Camera Technology for Monitoring Marine Biodiversity and Human Impact

REVIEWS REVIEWS REVIEWS 424 Camera technology for monitoring marine biodiversity and human impact

Anthony WJ Bicknell1,2*, Brendan J Godley1,2, Emma V Sheehan3, Stephen C Votier2, and Matthew J Witt2

Human activities have fundamentally altered the marine environment, creating a need for effective management in one of Earth’s most challenging habitats. Remote camera imagery has emerged as an essential tool for monitoring at all scales, from individuals to populations and communities up to entire marine ecosystems. Here we review the use of remote cameras to monitor the marine environment in relation to human activity, and consider emerging and potential future applications. Rapid technological advances in equipment and analytical tools influence where, why, and how remote camera imagery can be applied. We encourage the inclusion of cameras within multi-method and multi-sensor approaches to improve our understanding of ecosystems and help manage human activities and minimize impacts.

Front Ecol Environ 2016; 14(8): 424–432, doi: 10.1002/fee.1322

rom the early days of televised programs devoted to Mallet and Pelletier 2014). Moreover, their relative F marine natural history that informed the public ease of use and ever-decreasing cost have allowed a about previously unrecognized species and unexplored broader range of practitioners (eg government agencies, habitats, video has sparked the human imagination. non-governmental organizations [NGOs], environmen- Cameras continue to play an important role in marine tal consultants, universities) to utilize cameras for exploration and more recently have been used in envi- marine research. Here we assess the implementation of ronmental assessment (Figure 1; Bailey et al. 2007). cameras for examining human influence on marine Initially, camera equipment was complex, cumbersome, ecosystems and species – research that likely represents and expensive, but advances in the past two decades – a catalyst for driving future innovation in technology, including improvements in underwater housings, image methods, and analytical techniques. We review how quality, battery longevity, reduced cost, and data storage camera imagery can be applied to monitor ecosystems (Figure 1) – have driven a technological revolution. and populations, and examine individual responses to The potential for cameras to deliver highly repeatable human activities. Key challenges to the successful use of sampling over broad temporal (hours to years) and spa- videography are considered. Finally, we discuss potential (meters to kilometers) scales means they are now tial future developments in the application of cameras regularly employed in applied and theoretical research, and how multi-method and multi-sensor approaches to study behavior, species interactions, ecosystem func- will be vital in evaluating human impacts on marine tion, and resilience (detailed in WebTable 1; see also ecosystems.

In a nutshell: JJ Ecosystem- and population-level monitoring • Earth’s oceans are being transformed by the impacts of Monitoring change in marine ecosystems is challenging. anthropogenic activities, which can be difficult to monitor Water depth, pressure, and scale – at magnitudes typically effectively in the field • Camera imagery has become an important tool in under- not associated with other systems – restrict access to standing these impacts on individual animals, populations, locations or sampling sites and increase instrumentation and ecosystems costs. Remote camera technologies are helping to over- • Technological advances in equipment and deployment come some of these issues and provide non-destructive methods are stimulating novel applications and wider use methods for monitoring the effects and impact of human of cameras for impact assessment • As part of a multi-faceted monitoring approach, remote activities on habitats and species. These methods either cameras can help to improve our understanding of complement or replace traditional sampling techniques anthropogenic impacts in marine ecosystems and the used in the marine realm (eg benthic grabs, fishing management of marine resources trawls, box corers, diver surveys), many of which can be destructive or disturbing to marine life, unsuitable in certain habitats or locations, spatially restrictive, and/or 1Centre for Ecology and Conservation, University of Exeter, Penryn, prohibitively costly. Marine area protection and UK *([email protected]); 2Environment and Sustainability management of extractive activities are crucial to achieve Institute, University of Exeter, Penryn, UK; 3Marine Institute, healthy ecosystems and populations, and are enhanced Plymouth University, Plymouth, UK by the use of camera technologies.

www.frontiersinecology.org © The Ecological Society of America AWJ Bicknell et al. Camera technology in marine monitoring

