Colefax, et al. [
23
], but
eventually pilot fatigue is likely to reduce the accuracy of this approach similarly to typical
Drones 2021,5, 8 14 of 28
~6 h limits used in active acoustic tracking [
98
]. The only limits to numbers of active tracks
of sharks using drones are the number of available batteries, pilot fatigue and weather,
meaning that in ideal conditions with a battery recharge station and multiple pilots/drones
it is possible to obtain large datasets rapidly. Since the drone-based tracks are shorter than
active surface tracking, which can last for 24+ h [
99
], obtaining datasets that are sufficiently
large to infer habitat use and behaviour from drones thus require more individual tracks
than tagging approaches. However, the trade-off from using drones is obtaining higher
resolution data than tagging approaches, and the ability to directly examine behaviours.
Drones can also be used to assess the abundance and diversity of sharks and rays in
reef habitats. For example, by conducting line transects with drones, Kiszka, et al. [
3
] were
able to identify difference in densities of sharks and rays in various reef locations. Drones
were also effective at counting and identifying reef sharks and rays in impacted and pristine
lagoons [
10
]. In shallow reef environments, it should be possible to obtain morphometric
data on reef sharks and rays in a similar process as those used on whales [
45
] since shallow
depths are unlikely to affect measurements. The use of drones may thus provide an
alternative rapid approach for assessments of shark populations in shallow reefs, especially
compared to more conventional methods like BRUVS or capture and release programs that
impact behaviour and may come at a cost to the health of the animal. To date, most drone-
based shark population assessments have relied on multi-rotors that have comparatively
limited flight times relative to fixed-wing drones [
2
], and, for these applications fixed-wing
drones may be preferable for larger survey areas similar to conventional aircraft [100].
The quantitative movement or behaviour of sharks in these environments is a useful
metric, but haphazard flights can also reveal novel behaviours and inter-specific interac-
tions that open new avenues of research. For example, drone flights along shallow beaches
revealed that blacktip sharks use these shallow environments to seek refuge from great
hammerhead sharks known to predate on them [
47
]. Flights over a reef lagoon at low
tide revealed active foraging of blacktip reef sharks (C. melanopterus) and epaulette sharks
feeding on prey in bare sediment [
21
] (Figure 3f). Use of drones allows the safe observation
of large numbers of reef sharks including tiger sharks scavenging whale carcasses [
18
].
A greater availability and willingness to use drones during reef shark research projects
should allow new insights into their behaviour and ecology.
In these shallow reef environments, drones currently provide one of the only means
to study shark behaviour, distribution and abundance. The use of drones in these environ-
ments, however, is relatively novel and there are few long-term studies that have relied on
this approach to obtain conservation-relevant data. Satellite and acoustic tagging programs
initially encountered similar difficulties as sample sizes were low, yet global datasets are
now available that have allowed inferences to be drawn at scales never before possible.
As drones become more commonly used tools for shark research in shallow reefs, larger
drone-based datasets will allow broader examinations of ecological patterns of shark move-
ment, habitat use, and behaviour not obtainable with other approaches. For example, since
drone movement data are georeferenced at high resolutions, large drone-based datasets
could examine questions around the use of landmarks or ‘highways’ in shallow reefs to
move across reef lagoons, and whether movement of sharks in these areas has effects on
other species of shark and fishes that can be visible concurrently. The use of drones in these
environments could therefore be used to start making explicit links between behaviour,
movement and habitat, which has not been achievable using any single method before
drones. The power of drones to map shallow reefs and also provide bathymetric data at
very high resolutions (e.g., ~1 cm positioning error, [
101
103
] should facilitate this objective.
4. Enabling Technologies for Future Drone-Based Shark Research
4.1. Alternative Sensors on Drones for Shark Research
Cameras are rapidly increasing in functionality, obtaining higher resolutions and
progressively more compact and lightweight. For drone platforms, gimbal systems and
telemetry systems have also improved to allow advanced camera stabilisation and high-
Drones 2021,5, 8 15 of 28
definition transmission of video in real time. This has effectively increased the utility of
drones and subsequent use in ecology in recent years [
2
,
104
,
105
]. Similarly, the advances
in electronic component miniaturisation have allowed alternative sensors, such as thermal
infrared, multispectral (such as red edge and near infrared) and hyperspectral systems to
be mounted on small drones of less than 25 kg, and some micro-sensors on drones that are
less than 2 kg [
106
]. Such alternative sensors are available in different spatial resolutions,
just like their RGB counterparts. However, the spectral accuracy and resolution is also a
major consideration and usually scales with cost. Alternative sensors are typically used
when spectral information is desired that is also outside of the visible spectrum [
105
], or
when measurements of specific wavelengths are desired, such as for differentiating objects
or condition from spectral signatures [107,108].
The advantage of using drone-mounted platforms [
109
] as opposed to manned aircraft
or satellite is in the added flexibility of data acquisition timing, cost considerations, or
increased spatial resolution that comes with being nearer the ground [
97
,
110
]. This also
applies to the use of alternative sensors, especially considering the lower spatial resolutions
that are often associated. The vast majority of drone-based research on marine fauna is
currently done with RGB sensors. This is largely due to (1) the low signal to noise ratio
that occurs because of the attenuation properties of water, and the lack of transmission of
ultraviolet, near infrared and infrared wavelengths [
107
,
111
]; and (2) the additional cost
and expertise associated with integration, operation and data interpretation when using
alternative sensors. Consequently, the scope for which alternative sensors may offer to
marine fauna research, including observing sharks, has not yet been thoroughly researched.
Not surprisingly, most use cases of thermal infrared and multispectral sensors on
fauna have been in terrestrial environments. The use of thermal has enabled increased
detection rates compared with RGB when there is sufficient temperature difference between
target individuals and their surroundings [
112
,
113
]. Thermal infrared has also been used in
the marine environment; however, utility is restricted to animals breaking the surface [
114
].
In these cases, focus has been on detection at night or investigating temperature differentials
as indicators of animal health, such as in the case of whales [
114
,
115
]. However, because
of the submerged nature of sharks, using thermal infrared wavelengths for detection or
remotely investigating thermal properties of sharks will have little utility.
Alternatively, multispectral sensors have been reported as proof-of-concept to improve on
the detection rates of marine fauna offered from human spotters or RGB
cameras [116,117]
.
However, to our knowledge, no independent empirical assessments on the potential
increase in detection reliability have been made. Drone-mountable multispectral cameras
usually have a five or six band array, covering the colour range to red-edge or near-infrared,
and currently with full width at half maximum bandwidths typically between 10 and
40 nm, depending on the model. Fretwell, et al. [
118
] used imagery, comprising 8 colour
bands (one red-edge) and a panchromatic band, from the WorldView2 satellite, to detect
whales in Golfo Nuevo Bay, Argentina. They found that of the available bands, the coastal
band (wavelengths 400–450 nm) provided the best sub-surface features of whales due to
the better water penetration. Similarly, multispectral sensors may have utility in providing
enhanced detection of sharks from the air over RGB cameras, by selecting one or two of
the narrow colour bands. However, the ideal wavelengths and bandwidths likely differ
between locations, conditions and time of day.
Hyperspectral sensors have been used in shallow coral environments to create habitat
maps, differentiate between coral types and to assess coral health through empirically
assessing reflectance signatures [
109
,
110
]. Due to the light attenuation properties of water,
these studies typically use submersible hyperspectral units. From the air, the signal to noise
ratio decreases, however, airborne (drone-based) hyperspectral systems have proven useful
for detecting degrees of coral bleaching [
119
]. It is likely that submersible hyperspectral
sensors might have some utility on underwater drones for depicting features of demersal
elasmobranchs. Whether different features or species of elasmobranch have measurably
separable spectral signature remains uncertain. From the air, differentiating sharks from
Drones 2021,5, 8 16 of 28
other fauna based on spectral signatures might be difficult as the signal to noise ratio
decreases with the animal’s depth due to light attenuation. However, there is likely utility
in using hyperspectral sensors to define or isolate wavelengths that provide optimal sighta-
bility of sharks [
2
]. This approach is similar to that discussed with multispectral sensors,
however, due to the comparatively numerous wavelength channels on a hyperspectral
sensor, the identification of ideal wavelengths to increase detectability of sharks is likely to
be much more precise.
4.2. Artificial Intelligence for Shark Monitoring, Detection, and Alerting
Artificial intelligence (AI), encompassing systems using machine learning (ML), deep
learning and computer vision, is revolutionising ecology research across aquatic and land
environments. Particularly in recent years, the development of standardised algorithms
and the wide availability of automated sensing platforms have led to a paradigm change
in survey capability. It is now possible for AI systems to automate aspects of flight and the
detection and measurement of target species in all kinds of sensor data (e.g., photographs,
video, sound recordings and spectra). Such automated systems will make collection of
science-ready data significantly easier and cheaper, facilitate far-reaching citizen science
programs and assisting beach managers in reliably identifying potentially dangerous shark
species in real-time.
Ecologists focusing on terrestrial environments have been at the forefront of this new
wave. They have paired drone platforms and AI techniques to catch poachers [
120
,
121
],
count animals [
122
124
], detect invasive weeds [
125
], map forests [
126
,
127
] and monitor
plastic pollution [
128
]—amongst many other uses. By comparison, the adoption of AI
techniques in the marine context has been slower, likely because dynamic aquatic environ-
ments are more difficult to operate in and present greater challenges to the algorithms (see
Dujon and Schofield [129] for a recent review).
One of the earliest automated aerial surveys for marine fauna is by Maire et al. [130],
who used a convolutional neural network (CNN) to detect dugongs in aerial survey images.
By modern standards the network architecture was very simple, containing only three
convolutional blocks, which likely contributed to the high number of false positives they
report. Despite this, such deep learning techniques are now recognised as being state-of-
the-art for computer vision tasks. Modern CNNs are at the heart of mission-critical systems
like self-driving cars, e.g., [131,132] and healthcare diagnostic tools, e.g., [133].
CNN-based systems have been enthusiastically developed for the aquaculture indus-
try, where they have been deployed to identify fish species and measure their physical
properties—tasks directly relevant to shark research. Although usually applied to imagery
from underwater cameras, the underlying algorithms and training workflows are also
directly applicable to aerial imagery. For example, AI applied to underwater videos has
been very successfully demonstrated by Ditria et al. [
134
]. The authors used object de-
tection algorithms to track and identify fish species seen in video feeds from underwater
cameras at fixed locations. They demonstrated that the methods could be made to work
well on previously unseen data and on novel data from completely different sites. A key
innovation here was the development of an integrated active-learning software platform
called FishID (https://globalwetlandsproject.org/tools/fishid/) that allows new data to be
rapidly tagged and assimilated into an improved model. In another example, Fernandes
et al. [
135
] tackled a related challenge to segment the anatomy of fish species in order to
estimate body volume. This was largely accurate when applied in a controlled setting (fish
out of water on a uniformly lit background) but performance degraded when the system
was applied to other species, or in situ.
These exemplar studies and similar research showcase the possibilities of AI in a
marine setting. However, applying AI techniques to aerial imagery of ocean scenes is
generally more difficult because the dynamic air-water interface distorts sub-surface shapes.
In addition, the fidelity and quality of images are affected by weather conditions, water
Drones 2021,5, 8 17 of 28
turbidity and the presence of confusing objects (e.g., submerged reef, rocks and seaweed,
and floating foam).
The use of AI has shown immediate value to assist researchers with shark detection,
species identification and real-time alerting to beach users. There are several AI-based
shark detection systems in development globally and most of these focus on applying
computer vision techniques to aerial imagery. One striking exception is the work by Hughes
and Burghardt [
136
] who identify individual white sharks from imagery of their dorsal
fins (Figure 5). Their system applies computer-vision segmentation techniques to extract
candidate fin boundaries, which are then refined using a random forest ML algorithm. The
unique biometric notch patterns at the trailing edges of each fin are encoded and matched
to known individuals using a Bayesian nearest-neighbour classifier. The authors report an
average precision of 81% and expect the system to work well with other species of shark.
Another notable example is Clever Buoy, developed in a commercial collaboration between
Shark Mitigation Systems Ltd. and Tritech International Ltd. This system deploys sonar
transducers to detect moving objects in the water and uses AI to identify sharks over two
meters in length via their distinctive movement patterns.
Drones 2021, 5, x FOR PEER REVIEW 17 of 28
laboration between Shark Mitigation Systems Ltd.
and Tritech International Ltd. This sys-
tem deploys sonar transducers to detect moving objects in the water and uses AI to iden-
tify sharks over two meters in length via their distinctive movement patterns.
