What is Big BRUVver up to? Methods and uses of baited underwater video
Baited Remote Underwater Video Stations (BRUVS) is a popular technique to assess mobile nektonic and demersal assemblages, particularly for fish communities. The benefits of using BRUVS have been well documented, with their non-destructive and non-extractive nature, ease to replicate, relatively-cheap personnel costs, and low risk to personnel often cited. However, there is a wide variability in the set-up, experimental design, and implementation of this method. We performed a literature review of 161 peer-reviewed studies from all continents published from 1950 to 2016 to describe how BRUVS has been used by quantitatively assessing 24 variables, including camera set-up and orientation, soak time, bait quantity, type and preparation method, habitat and depth deployed in, and number of replicates used. Such information is critical to gauge the comparability of the results obtained across BRUVS studies. Generally, there was a wide variety in the location, deployment method, bait used, and for the purpose that BRUVS was deployed. In some studies, the methods were adequately described so that they included information on the 24 variables analysed, but there were 34 % of studies which failed to report three or more variables. We present a protocol for what minimal information to include in methods sections and urge authors to include all relevant information to ensure replicability and allow adequate comparisons to be made across studies.
KeywordsBRUVS Fish assemblages Nekton Non-destructive Behaviour Methodology
Information about marine ecosystems is becoming increasingly sought after as the understanding of their importance in ecosystem services, global processes, and economies increases (Costanza et al. 1997). For many of these services, fish and other nekton are particularly important and have been the main focus of several studies (e.g. Holmlund and Hammer 1999; Worm et al. 2006). Such studies have highlighted the need for methods which are capable of sampling a large portion of the population or community, are non-extractive, and allow for simultaneous counts of multiple taxa. There is also a growing desire for more behavioural data about fish species, along with less destructive methods suitable for protected areas, and for methods that are cheap, repeatable, and comparable. Baited underwater video (for the purpose of this review referred to as Baited Remote Underwater Video Stations or BRUVS) is a popular technique to assess mobile nektonic and demersal assemblages, particularly for fish communities and fits the above criteria.
BRUVS have been compared to many other commonly-used techniques for assessing fish assemblages with the most common comparison being between BRUVS and Underwater Visual Census (UVC) (e.g. Stobart et al. 2007; Colton and Swearer 2010; Lowry et al. 2012) or Diver Operated Video (DOV) (Watson et al. 2005; Langlois et al. 2010; Watson et al. 2010). Other comparisons include: BRUVS versus baited traps (Harvey et al. 2012a; Wakefield et al. 2013; Langlois et al. 2015); versus angling (Willis et al. 2000; Langlois et al. 2012a; Gardner and Struthers 2013); versus trawling (Cappo et al. 2004); versus seine netting (Whitmarsh 2012); versus longline surveys (Brooks et al. 2011; Santana-Garcon et al. 2014a; McLean et al. 2015); and versus Automated Underwater Vehicles (AUV) and towed video (Seiler 2013). These studies show that BRUVS are a useful tool with many benefits compared to more traditional techniques. Nevertheless, as each study has aims that vary, the appropriate method to use should be selected on a case-by-case basis (see Murphy and Jenkins (2010) or Mallet and Pelletier (2014) for a review of the benefits and biases of these methods in relation to BRUVS).
Over the last 15 years, as the available technology improved and the aims of studies using this equipment have broadened, the methods used when deploying BRUVS have progressively increased in variety. Factors that can vary from study to study include: the number and orientation of cameras; soak time (i.e. the amount of time the unit is left underwater); habitat(s) sampled; depth ranges of deployments; and the number of replicates used. The bait used can differ in terms of type, quantity, and preparation method. The type of video metric (i.e. how fish and other nekton are counted or measured) can also be different across studies. Standardisation in the use of BRUVS has previously been attempted (Cappo et al. 2007) to allow for a better comparison across studies, but modified or novel approaches to this technology are continually arising, increasing variability in methods used. We propose that authors should ensure that they provide enough information to allow comparisons between the different BRUVS set-ups used, instead of attempting to reach a level of standardisation that might not be achievable. The overall purpose of this literature review is to explore how and in which ways different studies have used BRUVS. We hope to highlight: the need for a comprehensive and descriptive method section; aspects which could be further investigated to improve the informational output of BRUVS; other unexplored applications of BRUVS; and ultimately suggest a protocol of information that authors should routinely include in the methods section.
A list of the 24 variables included in this review of baited studies that also will act as a protocol for factors to include in method sections
# of studies reported in (% out of 161)
When and where
1996, 2006, 2016
Location study was conducted in
Adelaide, South Australia, Australia
Temperate, tropical, polar
Marine, estuarine, freshwater
Seagrass, rocky reef
About the video system
Name of systems
Orientation of camera(s)
Horizontal, vertical (to substrate)
Number and type of cameras
1 or 2, GoPro Hero 3+, Panasonic HandyCam
Type of length measurement
Fork length using stereo-BRUVS
Max range visible
3 m, to bait bag
30, 60 min
Distance between reps
250, 500 m
About the bait
Sardines, Sardinops sagax
500, 1000 g
Crushed, whole, chopped
Mesh bag, perforated PVC bait container
About the deployment
3, 10 m
50, 25 m
Variation in depth (range)
47, 15 m
About the sampling design and analysis
Number of replicates
Video metric used
MaxN, T1st, etc.