Marine habitats and protected 425 areas

During the past half century, human impacts such as those related to overharvesting, pollution, and climate change have contributed to the global decline in areal coverage of tropical coral reefs (Bellwood et al. 2004). Consequently, research on reef mitigation and restoration has flourished, with underwater imagery as an important component. For example, by providing quantitative data on consumption rates of teleost fish on algae in coral reef systems (Vergés et al. 2012), in situ remote underwater cameras have demonstrated that reef resilience and recovery are dependent on herbi- vores, the populations of which can be directly affected by overfishing and climate change (Hughes et al. Figure 1. Timeline of important events in the development of camera imagery and its use 2007). Cameras have provided long- in studying and monitoring marine biodiversity (for full reference list, see WebPanel 1). term observational data in marine protected areas (MPAs) across habitats and depths to such as the deepwater coral Lophelia pertusa, as well as the detail reef change or recovery (eg McLean et al. 2010; ecological consequences (eg reduction in species diversity, Stevens et al. 2014; Bouchet and Meeuwig 2015). These function, resilience) resulting from this damage (Fosså et al. approaches have contributed to an understanding of 2002). Establishing deepwater MPA networks to conserve both direct (eg target population) and indirect (eg these habitats and species has now become a priority, lead- trophic cascades) effects of global change, as well as ing to further applications of camera systems (eg towed the timescales of these effects (Babcock et al. 2010). drop cameras) to capture imagery of deepwater epibenthic Static mounted cameras baited with dead fish, for assemblages, thereby assisting with habitat mapping and example, have revealed positive effects – related to site designation (Howell et al. 2010). abundance (Denny et al. 2004), density (Smith et al. 2014), size structure (Watson et al. 2009), and assem- Fish stocks and management blage structure (Malcolm et al. 2007) – of no-take reserves on target fish species in temperate and tropical Poorly managed capture fisheries – characterized by regions. unsustainable stock depletion (Myers and Worm 2003), Epibenthic communities and benthic habitats can be incidental mortality of non-target species (Lewison et al. assessed by non-destructive towed camera systems, which 2004), provision of unnatural food subsidies (Bicknell collect quantitative data over relatively large areas in a et al. 2013), and destruction of seabed habitat (Watling cost-effective manner (Sheehan et al. 2010). During a sur- and Norse 1998) – represent one of the most critical vey of an MPA, underwater video observations captured by threats to marine ecosystems. Nonetheless, fisheries are a benthic towed video camera provided the necessary a major source of livelihood and food for millions of replication and scale (spatial and temporal) to reveal that people worldwide and are therefore important in main- the functional extent of the reef habitat (feature) under taining global food security (Smith et al. 2010). To achieve protection was larger than thought, raising important ques- sustainable fish stocks, resource managers need to consider tions on the reliability of feature-based protection for sites multiple biotic, abiotic, and socioeconomic components within a seafloor habitat mosaic (Sheehan et al. 2013a). under ecosystem-based fisheries management (EBFM) Similarly, deepwater benthic habitats are being assessed (Pikitch et al. 2004). To realize the healthy and productive and monitored by camera systems, partly in response to the marine ecosystems that are necessary to support sustainable recent growth in bottom trawl fisheries (Norse et al. 2012) fisheries, we argue that the cumulative impacts of fishing and the potential for extraction of deep-sea mineral depos- must be evaluated and managed, which can be supported its (Hoagland et al. 2010). Cameras on remotely operated through the deployment of camera technology. vehicles (ROVs) have revealed the destructive effect of A combination of onboard video observations and deep-sea fisheries on important habitat-building species, in-trawl video systems can improve bycatch estimates

