TrapCam: an inexpensive camera system for studying deep-water animals


Correspondence author. E-mail:


1. Behavioural research in deep water (>40 m depth) has traditionally been expensive and logistically challenging, particularly because the light and sound produced by underwater vehicles make them unsuitably disruptive. Yet, understanding the behaviour of deep-water animals, especially those targeted by exploitation, is important for conservation. For example, understanding interactions between animals and deep-water fishing gear could inform the design of devices that minimize bycatch.

2. We describe the ‘TrapCam’, a self-contained, high-definition video system that requires neither the support of a vessel once deployed nor special equipment to deploy or retrieve. This system can record 13-h videos at 1080p resolution and is deployable on any substrata at depths of up to 100 m. The system is inexpensive (<$3000 USD), versatile and suited to the study of animal behaviour at depths inaccessible to scuba divers.

3. We evaluate the performance and cost effectiveness of TrapCam and analyse videos retrieved from pilot deployments to observe spot prawn (Pandalus platyceros) traps at 100 m depth. Preliminary analyses of animal–prawn trap interactions yield novel insights. We provide future directions for researchers to use this type of camera system to study deep water-dwelling species around the world.


It is difficult and costly to study animals that live in deep water. Much of our knowledge of species that live below depths accessible to scuba divers (i.e. deeper than 40 m) comes from destructive sampling as organisms are brought to the surface in fishing or sampling gear. While these methods of collection can provide information about the animals’ distribution, physiology and diet, they are inappropriate for the study of behaviour. As a result, we know relatively little about the in situ behaviour of deep water-dwelling organisms compared with shallow-living species.

Underwater cameras are a tool of choice for in situ observations of the behaviour of deep-water species. Dataloggers, pop-up satellite tags and other such equipments can provide information about selected behavioural aspects, such as habitat use and movement, of these animals but actual behaviour must often be inferred from tagging data, and many tags are unsuitable for small animals. In contrast, cameras have the unique ability to provide direct visual information of organisms, which often cannot be obtained in any other way. The use of camera technology can yield important information for conservation. In terrestrial systems, the use of (near-)continuous videography has been effectively used to gain insights about organisms that are difficult to access or phenomena that occur infrequently, such as predation in bird nests (Pietz & Granfors 2000; Kross & Nelson 2011), roosting behaviour in bat harems (Hoxeng et al. 2007) and even nocturnal foraging of spiders (Taylor & Bradley 2009). In the marine realm, however, most existing options for deep-water camera research are prohibitively expensive and require access to large vessels, expensive support technology and specially trained crew to operate. This is the case for submersibles, remotely operated vehicles (ROVs) and permanently cabled bottom-mounted observatories, which have all been used to study species that live below scuba-diving depths (e.g. Milliken & DeAlteris 2004; Piasente et al. 2004; Mills, Verdouw, & Frusher 2005; Woodroffe & Round 2008). In addition, the light and sound produced by mobile, and camera-bearing platforms can alter animal behaviour significantly (Popper 2003; Ryer et al. 2009), and it has been recently demonstrated that even crustacean behaviour can be greatly affected by sound transmission (Simpson et al. 2011). Surface-powered ‘drop cameras’ can be attached to boats or surface floats (e.g. Mills, Verdouw, & Frusher 2005) but these systems require either a stable vessel or a large electronics package on a surface float to be connected to the camera system for the duration of filming, which is logistically challenging for deep-water work. Baited Remote Underwater Video (BRUV) stations overcome some of these problems (Cappo, Speare, & De’ath 2004), but their use of full-spectrum illumination, reliance on bait as an attractant and often time-lapse photography essentially limits them to assessments of relative abundance and species distribution (Cappo, Harvey, & Shortis 2006). Cameras that are attached directly to animals, or ‘Crittercams,’ are capable of revealing behaviour on deep dives but are only suitable for animals that come to the surface and that are large enough to support a mounted camera (Herman & Bakhtiari 2007).