Figure 5. Evolution of artificial intelligence (AI) for shark identification: (a) automated evaluation of dorsal fins from video
collected developed by Hughes and Burghardt [126], (b) real time evaluation of shark analogues at beach (with reporting
to in-water users) demonstrated in Kiama, Australia through blimp and drone-based camera feeds by Gorkin et al. [16]
(University of Wollongong, May 2018), (c) automated marine animal detection based on images collected over Australian
beaches post-collection (Sharma et al. [127]) (UTS Nov 2018), (d) UTS system was reportedly deployed on Little Ripper
Drones in NSW beaches in Australia with Surf Life Saving as of 2019, (e) Sharkeye system reported by San Diego State
University and Sales Force—where videos collected from drones flown over Southern California could be analyzed and
reported to lifeguards (2019), and (f) in development AI on portable devices that can identify to species level in real time
by Macquarie University and NSW Department of Primary Industries research teams.
The majority of systems under development have focused on building AI systems
that learn salient features from a library of drone footage and apply this learned model to
detect sharks in live video streams. Candidate detections are annotated on visualisation
devices deployed in the field, either performing inference on-device or at a remote server.
The Little Ripper Group deployed one of the first such drone-based AI shark detection sys-
tems in Australia (Figure 5). Developed in partnership with the University of Technology
Sydney, the shark-spotting system operated on a live video feed from drones flown above
NSW beaches. Sharma et al. [127] reported on the initial development of the CNN-based
models used to detect and localise sharks, and other marine objects. In their non-field-
based tests the VGG16 network architecture delivered the most robust detections of
sharks (mAP > 90%) when applied to a 30% testing set drawn from the ensemble data.
Although accurate, VGG16 is a resource-intensive network and required desktop-class
computing hardware to achieve high frame rates (Sharma et al. [127] report a NVIDIA
Quadro P6000 GPU delivered an inference time of 0.130 s).
Other organisations have further expanded on these initial results, although these
have not yet been published as peer-reviewed papers. In the USA a team from UC Santa
Barbara’s Benioff Ocean Initiative, in collaboration with Salesforce AI Research
and com-
puter scientists at San Diego State University, have developed a system they call SharkEye.
According to their web page (https://www.sharkeye.org), data collection is done via RGB
Figure 5.
Evolution of artificial intelligence (AI) for shark identification: (
a
) automated evaluation of dorsal fins from video
collected developed by Hughes and Burghardt [
136
], (
b
) real time evaluation of shark analogues at beach (with reporting
to in-water users) demonstrated in Kiama, Australia through blimp and drone-based camera feeds by Gorkin et al. [
32
]
(University of Wollongong, May 2018), (
c
) automated marine animal detection based on images collected over Australian
beaches post-collection (Sharma et al. [
31
]) (UTS Nov 2018), (
d
) UTS system was reportedly deployed on Little Ripper
Drones in NSW beaches in Australia with Surf Life Saving as of 2019, (
e
) Sharkeye system reported by San Diego State
University and Sales Force—where videos collected from drones flown over Southern California could be analyzed and
reported to lifeguards (2019), and (
f
) in development AI on portable devices that can identify to species level in real time by
Macquarie University and NSW Department of Primary Industries research teams.
The majority of systems under development have focused on building AI systems that
learn salient features from a library of drone footage and apply this learned model to detect
sharks in live video streams. Candidate detections are annotated on visualisation devices
deployed in the field, either performing inference on-device or at a remote server. The
Little Ripper Group deployed one of the first such drone-based AI shark detection systems
Drones 2021,5, 8 18 of 28
in Australia (Figure 5). Developed in partnership with the University of Technology
Sydney, the shark-spotting system operated on a live video feed from drones flown above
NSW beaches. Sharma et al. [
31
] reported on the initial development of the CNN-based
models used to detect and localise sharks, and other marine objects. In their non-field-
based tests the VGG16 network architecture delivered the most robust detections of sharks
(
mAP >90%
) when applied to a 30% testing set drawn from the ensemble data. Although
accurate, VGG16 is a resource-intensive network and required desktop-class computing
hardware to achieve high frame rates (Sharma et al. [
31
] report a NVIDIA Quadro P6000
GPU delivered an inference time of 0.130 s).
Other organisations have further expanded on these initial results, although these
have not yet been published as peer-reviewed papers. In the USA a team from UC
Santa Barbara’s Benioff Ocean Initiative, in collaboration with Salesforce AI Research and
computer scientists at San Diego State University, have developed a system they call
SharkEye. According to their web page (https://www.sharkeye.org), data collection is done
via RGB cameras on drones flown over the beaches. The live video stream is processed on
remote servers, feeding a system that broadcasts real-time alerts to beach users.
In Australia, researchers from the University of Wollongong have developed a similar
end-to-end platform that operates with any aerial imagery (they present a case study
on a beach, based on blimp- and drone-mounted cameras). Also called SharkEye, the
system described in Gorkin et al. [
32
] performs inference in the cloud and delivers push
notifications to Apple mobile devices both on land, and critically, in the water, as the first
demonstration of personalized alerting to swimmers and surfers in real-time. Gorkin
et al. [
32
] report accuracies of 91.7%, 94.5% and 86.3% for sharks, stingrays, and surfers,
respectively. However, these metrics were derived from a limited sample of images and the
authors acknowledge that the purpose was to demonstrate the flexibility of the technology
platform, and like other groups with limited data collection, their specific detection models
will likely not generalise well to other locations, or environmental conditions. Finally,
several other small organisations have launched demonstrators of shark-detection systems.
For example, Greenroom Robotics (https://greenroomrobotics.com) built a proof-of-concept
drone-based system for a company on Reunion Island.
The greatest challenge for drone-based AI shark detection is the issue of generalisation.
Current systems are limited by: (1) the narrow range of environmental conditions sampled
in the training data, (2) the small and skewed distributions of species observed, (3) low
overall numbers of training images and (4) the small number of field sites sampled that are
not broadly representative. A generally useful shark detector would need to address all
these issues.
Supervised learning AI systems are trained using an iterative process that requires
accelerated computing hardware to progress in a reasonable time (typically 1–2 days).
High-quality annotated datasets are the key to building high-performing models and these
are extremely time-intensive to create. The data must reflect the full range of conditions
encountered in the field and (at least initially) human experts are required to manually
label each object of interest. However, the labelling process can also be assisted by AI
tools—once reasonable object detection models are available.
We note that the studies reported above assess performance by testing on a small
sub-sample split from their training data. Such tests can be misleading for the reasons
given above and reported accuracies are unlikely to hold when deployed in the field. The
tests essentially answer the question ‘How well does the AI know the current data?’ rather
than ‘How well does the AI solve the problem?’. In particular, the visual appearance of beaches
in different locations can vary dramatically, meaning that even a well-trained algorithm
would need tuning to avoid significant false positive detections. This problem is known as
‘domain shift’ and can be mitigated by using an active learning system that facilitates the
rapid assimilation of new data into the AI model.
Tools for active learning have started appearing in the literature recently. These
take the form of software that predict labels for new data and allow an ‘oracle’ user to
Drones 2021,5, 8 19 of 28
make corrections via an interactive graphical interface (e.g., ICON by Gonda et al. [
137
],
RootPainter by Smith et al. [
138
], FishID by Ditria et al. [
134
] and the Transfer Sampling
method by Kellenberger et al. [
139
]). A large trial (involving some of the authors of this
manuscript) of AI shark detection algorithms on drones has recently been completed in
NSW, Australia to systematically test the ability of ML algorithms to distinguish between
shark species, and to determine how well algorithms can generalise across a range of water
and environmental conditions. Over 4 TB of data was gathered at five beaches on the
NSW coast during March–June 2020, covering a wide range of environmental conditions.
The scientific team developed in-house active learning tools to assist with labelling the
dataset and expect to release results of the trial in early 2021. The goal is to deploy a robust
and accurate software tool to assist beach managers in confidently identifying potentially
dangerous sharks in real time. Further enhancements will aim to mitigate against surface
distortions, e.g., [
140
] and apply photogrammetric methods, e.g., [
141
] to measure shark
size, orientation and swim-parameters [8].
Taken as a whole, these capabilities have the potential to transform any intelligent
device into a data-gathering tool. For example, such automated systems could be the foun-
dation of a high-quality citizen science program. Members of the public could report shark
sightings using their personal drones, reporting consistent measurements and uncertainties.
Ecological surveys that once required expensive helicopters and manual labour could be
achieved with few resources and to greater reliability. Once properly calibrated, auto-
mated measurements could correct for confounding environmental conditions and supply
trustworthy quality-control flags and other essential metadata. Finally, drone systems are
already being deployed for spotting potentially dangerous sharks at public beaches. If AI
systems live up to their promise, then automated shark detection and tracking from drones
could be an unremarkable and trusted presence at public beaches in the near future.
4.3. The Potential of Underwater Drones
Over the past decade, underwater drones that use similar control software and hard-
ware as aerial drones have been developed to actively track marine animals while collecting
direct behavioural observations and environmental data [
142
146
]. There are two main
types of underwater drones: autonomous underwater vehicles (AUVs) and remotely oper-
ated vehicles (ROVs) (Figure 6). Autonomous underwater vehicles track an acoustically
tagged shark with or without any direct input from the pilot and without a link to the
surface, are typically torpedo-shaped, and as a result have 3 axes of movement (pitch, yaw,
roll) and only forward and reverse propulsion. Remotely operated vehicles (ROVs) for the
most part require manual operation and are tethered to the surface but can manoeuvre with
6 degrees of freedom in any direction or orientation. Autonomous underwater vehicles
have been used more often to study sharks, so we focus primarily on these machines here.
Drones 2021, 5, x FOR PEER REVIEW 19 of 28
Taken as a whole, these capabilities have the potential to transform any intelligent
device into a data-gathering tool. For example, such automated systems could be the foun-
dation of a high-quality citizen science program. Members of the public could report shark
sightings using their personal drones, reporting consistent measurements and uncertain-
ties. Ecological surveys that once required expensive helicopters and manual labour could
be achieved with few resources and to greater reliability. Once properly calibrated, auto-
mated measurements could correct for confounding environmental conditions and sup-
ply trustworthy quality-control flags and other essential metadata. Finally, drone systems
are already being deployed for spotting potentially dangerous sharks at public beaches.
If AI systems live up to their promise, then automated shark detection and tracking from
drones could be an unremarkable and trusted presence at public beaches in the near fu-
ture.
4.3. The Potential of Underwater Drones
Over the past decade, underwater drones that use similar control software and hard-
ware as aerial drones have been developed to actively track marine animals while collect-
ing direct behavioural observations and environmental data [133–137]. There are two
main types of underwater drones: autonomous underwater vehicles (AUVs) and remotely
operated vehicles (ROVs) (Figure 6). Autonomous underwater vehicles track an acousti-
cally tagged shark with or without any direct input from the pilot and without a link to
the surface, are typically torpedo-shaped, and as a result have 3 axes of movement (pitch,
yaw, roll) and only forward and reverse propulsion. Remotely operated vehicles (ROVs)
for the most part require manual operation and are tethered to the surface but can ma-
noeuvre with 6 degrees of freedom in any direction or orientation. Autonomous under-
water vehicles have been used more often to study sharks, so we focus primarily on these
machines here.
Figure 6. Examples of ‘underwater drones’ used for shark research. REMUS Automated Underwater Vehicle (AUV) used
to track acoustically-tagged white sharks (Carcharodon carcharias) (left: Image—G Skomal), and BlueRobotics BlueROV2, a
tethered, manually controlled remotely operated vehicle used to track untagged sharks post release (right: Image—T Gas-
ton).
The first attempts to autonomously track a fish with an underwater drone involved
the use of a 3 m kayak equipped with a hydrophone, receiver, and GPS that could follow
a simulated acoustically tagged fish from the surface for up to 24 h [135]. Subsequent stud-
ies used a self-propelled, subsurface AUV to passively collect data on acoustically tagged
Atlantic sturgeon (Acipenser oxyrinchus), winter flounder (Pseudopleuronectes americanus),
shortnose sturgeon (Acipenser brevirostrus), summer flounder (Paralichthys dentatus), and
sablefish (Anoplopoma fimbria) by canvassing an area along a pre-programmed path
[136,137].
Figure 6.
Examples of ‘underwater drones’ used for shark research. REMUS Automated Underwater Vehicle (AUV) used
to track acoustically-tagged white sharks (Carcharodon carcharias) (
left
: Image—G Skomal), and BlueRobotics BlueROV2, a
tethered, manually controlled remotely operated vehicle used to track untagged sharks post release (
right
: Image—T Gaston).
Drones 2021,5, 8 20 of 28
The first attempts to autonomously track a fish with an underwater drone involved the
use of a 3 m kayak equipped with a hydrophone, receiver, and GPS that could follow a simu-
lated acoustically tagged fish from the surface for up to 24 h [
144
]. Subsequent studies used
a self-propelled, subsurface AUV to passively collect data on acoustically tagged Atlantic
sturgeon (Acipenser oxyrinchus), winter flounder (Pseudopleuronectes americanus), shortnose
sturgeon (Acipenser brevirostrus), summer flounder (Paralichthys dentatus), and sablefish
(Anoplopoma fimbria) by canvassing an area along a pre-programmed path [145,146].