426 (Jaiteh et al. 2014), thereby providing crucial information The impact of trawl methods on non-target fishery stocks toward effective management, although this is compli- is another consideration, with implications for manage- cated in part by unseen underwater gear-related injury ment. In Newfoundland and Labrador, Canada, Nguyen and death to fish. Trawl fisheries in particular also discard et al. (2014) relied on net cameras to reveal the behavio- large quantities of undesirable (or unlandable) catch, ral responses of commercially important snow crabs which not only has direct negative consequences for tar- (Chionoecetes opilio) when encountering shrimp trawl get and non-target stocks but may also have ecosystem- fishing gear; this evidence helps to explain how the gear level effects (Bicknell et al. 2013). Discards are now being could be modified to mitigate future impacts. strictly managed by authorities in the US (www.nmfs. noaa.gov/by_catch/index.htm) and Europe (http://ec. Renewable and extractive energy industries europa.eu/fisheries/reform/index_en.htm), necessitating the monitoring of fishing activity to ensure compliance. As fossil-fuel reserves decline and the need to reduce Recently, video-based electronic surveillance has been global atmospheric emissions of carbon dioxide inten- piloted and implemented for compliance purposes in fish- sifies, the marine renewable energy (MRE) industry is eries in the US, New Zealand, Australia, and Europe likely to become more important in meeting future (eg McElderry 2008), in place of more traditional onboard global energy demands. Tidal, wave, and wind energy observer programs. Installation of a remote-sensing sys- project sites in marine habitats will be associated with tem – including closed-circuit cameras, a Global increased at-sea activities (eg shipping, construction) Positioning System (GPS) receiver, and sensors – on and the introduction of man-made structures into the fishing vessels can effectively monitor catch and discards, marine environment (eg monopiles, cables, turbines, providing a permanent data record (an improvement as lagoon walls). Cameras can provide long-term baseline compared with traditional logbooks) and allowing the and impact-related data for development sites and industry to engage in self-reporting. The latter is of energy converters (Broadhurst and Orme 2014). As increasing importance if fisheries want to be certified as part of ongoing monitoring of a MRE test facility site sustainable by eco-labelling organizations (eg the Marine off the north Cornwall coast in the UK, benthic towed Stewardship Council; www.msc.org/about-us/standards), (Figure 2) and static (Figure 3) remote underwater where responsible management and minimization of neg- imagery systems have been used to assess epibenthic ative environmental impact are part of the certification communities and habitats. The differences in benthic criteria. In the Greenland shrimp fishery, drop cameras assemblage composition and an increase in cuckoo are being deployed to assess gear-related impacts on ben- wrasse (Labrus mixtus) abundance observed around the thic habitats and communities, potentially encouraging seabed cable’s protective rock armoring (Sheehan et al. the fishery to consider more sustainable extraction meth- 2013b) revealed how structures act as artificial reefs ods and to minimize habitat degradation (Kemp and or fish aggregation devices (FADs; Langhamer 2012), Yesson 2013). Independent of fisheries certification, with potential positive effects for local fisheries. Cameras imagery in coastal waters around the Isle of Man (UK) mounted below or above the water can also monitor have increased our understanding of potential recovery the effect of operational energy convertors on fish, rates and the resilience of epibenthic communities to mammal, and bird behavior (eg Broadhurst et al. 2014). towed bottom-trawl fishing gear; there, estimated recov- In contrast to the MRE industry, extractive natural ery rates (<1 to >12 years) and identification of potential resource industries have a long and complicated history recovery drivers provide further insight into how destruc- with video imagery systems. Although initially developed tive fishing methods might be better managed (Lambert in association with first-generation ROVs for the offshore et al. 2014). petroleum sector, these systems are now employed to not Because fish behavior can introduce uncertainty or bias only discover new seabed areas suitable for underwater in catch estimates derived from fishery-independent sur- mining (Birney et al. 2006) but also to help understand veys, photographic evidence obtained from cameras may and/or predict the impact of extracting natural resources help to improve the accuracy of stock assessments by from the marine environment (Jones et al. 2007). Such providing behavioral information on fish in the vicinity