A more versatile tool to study deep-water species would ideally meet five requirements, each of which has implications for design. First, the system should be able to record for long periods of time (e.g. 12 or more hours) at great depths (e.g. >40 m), which suggests the need for autonomy from the surface and the use of adequate pressure casings. Second, to glean maximum information about behavioural interactions – particularly those that are rapid and infrequent, the system should be capable of recording continuously, rather than simply taking photographs or short video clips at intervals (e.g. Jury et al. 2001; Barber & Cobb 2009), thus requiring large storage capacity and power. Motion-activated video (e.g. Kross & Nelson 2011) is also not a viable option in underwater environments because of the near-constant movement of particles that would activate the video. Third, full-spectrum lighting, which can have a profound effect on animal behaviour in deep-water environments (Olla, Davis, & Rose 2000; Widder et al. 2005; Ryer & Barnett 2006; Ryer et al. 2009), should be avoided. Fourth, the system should be deployable on uneven substratum types without the assistance of divers, thus mandating a righting system or the ability to record in any position. Finally, the system should be inexpensive to build and deployable from a small vessel with no special equipment, to be appealing for behavioural and ecological research.

In this paper, we describe a novel, deep-water, in situ recording system that meets the above requirements. The development of this system was prompted by a call for the design of improved traps that selectively catch spot prawns (Pandalus platyceros) while excluding juvenile rockfish (Sebastes spp.), a group of species of conservation concern in British Columbia, Canada; thus, the custom-designed camera system was dubbed ‘TrapCam’. We review the capabilities of the system and provide results from a pilot study that suggest that the TrapCam can provide significant insights into interactions between deep-water animals and passive fishing gear beyond what catches can reveal. We then explain how modified versions of this basic design could be adapted to study deep-water animals at depths far exceeding those described in this study.


Design of camera apparatus

The camera and electronics for TrapCam were assembled with readily available parts from components marketed to semiprofessional film-makers (Table 1). Video was recorded with a Sony HDRXR550V Handycam, powered by an NPFV100 Sony battery pack. An additional external battery pack (PowerStream PST-MP3500-I with PST-MP3460 pack and Sony connectors) was connected to the camera’s DC-in slot. This camcorder’s lens was a 37 mm ‘G Lens™’, and video was recorded in 16:9 widescreen, giving a wide field of view that we further increased by attaching a screw-on Opteka 0·43× wide-angle lens to the camcorder. This system was capable of recording at 1080 pixels resolution (1920 × 1080) for 13 h. Firmware limitations on the camcorder prevented recording videos longer than 13 h, despite the fact that the camcorder has a 240 GB internal hard drive, which provides sufficient capacity for more than 24 h of continuous recording. The electronics were enclosed in a cylindrical anodized aluminium pressure case (15 cm diameter × 25 cm inside depth) with a 2-cm-thick scratch-resistant polycarbonate viewport at one end (custom manufactured by A.G.O Environmental Electronics, Victoria, British Columbia, Canada; Fig. S1). This case was certified by the manufacturer to sustain pressures of up to 11 atm (or 100 m depth). The camera and battery pack were held in place simply by contact with the inner sides of the case and by packing the empty parts of the case with ‘Pick-N-Pluck™’ foam (Fig. S1).

Table 1.   TrapCam components and estimated costs (USD in 2011)
ItemCost ($)
  1. The list does not include the fishing gear on which the TrapCam is mounted.