The first to use an underwater drone to study sharks was Clark et al. [
9
], who actively
tracked an acoustically tagged leopard shark (Triakis semifasciata) in a coastal lagoon. In
that study, the drone was constrained to the surface, was not equipped with cameras (for
behavioural observations), lacked the capacity to monitor animal depth, and resulted in
a coarse estimate of the shark’s horizontal movements. Packard et al. [
13
] were the first
to mount cameras on an underwater drone for the sole purpose of observing behaviour
while actively tracking sharks and collecting environmental data at depth. These authors
used a REMUS (Remote Environmental Monitoring UnitS; Woods Hole Oceanographic
Institution, Woods Hole, MA, USA) drone, which was developed as a platform for a wide
variety of oceanographic instrumentation and outfitted with a Global Positioning System
(GPS), wireless communication, iridium capabilities, an inertial navigation system, ring
laser gyroscopes to orient the vehicle spatially, and accelerometers to sense changes in
speed and velocity [
9
]. The drone also carried a variety of sensors including an acoustic
Doppler current profiler, a conductivity-temperature probe, magnetic heading sensor, and
pressure sensor. During this study, basking and white sharks were tracked at depth off
the coast of Cape Cod, MA and direct observations (video) and environmental data were
collected, thereby demonstrating that an AUV could actively and accurately track large
(>2 m) sharks in shallow waters (<20 m) [13].
With improvements to tracking capabilities, Skomal et al. [
6
] used this same drone
technology to investigate the behaviour, habitat use, and feeding ecology of white sharks off
Guadalupe Island, Mexico. In that study, six drone missions were conducted on four white
sharks, ranging in estimated total length from 3.9–5.7 m, for durations of 1.4–2.9 h. Based
on over 13 h of behavioural data, this study showed that the white sharks remained in the
area for the duration of each mission and moved through broad depth and temperature
ranges from the surface to 163.8 m and 27.1 to 7.9
C, respectively [
6
]. Video footage and
drone sensor data revealed that two of the white sharks being tracked and eight other white
sharks in the area approached, bumped, and/or bit the AUV during these tracks [
6
]. Not
only did this study demonstrate that a drone could be used to effectively track and observe
the behaviour of a large shark, but it also produced the first observations of subsurface
predatory behaviour for this species.
Using the behavioural and environmental datasets generated by these tracks and three
additional missions, Gabriel [
7
] conducted more detailed analyses related to underwater
drone performance, white shark behaviour, and environmental correlates. This study
concluded that the drone was not only able to track sharks more accurately, both horizon-
tally and vertically in the water column, than traditional vessel-based methods, but also
provided fine-scale, environmental datasets related to the tracks and direct observations
of white shark behaviour [
7
]. Using data collected by the drone, the first estimates of
swimming speed and course relative to water current direction/strength and changes in
depth were calculated for this species [
7
]. In addition, the influence of environmental
factors (e.g., temperature, time of day, tides, depth of thermocline, bathymetry, and wa-
ter current magnitude and direction) on swimming behaviour were also investigated [
7
].
Video observations collected by the on-board cameras allowed for the calculation of tailbeat
frequency, which remained consistent across tracks and swimming speeds [
7
]. Using a
manually piloted and tethered underwater drone controllable with 6 degrees of freedom
(remotely-operated vehicle (ROV), BlueRobotics BlueROV2) [
147
] obtained similar metrics
from small sharks (Cephaloscyllium laticeps,Squalus megalops) released into the ocean, albeit
over shorter durations than possible with AUVs.
Drones 2021,5, 8 21 of 28
Based on these studies, underwater drones can be used to monitor, follow, approach,
and image a randomly moving shark. These vehicles, which can be readily deployed in
waters inaccessible to, or unsafe, for divers (e.g., remote, rough seas, where sharks are
feeding), can produce high precision tracks [
6
,
147
] while collecting environmental data
and behavioural imagery over periods of several hours. Moreover, these vehicles are
versatile and highly customizable, and can take on different payloads to meet scientific
goals (e.g., underwater sonar, CTD probes, Niskin bottles). There is also evidence that
underwater drones may be less obtrusive than other commonplace sampling or active
tracking approaches [
148
] and may be superior for surveying abundance or behaviours
of sharks and rays than divers or snorkelers. However, like their aerial counterparts,
underwater drones might affect the natural behaviour of sharks. For example, the noise
from an AUV’s propeller might influence behaviour the same as a boat engine during active
tracking. It is anticipated that new advances in this field will ultimately be used to collect
observations over broader temporal and spatial scales and across many species [
15
,
34
], and
like aerial drones, the costs of these machines are decreasing rapidly.
5. Outlook and Conclusions
Drones have greatly enhanced the scope of research possible for scientists and man-
agers [
5
,
19
,
27
,
149
]. Sharks are inherently difficult to study due to their often migratory
nature, their sub-surface habits and the potential for negative human–shark interactions,
particularly in species that are potentially more dangerous to humans. Such data are partic-
ularly important for many species where our understanding of their ecological importance
in marine ecosystems remains rudimentary. Sharks that inhabit shallow coastal environ-
ments often occur in turbid waters or along breaking waves on beaches, making direct
observations difficult. Until recently, monitoring and observing the natural behaviour of
sharks has been limited to invasive tagging studies, animal-borne video, underwater diver
surveys, deployment of remote stationary cameras (baited or unbaited) or aerial surveys.
Tagging and animal borne video require capture and there is potential for injury and be-
havioural changes of the animals. These techniques also typically assess animal movement
on larger temporal and spatial scales. Scuba surveys are limited by the constraints of the
divers and the potential aggressiveness of the focal species, and remote cameras cannot
track individuals and are limited to whatever swims past a small field of view. Manned
aerial surveys allow vast distances to be covered quickly and can be effective for individual
shark tracking but are expensive, both financially and in labour. This review discusses the
roles that drones offer as non-intrusive and effective methods of surveying and tracking
sharks, and monitoring their behaviours.
Specific requirements and conditions need to be met to successfully deploy drones for
shark research. Once a researcher has established the appropriate equipment and training,
drones offer efficient, reliable and cost-effective ways to collect spatially explicit data that
have previously been unavailable through other methods. Drones offer direct visualisation
of sharks for surveying and mitigation strategies and technology is now underway that
allows managers to spot and monitor the behaviour of individuals to assess their risk of
interaction with humans in real time. Researchers have had the opportunity to document
direct predation events and associated behaviours of sharks in the wild and interactions
with each other without disturbance, sometimes for the first time, via drone technology.
Such behaviours are inherently difficult to document due to their brief and sporadic nature
and drones offer a highly opportunistic method for data collection in such events. Similarly,
ecological aspects of threatened species and those in complex habitats that are difficult to
monitor effectively via other technologies have been successfully assessed using drones.
The use of drones in shark research and management has rapidly risen in the past
five years and is projected for exponential increase, particularly if key technical challenges
are overcome. Many of these emerging technologies are already being developed, as
discussed in this review. Drone sensors, and camera quality and functionality need to be
further optimised to allow for higher definition video to be transmitted in real time and for
Drones 2021,5, 8 22 of 28
improved detection of sub-surface sharks. Drone design and battery improvements must
continue to allow for longer airtime and data collection. Advances in larger multirotor
systems such as hexacopters offer possibilities for refining lifting and deployment of heavy
equipment for shark-associated activities or interactive sampling and manipulation (i.e.,
drone imagery for 3d reconstruction of the body). Refinements to AI and ML are necessary
to optimise and automate processes for detection, identification and tracking sharks from
live video feed. This is particularly important considering the current bottleneck of inten-
sive data processing and video analysis and would pave the way for easier comparisons
between species in different ecosystems and environments. Finally, improvements in the
tracking capabilities, speed, manoeuvrability and broader depth ranges of autonomous
underwater vehicles are needed to jettison the capabilities of underwater drone use for
shark researchers studying deeper water or benthic species. Unique opportunities also
arise to pair aerial and underwater drone studies to allow for an integrated comparison of
shark behaviour from above and below.
It should be noted that regardless of the technological advances for drones and their
post video processing, ultimately the ubiquity of drone use for shark research comes down
to the expertise of the user, issues associated with flying beyond line-of-sight distance,
and with navigating a diversity of weather conditions. Increased training, situational
awareness and thorough flight protocols can all lessen the potential for error, but it is
difficult to imagine a scenario where human error will be eliminated from the process
entirely. Drone flights beyond line-of-sight are increasing in popularity as this allows
tracking of sharks as they venture further from the operator but also increases risk and
decreases safety. Finally, there are some weather conditions that make drone flight too
hazardous, regardless of the importance of the research at stake. As weather is inherently
changeable, it is prudent to have the ability to adequately forecast weather before and
during flight.
Our understanding of shark behaviour has been limited to date by difficulties in
obtaining direct observations of rare behaviours and fine-scale, spatially explicit data.
The use of drones in observing sharks allows scientists, managers, other stakeholders
and the broader community the opportunity to document real-time behavioural events in
relative safety. Documenting behaviours via video and imagery has the added advantage
of creating an historical record of an event that may form the foundations of future research
not considered at the time of data collection. This opportunistic footage taken by a drone
enthusiast may be valuable to researchers and managers, aligning interests of researchers
with those of the broader community. Such documentation can also be used to increase
public understanding of sharks and help challenge the inaccurate dogma of sharks being
killing machines that require lethal deterrent measures.
Author Contributions:
Conceptualization, P.A.B., V.R. and A.P.C.; methodology, all authors equally;
writing—original draft preparation, all authors equally; writing—review and editing, all authors
equally; project administration, P.A.B. All authors have read and agreed to the published version of
the manuscript.
Funding:
This research was funded by multiple sources related to individual authors and their
projects. These include (in alphabetical order): ARC LIEF Grant (LE170100007), Australian Govern-
ment, RTP Scholarship Centre de Recherche Insulaire et Observatoire de l’Environnement (CRIOBE),
Colgan Foundation, Department of Biological Science at Macquarie University, Discovery Commu-
nications, Jock Clough Marine Foundation (through the Oceans Institute Robson and Robertson
Award), NSW Department of Primary Industries (through the Shark Management Strategy), Save
our Seas Foundation (Small grant SOSF 283), Sci-eye, Sea World Research and Rescue Foundation
(SWR/13/2018), Southern Cross University and Woods Hole Oceanographic Institution.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Acknowledgments:
Thank you to the following people for their contributions to the projects encom-
passed in this review: Kevin Boswell, Craig Brand, Jose Carlos Castillo, Mike Heithaus Brendan
Drones 2021,5, 8 23 of 28
Kelaher, Kirk Gastrich, Jeremy Kiszka, Jessica Meeuwig, Guillaume Rieucau, Louise Tosetto and the
many pilots and observers that made these shark projects possible.
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Chapman, A. It’s okay to call them drones. J. Unmanned Veh. Syst. 2014,2, iii–v. [CrossRef]
2.
Colefax, A.P.; Butcher, P.A.; Kelaher, B.P. The potential for unmanned aerial vehicles (UAVs) to conduct marine fauna surveys in
place of manned aircraft. ICES J. Mar. Sci. 2018,75, 1–8. [CrossRef]
3.
Kiszka, J.J.; Mourier, J.; Gastrich, K.; Heithaus, M.R. Using unmanned aerial vehicles (UAVs) to investigate shark and ray densities
in a shallow coral lagoon. Mar. Ecol. Prog. Ser. 2016,560, 237–242. [CrossRef]
4. Rieucau, G.; Kiszka, J.J.; Castillo, J.C.; Mourier, J.; Boswell, K.M.; Heithaus, M.R. Using unmanned aerial vehicle (UAV) surveys
and image analysis in the study of large surface-associated marine species: A case study on reef sharks Carcharhinus melanopterus
shoaling behaviour. J. Fish Biol. 2018,93, 119–127. [CrossRef] [PubMed]
5.
Frixione, M.G.; García, M.D.; Gauger, M.F.W. Drone imaging of elasmobranchs: Whale sharks and golden cownose rays
co-occurrence in a zooplankton hot-spot in southwestern Sea of Cortez. Food Webs 2020,24, e00155. [CrossRef]
6.
Skomal, G.B.; Hoyos-Padilla, E.M.; Kukulya, A.; Stokey, R. Subsurface observations of white shark Carcharodon carcharias predatory
behavior using an autonomous underwater vehicle. J. Fish Biol. 2015,87, 1293–1312. [CrossRef] [PubMed]
7.
Gabriel, S. Using Autonomous Underwater Vehicles to Assess the Habitat Use and Swimming Behavior of White Sharks
(Carcharodon carcharias). Master’s Thesis, University of Massachusetts, Dartmouth, MA, USA, 2018.
8.