activities are coming under closer scrutiny over their of gear. For instance, cameras attached to a trawl net in direct and indirect environmental impacts (Ramirez- the Northeast Pacific revealed flatfish herding behavior Llodra et al. 2011), which has led to the development of associated with this gear type, which, if not considered long-term, multi-sensor, observatory systems (including during stock assessments, would lead to overestimation of imagery capture systems) to assess the effect on deepwater stock biomass (Bryan et al. 2014). When affixed to the ecosystems (Vardaro et al. 2013). entrance of nets, camera systems also offer the potential to gather in-trawl information on target and non-target JJ Individual-level monitoring species during surveys, allowing for a more detailed description of the catch, species distribution in the water Animal-borne remote sensors have revolutionized the column, and method efficiency (Rosen and Holst 2013). study of free-living, mobile species, providing insights

www.frontiersinecology.org © The Ecological Society of America AWJ Bicknell et al. Camera technology in marine monitoring into their physiology and behavior (Moll et al. 2007). (a) 427 For mobile and/or elusive marine species, they are the only available method to collect reliable data on individual movement and foraging, in addition to their encountered environment (Ropert-Coudert and Wilson 2005). Studying animal behavior in submarine systems poses unique challenges (eg high pressure and low light environments), which drove much of the early innovation in animal-borne equipment. The combined size and weight of cameras, batteries, and data storage capacity initially limited applications, but these technological barriers are being overcome. The first- described animal-borne video camera was attached (b) to loggerhead (Caretta caretta) and leatherback (Dermochelys coriacea) sea turtles (early National Geographic CRITTERCAM; Marshall 1990). Field applications soon extended to a wide range of marine vertebrates and invertebrates, not only to study foraging behavior, social interactions, and habitat preference, but also to gather empirical data to test the validity of activities and behavior that have previously been indirectly inferred from tracking or accelerometry data (WebTable 2).

Foraging behavior and habitat use (c)

Animal-borne cameras can be combined with movement sensors (eg GPS receivers, time depth recorders, ac- celerometers) to gain novel perspectives on behavior and habitat use (reviewed in Moll et al. 2007). For example, imagery combined with time depth recorders and satellite tags revealed previously unknown omnivory in green sea turtles (Chelonia mydas) (reviewed in Hochscheid 2014), which has led to the identification of new foraging habitats and consequently a change in the approach to conservation of key areas. Moreover, images and GPS data from bird-borne devices revealed Figure 2. (a–c) Still frame grabs from benthic towed video that while northern gannets (Morus bassanus) followed imagery systems at a marine renewable energy test facility site in fishing vessels more than was previous thought north Cornwall, UK. (a) Two red gurnard and a lesser octopus, (Figure 4), they also switched between scavenging and (b) Ross coral, and (c) a high resolution frame grab to enable natural foraging (Votier et al. 2013). These findings identification of smaller organisms. are important as they provide a more thorough understanding of the way in which birds exploit fisheries discards, which may change in the face of global fish- for example, to monitor seabird mortality from collisions eries management reform (Bicknell et al. 2013). with offshore wind turbines, and birdstrike vulnerability has been assessed or modeled mainly through known Inter- and intra-specific interactions flight heights (Cleasbyet al. 2015). Capturing imagery to quantify intraspecific (eg individual, sex, age) and Individual interactions with man-made objects or with interspecific variation in behavior during encounters human activities in the marine environment (eg fishing with wind turbines (eg avoidance, reduction of flight boats/gear, renewable energy converters, oil rigs) are height, use for resting or foraging) would greatly im- currently inferred through observations (eg bycatch prove not only predictions regarding species’ collision rates, boat-based surveys) or tracking devices. Gathering risks and potential population-level impacts but also evidence of how different individuals, sexes, and age evaluation of future turbine sites and mitigation efforts. classes respond during these encounters may help im- A similar rationale could be applied to monitoring prove predictions of where, when, and how impacts bycatch in fishing gear (eg sea turtles in gillnets), since will manifest in populations. It is logistically difficult, little is currently known about entrapment rates in