Pressure case (Anodized aluminium, 100 m rating)1250
Sony HDR-XR550V camcorder1000
Sony NPFV-100 battery pack130
PowerStream PST-MP3500-I with PST-MP3460 battery pack270
Four Princeton Tec Torrent LED Scuba lights240
Four high output red LEDs20
ABS pipes and cement60

The pressure case was mounted atop a frame constructed of 7·62-cm-diameter ABS pipe, which, for our purposes, was connected to a standard commercial truncated cone prawn trap (as described in, Rutherford, Nguyen, & Gillespie 2004; Fig. 1, Model S1), but could be attached to any other structure or frame. The ABS pieces were secured using ABS cement, and the pipes were attached to the prawn trap using plastic cable ties. The case was locked into place by a steel bar placed through the top of the frame. The camera was oriented facing downward about one metre above the trap, giving a top-down view of the trap with a field of view of 110 cm by 80 cm. This view enabled us to see the contents of the trap as well as the area immediately surrounding the trap. To maintain an upright orientation when deployed, a cross-shaped ABS pipe frame was attached to the bottom of the trap. The trap was weighted internally with three standard red bricks (approximate total weight: 2·5 kg), to add mass close to the centre of gravity of the apparatus and thus minimize tipping once deployed.

Figure 1.

 Labelled diagrams (not to scale) of (a) the TrapCam apparatus and (b) a schematic of how the apparatus is deployed at the centre of a 3-trap string, which is anchored to the bottom by a cinder block at either end. Each block anchors a surface buoy.


Little ambient light is present at 100 m depths, and as a result, an external lighting system was necessary. The retinal pigments in deep-sea fishes and many crustaceans tend to be insensitive to red light (Goldsmith & Fernandez 1968; Meyer-Rochow & Tiang 1984; Douglas, Partridge, & Hope 1995), making such light a good tool for observations of natural behaviour. However, our initial attempts to use infrared light systems were unsuccessful as they did not provide sufficient illumination, owing to the high attenuation of infrared and red light in water (Pegau, Gray, & Zaneveld 1997). Instead, we attached four Princeton Tec-Torrent LED scuba lights modified to use red LEDs in lieu of the full-spectrum illuminators that the lights are usually equipped with. We attached each light (powered by eight AA rechargeable batteries each) to the ABS frame to illuminate the field of view (Fig. 1).

Gear deployment

The camera system was deployed for a pilot study in Howe Sound, British Columbia (49°25′30′′N, 123°20′00′′W), in July and August 2010, using the Simon Fraser University research vessel CJ Walters (length: 9·8 m; beam: 3·7 m). Based on published locations of regular prawn surveys by Canada’s federal Department of Fisheries and Oceans and on experience (by B. Favaro) on commercial vessels in the area (Fig. S2), we selected deployment sites that would maximize prawn catch and/or bycatch, thus maximizing potential interactions between prawns, rockfish and traps. Gear was deployed in strings (i.e. multiple traps connected to a single line weighted with one cinder block at either end) as in the commercial prawn fishery (Fig. 1b, Model S1, Video S1). In our case, the string consisted of the camera-equipped trap paired with two unmodified traps, one of either side of the camera, to evaluate any difference in catch between traps with and without cameras. We deployed the 3-trap string on 16 occasions, and the deployments always commenced during daylight hours, recording into the night. We recorded the GPS coordinates as well as the depth of deployment (mean ± 1 SD = 88 ± 10 m, range = 60–100 m), taken from the vessel’s depth sounder, for each deployment.

We left the strings to soak for 17–45 h (mean = 24 h). All but three strings were recovered the day after they were set. We retrieved the strings using an electric anchor puller. We counted the number of individuals of each species caught in the traps, identified to the lowest possible taxonomic level. To compare the number of prawns caught in traps with and without cameras, we performed a nested analysis of variance (anova), with string nested within camera treatment to test for variability in catch between strings.

Video analysis

We recorded videos at a resolution of 1920 × 1080 pixels using the ‘night mode’ and ‘low lux’ settings on the camcorder. Each 13-h video was 100 GB in size, and the files were stored on external USB hard drives. We watched videos using Elecard AVC HD Player software on 24-inch 16:9 (widescreen) flat screen monitors.