Ho, C.; Joly, K.; Nosal, A.P.; Lowe, C.G.; Clark, C.M. Predicting Coordinated Group Movements of Sharks with Limited
Observations using AUVs. In Proceedings of the Symposium on Applied Computing, Marrakech, Morocco, 3–7 April 2017; pp.
289–296.
9.
Clark, C.M.; Forney, C.; Manii, E.; Shinzaki, D.; Gage, C.; Farris, M.; Lowe, C.G.; Moline, M. Tracking and following a tagged
leopard shark with an autonomous underwater vehicle. J. Field Robot. 2013,30, 309–322. [CrossRef]
10.
Hensel, E.; Wenclawski, S.; Layman, C.A. Using a small, consumer-grade drone to identify and count marine megafauna in
shallow habitats. Lat. Am. J. Aquat. Res. 2018,46, 1025–1033. [CrossRef]
11.
Doan, M.D.; Kajiura, S.M. Adult blacktip sharks (Carcharhinus limbatus) use shallow water as a refuge from great hammerheads
(Sphyrna mokarran). J. Fish Biol. 2020,96, 1530–1533. [CrossRef]
12.
Benavides, M.T.; Fodrie, F.J.; Johnston, D.W. Shark detection probability from aerial drone surveys within a temperate estuary. J.
Unmanned Veh. Syst. 2020,8, 44–56. [CrossRef]
13.
Packard, G.E.; Kukulya, A.; Austin, T.; Dennett, M.; Littlefield, R.; Packard, G.; Purcell, M.; Stokey, R. Continuous autonomous
tracking and imaging of white sharks and basking sharks using a REMUS-100 AUV. In Proceedings of the 2013 Ocean Sciences
Meeting, San Diego, CA, USA, 23–27 September 2013; pp. 1–5.
14.
Fortuna, J.; Ferreira, F.; Gomes, R.; Ferreira, S.; Sousa, J. Using low cost open source UAVs for marine wild life monitoring—Field
report. IFAC Proc. 2013,2, 291–295. [CrossRef]
15.
Hawkes, L.A.; Exeter, O.; Henderson, S.M.; Kerry, C.; Kukulya, A.; Rudd, J.; Whelan, S.; Yoder, N.; Witt, M.J. Autonomous
underwater videography and tracking of basking sharks. Anim. Biotelem. 2020,8, 29. [CrossRef]
16.
Gore, M.; Abels, L.; Wasik, S.; Saddler, L.; Ormond, R. Are close-following and breaching behaviours by basking sharks at
aggregation sites related to courtship? J. Mar. Biol. Assoc. UK 2019,99, 681–693. [CrossRef]
17.
Dines, S.; Gennari, E. First observations of white sharks (Carcharodon carcharias) attacking a live humpback whale (Megaptera
novaeangliae). Mar. Freshw. Res. 2020,71, 1205–1210. [CrossRef]
18.
Lea, J.S.E.; Daly, R.; Leon, C.; Daly, C.A.K.; Clarke, C.R. Life after death: Behaviour of multiple shark species scavenging a whale
carcass. Mar. Freshw. Res. 2019,70, 302–306. [CrossRef]
19.
López, N.A.; McAuley, R.; Meeuwig, J. Identification of the southernmost aggregation of juvenile scalloped hammerhead sharks
(Sphyrna lewini) in Australia. 2021; in prepare.
20.
Gallagher, A.J.; Papastamatiou, Y.P.; Barnett, A. Apex predatory sharks and crocodiles simultaneously scavenge a whale carcass.
J. Ethol. 2018,36, 205–209. [CrossRef]
21.
Raoult, V.; Tosetto, L.; Williamson, J.E. Drone-Based High-Resolution Tracking of Aquatic Vertebrates. Drones
2018
,2, 37.
[CrossRef]
22.
Tucker, J.P.; Vercoe, B.; Santos, I.R.; Dujmovic, M.; Butcher, P.A. Whale carcass scavenging by sharks. Glob. Ecol. Conserv.
2019
,19,
e00655. [CrossRef]
23.
Colefax, A.P.; Kelaher, B.P.; Pagendam, D.E.; Butcher, P.A. Assessing white shark (Carcharodon carcharias) behaviour along coastal
beaches for conservation-focused shark mitigation. Front. Mar. Sci. 2020,7, 268. [CrossRef]
24.
Tucker, J.P.; Colefax, A.P.; Santos, I.R.; Kelaher, B.P.; Pagendam, D.E.; Butcher, P.A. White shark behaviour altered by stranded
whale carcasses: Insights from drones and implications for beach management Ocean Coast.Manag. 2021,200, 105477.
25.
Colefax, A.P.; Butcher, P.A.; Pagendam, D.E.; Kelaher, B.P. Reliability of marine faunal detections in drone-based monitoring.
Ocean Coast. Manag. 2019,174, 108–115. [CrossRef]
26.
Colefax, A.P.; Butcher, P.A.; Pagendam, D.E.; Kelaher, B.P. Comparisons of localised distributions of white, bull, and tiger sharks
using three tech-based methods. Ocean Coast. Manag. 2020,198, 105366. [CrossRef]
Drones 2021,5, 8 24 of 28
27.
Colefax, A.P.; Kelaher, B.P.; Walsh, A.J.; Purcell, C.R.; Pagendam, D.E.; Cagnazzi, D.D.B.; Butcher, P.A. Utility of spectral band selection
from drone-based hyperspectral imagery for improving detectability of submerged marine fauna. Biol. Conserv. 2021, submitted.
28.
Kelaher, B.P.; Colefax, A.P.; Tagliafico, A.; Bishop, M.J.; Giles, A.; Butcher, P.A. Assessing variation in assemblages of large marine
fauna off ocean beaches using drones. Mar. Freshw. Res. 2019,71, 68–77. [CrossRef]
29.
Kelaher, B.P.; Peddemors, V.M.; Hoade, B.; Colefax, A.P.; Butcher, P.A. Comparison of sampling precision for nearshore marine
wildlife using unmanned and manned aerial surveys. J. Unmanned Veh. Syst. 2019,8, 30–43. [CrossRef]
30.
Saqib, M.; Khan, S.D.; Sharma, N.; Scully-Power, P.; Butcher, P.; Colefax, A.; Blumenstein, M. Real-time drone surveillance and
population estimation of marine animals from aerial imagery. In Proceedings of the 2018 International Conference on Image and
Vision Computing New Zealand, Auckland, New Zealand, 19–21 November 2018; pp. 1–6.
31.
Sharma, N.; Scully-Power, P.; Blumenstein, M. Shark Detection from Aerial Imagery Using Region-Based CNN, a Study; Mitrovic, T.,
Xue, B., Li, X., Eds.; AI 2018: Advances in Artificial Intelligence; Springer International Publishing: Cham, Switzerland, 2018; pp.
224–236.
32.
Gorkin, R., III; Adams, K.; Berryman, M.J.; Aubin, S.; Li, W.; Davis, A.R.; Barthelemy, J. Sharkeye: Real-Time Autonomous
Personal Shark Alerting via Aerial Surveillance. Drones 2020,4, 18. [CrossRef]
33.
Butcher, P.; Piddock, T.; Colefax, A.; Hoade, B.; Peddemors, V.; Borg, L.; Cullis, B. Beach safety: Can drones provide a platform for
sighting sharks? Wildl. Res. 2019,46, 701–712. [CrossRef]
34.
Raoult, V.; Tosetto, L.; Harvey, C.; Nelson, T.M.; Reed, J.; Parikh, A.; Chan, A.J.; Smith, T.M.; Williamson, J.E. Remotely operated
vehicles as alternatives to snorkellers for video-based marine research. J. Exp. Mar. Biol. Ecol. 2020,522, 1–10. [CrossRef]
35.
Dulvy, N.K.; Fowler, S.L.; Musick, J.A.; Cavanagh, R.D.; Kyne, P.M.; Harrison, L.R.; Carlson, J.K.; Davidson, L.N.; Fordham, S.V.;
Francis, M.P.; et al. Extinction risk and conservation of the world’s sharks and rays. eLife 2014,3, 1–34. [CrossRef]
36.
Roff, G.; Brown, C.J.; Priest, M.A.; Mumby, P.J. Decline of coastal apex shark populations over the past half century. Commun. Biol.
2018,1, 1–11. [CrossRef]
37.
Pepin-Neff, C.L.; Wynter, T. Reducing fear to influence policy preferences: An experiment with sharks and beach safety policy
options. Mar. Policy 2018,88, 222–229. [CrossRef]
38.
Chirayath, V.; Earle, S.A. Drones that see through waves—Preliminary results from airborne fluid lensing for centimetre-scale
aquatic conservation. Aquat. Conserv. Mar. Freshw. Ecosyst. 2016,26 (Suppl. 2), 237–250. [CrossRef]
39.
Ferguson, M.C.; Angliss, R.P.; Kennedy, A.; Lynch, B.; Willoughby, A.; Helker, V.; Brower, A.A.; Clarke, J.T. Performance of
manned and unmanned aerial surveys to collect visual data and imagery for estimating arctic cetacean density and associated
uncertainty. J. Unmanned Veh. Syst. 2018,6, 128–154. [CrossRef]
40.
Robbins, W.D.; Peddemors, V.M.; Kennelly, S.J.; Ives, M.C. Experimental evaluation of shark detection rates by aerial observers.
PLoS ONE 2014,9, e83456. [CrossRef] [PubMed]
41.
Stokes, D.; Apps, K.; Butcher, P.A.; Weiler, B.; Luke, H.; Colefax, A.P. Beach-user perceptions and attitudes towards drone
surveillance as a shark mitigation tool. Mar. Policy 2020,120, 104127. [CrossRef]
42.
Provost, E.J.; Butcher, P.A.; Colefax, A.P.; Coleman, M.A.; Curley, B.G.; Kelaher, B.P. Using drones to quantify beach users across a
range of environmental conditions. J. Coast. Conserv. 2019,23, 633–642. [CrossRef]
43.
Giles, A.B.; Butcher, P.A.; Colefax, A.P.; Pagendam, D.E.; Kelaher, B.P. Responses of bottlenose dolphins (Tursiops spp.) to small
drones. Aquat. Conserv. Mar. Freshw. Ecosyst. 2020, 1–8. [CrossRef]
44.
Klimley, A.P.; Anderson, S.D.; Pyle, P.; Henderson, R.P. Spatiotemporal Patterns of White Shark (Carcharodon carcharias) Predation
at the South Farallon Islands, California. Copeia 1992,3, 680–690. [CrossRef]
45.
Christiansen, F.; Rojano-Doñate, L.; Madsen, P.T.; Bejder, L. Noise Levels of Multi-Rotor Unmanned Aerial Vehicles with
Implications for Potential Underwater Impacts on Marine Mammals. Front. Mar. Sci. 2016,3, 277. [CrossRef]
46.
Porter, M.E.; Ruddy, B.R.; Kajiura, S.M. Volitional Swimming Kinematics of Blacktip Sharks, Carcharhinus limbatus, in the Wild.
Drones 2020,4, 78. [CrossRef]
47.
Kajiura, S.M.; Tellman, S.L. Quantification of massive seasonal aggregations of blacktip sharks (Carcharhinus limbatus) in southeast
Florida. PLoS ONE 2016,11, e0150911. [CrossRef] [PubMed]
48.
Raoult, V.; Broadhurst, M.K.; Peddemors, V.M.; Williamson, J.E.; Gaston, T.F. Resource use of great hammerhead sharks (Sphyrna
mokarran) off eastern Australia. J. Fish Biol. 2019,95, 1430–1440. [CrossRef] [PubMed]
49.
Tagliafico, A.; Butcher, P.A.; Colefax, A.P.; Clark, G.F.; Kelaher, B.P. Variation in cownose ray Rhinoptera neglecta abundance and
group size on the central east coast of Australia. J. Fish Biol. 2020,96, 427–433. [CrossRef] [PubMed]
50.
Torney, C.J.; Lamont, M.; Debell, L.; Angohiatok, R.J.; Leclerc, L.-M.; Berdahl, A.M. Inferring the rules of social interaction in
migrating caribou. Philos. Trans. R. Soc. B Biol. Sci. 2018,373, 20170385. [CrossRef]
51.
Harris, J.M.; Nelson, J.A.; Rieucau, G.; Broussard, W.P. Use of Drones in Fishery Science. Trans. Am. Fish. Soc.
2019
,148, 687–697.
[CrossRef]
52.
Spaet, J.L.Y.; Patterson, T.A.; Bradford, R.W.; Butcher, P.A. Spatiotemporal distribution patterns of immature Australasian white
sharks (Carcharodon carcharias). Sci. Rep. 2020,10, 10169. [CrossRef]
53.
Curtis, T.H.; Kelly, J.; Menard, K.; Laroche, R.; Jones, R.; Klimley, A.P. Observations on the behavior of White Sharks scavenging
from a Whale carcass at Point Reyes, California. Calif. Fish Game 2006,92, 113–124.