428 (a) JJ Emerging applications and future developments As technologies have become more sophisticated (eg inertial navigation systems, Fastloc™ GPS receivers, wireless network transmission) and as equipment costs have decreased, there are more opportunities to apply and expand the capacity of camera-based methods to ecological research.

Aerial, surface, and sub-surface vehicles

Autonomous and unmanned vehicles are being used (b) for remote sensing in both sea and air (Williams et al. 2012; Anderson and Gaston 2013). Miniaturized and lightweight cameras installed on unmanned aerial vehicles (UAVs) have been deployed to survey marine vertebrates on land (Pacific walrus Odobenus[ rosmarus divergens]; Monson et al. 2013) and in water (dugongs [Dugong dugon]; Hodgson et al. 2013). Aerial imagery techniques have previously been utilized for population estimates (see Buckland et al. 2012), Environmental Impact Assessments (Thaxter and Burton 2009), and coastal habitat mapping (Lathrop et al. 2006), but the improved spatial and temporal resolution provided by (c) UAVs offer many other applications in marine systems (Anderson and Gaston 2013). Autonomous underwater vehicles (AUVs) and unmanned surface vehicles are not constrained by weather or sea state, and can gather high-resolution imagery and environmental data from previously inaccessible habitats (eg deep-sea benthos) (Singh et al. 2004). By affixing cameras to AUVs to monitor the health of benthic habitats and communities, Smale et al. (2012) captured information directly relevant to EBFM. A more recent technique to control the invasive crown-of- thorns sea star (Acanthaster planci) in reef systems uses AUVs to visually identify indi- Figure 3. (a–c) Still frame grabs from benthic baited (static) viduals and subsequently administer a lethal injection video imagery systems at a marine renewable energy test facility of bile salts (Dayoub et al. 2015). Despite their high site in north Cornwall, UK. (a) Reef structure with encrusting purchase costs (circa £30k–300k, €37k–370k, organisms and sea urchin, (b) large- and small-spotted cat $50k–500k), AUVs offer great potential for combining sharks, and (c) a European lobster. multiple sensors, data streams (eg hydrography and environmental monitoring), and physical equipment. relation to encounters. For instance, important turtle nesting beaches close to high-density gillnet fisheries Underwater camera-traps could be targeted and pop-off cameras (that detach automatically after a set period) deployed on females For decades, camera-traps (ie remotely triggered cameras (on the beaches) and males (caught offshore) to detail that automatically take images of subjects passing in front intra- and interspecific differences in behavior when of them) have been successfully deployed in terrestrial encountering nets, and the ratio of encounters to systems to measure abundance, species diversity, and entrapment. Technological advances are still required behavior, as well as to observe rare species (see Burton to enable extended periods of animal-borne video et al. 2015), but only recently has the concept been capture, but cameras have greatly improved our view developed for marine systems. Williams et al.’s (2014) of animal behavior and interactions with human TrigCam is a low-cost, low-power, motion-triggered un- activities. The integration of tracking technologies that derwater system that employs far red illumination and allow remote downloading between individual animals stereo images to reveal the number, size, position, and (including proximity data), and to base stations, will orientation of animals that activate the camera. By using provide further insights. novel hardware and open-source software, the authors

www.frontiersinecology.org © The Ecological Society of America AWJ Bicknell et al. Camera technology in marine monitoring have enabled further, independent system development (a) 429 by individuals or research groups to help expand its use in various marine environments. The rapid adoption of camera-traps within terrestrial research provided valuable new data and insightful images but also highlighted in- herent challenges associated with many remote camera systems (eg imperfect detection, effective sampling area, multispecies inference; Burton et al. 2015). To enable robust ecological inquiry using camera systems, those who develop survey methods/designs need to explicitly focus on the ecological processes (eg abundance, movement) they intend to measure and the assumptions necessary to relate camera detections to those processes. Heeding the lessons learned from camera-trap usage in terrestrial systems will help facilitate a successful transition of this method to the marine environment. (b)