EventMeasure, VLC etc.
Teleost, Chondrichthyes, Cephalopoda, Crustacea
% to species level
75 % able to be identified to species level
Results and discussion
A comparison of methods used in baited video studies
Description of the study: when and where
We found that BRUVS studies had been performed in all continents as well as within international waters (Fig. 1b). Oceania was by far the most popular (70 %) region for studies to be conducted in, with Australia alone contributing 61 %. Other continents had generally fewer studies conducted there, especially Africa (6), Asia (6), and South America (3) along with Antarctica (3). Geographically, temperate areas were the main focus areas at 42 %; however, tropical and sub-tropical areas were still well-represented at 26 and 35 %, respectively. Polar areas received less attention and were only investigated in 2 % of studies. The majority of the studies included in this review were exclusively in marine ecosystems (94 %), while 4 % were in estuarine and only 2 % in freshwater ecosystems (e.g. Ebner and Morgan 2013; Ebner et al. 2015).
The most common habitat in which BRUVS were deployed was reef areas, with coral and rocky reefs together accounting for 43 % of the habitats studied (Fig. 1c). Studies in multiple habitats (23 %) also commonly used reef habitats as one of their sampled areas. Rocky reefs were more commonly sampled than coral reefs, which may be due to the prevalence of rocky reef areas within the temperate regions of Australia, where a large proportion of BRUVS studies are conducted. Pelagic (7 %; e.g. Rees et al. 2015) and deep-water (12 %; e.g. Collins et al. 2002) habitats were also studied. Seagrass and ‘other’ habitats were less common with only 2 % for seagrass (e.g. Whitmarsh et al. 2014) and 9 % for the ‘other’ category, which included soft sediments (e.g. Howarth et al. 2015) and restricted habitats such as intertidal rock pools (e.g. Harasti et al. 2014). The prevalence of use in reef habitats is most likely a factor of increased visibility through the water column compared to some other benthic (soft bottom) habitats. Reef areas are often home to commercially-targeted fish species and are prime areas for tourism such as snorkelling and diving, which makes these areas of high commercial interest. Ecologically, reef areas support a wide range of species and usually have high biodiversity (Malcolm et al. 2007) leading them to be targeted by researchers and managers.
Variables relating to the video system
The terminology surrounding BRUVS was widely variable, with the name of the unit falling into more than 17 categories (Fig. 1d). Three very similar unit names dominated the literature: BRUV, BRUVS and BRUVs; all acronyms standing for Baited Remote Underwater Video Stations or Systems (Fig. 1d). BRUV(S/s) as an acronym appears to have first been published by Cappo et al. (2001). Other common names include Baited Underwater Video (BUV) and more general names such as baited landers or baited video. Some authors have developed individual names for their systems such as DeepBRUVS (Marouchos et al. 2011) and BotCam (Merritt et al. 2011). Generally, having multiple names can be a problem because it leads to confusion and allows for ambiguity about the method. Multiple names also make literature searches more difficult and may confuse non-specialists. The name BRUVS has been trademarked by AIMS, but it is not linked to any patent of the design and AIMS does not enforce the use of the trademarked name in peer-reviewed publications (M. Cappo, pers. comm.). Since variations on BRUV(S/s) are the most commonly published names for this method, we urge that a standard form of this name be chosen and used. We are recommending BRUVS as this name has most prevalent use (Fig. 1d).
The orientation of the camera(s) is an important aspect to consider when setting up the BRUVS arrangement. The majority (85 %) of BRUVS set-ups used a horizontal camera arrangement, while 14 % had a vertical orientation pointing down towards the seafloor; the remaining (1 %) studies did not specify the camera orientation. The orientation of the camera can affect the number of organisms that can be observed or reliably identified. For example, Langlois et al. (2006) showed that a horizontal set-up recorded 14 species versus four for a vertical set-up, with some species appearing shy of entering the vertical field of view, most likely due to the perceived confined space under the camera. A major benefit of vertical set-ups, however, is the ability to measure fish size with single cameras using the known fixed height above the substrate and a ruler to measure fish. Vertical BRUVS were used first in the early 2000s (e.g. Willis and Babcock 2000) but have had limited use across the years, with one third of vertical set-ups occurring in deep habitats. The prevalence of horizontal BRUVS is likely because of the increased field of view (depending on water clarity) and the ease of identification of many fish species from a side-on perspective.