54.
Clua, E.; Chauvet, C.; Read, T.; Werry, J.M.; Lee, S.Y. Behavioural patterns of a Tiger Shark (Galeocerdo cuvier) feeding aggregation
at a Blue Whale carcass in Prony Bay, New Caledonia. Mar. Freshw. Behav. Physiol. 2013,46, 1–20. [CrossRef]
Drones 2021,5, 8 25 of 28
55.
Dicken, M. First observations of young of the year and juvenile Great White Sharks (Carcharodon carcharias) scavenging from a
whale carcass. Mar. Freshw. Res. 2008,59, 596–602. [CrossRef]
56.
Tucker, J.P.; Santos, I.R.; Crocetti, S.; Butcher, P. Whale carcass strandings on beaches: Management challenges, research needs,
and examples from Australia. Ocean Coast. Manag. 2018,163, 323–338. [CrossRef]
57.
Tucker, J.P.; Santos, I.R.; Davis, K.L.; Butcher, P.A. Whale carcass leachate plumes in beach groundwater: A potential shark
attractant to the surf? Mar. Pollut. Bull. 2019,140, 219–226. [CrossRef]
58. Fowler, S. The Conservation Status of Migratory Sharks; UNEP/CMS Secretariat: Bonn, Germany, 2014; 30p.
59.
Gallagher, A.J.; Hammerschlag, N.; Danylchuk, A.J.; Cooke, S.J. Shark recreational fisheries: Status, challenges, and research
needs. Ambio 2017,46, 385–398. [CrossRef]
60.
Fields, A.T.; Fischer, G.A.; Shea, S.K.H.; Zhang, H.; Abercrombie, D.L.; Feldheim, K.A.; Babcock, E.A.; Chapman, D.D. Species
composition of the international shark fin trade assessed through a retail-market survey in Hong Kong. Conserv. Biol.
2018
,32,
376–389. [CrossRef]
61.
Dent, F.; Clarke, S. State of the global market for shark products. In FAO Fisheries and Aquaculture Technical Paper No. 590; FAO:
Rome, Italy, 2015; 187p.
62.
Ferretti, F.; Myers, R.A.; Serena, F.; Lotze, H.K. Loss of large predatory sharks from the Mediterranean Sea. Conserv. Biol.
2008
,22,
952–964. [CrossRef] [PubMed]
63.
Hayes, C.G.; Jiao, Y.; Cortés, E. Stock Assessment of Scalloped Hammerheads in the Western North Atlantic Ocean and Gulf of
Mexico. North Am. J. Fish. Manag. 2009,29, 1406–1417. [CrossRef]
64.
Hutchings, J.A.; Myers, R.A.; García, V.B.; Lucifora, L.O.; Kuparinen, A. Life-history correlates of extinction risk and recovery
potential. Ecol. Appl. 2012,22, 1061–1067. [CrossRef]
65.
Dulvy, N.K.; Baum, J.K.; Clarke, S.; Compagno, L.J.V.; Cortés, E.; Domingo, A.; Fordham, S.; Fowler, S.; Francis, M.P.; Gibson, C.;
et al. You can swim but you can’t hide: The global status and conservation of oceanic pelagic sharks and rays. Aquat. Conserv.
Mar. Freshw. Ecosyst. 2008,18, 459–482. [CrossRef]
66.
Ketchum, J.T.; Galván-Magaña, F.; Klimley, A.P. Segregation and foraging ecology of whale sharks, Rhincodon typus, in the
southwestern Gulf of California. Environ. Biol. Fishes 2013,96, 779–795. [CrossRef]
67.
Simpfendorfer, C.A.; Mildward, N.E. Utilisation of a tropical bay as a nursery area by sharks of the families Carcharhinidae and
Sphyrnidae. Environ. Biol. Fishes 1993,37, 337–345. [CrossRef]
68.
Heupel, M.R.; Simpfendorfer, C.A. Quantitative analysis of aggregation behavior in juvenile blacktip sharks. Mar. Biol.
2005
,147,
1239–1249. [CrossRef]
69.
Rowat, D.; Brooks, K.; March, A.; McCarten, C.; Jouannet, D.; Riley, L.; Jeffreys, G.; Perri, M.; Vely, M.; Pardigon, B. Long-term
membership of whale sharks (Rhincodon typus) in coastal aggregations in Seychelles and Djibouti. Mar. Freshw. Res.
2011
,62,
621–627. [CrossRef]
70.
Nalesso, E.; Hearn, A.; Sosa-Nishizaki, O.; Steiner, T.; Antoniou, A.; Reid, A.; Bessudo, S.; Soler, G.; Klimley, P.; Lara, F.; et al.
Movements of scalloped hammerhead sharks (Sphyrna lewini) at Cocos Island, Costa Rica and between oceanic islands in the
Eastern Tropical Pacific. PLoS ONE 2019,14, e0213741. [CrossRef]
71.
Mucientes, G.R.; Queiroz, N.; Sousa, L.L.; Tarroso, P.; Sims, D.W. Sexual segregation of pelagic sharks and the potential threat
from fisheries. Biol. Lett. 2009,5, 156–159. [CrossRef]
72.
Clarke, S.C.; McAllister, M.K.; Milner-Gulland, E.J.; Kirkwood, G.P.; Michielsens, C.G.J.; Agnew, D.J.; Pikitch, E.K.;
Nakano, H.
;
Shivji, M.S. Global estimates of shark catches using trade records from commercial markets. Ecol. Lett.
2006
,9, 1115–1126.
[CrossRef]
73.
Graham, F.; Rynne, P.; Estevanez, M.; Luo, J.; Ault, J.S.; Hammerschlag, N. Use of marine protected areas and exclusive economic
zones in the subtropical western North Atlantic Ocean by large highly mobile sharks. Divers. Distrib.
2016
,22, 534–546. [CrossRef]
74.
Queiroz, N.; Humphries, N.E.; Mucientes, G.; Hammerschlag, N.; Lima, F.P.; Scales, K.L.; Miller, P.I.; Sousa, L.L.; Seabra, R.; Sims,
D.W. Ocean-wide tracking of pelagic sharks reveals extent of overlap with longline fishing hotspots. Proc. Natl. Acad. Sci. USA
2016,113, 1582–1587. [CrossRef]
75. Compagno, L.J.V. Sharks of the World; Princeton University Press: Princeton, NJ, USA, 2005.
76.
Gallagher, A.J.; Klimley, A.P. The biology and conservation status of the large hammerhead shark complex: The great, scalloped,
and smooth hammerheads. Rev. Fish Biol. Fish. 2018,28, 777–794. [CrossRef]
77.
Bessudo, S.; Soler, G.A.; Klimley, P.A.; Ketchum, J.; Arauz, R.; Hearn, A.; Guzmán, A.; Calmettes, B. Vertical and horizontal
movements of the scalloped hammerhead shark (Sphyrna lewini) round Malpelo and Cocos Islands (Tropical Eastern Pacific)
using satellite telemetry. Bull. Mar. Coast. Res. 2011,40, 91–106.
78.
Hammerschlag, N.; Gallagher, A.J.; Lazarre, D.M.; Slonim, C. Range extension of the endangered great hammerhead shark
Sphyrna mokarran in the Northwest Atlantic: Preliminary data and significance for conservation. Endanger. Species Res.
2011
,13,
111–116. [CrossRef]
79.
Santos, C.; Coehlo, R. Migrations and habitat use of the smooth hammerhead shark (Sphyrna zygaena) in the Atlantic Ocean. PLoS
ONE 2018,13, e0198664. [CrossRef]
80.
Roemer, R.P.; Gallagher, A.J.; Hammerschlag, N. Shallow water tidal flat use and associated specialized foraging behavior of the
great hammerhead shark (Sphyrna mokarran). Mar. Freshw. Behav. Physiol. 2016,49, 235–249. [CrossRef]
Drones 2021,5, 8 26 of 28
81.
Hearn, A.; Ketchum, J.; Klimley, A.P.; Espinoza, E.; Peñaherrera, C. Hotspots within hotspots? Hammerhead shark movements
around Wolf Island, Galapagos Marine Reserve. Mar. Biol. 2010,157, 1899–1915. [CrossRef] [PubMed]
82.
Francis, M.P. Distribution, habitat and movement of juvenile smooth hammerhead sharks (Sphyrna zygaena) in northern New
Zealand. N. Z. J. Mar. Freshw. Res. 2016,50, 506–525. [CrossRef]
83.
Brown, K.T.; Seeto, J.; Lal, M.M.; Miller, C.E. Discovery of an important aggregation area for endangered scalloped hammerhead
sharks, Sphyrna lewini, in the Rewa River estuary, Fiji Islands. Pac. Conserv. Biol. 2016,22, 242–248. [CrossRef]
84.
Duncan, K.M.; Holland, K.N. Habitat use, growth rates and dispersal patterns of juvenile scalloped hammerhead sharks Sphyrna
lewini in a nursery habitat. Mar. Ecol. Prog. Ser. 2006,312, 211–221. [CrossRef]
85. Jennings, R.D. Seasonal abundance of hammerhead sharks off Cape Canaveral, Florida. Copeia 1985, 223–225. [CrossRef]
86.
Kenney, R.D.; Owen, R.E.; Winn, H.E. Shark distributions off the Northeast United States from Marine Mammal Surveys. Copeia
1985,1985, 220–223. [CrossRef]
87.
Dicken, M.L.; Booth, A.J. Surveys of white sharks (Carcharodon carcharias) off bathing beaches in Algoa Bay, South Africa. Mar.
Freshw. Res. 2013,64, 530–539. [CrossRef]
88.
Laran, S.; Authier, M.; Van Canneyt, O.; Dorémus, G.; Watremez, P.; Ridoux, V. A comprehensive survey of pelagic megafauna:
Their distribution, densities, and taxonomic richness in the tropical Southwest Indian ocean. Front. Mar. Sci.
2017
,4, 139.
[CrossRef]
89.
Ducatez, S. Which sharks attract research? Analyses of the distribution of research effort in sharks reveal significant non-random
knowledge biases. Rev. Fish Biol. Fish. 2019,29, 355–367. [CrossRef]
90.
Osgood, G.; Baum, J. Reef sharks: Recent advances in ecological understanding to inform conservation. J. Fish Biol.
2015
,87,
1489–1523. [CrossRef] [PubMed]
91.
Heupel, M.; Simpfendorfer, C. Using acoustic monitoring to evaluate MPAs for shark nursery areas: The importance of long-term
data. Mar. Technol. Soc. J. 2005,39, 10–18. [CrossRef]
92.
Heupel, M.R.; Lédée, E.J.; Simpfendorfer, C.A. Telemetry reveals spatial separation of co-occurring reef sharks. Mar. Ecol. Prog.
Ser. 2018,589, 179–192. [CrossRef]
93.
Cagua, E.F.; Berumen, M.L.; Tyler, E. Topography and biological noise determine acoustic detectability on coral reefs. Coral Reefs
2013,32, 1123–1134. [CrossRef]
94.
Whitmarsh, S.K.; Fairweather, P.G.; Huveneers, C. What is Big BRUVver up to? Methods and uses of baited underwater video.
Rev. Fish Biol. Fish. 2017,27, 53–73. [CrossRef]
95.
Barker, S.M.; Peddemors, V.M.; Williamson, J.E. A video and photographic study of aggregation, swimming and respiratory
behaviour changes in the Grey Nurse Shark (Carcharias taurus) in response to the presence of SCUBA divers. Mar. Freshw. Behav.
Physiol. 2011,44, 75–92. [CrossRef]
96.
Smith, K.; Scarr, M.; Scarpaci, C. Grey nurse shark (Carcharias taurus) diving tourism: Tourist compliance and shark behaviour at
Fish Rock, Australia. Environ. Manag. 2010,46, 699–710. [CrossRef]
97.
Joyce, K.; Duce, S.; Leahy, S.; Leon, J.X.; Maier, S. Principles and practice of acquiring drone based image data in marine
environments. Mar. Freshw. Res. 2018,70, 952–963. [CrossRef]
98.
Carlisle, A.B.; Starr, R.M. Habitat use, residency, and seasonal distribution of female leopard sharks Triakis semifasciata in Elkhorn
Slough, California. Mar. Ecol. Prog. Ser. 2009,380, 213–228. [CrossRef]
99.
Nakano, H.; Matsunaga, H.; Okamoto, H.; Okazaki, M. Acoustic tracking of bigeye thresher shark Alopias superciliosus in the
eastern Pacific Ocean. Mar. Ecol. Prog. Ser. 2003,265, 255–261. [CrossRef]
100.
Kessel, S.T.; Gruber, S.; Gledhill, K.; Bond, M.; Perkins, R.G. Aerial survey as a tool to estimate abundance and describe distribution
of a carcharhinid species, the lemon shark, Negaprion brevirostris.J. Mar. Biol. 2013,2013, 1–10. [CrossRef]
101.