Accessibility and capacity

The production of quality, mass-produced, high-definition (HD) cameras (eg GoPro©) is changing how imagery can be applied to study marine biodiversity. Because of their greater affordability (circa £100–£400, €150–490, $250–660), reduced size, and improved underwater hous- ing, cameras can now be deployed in places that were previously unfeasible and by a much wider group of investigators (eg students, NGOs, citizen scientists). Research and monitoring involving these types of cameras will undoubtedly increase, especially with innova- tions in extending battery longevity and expanding the (c) storage capacity for captured images. Fisheries interactions (eg post-catch behavior on longlines, habitat impact of potting), effects of MRE devices (eg FAD effects, seabird feeding), and monitoring marinas or commercial ports for introduced species are examples where this equipment could provide invaluable data to help understand potential anthropogenic impacts. While logistically and scientifically challenging (Bonneyet al. 2009), conducting relatively low-cost and long-term research incorporating networks of cameras with multiple stakeholders is a promising prospect in both marine and terrestrial systems (Dickinson et al. 2012; McShea et al. 2015). With suitable support, training and management of stakeholders/volunteers, and development of e-technologies to overcome data management and storage issues, these Figure 4. (a–c) Gannet-borne camera images of interactions projects could substantially change how cameras are with fishing vessels in the Celtic Sea. utilized in ecological monitoring (eg McShea et al. 2015; Wild.Id software, www.teamnetwork.org/solution). the process remains a major challenge. To this end, Automated classification, visual recognition, and motion detection software for video files has been writ- image enhancement ten, which can identify moving objects and therefore help categorize useful and redundant sections of video Despite advances in imagery collection, the development for analysis (Weinstein 2015). Edgington et al. (2006) of methods to extract biological or ecologically important have also developed a visual recognition system to iden- information from photographs or video lags behind. The tify a suite of fish and echinoderms independently (without conversion of imagery into quantitative data can still human observers), and Lukeman et al. (2010) used dig- be painstakingly slow, and streamlining and automating itized images to study collective behavior in seaducks.

430 Panel 1. Limitations and challenges when using camera imagery in ecological monitoring Technology • Species, individuals, sexes, and age classes may respond • Battery life for cameras and/or lights, as well as data storage, have differently to attractants. advanced but are still major limitations to long-term monitoring • Species or individuals may become habituated to involving remote camera systems or animal-borne devices. attractants. • Size, mass, and hydrodynamic shape currently limit the use • Attractants may increase the chance of predation for target of animal-borne cameras to larger species. species. Device effects Video or stills analysis • Necessary attention/assessment is required to avoid device • Automation of video or photo analysis is challenging in effects on individual animals that might harm them and/or marine systems, and human intervention will always be alter their “natural” behavior. required, making it a time-consuming process. • Species or individuals may show avoidance behavior to Experimental design and sampling underwater camera systems. • There may be imperfect detection of species and/or individ- Visibility uals by camera systems. • Reduced water clarity or reduced light can prevent species • Reasonable levels of redundancy in the experimental design identification and abundance estimates, in some cases are required to account for device failure or unusable rendering underwater footage unusable. footage, particularly required in dynamic environments Attractants and olfactory plume (eg temperate marine systems) • For baited camera systems, the type of bait and the range of • Typically, animal-borne camera studies have small sample the olfactory plume need to be considered. sizes.

Automated classification of benthic habitats is improving ethical and legal issues is needed before conducting through advances in multi-spectral underwater image research or regulatory agencies permitting activities. processing and segmentation algorithms, which have been employed to discriminate coral and algal cover in reef JJ Conclusions systems (Gleason et al. 2007), not possible with standard color imagery. This method (and others) will be further By offering access to previously inaccessible marine hab- enhanced in many ecosystems by the inclusion of optical itats and novel perspectives on animal behavior, remote back-scatter filters to improve image clarity (Mortazavi camera systems can help provide a better understanding et al. 2013). Such methods are unlikely to entirely exclude of how animals and habitats are affected by human the need for human analysis, especially in low-contrast activities. However, imagery has limitations (Panel 1) underwater environments, but it is hoped that the de- and requires complementary sensor data and/or direct veloping techniques will substantially decrease observer sampling to realize its full potential. For example, animal- time in the future and therefore further propel the utility borne images have limited value in the absence of other of videography. data such as location (eg latitude and longitude) and behavior (eg Votier et al. 2013). Similarly, although JJ Welfare and environmental ethics remote underwater images allow mobile species to be seen in context of habitats and sessile organisms, iden- Widespread application of cameras to study marine eco- tifying small and cryptic species can be challenging and systems and/or organisms raises some ethical and legal requires direct physical sampling. questions. Animal welfare is a primary concern for Accurately assessing marine ecosystem health requires animal-borne systems, and in many countries is currently information on both biotic and abiotic components, regulated and managed by national agencies. Elsewhere, including animal behavior, species diversity, genetic however, such regulations are limited or absent, and structure, bathymetry, and water chemistry. It also relies on the increased accessibility of small, low-cost, HD cameras a detailed understanding of the underlying ecological per- could lead to unnecessary and/or harmful applications. turbations, which may not be available for many aquatic A potential but often overlooked issue with the use of systems (Friberg et al. 2011). A prerequisite to monitor any equipment in the marine environment is littering change is the compilation of baseline data of these compo- and pollution. Cumulatively, numerous sources of waste nents, which requires a multi-method and sensor approach (eg shipping, discharges) have already contributed to over relevant spatial and temporal scales. Robust experi- extensive pollution in the oceans. Privacy issues for mental design, survey methods, and sampling are also citizens are already a growing concern with respect to essential for detecting changes caused by human activities. cameras in terrestrial systems (Butler and Meek 2013). Through the use of cameras, passive and active acoustic Although unlikely to be problematic in underwater set- systems, oceanographic sensors (eg those measuring tem- tings, such concerns may be valid to people in developed perature, salinity, fluorescence, and conductivity), track- coastal areas. Proper understanding and consideration of ing technology, next-generation genetic sequencing,