BRUVS are predominantly used with one (single) or two (stereo) cameras. Single-BRUVS consist of one camera usually mounted directly behind or above the bait arm. Stereo-BRUVS consist of two cameras mounted at specific angles (usually 7°–8°) to each side of the bait of the arm and are calibrated to allow for accurate fish measurements (see Harvey et al. 2002a for more details). Single-BRUVS are smaller, lighter, cheaper, and take less time to set-up (prior to and during field work) than stereo-BRUVS. Stereo-BRUVS take up more boat space, require more specialised gear for retrieval and may be a limiting factor for replicate numbers when using smaller vessels or make field costs higher by requiring more days in the field than single systems. The calibration of stereo-BRUVS also adds to preparation and analysis time. Based on our literature search, the majority of studies (60 %) used a single camera compared to only 36 % using stereo-BRUVS and four studies using a combination of both systems and one failing to specify the number of cameras used. It is likely that the prevalence of single-BRUVS is the result of their ease of use, affordability, and space constraints. Overall, the question of whether to use single- or stereo-BRUVS may come down to a number of factors (e.g. money, space, and time) but ultimately should be decided depending on whether there is a need for accurate length measurements to fulfil the proposed aims of the study.
Length measurements can be used to estimate biomass, gain an understanding of population and recuitment dynamics, and estimate fecundity (Ricker 1975). It can also be particularly useful in protected areas, where there is an expectation that fishing influences the size of fishes and that there will be a different size distribution in protected areas compared to unprotected areas (e.g. Watson et al. 2009). There were 65 studies that either used stereo-BRUVS or mentioned length as a variable for their study. Out of those 65, 38 % failed to present or use any of the estimated length data. Typically, studies which did present the length data were evaluating the use of length data under a range of circumstances, e.g. for precision (Merritt et al. 2011), with new technology (Letessier et al. 2015) or over different soak times (Misa et al. 2016), comparing lengths or biomass between protected and unprotected areas, or across different methods (e.g. Langlois et al. 2015), habitats (e.g. Fitzpatrick et al. 2012), or other impacts (e.g. seasons (McIlwain et al. 2011). Where length data were presented, it was most often measured using stereo-BRUVS (41 studies) and within those, 15 studies presented fork length data. There were 20 studies that estimated fish length using single cameras, most often using a reference object or ruler within the field of view (e.g. known length of bait bag) to gauge fish length. Stereo-BRUVS (or single-BRUVS with the ability to accurately measure fishes (see below)) are necessary to answer specific questions where fish size is a critical variable, but where this information is not required and in 38 % of studies not even presented, we believe that sampling effort and the additional cameras required for stereo could be better spent on increasing replication.
Studies have compared the accuracy of stereo-BRUVS versus single-BRUVS and shown that the accuracy of fish length measurement using single-BRUVS deteriorated with distance from the measuring scale (±2 m) and angle of view (>50°), while stereo gave a good estimate of length at a variety of angles and distances within 7 m (Harvey et al. 2002b). Some work has been done more recently to improve the accuracy of measurements taken from single-BRUVS, such as the development of mirrored surfaces allowing for a more exact positioning of fish in vertical set-ups, leading to more accurate measurements (Trobbiani and Venerus 2015). It is also possible to obtain accurate length data using on a known ratio of eye to head height predetermined for each fish species (Richardson et al. 2015). This method could be especially useful for targeted studies that are focusing on a few species only, as the proportion of eye size to head height has to be calculated for each fish species prior to BRUVS deployments. Developments such as these continue to improve BRUVS as a method and make it more accessible by providing ways to gain additional accurate information from single-BRUVS.
Soak time differed greatly across studies (Fig. 2b), with an apparent trimodal distribution. Peaks occurred around 30, 60, and >90 min. Few studies used times of less than 30 min, but 17 % of studies ran for more than 90 min, which often involved the use of additional power sources or extended batteries (e.g. Jamieson et al. 2006). Four studies (Gladstone et al. 2012; Santana-Garcon et al. 2014c; Harasti et al. 2015; Misa et al. 2016) specifically compared different soak times and found 60–90 min to be optimal for an estuarine environment (Gladstone et al. 2012), 120 min optimal for pelagic habitats (Santana-Garcon et al. 2014c), and 30 min was found to be sufficient in rocky reef habitats (Harasti et al. 2015). Misa et al. (2016) found shorter soak times (15 min) sufficient for snap-shot abundance estimates of Hawaiian bottomfish assemblages. Furthermore, some studies have included pilot studies of longer soak times to determine species and abundance accumulation curves, which justified a shorter time to be used subsequently in the main study (e.g. Unsworth et al. 2014).