Bennett, M.K.; Younes, N.; Joyce, K. Automating Drone Image Processing to Map Coral Reef Substrates Using Google Earth
Engine. Drones 2020,4, 50. [CrossRef]
102.
Casella, E.; Collin, A.; Harris, D.; Ferse, S.; Bejarano, S.; Parravicini, V.; Hench, J.L.; Rovere, A. Mapping coral reefs using
consumer-grade drones and structure from motion photogrammetry techniques. Coral Reefs 2017,36, 269–275. [CrossRef]
103.
Kabiri, K.; Rezai, H.; Moradi, M. A drone-based method for mapping the coral reefs in the shallow coastal waters–case study:
Kish Island, Persian gulf. Earth Sci. Inform. 2020,13, 1265–1274. [CrossRef]
104.
Chabot, D. Trends in drone research and applications as the Journal of Unmanned Vehicle Systems turns five. J. Unmanned Veh.
Syst. 2018,6, vi–xv. [CrossRef]
105.
Hardin, P.J.; Lulla, V.; Jensen, R.R.; Jensen, J.R. Small Unmanned Aerial Systems (sUAS) for environmental remote sensing:
Challenges and opportunities revisited. GISci. Remote Sens. 2018,56, 309–322. [CrossRef]
106.
Johnston, D.W. Unoccupied aircraft systems in marine science and conservation. Annu. Rev. Mar. Sci.
2019
,11, 439–463.
[CrossRef]
107.
Letnes, P.A.; Hansen, I.M.; Aas, L.M.S.; Eide, I.; Pettersen, R.; Tassara, L.; Receveur, J.; le Floch, S.; Guyomarch, J.; Camus, L.;
et al. Underwater hyperspectral classification of deep sea corals exposed to 2-methylnaphthalene. PLoS ONE
2019
,14, e0209960.
[CrossRef]
108.
Chennu, A.; Farber, P.; De’ath, G.; de Beer, D.; Fabricius, K.E. A diver-operated hyperspectral imaging and topographic surveying
system for automated mapping of benthic habitats. Sci. Rep. 2017,7, 7122. [CrossRef]
Drones 2021,5, 8 27 of 28
109.
Colefax, A. Developing the Use of Drones for Non-Destructive Shark Management and Beach Safety. Ph.D. Thesis, Southern
Cross University, Lismore, Australia, 2020.
110.
Hodgson, J.C.; Baylis, S.M.; Mott, R.; Herrod, A.; Clarke, R.H. Precision wildlife monitoring using unmanned aerial vehicles. Sci.
Rep. 2016,6, 22574. [CrossRef]
111.
Pope, R.M.; Fry, E.S. Absorption spectrum (380–700 nm) of pure water. II. Integrating cavity measurements. Appl. Opt.
1997
,36,
8710–8723. [CrossRef] [PubMed]
112.
Seymour, A.C.; Dale, J.; Hammill, M.; Halpin, P.N.; Johnston, D.W. Automated detection and enumeration of marine wildlife
using unmanned aircraft systems (UAS) and thermal imagery. Sci. Rep. 2017,7, 45127. [CrossRef] [PubMed]
113.
Spaan, D.; Burke, C.; McAree, O.; Aureli, F.; Rangel-Rivera, C.E.; Hutschenreiter, A.; Longmore, S.N.; McWhirter, P.R.; Wich, S.A.
Thermal Infrared Imaging from Drones Offers a Major Advance for Spider Monkey Surveys. Drones 2019,3, 34. [CrossRef]
114.
Horton, T.W.; Hauser, N.; Cassel, S.; Klaus, K.F.; Fettermann, T.; Key, N. Doctor Drone: Non-invasive Measurement of Humpback
Whale Vital Signs Using Unoccupied Aerial System Infrared Thermography. Front. Mar. Sci. 2019,6, 466. [CrossRef]
115. Thomas, G.L.; Thorne, R.E. Night-time predation by Steller sea lions. Nature 2001,411, 1013. [CrossRef] [PubMed]
116.
Schoonmaker, J.S.; Podobna, Y.; Boucher, C.D. Electro-optical approach for airborne marine mammal surveys and density
estimations. U.S. Navy J. Underw. Acoust. 2011,61, 668–985.
117.
Blount, C.; Schoonmaker, J.; Saggese, S.; Oakley, D. An Innovative Method for Obtaining High Detection Rates of Sharks on Ocean
Beaches; A Report for Shark Alert Pty Ltd.; Cardno: Sydney, NSW, Australia, 2016; 35p.
118.
Fretwell, P.T.; Staniland, I.J.; Forcada, J. Whales from space: Counting southern right whales by satellite. PLoS ONE
2014
,9,
e88655. [CrossRef]
119.
Parsons, M.; Bratanov, D.; Gaston, K.J.; Gonzalez, F. UAVs, Hyperspectral Remote Sensing, and Machine Learning Revolutionizing
Reef Monitoring. Sensors 2018,18, 2026. [CrossRef]
120.
Burke, C.; Rashman, M.F.; McAree, O.; Hambrecht, L.; Longmore, S.N.; Piel, A.K.; Wich, S.A. Addressing environmental and
atmospheric challenges for capturing high-precision thermal infrared data in the field of astro-ecology. In Proceedings Volume
10709, High Energy, Optical, and Infrared Detectors for Astronomy VIII; SPIE Astronomical Telescopes + Instrumentation: Austin, TX,
USA, 2018.
121.
Hambrecht, L.; Brown, R.P.C.; Piel, A.K.; Wich, S.A. Detecting ‘poachers’ with drones: Factors influencing the probability of
detection with TIR and RGB imaging in miombo woodlands, Tanzania. Biol. Conserv. 2019,233, 109–117. [CrossRef]
122.
Hodgson, J.; Mott, R.; Baylis, S.; Pham, T.; Wotherspoon, S.; Kilpatrick, A.; Raja Segaran, R.; Reid, I.; Terauds, A.; Koh, L. Drones
count wildlife more accurately and precisely than humans. Methods Ecol. Evol. 2018,9, 1160–1167. [CrossRef]
123.
Burr, P.; Samiappan, S.; Hathcock, L.; Moorhead, R.; Dorr, B. Estimating waterbird abundance on catfish aquaculture ponds using
an unmanned aerial system. Hum. Wildl. Interact. 2019,13. [CrossRef]
124.
Eikelboom, J.; Wind, J.; Ven, E.; Kenana, M.; Schroder, B.; Knegt, H.; Langevelde, F.; Prins, H. Improving the precision and
accuracy of animal population estimates with aerial image object detection. Methods Ecol. Evol. 2019,10, 1875–1887. [CrossRef]
125.
Sandino, J.; Gonzalez, F. A novel approach for invasive weeds and vegetation surveys using UAS and Artificial Intelligence. In
Proceedings of the 2018 23rd International Conference on Methods Models in Automation Robotics (MMAR), Mi˛edzyzdroje,
Poland, 27–30 August 2018; pp. 515–520.
126.
Nevalainen, O.; Honkavaara, E.; Tuominen, S.; Viljanen, N.; Hakala, T.; Yu, X.; Hyyppä, J.; Saari, H.; Pölönen, I.; Imai, N.; et al.
Individual tree detection and classification with UAV-based photogrammetric point clouds and hyperspectral imaging. Remote
Sens. 2017,9, 185. [CrossRef]
127.
Sandino, J.; Pegg, G.; Gonzalez, F.; Smith, G. Aerial Mapping of forests affected by pathogens using UAVs, hyperspectral sensors,
and artificial intelligence. Sensors 2018,18, 944. [CrossRef] [PubMed]
128.
Geraeds, M.; van Emmerik, T.; de Vries, R.; Ab Razak, M.S. Riverine plastic litter monitoring using unmanned aerial vehicles
(UAVs). Remote Sens. 2019,11, 2045. [CrossRef]
129.
Dujon, A.; Schofield, G. Importance of machine learning for enhancing ecological studies using information-rich imagery.
Endanger. Species Res. 2019,39, 91–104. [CrossRef]
130.
Maire, F.; Alvarez, L.M.; Hodgson, A. Automating marine mammal detection in aerial images captured during wildlife surveys:
A deep learning approach. In AI 2015: Advances in Artificial Intelligence; Pfahringer, B., Renz, J., Eds.; Springer International
Publishing: Cham, Switzerland, 2015; pp. 379–385.
131.
Dharmawan, W.; Nambo, H. End-to-End Xception model implementation on Carla Self Driving Car in moderate dense environ-
ment. In Proceedings of the 2019 2nd Artificial Intelligence and Cloud Computing Conference, AICCC 2019, Kobe, Japan, 21–23
December 2019; Association for Computing Machinery: New York, NY, USA, 2019; pp. 139–143.
132.
Sanil, N.; Rakesh, V.; Mallapur, R.; Ahmed, M.R. Deep learning techniques for obstacle detection and avoidance in driverless cars.
In Proceedings of the 2020 International Conference on Artificial Intelligence and Signal Processing (AISP), Vellore, India, 10–12
January 2020; pp. 1–4.
133.
Ismail, W.N.; Hassan, M.M.; Alsalamah, H.A.; Fortino, G. CNN-Based health model for regular health factors analysis in
internet-of-medical things environment. IEEE Access 2020,8, 52541–52549. [CrossRef]
134.
Ditria, E.M.; Lopez-Marcano, S.; Sievers, M.; Jinks, E.L.; Brown, C.J.; Connolly, R.M. Automating the analysis of fish abundance
using object detection: Optimizing animal ecology with deep learning. Front. Mar. Sci. 2020,7, 429. [CrossRef]
Drones 2021,5, 8 28 of 28
135.
Fernandes, A.F.; Turra, E.M.; de Alvarenga, É.R.; Passafaro, T.L.; Lopes, F.B.; Alves, G.F.; Singh, V.; Rosa, G.J. Deep Learning
image segmentation for extraction of fish body measurements and prediction of body weight and carcass traits in Nile tilapia.
Comput. Electron. Agric. 2020,170, 105274. [CrossRef]
136.
Hughes, B.; Burghardt, T. Automated visual fin identification of individual great white sharks. Int. J. Comput. Vis.
2017
,122, 542.
[CrossRef]
137.
Gonda, F.; Kaynig, V.; Jones, T.R.; Haehn, D.; Lichtman, J.W.; Parag, T.; Pfister, H. ICON: An Interactive Approach to Train Deep
Neural Networks for Segmentation of Neuronal Structures. In Proceedings of the 2017 IEEE 14th International Symposium on
Biomedical Imaging (ISBI 2017), Melbourne, Australia, 18–21 April 2017; pp. 327–331.
138.
Smith, A.G.; Han, E.; Petersen, J.; Olsen, N.A.F.; Giese, C.; Athmann, M.; Dresbøll, D.B.; Thorup-Kristensen, K. RootPainter: Deep
learning segmentation of biological images with corrective annotation. bioRxiv 2020. [CrossRef]
139.
Kellenberger, B.; Marcos, D.; Lobry, S.; Tuia, D. Half a percent of labels is enough: Efficient animal detection in UAV imagery
using deep CNNs and active learning. IEEE Trans. Geosci. Remote Sens. 2019,57, 9524–9533. [CrossRef]
140. Chirayath, V.; Li, A. Next-Generation optical sensing technologies for exploring ocean worlds—NASA FluidCam, MiDAR, and
NeMO-Net. Front. Mar. Sci. 2019,6, 521. [CrossRef]
141.
Gray, P.C.; Bierlich, K.C.; Mantell, S.A.; Friedlaender, A.S.; Goldbogen, J.A.; Johnston, D.W. Drones and convolutional neural
networks facilitate automated and accurate cetacean species identification and photogrammetry. Methods Ecol. Evol.
2019
,10,
1490–1500. [CrossRef]
142.
Lowe, C.G.; White, C.F.; Clark, C.M. Use of autonomous vehicles for tracking and surveying of acoustically tagged elasmobranchs.
In Shark Research: Emerging Technologies and Applications for the Field and Laboratory; Carrier, J., Heithaus, M., Simpfendorfer, C.,
Eds.; CRC Press: Boca Raton, FL, USA, 2018.
143.
Eiler, J.H.; Grothues, T.M.; Dobarro, J.A.; Masuda, M.M. Comparing autonomous underwater vehicle (AUV) and vessel-based
tracking performance for locating acoustically tagged fish. Mar. Fish. Rev. 2013,75, 27–42. [CrossRef]
144.
Goudey, C.A.; Consi, T.; Manley, J.; Graham, M.; Donovan, B.; Kiley, L. A robotic boat for autonomous fish tracking. Mar. Technol.
Soc. J. 1998,32, 47.
145.
Grothues, T.; Dobarro, J.; Eiler, J. Collecting, interpreting, and merging fish telemetry data from an AUV: Remote sensing from an
already remote platform. In Proceedings of the 2010 Autonomous Underwater Vehicles Symposium, Monterey, CA, USA, 1–3
September 2010; Volume 136, pp. 1–9.