www.frontiersinecology.org © The Ecological Society of America AWJ Bicknell et al. Camera technology in marine monitoring remote sensing, and direct sampling, there is the poten- Dayoub F, Dunbabin M, and Corke P. 2015. Robotic detection and 431 tial to construct a comprehensive, multi-level depiction tracking of crown-of-thorns starfish. IEEE/RSJ International Conference on Intelligent Robots and Systems; 28 Sep–2 Oct of a marine ecosystem. This is an imposing task, but 2015. Hamburg, Germany. http://bit.ly/2aeq9Zr. Viewed 25 Jul monitoring aspects in isolation will not provide the 2016. details required to truly understand the component Denny CM, Willis TJ, and Babcock RC. 2004. Rapid recolonisa- interactions and associations that can ultimately influ- tion of snapper Pagrus auratus: Sparidae within an offshore ence the health and productivity of a marine system, and island marine reserve after implementation of no-take status. the marked or subtle effects of human activities. Mar Ecol-Prog Ser 272: 183–90. Dickinson JL, Shirk J, Bonter D, et al. 2012. The current state of citizen science as a tool for ecological research and public JJ Acknowledgements engagement. Front Ecol Environ 10: 291–97. Edgington DR, Cline DE, Davis D, et al. 2006. Detecting, tracking This work was funded by a Natural Environment and classifying animals in underwater video. Proceedings of Research Grant (NE/J012319/1). OCEANS 2006; 18–21 Sep 2006; Boston, MA. IEEE. doi:10.1109/OCEANS.2006.306878. Fosså JH, Mortensen PB, and Furevik DM. 2002. The deep-water JJ References coral Lophelia pertusa in Norwegian waters: distribution and Anderson K and Gaston KJ. 2013. Lightweight unmanned aerial fishery impacts. Hydrobiologia 471: 1–12. vehicles will revolutionize spatial ecology. Front Ecol Environ Friberg N, Bonada N, Bradley DC, et al. 2011. Biomonitoring of 11: 138–46. human impacts in freshwater ecosystems: the good, the bad and Babcock RC, Shears NT, Alcala AC, et al. 2010. Decadal trends in the ugly. Adv Ecol Res 44: 1–68. marine reserves reveal differential rates of change in direct and Gleason ACR, Reid RP, and Voss KJ. 2007. Automated classification indirect effects. P Natl Acad Sci USA 107: 18256–61. of underwater multispectral imagery for coral reef monitoring. Bailey DM, King NJ, and Priede IG. 2007. Cameras and carcasses: Proceedings of OCEANS 2007; 29 Sep–4 Oct 2007; Vancouver, historical and current methods for using artificial food falls to Canada. IEEE. doi:10.1109/OCEANS.2007.4449394. study deep-water animals. Mar Ecol-Prog Ser 350: 179–91. Hoagland P, Beaulieu S, Tivey MA, et al. 2010. Deep-sea mining of Bellwood DR, Hughes TP, Folke C, et al. 2004. Confronting the seafloor massive sulfides. Mar Policy 34: 728–32. coral reef crisis. Nature 429: 827–33. Hochscheid S. 2014. Why we mind sea turtles’ underwater busi- Bicknell AWJ, Oro D, Camphuysen K, et al. 2013. Potential conse- ness: a review on the study of diving behavior. J Exp Mar Biol quences of discard reform for seabird communities. J Appl Ecol Ecol 450: 118–36. 50: 649–58. Hodgson A, Kelly N, and Peel D. 2013. Unmanned aerial vehicles Birney K, Griffin A, Gwiazda J, et al. 2006. Potential deep-sea min- (UAVs) for surveying marine fauna: a dugong case study. PLoS ing of seafloor massive sulfides: a case study in Papua New ONE 8: e79556. Guinea. Santa Barbara, CA: University of California–Santa Howell KL, Davies JS, and Narayanaswamy BE. 2010. Identifying Barbara. Technical report. http://bit.ly/2a8n0Ge. Viewed 25 deep-sea megafaunal epibenthic assemblages for use in habitat Jul 2016. mapping and marine protected area network design. J Mar Biol Bonney R, Ballard H, Jordan R, et al. 2009. Public participation in Assoc UK 90: 33–68. scientific research: defining the field and assessing its potential Hughes TP, Rodrigues MJ, Bellwood DR, et al. 2007. Phase shifts, for informal science education. A CAISE Inquiry Group herbivory, and the resilience of coral reefs to climate change. Report. Washington, DC: Center for Advancement of Informal Curr Biol 17: 360–65. Science Education (CAISE). Jaiteh VF, Allen SJ, Meeuwig JJ, et al. 2014. Combining in-trawl Bouchet PJ and Meeuwig JJ. 2015. Drifting baited stereo- video with observer coverage improves understanding of pro- videography: a novel sampling tool for surveying pelagic wild- tected and vulnerable species by-catch in trawl fisheries. Mar life in offshore marine reserves. Ecosphere 6: art137. Freshwater Res 65: 830–37. Broadhurst M and Orme CDL. 2014. Spatial and temporal ben- Jones DOB, Wigham BD, Hudson IR, et al. 2007. Anthropogenic thic species assemblage responses with a deployed marine tidal disturbance of deep-sea megabenthic assemblages: a study with energy device: a small scaled study. Mar Environ Res 99: remotely operated vehicles in the Faroe-Shetland Channel, NE 76–84. Atlantic. Mar Biol 151: 1731–41. Broadhurst M, Barr S, and Orme CDL. 2014. In-situ ecological Kemp K and Yesson C. 2013. Greenland benthic assessment, 2nd interactions with a deployed tidal energy device: an observa- annual report. London, UK: Institute of Zoology, Zoological tional pilot study. Ocean Coast Manage 99: 31–38. Society of London. Bryan DR, Bosley KL, Hicks AC, et al. 2014. Quantitative video Lambert GI, Jennings S, Kaiser MJ, et al. 2014. Quantifying recov- analysis of flatfish herding behavior and impact on effective ery rates and resilience of seabed habitats impacted by bottom area swept of a survey trawl. Fish Res 154: 120–26. fishing. J Appl Ecol 51: 1326–36. Buckland ST, Burt ML, Rexstad EA, et al. 2012. Aerial surveys of Langhamer O. 2012. Artificial reef effect in relation to offshore seabirds: the advent of digital methods. J Appl Ecol 49: 960– renewable energy conversion: state of the art. Sci World J 2012: 67. art386713. Burton AC, Neilson E, Moreira D, et al. 2015. Wildlife camera Lathrop RG, Montesano P, and Haag S. 2006. A multi-scale trapping: a review and recommendations for linking surveys to segmentation approach to mapping seagrass habitats using ecological processes. J Appl Ecol 52: 675–85. airborne digital camera imagery. Photogramm Eng Rem S 72: Butler DA and Meek P. 2013. Camera trapping and invasions 665–75. of privacy: an Australian legal perspective. Torts Law J 20: Lewison RL, Crowder LB, Read AJ, et al. 2004. Understanding 235–64. impacts of fisheries bycatch on marine megafauna. Trends Ecol Cleasby IR, Wakefield ED, Bearhop S, et al. 2015. Three- Evol 19: 598–604. dimensional tracking of a wide-ranging marine predator: flight Lukeman R, Li Y-X, and Edelstein-Keshet L. 2010. Inferring indi- heights and vulnerability to offshore wind farms. J Appl Ecol 52: vidual rules from collective behavior. P Natl Acad Sci USA 107: 1474–82. 12576–80.