Variables relating to the bait
The use of bait compared to unbaited systems has been specifically investigated by four studies (Harvey et al. 2007; Bernard and Götz 2012; Dorman et al. 2012; Hannah and Blume 2014). Bait increased the similarity between replicates providing better statistical power (Harvey et al. 2007; Bernard and Götz 2012; Dorman et al. 2012). Bait also increased the number of predatory and scavenging species, while not affecting the numbers of herbivorous and omnivorous fishes seen, and baited replicates were better able to detect changes between habitat types (Harvey et al. 2007; Bernard and Götz 2012; Dorman et al. 2012). Hannah and Blume (2014) showed that bait increased the abundance of deep-water demersal fishes by 47 % and lured the fish closer to the unit allowing for more accurate length measurements and species identification. Overall, these studies conclude that the benefits of using bait in marine environments appear to outweigh any perceived costs. There has been concern, however, about the ability of bait to attract fish from a large area potentially leading to inflated densities (Taylor et al. 2013). Variability in currents, winds, and turbidity across replicate deployments can lead to large changes in plume dispersal and significantly alter interpretations (Taylor et al. 2013). Studies rarely considered this factor and few studies implemented current measuring devices, such as current meters or drogues (Taylor et al. 2013).
Various quantities of bait have been used when deploying baited video ranging from 50 g to over 2 kg (Fig. 4). The most common bait quantity was within the 801–1000 g category (with 27 % of studies; Fig. 4), with a majority of these using approximately 1000 g (Online Resource 1); 501–800 g was the next most popular category with 20 % of studies using a quantity within this range (most commonly 800 g; Online Resource 1). Bait quantity was thus more varied across studies than the type or preparation method used (Fig. 4). Only one study tested whether varying the quantity of bait affected the observed fish assemblages (Hardinge et al. 2013). There were no significant differences in fish diversity between 200, 1000, or 2000 g of bait but there were some individual species differences, with the moray eel, Gymnothorax woodwardi, being significantly more abundant with 2000 g of bait than with 200 g (Hardinge et al. 2013). The lack of differences in fish diversity may have been caused by the limited bait depletion (i.e. low bait predation) leading to fish being equally attracted to the baited video throughout the deployment regardless of the quantity of bait used (Harvey et al. 2007). There is, however, a need to spatially replicate the study in areas likely to have high bait depletion, such as those with high fish abundance or areas with different water temperatures that may affect the foraging rates of fishes. There was a variable positive correlation between bait quantity and soak time (Fig. 3b; Pearson correlation p = 0.221, 2-tailed probability = 0.015), which only suggests a weak trend for the deployments with longer soak times (>90 min) to also use more bait (> 1000 g). Overall, there appeared to be little consideration of the appropriate quantity of bait to choose and this remains an area for improvement and future research.
Despite the copious literature available on BRUVS, there appears to be no studies investigating the effects of the preparation method for the bait. Many authors assume that crushing the bait (particularly sardines) enables a more even plume dispersal (e.g. Watson et al. 2009), but it may be worthwhile for a future study to test this hypothesis. From the literature analysed, bait was prepared in a variety of ways, with crushing being the most common method (55 % of studies). Chopped and whole-bait preparations were less common at 15 and 12 %, respectively, while 16 % of studies did not specify the way the bait was prepared (Table 1).
Deployment method for the bait also varied across studies (Fig. 2d) with a majority of studies (55 %) using some form of mesh bag in which bait was likely to be somewhat accessible to taxa for feeding. The use of a perforated container (usually PVC) was the next most common method (19 %), which served to disperse the bait plume but restricted the access to taxa for feeding on the bait. Some studies (2 %) in the deep-water habitat choose a timed-release method using canisters to enable fresh bait to be released periodically. Six percent of studies used no form of vessel for bait deployment. The remaining 18 % of studies did not specify the bait deployment method. It is possible that the delivery method for the bait may influence assemblages observed as the ability to physically feed on the bait (more likely with a mesh bag) may lengthen the amount of time individuals remain around the BRUVS and hence inflate MaxN, or attract/deter other species. This area of study has not been investigated, but warrants future investigation.
Variables relating to the deployment
Variables relating to sampling design and analysis
The number of replicates used when deploying BRUVS was variable (Fig. 5b). Thirty-two percent of the studies reviewed used four to six replicates at each location. We were unable to determine the number of replicates used in 9 % of the reviewed studies, while 9 % were unreplicated (all of which were in deep-water or ‘other’ habitats). These studies are likely to be un-replicated due to the large cost involved in sampling at such great depths. Some studies (8 %) also chose a stratified sampling design. This design type involves the deployment of BRUVS across a large area while ensuring that replicates are representative of all strata, e.g., habitats or depths, represented within that area (e.g. Moore et al. 2010).
Images from BRUVS videos can be measured in a variety of ways, depending on the aims of the study. The most common metric recorded was MaxN (the maximum number of a particular species seen in any one video frame across the duration of the video record). MaxN was used in some form, either over a set time period or across the whole video, in 81 % of the reviewed studies. MaxN can be used in conjunction with other metrics such as time of first arrival (T1st) or time first fed (Fig. 5c). However, this was done in less than 20 % of cases. MaxN can be modified slightly to include the maximum number seen over a set time period, e.g. 30 s, or can be estimated at specific intervals, e.g. every 5 min. Some studies (10 %) focused upon species-specific metrics such as identifying individual sharks (e.g. Bond et al. 2012; Ryan et al. 2015) or Nautilis spp. based on skin or shell colouration patterns (e.g. Dunstan et al. 2011). The ‘Other’ category includes observation of specific behaviour (e.g. Bailey et al. 2007), total species counts (e.g. Craig et al. 2011), and residence time (i.e. how long animals stayed at the bait; e.g. Smale et al. 2007) metrics.