146.
Grothues, T.; Dobarro, J.; Ladd, J.; Higgs, A.; Niezgoda, G.; Miller, D. Use of a multi-sensored AUV to telemeter tagged Atlantic
sturgeon and map their spawning habitat in the Hudson River, USA. In Proceedings of the 2008 Autonomous Underwater
Vehicles Symposium, Woods Hole, MA, USA, 13–14 October 2008; pp. 1–7.
147.
Raoult, V.; Williamson, J.E.; Smith, T.M.; Gaston, T.F. Effects of on-deck holding conditions and air exposure on post-release
behaviours of sharks revealed by a remote operated vehicle. J. Exp. Mar. Biol. Ecol. 2019,511, 10–18. [CrossRef]
148.
White, C.F.; Lin, Y.; Clark, C.M.; Lowe, C.G. Human vs robot: Comparing the viability and utility of autonomous underwater
vehicles for the acoustic telemetry tracking of marine organisms. J. Exp. Mar. Biol. Ecol. 2016,485, 112–118. [CrossRef]
149.
Raoult, V.; Colefax, A.P.; Allan, B.M.; Cagnazzi, D.; Castelblanco-Martínez, N.; Ierodiaconou, D.; Johnston, D.W.; Landeo-Yauri,
S.; Lyons, M.; Pirotta, V.; et al. Operational protocols for the Use of Drones in Marine Animal Research. J. Drones
2020
,4, 64.
[CrossRef]
... Integrating measures of detection error within a surveyed area and identifying habitat covariates strongly correlated with the population density can greatly improve the accuracy of estimates when extrapolating density to larger spatial scales [3]. Only in the last decade, UAVs have been tested and successfully applied as a cost-effective alternative to traditional surveys to estimate abundance of wildlife, especially in gregarious species [22,23], but also for solitary animals [9,24]. ...
... The independent ground TC gave the most precise estimates (CVs range: 0.07-0. 24 Figure 5). Table 1. ...
... This is the case with counting rare deer in dense forests, where ground counts are ineffective due to forest cover and low densities of deer [8], where aerial imagery may provide better overview or spatial coverage. This may also be the case when there are challenges detecting marine animals at the sea surface from boats [9,10,24]. In our study, the ground DS survey had a maximum line of sight of about 900 metres, three times more compared to other study systems with lower linear detection rates, such as DS conducted on deer in woody, heterogenous terrain (250 m; [57]). ...
Preprint
Conservation of wildlife depends on precise and unbiased knowledge on the abundance and distribution of species. It is challenging to choose appropriate methods to obtain a sufficiently high detectability and spatial coverage matching the species characteristics and spatiotemporal use of the landscape. In remote regions, such as in the Arctic, monitoring efforts are often resource-intensive and there is a need for cheap and precise alternative methods. Here, we compare an uncrewed aerial vehicle (UAV; quadcopter) pilot-survey of the non-gregarious Svalbard reindeer to traditional population abundance surveys from ground and helicopter to investigate whether UAVs can be an efficient alternative technology. We found that the UAV survey underestimated reindeer abundance compared to the traditional abundance surveys when used at management relevant spatial scales. Observer variation in reindeer detection on UAV imagery was influenced by the RGB greenness index and mean blue channel. In future studies, we suggest to test long-range fixed-wing UAVs to increase the sample size of reindeer and area coverage and incorporate detection probability in animal density models from UAV imagery. In addition, we encourage focus on more efficient post-processing techniques, including automatic animal object identification with machine learning and analytical methods that account for uncertainties.
... Rapid advancements in computing combined with the developments associated with unmanned aerial vehicles (UAVs), commonly referred to as drones (Chapman, 2014), promise to continue to revolutionise beach monitoring and marine ecology (e.g., see Chabot, 2018;Li et al., 2020;Raoult et al., 2020;Butcher et al., 2021). Readily accessible UAVs have the capacity to autonomously follow fixed search patterns and can deliver high-resolution imagery in post and real-time, which is often crucial for making robust fauna identifications and assessments (Burke et al., 2019;Colefax et al., 2019). ...
... These authors analyse the effect of factors such as spacing, animal morphology and depth, water turbidity and sun glitter. See also Butcher et al. (2021) for a general review covering the use of UAVs in shark research. ...
... We believe that ML-driven shark species detectors will be excellent decision-support tools for beach managers in the very near future. The technology is already changing the way ecologists survey marine fauna (e.g., Butcher et al., 2021;Dujon et al., 2021;Jenrette et al., 2022;Marrable et al., 2022;Zhang et al., 2022;Shi et al., 2022). Humans will always need to be included in the decision loop, but the human role will change as the ML models become more reliable and flight systems become more automated. ...
Article
Full-text available
Over the last five years remotely piloted drones have become the tool of choice to spot potentially dangerous sharks in New South Wales, Australia. They have proven to be a more effective, accessible and cheaper solution compared to crewed aircraft. However, the ability to reliably detect and identify marine fauna is closely tied to pilot skill, experience and level of fatigue. Modern computer vision technology offers the possibility of improving detection reliability and even automating the surveillance process in the future. In this work we investigate the ability of commodity deep learning algorithms to detect marine objects in video footage from drones, with a focus on distinguishing between shark species. This study was enabled by the large archive of video footage gathered during the NSW Department of Primary Industries Drone Trials since 2016. We used this data to train two neural networks, based on the ResNet-50 and MobileNet V1 architectures, to detect and identify ten classes of marine object in 1080p resolution video footage. Both networks are capable of reliably detecting dangerous sharks: 80% accuracy for RetinaNet-50 and 78% for MobileNet V1 when tested on a challenging external dataset, which compares well to human observers. The object detection models correctly detect and localise most objects, produce few false-positive detections and can successfully distinguish between species of marine fauna in good conditions. We find that shallower network architectures, like MobileNet V1, tend to perform slightly worse on smaller objects, so care is needed when selecting a network to match deployment needs. We show that inherent biases in the training set have the largest effect on reliability. Some of these biases can be mitigated by pre-processing the data prior to training, however, this requires a large store of high resolution images that supports augmentation. A key finding is that models need to be carefully tuned for new locations and water conditions. Finally, we built an Android mobile application to run inference on real-time streaming video and demonstrated a working prototype during fields trials run in partnership with Surf Life Saving NSW.
... Environmental conditions, including water clarity and sea state, can affect shark detectability during drone surveys [30,31,73]. Observations occurred across all classifications of water clarity; however, most observations occurred in good to poor water clarity ranges. ...
... These limitations are, however, being overcome by rapidly improving drone technology and the easing of aviation regulations. For example, developments in deep-learning convoluted neural networks are improving computer-based shark identification in real-time [73][74][75]. Additionally, hybrid petrol-electric multi-rotor drones are routinely exceeding 2-h flight times, and some civil aviation authorities are making beyond-visual-line-of-sight operations (e.g., beyond 2.0 km from the ground control station) more cost-effective [31]. ...
Article
Full-text available
There is still limited information about the diversity, distribution, and abundance of sharks in and around the surf zones of ocean beaches. We used long-term and large-scale drone surveying techniques to test hypotheses about the relative abundance and occurrence of sharks off ocean beaches of New South Wales, Australia. We quantified sharks in 36,384 drone flights across 42 ocean beaches from 2017 to 2021. Overall, there were 347 chondrichthyans recorded, comprising 281 (81.0%) sharks, with observations occurring in <1% of flights. Whaler sharks (Carcharhinus spp.) had the highest number of observations (n = 158) recorded. There were 34 individuals observed for both white sharks (Carcharodon carcharias) and critically endangered greynurse sharks (Carcharias taurus). Bull sharks (Carcharhinus leucas), leopard sharks (Stegostoma tigrinum) and hammerhead species (Sphyrna spp.) recorded 29, eight and three individuals, respectively. Generalised additive models were used to identify environmental drivers for detection probability of white, bull, greynurse, and whaler sharks. Distances to the nearest estuary, headland, and island, as well as water temperature and wave height, were significant predictors of shark occurrence; however, this varied among species. Overall, we provide valuable information for evidence-based species-specific conservation and management strategies for coastal sharks.
... Under such circumstances, efficient data processing and management becomes a challenge. Lately, however, rapid advances in automated pattern-recognition and machine-learning algorithms have been catching up with current state-of-the-art digital photography, opening up this field of research to new possibilities and ideas of how to further improve the efficiency of field data collection (e.g., Schofield et al. 2019;Butcher et al. 2021;Machado and Cantor 2022), broaden the type, array and geographic scale of data collected during any given field day (e.g., Rieucau et al. 2018;Schneider et al 2019;Khan et al. 2022), and-importantly-improve the efficiency of processing the individual-ID data immediately upon the completion of a field day (e.g., Lahiri et al. 2011;Guo et al 2020;) and ensure the reliability of identification systems across long time-scales (e.g., Bodesheim et al. 2022;Cheeseman et al. 2022;Tyson Moore et al 2022). Other photographic applications also received a fresh breath of new ideas, such as photogrammetry (e.g., Galimberti et al. 2019;Gray et al. 2019;Shirane et al. 2020;O'Connell-Rodwell et al. 2022;Richardson et al. 2022) or biodiversity and monitoring surveys (e.g., Miao et al. 2019;Tabak et al. 2019;Howell et al. 2022;Sun et al. 2022;Rydell et al. 2022), broadening their applications well beyond their customary use. ...
... Some of this high-tech camera equipment is non-standard in published scientific research and was used in these episodes to obtain high-quality imagery for television, rather than scientific research purposes. While we note that examples of published studies using high-tech camera equipment do exist (as reviewed in [77]), the definition of "research method" used here was inclusive of activities that would not meet the scientific threshold for "research". 18.3% of Shark Week episodes featured no research that met this criterion. ...
Article
Full-text available
Despite evidence of their importance to marine ecosystems, at least 32% of all chondrichthyan species are estimated or assessed as threatened with extinction. In addition to the logistical difficulties of effectively conserving wide-ranging marine species, shark conservation is believed to have been hindered in the past by public perceptions of sharks as dangerous to humans. Shark Week is a high-profile, international programming event that has potentially enormous influence on public perceptions of sharks, shark research, shark researchers, and shark conservation. However, Shark Week has received regular criticism for poor factual accuracy, fearmongering, bias, and inaccurate representations of science and scientists. This research analyzes the content and titles of Shark Week episodes across its entire 32 years of programming to determine if there are trends in species covered, research techniques featured, expert identity, conservation messaging, type of programming, and portrayal of sharks. We analyzed titles from 272 episodes (100%) of Shark Week programming and the content of all available (201; 73.9%) episodes. Our data demonstrate that the majority of episodes are not focused on shark bites, although such shows are common and many Shark Week programs frame sharks around fear, risk, and adrenaline. While criticisms of disproportionate attention to particular charismatic species (e.g. great whites, bull sharks, and tiger sharks) are accurate and supported by data, 79 shark species have been featured briefly at least once. Shark Week’s depictions of research and of experts are biased towards a small set of (typically visual and expensive) research methodologies and (mostly white, mostly male) experts, including presentation of many white male non-scientists as scientific experts. While sharks are more often portrayed negatively than positively, limited conservation messaging does appear in 53% of episodes analyzed. Results suggest that as a whole, while Shark Week is likely contributing to the collective public perception of sharks as bad, even relatively small alterations to programming decisions could substantially improve the presentation of sharks and shark science and conservation issues.
... There is potential for advanced camera technologies (e.g., hyper or multispectral cameras) to improve the detection of sharks when conditions are suboptimal, such as when turbidity or glare are higher, by selecting wavelengths that improve water penetration and contrast of sharks [8,49]. Although, the optimal wavelength will likely vary depending on the conditions and locations [8]. ...
Article
Full-text available
Drones enable the monitoring for sharks in real-time, enhancing the safety of ocean users with minimal impact on marine life. Yet, the effectiveness of drones for detecting sharks (especially potentially dangerous sharks; i.e., white shark, tiger shark, bull shark) has not yet been tested at Queensland beaches. To determine effectiveness, it is necessary to understand how environmental and operational factors affect the ability of drones to detect sharks. To assess this, we utilised data from the Queensland SharkSmart drone trial, which operated at five southeast Queensland beaches for 12 months in 2020–2021. The trial conducted 3369 flights, covering 1348 km and sighting 174 sharks (48 of which were >2 m in length). Of these, eight bull sharks and one white shark were detected, leading to four beach evacuations. The shark sighting rate was 3% when averaged across all beaches, with North Stradbroke Island (NSI) having the highest sighting rate (17.9%) and Coolum North the lowest (0%). Drone pilots were able to differentiate between key shark species, including white, bull and whaler sharks, and estimate total length of the sharks. Statistical analysis indicated that location, the sighting of other fauna, season and flight number (proxy for time of day) influenced the probability of sighting sharks.