432 Malcolm HA, Gladstone W, Lindfield S, et al. 2007. Spatial and Sheehan EV, Witt MJ, Cousens SL, et al. 2013b. Benthic interac- temporal variation in reef fish assemblages of marine parks in tions with renewable energy installations in a temperate New South Wales, Australia: baited video observations. Mar ecosystem. Proceedings of the Twenty-third International Ecol-Prog Ser 350: 277–90. Offshore and Polar Engineering Conference; Anchorage, AK; Mallet D and Pelletier D. 2014. Underwater video techniques for 30 Jun–5 Jul 2013. International Society of Offshore and Polar observing coastal marine biodiversity: a review of sixty years of Engineers. ISOPE-I-13-159. publications (1952–2012). Fish Res 154: 44–62. Singh H, Can A, Eustice R, et al. 2004. Seabed AUV offers new Marshall G. 1990. A video-collar to study aquatic fauna: a view platform for high-resolution imaging. Eos T Am Geophys Un from the animal’s back. In: Reed D (Ed). Spirit of Enterprise: 85: 289–96. The 1990 Rolex Awards. Bern, Switzerland: Buri International. Smale DA, Kendrick GA, Harvey ES, et al. 2012. Regional-scale McElderry H. 2008. At-sea observing using video-based electronic benthic monitoring for ecosystem-based fisheries management monitoring. Electronic Monitoring Workshop; Seattle, WA; (EBFM) using an autonomous underwater vehicle (AUV). 29–30 Jul 2008. Victoria, Canada: Archipelago Marine Rese ICES J Mar Sci 69: 1108–18. arch Ltd. Smith ANH, Anderson MJ, Millar RB, et al. 2014. Effects of McLean DL, Harvey ES, Fairclough DV, et al. 2010. Large decline marine reserves in the context of spatial and temporal varia- in the abundance of a targeted tropical lethrinid in areas open tion: an analysis using Bayesian zero-inflated mixed models. and closed to fishing. Mar Ecol-Prog Ser 418: 189–99. Mar Ecol-Prog Ser 499: 203–16. McShea WJ, Forrester T, Costello R, et al. 2015. Volunteer-run Smith MD, Roheim CA, Crowder LB, et al. 2010. Sustainability cameras as distributed sensors for macrosystem mammal and global seafood. Science 327: 784–86. research. Landscape Ecol 31: 55–66. Stevens TF, Sheehan EV, Gall SC, et al. 2014. Monitoring benthic Moll RJ, Millspaugh JJ, Beringer J, et al. 2007. A new ‘view’ of ecol- biodiversity restoration in Lyme Bay marine protected area: ogy and conservation through animal-borne video systems. design, sampling and analysis. Mar Policy 45: 310–17. Trends Ecol Evol 22: 660–68. Thaxter CB and Burton NHK. 2009. High definition imagery for Monson DH, Udevitz MS, and Jay CV. 2013. Estimating age ratios surveying seabirds and marine mammals: a review of recent and size of Pacific walrus herds on coastal haulouts using video trials and development of protocols. Thetford, UK: BTO. imaging. PLoS ONE 8: e69806. Vardaro MF, Bagley PM, Bailey DM, et al. 2013. A Southeast Mortazavi H, Oakley JP, and Barkat B. 2013. Mitigating the effect Atlantic deep-ocean observatory: first experiences and results. of optical back-scatter in multispectral underwater imaging. Limnol Oceanogr-Meth 11: 304–15. Meas Sci Technol 24: 074025. Vergés A, Bennett S, and Bellwood DR. 2012. Diversity among Myers RA and Worm B. 2003. Rapid worldwide depletion of pred- macroalgae-consuming fishes on coral reefs: a transcontinental atory fish communities. Nature 423: 280–83. comparison. PLoS ONE 7: e45543. Nguyen TX, Winger PD, Legge G, et al. 2014. Underwater observa- Votier SC, Bicknell A, Cox SL, et al. 2013. A bird’s eye view of tions of the behaviour of snow crab (Chionoecetes opilio) discard reforms: bird-borne cameras reveal seabird/fishery inter- encountering a shrimp trawl off northeast Newfoundland. Fish actions. PLoS ONE 8: e57376. Res 156: 9–13. Watling L and Norse EA. 1998. Disturbance of the seabed by Norse EA, Brooke S, Cheung WW, et al. 2012. Sustainability of mobile fishing gear: a comparison to forest clearcutting. deep-sea fisheries. Mar Policy 36: 307–20. Conserv Biol 12: 1180–97. Pikitch EK, Santora C, Babcock EA, et al. 2004. Ecosystem-based Watson DL, Anderson MJ, Kendrick GA, et al. 2009. Effects of fishery management. Science 305: 346–47. protection from fishing on the lengths of targeted and non- Ramirez-Llodra E, Tyler PA, Baker MC, et al. 2011. Man and the targeted fish species at the Houtman Abrolhos Islands, Western last great wilderness: human impact on the deep sea. PLoS Australia. Mar Ecol-Prog Ser 384: 241–49. ONE 6: e22588. Weinstein BG. 2015. MotionMeerkat: integrating motion video Ropert-Coudert Y and Wilson R. 2005. Trends and perspectives detection and ecological monitoring. Methods Ecol Evol 6: in animal-attached remote sensing. Front Ecol Environ 3: 357–62. 437–44. Williams K, De Robertis A, Berkowitz Z, et al. 2014. An underwa- Rosen S and Holst JC. 2013. DeepVision in-trawl imaging: sam- ter stereo-camera trap. Methods Oceanogr 11: 1–12. pling the water column in four dimensions. Fish Res 148: Williams SB, Pizarro OR, Jakuba MV, et al. 2012. Monitoring of 64–73. benthic reference sites: using an autonomous underwater vehi- Sheehan EV, Stevens TF, and Attrill MJ. 2010. A quantitative, cle. IEEE Robot Autom Mag 19: 73–84. non-destructive methodology for habitat characterisation and benthic monitoring at offshore renewable energy develop- JJ ments. PLoS ONE 5: e14461. Supporting Information Sheehan EV, Cousens SL, Nancollas SJ, et al. 2013a. Drawing lines at the sand: evidence for functional vs visual reef boundaries in Additional, web-only material may be found in the temperate Marine Protected Areas. Mar Pollut Bull 76: online version of this article at http://onlinelibrary. 194–202. wiley.com/doi/10.1002/fee.1322/suppinfo