MaxN can be used to assess the relative abundance of organisms. It is often considered a conservative estimate as more individual organisms may be present around the BRUVS but remain uncounted because they do not appear in the field of view at the same time. This relative abundance measure can be used to assess and compare spatio-temporal differences in aquatic assemblages. Saturation may, however, occur when a high number of individuals obscure the field of view to the point that additional individuals cannot be seen (Schobernd et al. 2014; Stobart et al. 2015). Such saturation can result in the inability to detect differences between locations when fish abundance is high (Stobart et al. 2015) and results in MaxN being non-linearly related to true abundance (Schobernd et al. 2014).
Recently, MeanCount (used in 2 % of studies; Fig. 5c) has been suggested as an alternative to MaxN that can be linearly related to true abundance (Schobernd et al. 2014). MeanCount uses either systematically or randomly selected individual frames from across the video which are subsequently counted and then the mean is calculated. As the entirety of the video is not viewed, MeanCount has a tendency to over-inflate zero observations and is less precise than MaxN (Campbell et al. 2015; Stobart et al. 2015).
T1st is a measure of how fast species are first observed in the field of view. In some cases, there has been a negative correlation shown between T1st and MaxN, meaning that if a species where to arrive quickly to the bait, the species is often highly abundant (Stobart et al. 2015). T1st can also be used to infer the distance a species may have travelled to get to the BRUVS. However, as T1st is influenced by both the distance the fishes are away from the BRUVS as well as the behavioural response to the bait used (i.e. how attracted they are to the bait) which can vary between species, it can be difficult to disentangle what T1st is really showing.
Purpose of studies: what is BRUVS used for?
BRUVS has been used for answering a wide variety of scientific questions (Fig. 6b). The most frequent reason (34 %) for deploying BRUVS was in relation to assessing the effects of marine protected areas (MPAs; e.g. Bornt et al. 2015; Coleman et al. 2015). The large number of studies using BRUVS to study MPAs is likely related to the non-destructive and non-extractive nature of BRUVS, making it a suitable alternative to more traditional methods. Studies that looked at particular species or behaviours (24 %; e.g. Denny et al. 2004; Gutteridge et al. 2011) and those which assessed changes in fish assemblages along a gradient or between habitats (25 %; e.g. Gomelyuk 2009; Langlois et al. 2012b) were the second- and third-most common study aims. Method comparisons both within BRUVS (e.g. different soak times; Gladstone et al. 2012) and between BRUVS and other methods (e.g. BRUVS vs. longlines; Brooks et al. 2011), were also popular with 19 % of studies choosing to focus upon within-BRUVS method comparisons and 18 % on comparisons with other methods. This perhaps reflects a view in many minds that BRUVS is still developing and their use needs justification. There were also studies that investigated day-to-day (Birt et al. 2012) or day-to-night (e.g. Svane et al. 2008) variation and variability in night-time (e.g. Fitzpatrick et al. 2013) assemblages, which accounted for 5 % of the total.
We were able to determine the percentage of taxa putatively identified to species level in 65 % of studies (Fig. 7b). Ten percent of all studies were able to identify all taxa to species level while only 2 % of studies had greater than 30 % unable to be identified to species level and none greater than 40 %. Generally, species that could not be identified to species level were small, cryptic or rare species, which is likely to result in a bias against such species. Visibility may also affect how well species are able to be identified. The type of organism targeted for the study can also affect rates of identification. For example, fish species can generally be reliably identified from video footage, but other smaller mobile animals e.g. crustaceans, echinoderms, and cephalopods can be more difficult to identify. Despite this, since these types of animals are generally less well-studied, any information gathered about them can be useful.
A possibility for assisting newcomers to BRUVS and improving the ease of identification for existing persons could be for more routine and shared image archives. Mentions of image archives are not prominent within the published literature, but archives are likely to exist for many BRUVS teams and the sharing and concatenation of such archives would assist in ensuring the accuracy of identified species.
Out of the 161 studies, 110 (68 %) were considered standard and 51 (32 %) novel. The main novel developments for BRUVS were the extensions into pelagic habitats, modification for deep-water deployments, and night-time uses.
Daytime deployment on the seafloor, in subtidal, shallow (<120 m deep) habitat, single or stereo (2 cameras maximum) camera(s) facing towards the bait bag either horizontal or downwards, a single bait arm with a mesh bait bag attached, single use bait, and with a video length of no more than 90 min.
Overall, only 6 % of the studies analysed had detailed method sections that stated all of the 24 main variables in this literature review. However, 60 % of the studies were only missing values for 2 or less of these categories (Fig. 5d; Table 1). The most commonly unreported variables included the maximum visible range (reported in only 46 % of studies), the software used for analysis (54 %), the number of species able to be identified to species level (55 %) and the distance between replicates (65 %; Table 1).