For the purposes of the present analysis, we scored three of the videos (39 h). We noted in detail the actions of all animals caught on film. An approach was recorded whenever an animal entered the field of view of the camera. We recorded data in 30-s bins across the entire video to facilitate the scoring of rapid events while preserving the chronological order of such events. As most organisms were not individually identifiable, individuals were undoubtedly counted multiple times as they left and re-entered the field of view. We recorded entries into the trap, as well as how the animal entered the trap (i.e. through a tunnel or through the mesh). Prawns, other crustaceans, rockfish, fish other than rockfish and other animals were all recorded separately. When possible, we identified the organism in the field of view to species.


Camera performance

We recorded a total of 208 h of video across the 16 camera deployments. We inferred from the videos that the apparatus was deployed upright for all but one of the 16 trials. The technique of attaching a camera-bearing trap to a string as with normal prawn traps was effective, and it was easy to both set and recover the gear. Image quality was excellent (Fig. 2), and the lens was never affected by fog inside the case. Many species could be clearly identified (Video S2). The lighting was effective but dim at the periphery of the viewable area. In addition, images at the periphery appeared slightly distorted because of the use of the wide-angle lens, which made species identification more difficult. Illumination decreased slightly over the course of the 13-h videos but did not inhibit conducting observations. However, at the end of each 24-h deployment, light output would have been insufficient to illuminate the trap. An improved battery system for the lights would therefore be required to record videos longer than 13 h.

Figure 2.

 Sample TrapCam image. In this image, a top-down view of the prawn trap and its immediate surroundings are visible. The three openings are marked by (A). Prawns on top of the trap (B) and inside the trap (C) are visible in the image. This image was recorded 7 h into a deployment.

Effect of camera on catch

Spot prawns (Pandalus platyceros) were present in 94% of traps (mean ± 1 SD = 22·5 ± 21·4 prawns trap−1, n = 48). The trap equipped with a camera caught similar numbers of prawns as traps without cameras (mean ± 1 SD; camera = 29·7 ± 27·9 prawns trap−1, n = 16 deployments, no camera = 18·9 ± 16·6 prawns trap−1, n = 32 deployments; nested anova, camera effect: F1,44 = 2·75, P = 0·11). In addition, there was no difference in prawn catch between strings within each trap type (F2,44 = 0·23, P = 0·79), suggesting that all strings caught comparable numbers of prawns across our study.

Highlights of in situ observations of animal–trap interactions

Spot prawns comprised 95·8% of all approaches to the trap (Table 2). Across the three camera-trap deployments, an average of only 12·6 ± 12·9% of prawns that approached the trap actually entered it. Although only spot prawns were caught in the three deployments analysed (Table 2), we witnessed 65 interactions (1·7 ± 14·4 approaches per hour) between nontarget species and prawn traps and an average of 4·6 ± 3·5 identifiable taxa (range = 1–8) in or on the camera-bearing trap during three deployments (Table 2).

Table 2.   Species recorded in three 13-h videos taken during deployments of a camera attached to a prawn trap at 100 m depth
Common nameScientific nameSourceNumber of approachesNumber of observed trap entriesFinal number caught
  1. Taxa are listed in order of decreasing frequency of approaches. Note that the final number caught when traps were retrieved can exceed the number of entries because only the first 13 h of each soak were recorded.

Spot prawnPandalus platycerosBrandt, 18511426180198
Quillback rockfishSebastes maliger(Jordan & Gilbert, 1880)1610
Fish-eating starStylasterias forreri(de Loriol, 1887)1000
Cancrid crabsFamily Cancridae 900
FlatfishOrder Pleuronectiformes 800
Dungeness crabMetacarcinus magister(Dana, 1852)700
Red rock crabCancer productusRandall, 1839600
Tanner crabGenus Chionoecetes 400
Pink shrimpPandalus eousMakarov, 1935300
Rockfish spp.Genus Sebastes 200

We captured on film a quillback rockfish that entered the trap approximately 6 h after deployment and remained in the trap until the end of the 13-h video. The fish successfully exited before the gear was retrieved. While in the trap, this rockfish attempted to escape by swimming upward against the top of the trap (Video S3).