... RPVs have been used to capture environmental data as early as the 1990s but recent advancements in RPV technology, both the RPV itself and the camera/equipment attached to the RPV, have expanded application (Butcher et al., 2021;Nowak, Dzi ob, & Bogawski, 2019). ...
Article
Full-text available
An integral part of population monitoring within fisheries is ground‐based surveys of fish redds. Remotely piloted vehicles or drones (RPVs) could provide a complementary method but need verification due to a host of methodological differences. To compare methods, we counted summer Chinook redds (Oncorhynchus tshawytscha) (~6 m2 in size) using RPVs and compared them to ground‐based counts in the Wenatchee River (WA, USA). We found individual aerial counts were many times twice the corresponding ground counts. We also found large inter‐observer variability among aerial counters. The coefficient of variation among multiple aerial counts were 37%, 38%, and 50% across three sites, which are comparable to published variation in ground counts. We attribute inter‐observer variability to inherent uncertainties in redd identification similar to ground counting, and importantly, we did not see evidence that the clarity of substrate in the image influenced observer bias. Overall, our data suggest that redd counting using RPVs is an effective method, particularly in high‐density spawning locations. We conclude that RPV imagery accurately identifies redds in a clear, relatively wide (60 m) river, but suggest continuing research into increasing precision, limiting observer variability, and assessing the accuracy across methods and locations.
Article
While personal electric deterrents can reduce the risk of shark bites, evidence for the efficacy of other products is limited. We assessed two versions of a novel electric deterrent-80 and 150 volts (V)-designed to protect a large area (8 m deep × 6 m wide) or to be linked together for greater spatial coverage. We did 116 experimental trials on 43 white sharks (Carcharodon carcharias) to assess: (a) percentage of baits taken; (b) distance between bait and shark; (c) number of passes; and (d) whether sharks reacted to the deterrent. The proportion of baits taken was reduced by 24% (80 V) and 48% (150 V), although the high variance of the effect coefficient precluded statistical differentiation. Only the 150-V deterrent increased the distance between bait and shark (control: 1.59 ± 0.28 m versus active deterrent: 3.33 ± 0.33 m), but both versions increased the likelihood of a reaction (average reaction distance: 1.88 ± 0.14 m). Results were similar whether we measured distances using stereo-cameras or estimated them in situ, suggesting that stereo-cameras might not be necessary to quantify distances between sharks and baits. Our findings provide more evidence that electric deterrents can reduce the risk of shark bite, but the restricted efficacy limits the suitability of this device.
Article
Full-text available
There is increasing support for shark bite mitigation measures, such as SMART (Shark-Management-Alert-in-Real-Time) drumlines that minimise impacts on target sharks and other marine animals. On the east coast of Australia, SMART drumlines are used in a shark management program to catch and relocate white (Carcharodon carcharias), tiger (Galeocerdo cuvier), and bull sharks (Carcharhinus leucas; herein referred to as target sharks). This study examines the effect of bait position relative to the seabed on SMART drumline catches in eastern Australian waters, aiming to optimise catches of target sharks while reducing bycatch. Over 17 months, SMART drumlines were deployed at Ballina and Evans Head, New South Wales. Trace extensions were attached to 3.2 m standard traces to test the effect of bait height above the seabed on shark catch in an experimental design that alternated bait position every fortnight. White and tiger shark catches accounted for 67% of the total catch, whereas bull sharks were infrequently caught (3%). Bait position above the seabed did not significantly influence catch probability of white and tiger sharks. However, catches of Critically Endangered grey nurse sharks (Carcharias taurus) and false alarm events significantly increased when baits were closer to the seabed. Catches of white and tiger sharks varied throughout the year and were linked to seasonal water temperature changes. The standard traces effectively caught target shark species whilst reducing catches of grey nurse sharks and false alarm events, highlighting that the trace length currently used for NSW SMART drumline deployments is optimal.
Article
Full-text available
This article seeks to situate drone imagery within a more extensive lineage of practice by focusing on one particular form with which it is comparable: pigeon photography. Using a combination of visual social semiotic analysis, literature from Drone Studies, and archival research, it highlights four overarching characteristics shared between photographs taken by pigeons between 1908 and 1912 and contemporary drone visuals produced by hobbyists: verticality, geographical reimaginations, access to inaccessible places, and aerial self-portraits. In doing so, it aims to develop a better understanding of the social and material affordances/constraints of aerial photography, its meaning potentials and how they may have changed across space and time, and the social relations that are reflected in and shaped by its images. The article concludes by suggesting a nuanced perspective into the relationship between “new” and “old” media, arguing that images taken by drones and pigeons have similarities in their forms and functions, but their creation is guided by different ideological values and bounded by the potentials, norms, and traditions of the time. This perspective builds upon the recent turn in media studies toward transhistorical approaches to place seemingly novel contemporary communication technology within historical patterns of practice and use.
Article
Full-text available
Scalloped hammerheads (Sphyrna lewini) occur in tropical to subtropical waters across all ocean basins and are globally assessed as Critically Endangered by the International Union for the Conservation of Nature. In Australia, scalloped hammerheads range from Sydney, New South Wales (34° S; 151° E), around northern Australia, down to Geographe Bay, Western Australia (33° S; 115° E). However, in Western Australia, the species has rarely been recorded south of Jurien Bay (30° S; 115° E). We report a recurrent aggregation of scalloped hammerheads within the Shoalwater Islands Marine Park (32° S; 115° E), 240 km south of Jurien Bay, observed from drone footage collected during the 2019 and 2020 Austral summers. These new records challenge previous understanding of the distributional range of this Critically Endangered species and prompt questions about the adequacy of current protection measures. Scalloped hammerheads are amongst the most threatened of vertebrates globally and are listed as Critically Endangered by the International Union for the Conservation of Nature. Using aerial drones we report the southernmost aggregation of scalloped hammerheads in Australia, potentially extending the known distribution of the species. These new records challenge previous understanding of the distributional range of the species and prompt questions about the adequacy of current protection measures.
Article
Full-text available
The rapid expansion of human activities threatens ocean-wide biodiversity. Numerous marine animal populations have declined, yet it remains unclear whether these trends are symptomatic of a chronic accumulation of global marine extinction risk. We present the first systematic analysis of threat for a globally distributed lineage of 1,041 chondrichthyan fishes-sharks, rays, and chimaeras. We estimate that one-quarter are threatened according to IUCN Red List criteria due to overfishing (targeted and incidental). Large-bodied, shallow-water species are at greatest risk and five out of the seven most threatened families are rays. Overall chondrichthyan extinction risk is substantially higher than for most other vertebrates, and only one-third of species are considered safe. Population depletion has occurred throughout the world's ice-free waters, but is particularly prevalent in the Indo-Pacific Biodiversity Triangle and Mediterranean Sea. Improved management of fisheries and trade is urgently needed to avoid extinctions and promote population recovery.
Article
Full-text available
Recent work showed that two species of hammerhead sharks operated as a double oscillating system, where frequency and amplitude differed in the anterior and posterior parts of the body. We hypothesized that a double oscillating system would be present in a large, volitionally swimming, conventionally shaped carcharhinid shark. Swimming kinematics analyses provide quantification to mechanistically examine swimming within and among species. Here, we quantify blacktip shark (Carcharhinus limbatus) volitional swimming kinematics under natural conditions to assess variation between anterior and posterior body regions and demonstrate the presence of a double oscillating system. We captured footage of 80 individual blacktips swimming in the wild using a DJI Phantom 4 Pro aerial drone. The widespread accessibility of aerial drone technology has allowed for greater observation of wild marine megafauna. We used Loggerpro motion tracking software to track five anatomical landmarks frame by frame to calculate tailbeat frequency, tailbeat amplitude, speed, and anterior/posterior variables: amplitude and frequency of the head and tail, and the body curvature measured as anterior and posterior flexion. We found significant increases in tailbeat frequency and amplitude with increasing swimming speed. Tailbeat frequency decreased and tailbeat amplitude increased as posterior flexion amplitude increased. We found significant differences between anterior and posterior amplitudes and frequencies, suggesting a double oscillating modality of wave propagation. These data support previous work that hypothesized the importance of a double oscillating system for increased sensory perception. These methods demonstrate the utility of quantifying swimming kinematics of wild animals through direct observation, with the potential to apply a biomechanical perspective to movement ecology paradigms.
Article
Full-text available
The use of drones to study marine animals shows promise for the examination of numerous aspects of their ecology, behaviour, health and movement patterns. However, the responses of some marine phyla to the presence of drones varies broadly, as do the general operational protocols used to study them. Inconsistent methodological approaches could lead to difficulties comparing studies and can call into question the repeatability of research. This review draws on current literature and researchers with a wealth of practical experience to outline the idiosyncrasies of studying various marine taxa with drones. We also outline current best practice for drone operation in marine environments based on the literature and our practical experience in the field. The protocols outlined herein will be of use to researchers interested in incorporating drones as a tool into their research on marine animals and will help form consistent approaches for drone-based studies in the future.
Article
Full-text available
While coral reef ecosystems hold immense biological, ecological, and economic value, frequent anthropogenic and environmental disturbances have caused these ecosystems to decline globally. Current coral reef monitoring methods include in situ surveys and analyzing remotely sensed data from satellites. However, in situ methods are often expensive and inconsistent in terms of time and space. High-resolution satellite imagery can also be expensive to acquire and subject to environmental conditions that conceal target features. High-resolution imagery gathered from remotely piloted aircraft systems (RPAS or drones) is an inexpensive alternative; however, processing drone imagery for analysis is time-consuming and complex. This study presents the first semi-automatic workflow for drone image processing with Google Earth Engine (GEE) and free and open source software (FOSS). With this workflow, we processed 230 drone images of Heron Reef, Australia and classified coral, sand, and rock/dead coral substrates with the Random Forest classifier. Our classification achieved an overall accuracy of 86% and mapped live coral cover with 92% accuracy. The presented methods enable efficient processing of drone imagery of any environment and can be useful when processing drone imagery for calibrating and validating satellite imagery.
Article
Full-text available
The present study aimed to examine the capabilities of a low-cost and standard drone (namely, the DJI™ Phantom 4 Pro) for mapping of coral reefs in shallow coastal waters. To this end, a coral site in Kish Island, located in the northern Persian Gulf, was selected as the study area, wherein other methods had been previously practiced to generate maps from the corals. The drone flight was operated by the Pix4Dcapture® mobile application, due to its simplicity and high capabilities for performing an automated flight based on user’s settings. The flight altitude was set at 50 m, covering the ~6 ha (200 × 300 m) area in ~5.5 min. Prior to the drone flight, 11 diving buoys were positioned as ground control points (GCPs) on the study area and their coordinates were subsequently measured three times via a handheld GPS. Afterwards, the Agisoft™ Metashape software was utilized to create the orthophoto mosaic from the total number of 121 overlapped taken photos. The spatial resolution of the final orthophoto mosaic was by ~2 cm, contributing to the identification of the features with the minimum size of 20 cm. Although the produced mosaic was geo-registered, it was geo-referenced once again based on the above-mentioned GCPs in the ENVI™ 5.3 software to increase planimetric accuracy. Thereafter, the corrected mosaic was digitized by on-screen digitizing in the Autodesk® AutoCAD Map software, so that an identifier (ID) was assigned to each polygon with reference to the types of corals or other substrate features such as the rocks. Consequently, a topological map was generated in the format of the ESRI™ shapefile. The results demonstrated that the proposed technique was capable to differentiate the types of corals (including bleached ones) and other substrate features. In comparison with other alternative methods, the given technique cost lower than field observations and its results were much more accurate compared with those from satellite imagery. The capabilities of the proposed technique may be significantly improved if drones equipped with a multispectral camera were used, but then again, the costs will be increasingly higher.
Article
White sharks (Carcharodon carcharias) are attracted to and scavenge on floating whale carcasses. However, little is known about how stranded whale carcasses may affect their behaviour. With increasing whale populations and beach stranding events, sharks may be attracted to nearshore waters at carcass sites, increasing the potential conflict with human use. Here, we used aerial drones to assess whether white shark behaviour around stranded whale carcasses differs from their behaviour away from carcasses. We quantified white shark behaviour by measuring swim speed, net velocity, straightness and sinuosity of shark tracks, as well as the total length of each shark. White sharks in the vicinity of whale carcasses travelled at 0.46 m s⁻¹ (±0.06 SD) faster, were 0.26 m (±0.15 SD) longer, swam tracks that were 0.15 (±0.11 SD) lower on the straightness index, and showed more sinuous tracks by 0.07 (±0.02 SD), compared to sharks away from a carcass. The presence of a stranded whale carcass may, therefore, significantly altered the behaviour and size of white sharks close to shore. As white shark activity increases in a relatively small nearshore area, which was indicated by decreased straightness and increased sinuosity, there may be an elevated risk of shark interactions with water users in the vicinity of stranded whale carcasses.