Novel method development of BRUVS
The use of BRUVS in the pelagic environment is a relatively recent development, with only two studies published using this method up to and including 2012 (Heagney et al. 2007; Robbins et al. 2011). Since 2012, it has increased in popularity with an additional nine studies using BRUVS in the pelagic environment (Letessier et al. 2013; Santana-Garcon et al. 2014a, b, c, d; Anderson and Santana-Garcon 2015; Bouchet and Meeuwig 2015; Rees et al. 2015; Scott et al. 2015). This method involves changing the focus of BRUVS from the traditional demersal setting to suspending the unit within the water column to better sample pelagic fishes. The pelagic BRUVS are horizontally set up and usually allowed to float at a specific depth below the surface (e.g. 10 m; Heagney et al. 2007), as opposed to resting on or near the seafloor in standard use, although some studies set a specific distance above the substrate (e.g. 10 m above the bottom; Santana-Garcon et al. 2014a). Other major modifications include the use of additional floats, ropes, and weights to allow for a stable mid-water deployment. Recently, developments have been made to allow for a drifting pelagic set-up (Bouchet and Meeuwig 2015) that can cover broad stretches of ocean space that in that study had an average transect length of 4.9 km during a 165 min deployment.
Bait plume dispersal has been highlighted as a major factor that could affect the fish assemblages observed via pelagic BRUVS in particular due to the sparse and heterogeneous nature of fish assemblages within this environment (Heagney et al. 2007). Heagney et al. (2007) recommended the addition of a current meter to assist in determining the likely plume dispersal. Taylor et al. (2013) also recommended this or similar current-measuring devices to be used for benthic deployments. Furthermore, an increased soak time (Letessier et al. 2013) and replication (minimum of 8 in tropical environments) is needed to account for the highly heterogeneous distribution of pelagic species (Santana-Garcon et al. 2014c). There has also been evidence for additional attractants to be used alongside traditional bait in the pelagic environment, such as those based on sound (recordings of bait fish) and sight (metallic reflectors; Rees et al. 2015). Rees et al. (2015) compared these different attractant methods and found that the combination of all three attractants was more effective at attracting consistent numbers of fish than the individual components alone.
Six of the 11 studies using pelagic BRUVS (55 %) were focussed on developing and assessing the validity of the method, while, of the others, two looked at behaviour of a particular species (Robbins et al. 2011; Santana-Garcon et al. 2014b), one looked at the the impacts of artificial reefs (Scott et al. 2015), another used the data specifically to demonstrate a novel statistical analysis technique (Anderson and Santana-Garcon 2015) and one focussed upon using BRUVS to determine the effects of MPAs on pelagic species (Santana-Garcon et al. 2014d).
While the use of still photography in the deep sea has occurred since the 1960s (Gage and Tyler 1991), the use of BRUVS in deep-water habitats has only begun in the last 14 years, with the first published articles appearing in 2002 (e.g. Collins et al. 2002; Yau et al. 2002). There are numerous challenges to using BRUVS within the deep sea that are not present in shallower environments, such as increased pressure resulting in the need for sturdier housing for the cameras, reduced light resulting in the need for external lighting sources (and consequently powered by batteries), reduced diversity and abundance resulting in a need for longer soak time and potentially more replication being necessary, which is also compounded by the long descent time from the surface. There are also depths where ropes and surface floats become impractical leading to the need for remote release mechanisms to allow gear recovery. Some deep-water studies also used larger baits, such as pig carcasses, and leave them out for extended periods (days-months; Anderson and Bell 2014). The additional cost for the these features along with increased general field costs associated with working in deep-water habitats means that sampling becomes very expensive, which could be a reason why there is little to no replication with deep-water studies (70 % with none or unknown) and also why 60 % have a soak time longer than 90 min (e.g. Bailey et al. 2007).
The optimisation of BRUVS for use specifically at night has begun recently with studies such as Fitzpatrick et al. (2013), although the use of BRUVS in deep-water habitats has occurred for a longer period and has some of the same challenges (e.g. use of lights). To observe the impact on fish assemblages, Fitzpatrick et al. (2013) examined three different light colours (red, white, and blue) in a range of habitats both inside and outside protected areas. They found that each light affected fish assemblages differently and suggested that this was most likely due to differences in fish behaviour or physiology towards different light sources. The wavelength of red light (620–630 nm), like that of infrared (<700 nm), is below the spectrum that fish are sensitive to but is rapidly attenuated in the water column compared to white and blue light, which can be seen for a greater distance but may attract or disturb some species. Fitzpatrick et al. (2013) found that red light sampled the highest abundance of fish of the three light colours and was particularly good at sampling non-commercial species; however, it illuminated the smallest area due to the attenuation of red light in seawater. White and blue light sampled similar fish assemblages but had higher abundances of some commercially-targeted species such as snapper, Chrysophrys auratus, compared to red, and also illuminated a greater area. The authors recommended further studies into the impacts of light colour on fish assemblages. These results are somewhat different from those found from another study by Harvey et al. (2012b), where white light sampled a greater number of individual fish compared to red but was not able to distinguish between six different benthic habitat types as well as the red light could.