We assembled and field-tested a new, inexpensive camera system, suitable for in situ behavioural studies of aquatic organisms. Various camera technologies have been previously used for similar or related purposes. A brief review of the most recent technologies (Table 3) highlights the unique niche of TrapCam in terms of cost, working depth and ease of deployment. Pilot deployments of TrapCam, performed as a proof of concept, yielded significant insights about interactions between deep-water animals and fishing gear.

Table 3.   Comparison of selected camera technologies that can be used to study animal behaviour underwater
Camera typeSourceDeployment depth (m)Deployment methodRecording typeVideo duration, resolution and lightingPlatform typeRequires surface support?Primary useRelative cost
  1. Improved versions of the systems described may exist but have not yet been used in published studies. Relative cost is indicated by the number of dollar signs ($ = USD 0–5000; $$ = USD 5000–20 000; $$$ = >USD 20 000).

TrapCamThis study100Self-righting system that descends uncontrolledContinuous (30 fps)13 h, 1920 × 1080, red LED (modified SCUBA lights)Trap mounted – adaptable to any weighted framesNoBehaviour – prawn traps at various depths or other purposes requiring top-down view$
Lobster TVJury et al. (2001)6Controlled descent from boatTime-lapse VHS (2 fps)12 h, 480 × 300, red LEDAttached directly to lobster trapNoBehaviour – lobster traps in shallow water$
Crab video systemBarber & Cobb (2009)5Controlled descent from boat2 s record per 20 s interval8–11 h, 720 × 480, unlitAttached directly to crab trapNoBehaviour – crab traps in shallow water$
Multi-camera systemMills, Verdouw, & Frusher (2005)30Deployable anywhere where three anchors can be set, away from boat trafficContinuous recording with up to eight cameras, variable fps24 h at 1 frame per second, 480 × 300, red LED or infraredFloating pontoon with electronics package, secured by three anchors, with camera as deep as 30 mYes – tethered to floating platform – camera signals transmitted via microwave to separate station up to 10 km awayBehaviour – traps on reefs and other substrata$$
University of Rhode Island underwater cameraMilliken & DeAlteris (2004)Up to 1000Attached to trawl net, pulled with trawling boatContinuous2 h, 320 × 240, no external lightingTrawl mountedYes – cable connects camera to vesselBehaviour – trawling$$
VENUS (CMAP Systems Cyclops Digital Camera)Woodroffe & Round (2008)103Permanent observatory, camera mounted onto structure using remotely operated vehicleContinuous (30 fps)Unlimited duration, 720 × 480, 100 W incandescent lightsPermanent stationary platform, orientation controllable remotelyYes – cable connection to on-ground observatoryCommunity composition around observatory$$$
Baited Remote Underwater Video stationReviewed in Cappo, Harvey, & Shortis (2006)Up to 3420Uncontrolled descentContinuous or time-lapse (30 fps to fewer than 1 frame per minute)Depends on camera used – minutes to hours, various resolutions, generally full-spectrum lightCamera mounted to frame, giving horizontal look-outward viewNoAbundance estimation and community composition$ to $$
CrittercamHerman & Bakhtiari (2007)298Mounted on large animalsContinuous (30 fps)180 min, 720 × 480, no external lightingAnimal mountedNo – after time, device detaches from animal, recovered by boatBehaviour$$