Another study used infrared light to assess nocturnal fish assemblages and compared these results to those from UVC (Bassett and Montgomery 2011). A higher abundance of olfactory specialists, species which rely heavily on sense of smell (e.g. yellow moray eels, northern conger eels, southern bastard cod) were observed from infrared BRUVS compared to UVC, and these species consistently arrived at the bait quicker than non-olfactory specialists. Studies have also used BRUVS to compare assemblages between day and night (e.g. Svane et al. 2008; Svane and Barnett 2008), and found that BRUVS can effectively discern changes between day-time and night-time behaviours, such as an increased consumption of bait at night.
Other novel uses of BRUVS include the development of ‘miniBRUVS’ for use in rockpool environments (Harasti et al. 2014), which is also the only intertidal use of BRUVS that has occurred so far. This development was successfully used to assess the abundance and distribution of a threatened and otherwise hard-to-study rockpool-specialist fish, the black rock-cod, Epinephelus daemelii.
Optimisation versus standardisation: developing a protocol for reporting methods
Optimisation is the trialling of different variables to ensure the best use of resources (time, effort and money) to deliver benefits (e.g. detect increased abundance or diversity or maximise ability to discriminate between factors). There have been several studies that have focussed on the optimisation of BRUVS (e.g. Gladstone et al. 2012; Harasti et al. 2015), with all studies falling into the method development within BRUVS (19 %) considered as working towards optimisation. However, few studies have compared method optimisation between locations or habitats. Different areas even within similar habitat types, such as temperate reefs, still seem to display different values for each optimal scenario, such as seen in a study by Harasti et al. (2015), which showed that, in eastern Australia, the MaxN for many reef species occurred within 12.5 min making a soak time of 30 min quite practical. In contrast, MaxN in South Australia took longer to be reached (30–40 min; Whitmarsh et al. unpublished data) meaning that a soak time of 60 min is more applicable. Both studies used similar methods with the exception that Harasti et al. (2015) used slightly more bait at 1000 g compared to 800 g. Generally, we urge caution when assuming optimal scenarios still apply in different areas or habitats and advise authors to conduct their own pilot studies if possible.
In general, it is easy to deviate from the ‘standard’ use of BRUVS to tailor to specific objectives such as studies of Nautilis sp. using chicken as bait with a soak time of 12 h (e.g. Barord et al. 2014) or modifying the system to work in small rock-pool environments (e.g. Harasti et al. 2014). There is, however, no consensus about whether it is better to tailor the method to each specific scenario being tested or to strive towards standardisation to better enable valid comparisons across studies. The goal of standardisation of BRUVS as a method may be worthwhile but is ultimately, we believe, unachievable and may in fact negatively impact novel developments and methodological breakthroughs. Currently, if comparisons amongst studies are attempted, some authors fail to specify enough details in their paper’s methods section for the differences to be accurately accounted for. We suggest a standard protocol of what information to be included within the literature (Table 1), rather than a standard protocol for use.
We have identified some gaps in the current knowledge base such as the effects of distance between replicates, bait amount, preparation, and deployment method, continued lack of studies accounting for plume effects and using current meters, further impacts of light colours on nocturnal or deep fish assemblages, appropriate soak times under a range of habitats and conditions, and the appropriate numbers of replicates to account for the variable nature of fish assemblages.
One key aspect of method deployment not often covered in the literature is the effect of bait preparation on fish assemblages observed. Although it is unlikely that large differences in assemblages would be observed from using chopped versus crushed sardines, it is reasonable to assume that some differences may result from a comparison of whole versus crushed sardines, if there is any increased areal coverage of plume dispersal coming from crushed bait.
Future research using BRUVS could focus on gaining additional data from the video metrics in addition to MaxN. For example, behavioural data could enhance our knowledge of how species interact with themselves, other species, and bait, while oceanographic data (e.g. temperature, salinity) through attachment of sensors to the unit would provide a way to investigate the influence that these factors have on fish and other nekton. A more formal description of habitat features seen from the images and better use of fish arrival or departure times and hence length of stays could also increase our knowledge of fish assemblages. There is also scope to increase the use of BRUVS outside of reef areas, with some studies showing that it is an effective method for soft-bottom (Gladstone et al. 2012), seagrass (Whitmarsh et al. 2014), pelagic (Rees et al. 2015), and deep-water (D’Onghia et al. 2015a, b) environments.
The other major area for potential growth in BRUVS is to focus on other nektonic species rather than fishes. Combinations of different unit designs and bait may enable BRUVS to be tailored to any number of mobile species including cephalopods, marine birds, marine mammals, marine reptiles, crustaceans, and other benthic mega-invertebrates (e.g. sea stars, sea cucumbers and large gastropods).
Overall, BRUVS is a widely-used method for assessing nektonic assemblages and their behaviour. This review shows the robust and flexible nature of BRUVS and its widely applicable uses from cataloguing the behaviour of particular species to broader changes in mobile communities within a wide variety of depths and habitats. Its use over the last two decades has led to further developments to the method, including the introduction of stereo-BRUVS, pelagic BRUVS, and night-time BRUVS. Several studies have also focused on optimising or standardising the use of BRUVS. To enable more accurate comparisons across studies while still allowing novel and specialised use, we recommend a protocol that authors can follow to allow sufficient detail to be included in methods sections.
We thank two anonymous reviewers for their comments which improved our manuscript.
- Barord GJ, Dooley F, Dunstan A, Ilano A, Keister KN, Neumeister H, Preuss T, Schoepfer S, Ward PD (2014) Comparative population assessments of Nautilus sp. in the Philippines, Australia, Fiji, and American Samoa using baited remote underwater video systems. PLoS One 9:e100799CrossRefPubMedPubMedCentralGoogle Scholar
- Cappo M, Speare P, Wassenberg TJ, Harvey E, Rees M, Heyward A, Pitcher R (2001) Use of Baited Remote Underwater Video Stations (BRUVS) to survey demersal fish—how deep and meaningful? In: Harvey ES, Cappo M (eds) Direct sensing of the size frequency and abundance of target and non-target fauna in Australian Fisheries, 4–7 September 2000, Rottnest Island, Western Australia. Fisheries Research and Development Corporation, pp 63–71, ISBN 1 74052 057 2Google Scholar
- Cappo M, Harvey E, Shortis M (2007) Counting and measuring fish with baited video techniques—an overview. In: Lyle JM, Furlani DM, Buxton CD (eds) Proceedings of the 2006 Australian society of fish biology conference and workshop cutting edge technologies in fish and fisheries science. ASFB, Hobart, Hobart, August 2006, pp 101–114Google Scholar
- D’Onghia G, Capezzuto F, Cardone F, Carlucci R, Carluccio A, Chimienti G, Corriero G, Longo C, Maiorano P, Mastrototaro F, Panetta P, Rosso A, Sanfilippo R, Sion L, Tursi A (2015a) Macro- and megafauna recorded in the submarine Bari Canyon (southern Adriatic, Mediterranean Sea) using different tools. Mediterr Mar Sci 16:180–196Google Scholar
- D’Onghia G, Capezzuto F, Carluccio A, Carlucci R, Giove A, Mastrototaro F, Panza M, Sion L, Tursi A, Maiorano P (2015b) Exploring composition and behaviour of fish fauna by in situ observations in the Bari Canyon (Southern Adriatic Sea, Central Mediterranean). Mar Ecol 36:541–556CrossRefGoogle Scholar
- Ellis D, DeMartini E (1995) Evaluation of a video camera technique for indexing abundances of juvenile pink snapper, Pristipomoides filamentosus, and other Hawaiian insular shelf fishes. Oceanogr Lit Rev 9:786Google Scholar
- Langlois T, Chabanet P, Pelletier D, Harvey E (2006) Baited underwater video for assessing reef fish populations in marine reserves. Fish Newslett S Pac Comm 118:53Google Scholar
- Langlois TJ, Fitzpatrick BR, Fairclough DV, Wakefield CB, Hesp SA, McLean DL, Harvey ES, Meeuwig JJ (2012a) Similarities between line fishing and baited stereo-video estimations of length-frequency: novel application of kernel density estimates. PLoS One 7:e45973CrossRefPubMedPubMedCentralGoogle Scholar
- Marouchos A, Sherlock M, Barker B, Williams A (2011). Development of a stereo deepwater baited remote underwater video system (DeepBRUVS), OCEANS, 2011 IEEE, Spain, pp 1–5Google Scholar
- Merritt D, Donovan MK, Kelley C, Waterhouse L, Parke M, Wong K, Drazen JC (2011) BotCam: a baited camera system for nonextractive monitoring of bottomfish species. Fish Bull 109:56–67Google Scholar
- Ricker WE (1975) Computation and interpretation of biological statistics of fish populations. J Fish Res Board Can 191:1–367Google Scholar
- Seiler J (2013). Testing and evaluating non-extractive sampling platforms to assess deep-water rocky reef ecosystems on the continental shelf. PhD Thesis, University of Tasmania, HobartGoogle Scholar
- Stobart B, García-Charton JA, Espejo C, Rochel E, Goñi R, Reñones O, Herrero A, Crec’hriou R, Polti S, Marcos C, Planes S, Pérez-Ruzafa A (2007) A baited underwater video technique to assess shallow-water Mediterranean fish assemblages: methodological evaluation. J Exp Mar Biol Ecol 345:158–174CrossRefGoogle Scholar
- Whitmarsh S (2012) Methods of monitoring shallow seagrass communities of Kangaroo Island, South Australia, Honours Thesis, School of Biological Sciences. Flinders University, AdelaideGoogle Scholar