Cost effectiveness and system flexibility

TrapCam has proven to be cost efficient and more capable in many ways than other camera systems (Table 3). In terms of procurement cost, TrapCam was substantially less expensive than a mid-sized ROV, notwithstanding the fact that ROVs would have been much more invasive and not permitted the bottom time necessary for behavioural research on traps. A mid-sized ROV can cost between US$50 000 and US$100 000 to procure and requires a large, crewed vessel with a crane to operate. Usage fees for a mid-sized ROV can approximate US$3000 per day inclusive of vessel and crew (J. Martin, Simon Fraser University research vessel/ROV operations manager, pers. comm. March 2010). In contrast, each TrapCam unit costs approximately US$3000 to procure and build (Table 1). The cost of each deployment is then limited to the cost of operating a small vessel. The only equipment required on the vessel is a winch or electric puller to assist in hauling the gear, although manual retrieval of the gear is possible (though physically taxing). The size of vessel required to deploy and retrieve the system would depend on the environmental conditions at the deployment site, but in calm seas we believe that our system could be safely deployed from a vessel as small as 6 m in length.

A low-cost, self-contained camera system like the one implemented here could be used to study any trap fishery. Crab, sablefish, Nephrops norvegicus, and other trap fisheries would benefit from the behavioural insights gained, both in terms of designing more effective fishing gear and in studying the behavioural dynamics of species interacting with traps and each other at depth. In addition, an apparatus like TrapCam could be effective for visualizing the impacts of trap gear on habitat (i.e. by observing habitat destruction when the gear impacts the bottom). We could also envisage deploying TrapCams on short, scaffold-like frames that could be baited, in a deep-water, long-lasting version of BRUVs, to obtain estimates of relative species abundance and information on foraging behaviour and/or competitive interactions among species. The principle of a self-contained camera system could theoretically be applied to any depth, with deeper systems simply requiring a more robust pressure case.

Successful practical application of TrapCam

TrapCam successfully permitted long duration, in situ observations of animals in and around prawns traps deployed in deep water. Importantly, the camera apparatus did not appear to affect the behaviour of prawns and other organisms observed. Indeed, there was no evidence of attraction or avoidance of the trap or camera from the behaviour of animals captured on video. In addition, the final number of prawns caught was similar between camera-bearing and control prawn traps.

Analysis of videos taken from only three deployments of our camera-bearing trap has shed light on three aspects of interactions between target and nontarget species and passive fishing gear. First, a larger diversity of organisms is present in and around traps at depth than is observed in trap catches. Although up to eight identifiable taxa were present around prawn traps, only prawns were caught when the gear was retrieved, suggesting that prawn traps are highly effective at excluding or facilitating the exit of nontarget species and that high catch specificity was not simply a result of low species diversity at the deployment sites. Second, few of the prawns attracted to traps actually enter them, suggesting that prawn catchability may be quite low, potentially informing our assessment of the relationship between fishing effort and fishing mortality in this system. Finally, the entry into the trap and attempted escape by the rockfish provide important information, which could help us redesign prawn traps to minimize rockfish bycatch. The rockfish in the trap attempted to escape by repeatedly swimming upward against the mesh on the top of the trap (Video S3). If this behaviour is characteristic of trapped rockfish, then adding exit panels to the top of traps might facilitate fish exit.


The main goal of our study, i.e. developing a low-cost tool to perform minimally invasive observations of animal behaviour in deep water, was achieved. We were able to gather many hours of useful data, and preliminary analysis of the videos has already yielded results that could not have been obtained through examination of catch data alone. Although TrapCam was designed for the purpose of devising bycatch reduction modifications, this basic design could be employed by any researcher interested in the behavioural dynamics of animals in deep water and is of particular use for studies of fish and shellfish traps. The use of TrapCam can enable researchers to perform minimally invasive observations on animals that were previously effectively inaccessible for behavioural study.


Funding was provided by the Natural Science and Engineering Research Council of Canada (NSERC) (CRDPJ 341335-06) to SD, the Pacific Prawn Fishers’ Association, MITACS, the PADI Foundation, and the Oceans and Marine Fisheries Branch of the Province of BC. BF was supported by an NSERC Post-Graduate Scholarship, a Garfield Weston/BC Packers Graduate Scholarship and a J. Abbott/M. Fretwell Graduate Fellowship. Thank you to Vin Fogarty for assistance in the field.