The use of cameras on commercial deep-water trawls to investigate habitat associations and scampi trawl efficiency in New Zealand scampi Metanephrops challengeri

This study assessed the effectiveness of using video cameras mounted on commercial trawls for a deep-sea lobster, the New Zealand scampi Metanephrops challengeri, for acquiring information on habitat associations of demersal species and fishing efficiency. Randomly selected frames from 1 h of recording replicated over 12 trawls and conducted over three fisheries management areas (FMA), were analysed for counts of visible scampi, scampi burrows, visible topographical features and visible epifauna and fish taxa. Poisson regression analysis indicated a positive relationship between the distribution of scampi burrows (a proxy of scampi abundance) and conical mounds (formed by bioturbators), but a negative relationship with smooth topography. Counts of scampi burrows were also positively associated with counts of visible scampi for four trawls, while significant relationships were seldom found between counts of scampi burrows and other commonly observed individual taxa. Scampi trawl efficiency was low, with only 56 of the 186 observed scampi (30%) passing over the footrope and into the trawl net. The results confirm the effectiveness of video sampling from commercial trawls to improve our understanding of habitats and fishing efficiency in deep-sea fisheries.


Habitat associations
Associations between species and habitats are one of the foundational principles of ecology (Aarts et al. 2013).Identifying such associations is important for predicting a species distribution to help inform conservation or management efforts (Ando et al. 1998;Guisan and Zimmermann 2000).While broad-scale patterns of association offer insights into a species' potential habitat range (Pimm et al. 2014), the broad-scale distribution patterns may not account for finer scale variation in occurrence (Palmer and Dixon 1990).
For epifaunal and demersal species in deep-sea environments, acquiring finer-scale information on species-habitat associations can be challenging due to the difficulty of gathering data at depth (Robison 2004).As a consequence, the effective management of deep-sea fisheries is frequently hampered by a lack of availability of biological information about the species and the fishing activity (Mengerink et al. 2014).New Zealand scampi Metanephrops challengeri is one such deep-sea species for which biological information is sorely lacking, and its distribution is primarily understood through broadscale patterns derived from commercial trawl catch records (Cryer et al. 2005;Tuck 2009).
Scampi is a commercially important species of clawed deep-sea lobster that is widely distributed on the continental shelf of New Zealand at depths of 200-600 m (Cryer Fisheries et al. 2005;Tuck 2009;Tuck et al. 2015b).The entire scampi fishing fleet comprises around 10 vessels of 20-40 m in length that utilise benthic trawls that are typically a double or triple net configuration with total annual landings of 550-893 t (MPI 2021).The fishery is divided into 10 spatial FMAs (officially referred to as SCI-scampi fishery management area; i.e., SCI1-10) based only on arbitrary geographic boundaries.With some of the most productive scampi FMAs located in distant offshore waters, the fishing vessels will frequently travel for over 24 h to reach fishing grounds and stay at sea fishing for many weeks, particularly on the Mernoo Bank (SCI 3), Chatham Islands (SCI 4), and in the Auckland Islands (SCI 6A) which is in the Sub-Antarctic region to the south of New Zealand.These long travel times and trawl towing distances (up to 50 km) all contribute to the comparatively high fuel demands of the New Zealand scampi fishery.
Scampi build burrows in the sediment and are thought to spend a considerable proportion of time either in their burrows or foraging near them (Tuck 2010;Katoh et al. 2013;Major et al. 2017a;van der Reis et al. 2018van der Reis et al. , 2020)).Their sedentary nature and a strong behavioural preference to stay near burrows suggests that any dispersal in this species most likely only occurs during the short pelagic larval phase (Tuck 2010;Bell et al. 2013;Heasman and Jeffs 2019;van der Reis et al. 2022).While the main fishing grounds are well established, a number of exploratory trawls have discovered that scampi are also present along the west coast of the North and South Islands, including along the Challenger Plateau, although not in commercially exploitable quantities (Tuck 2009).As scampi are dependent on a muddy to sandy seabed habitat, their overall geographical distribution is discontinuous (Bell et al. 2013).Their presence across deep water continental shelf plateaus in New Zealand suggests that sediment type, suitable for burrow construction, and water temperatures may be important limiting factors in their distribution (Tuck 2010).A better understanding of scampi-habitat associations at a range of scales may give insight into the general role of environmental factors in determining scampi density as well as allowing for more targeted fishing.However, scampi distribution patterns at finer scales, as well as their ecosystem roles and interactions with species that are not fisheries targets remain poorly understood (Johnson et al. 2013).Investigations of habitat associations at these finer scales often require techniques or technology that allow for direct observations of species in their habitat.For example, deep water drop cameras are a highly inefficient sampling method because observations can only be taken for a very small area of seabed at a time, making it challenging for assessing species at low densities, such as scampi (Cryer et al. 2001(Cryer et al. , 2004)).

Scampi trawl efficiency
Crustacean trawl fisheries are associated with some of the highest bycatch and discard rates in the world (Alverson et al. 1994;Kelleher 2005;Catchpole et al. 2008).This trend holds true for the New Zealand scampi trawl fishery, with scampi accounting for only 17% of the total catch biomass for all observed scampi-targeted trawls from 1990 to 2010 (Ballara and Anderson 2009;Anderson 2012).The remainder of the observed catch comprises over 450 different finfish and benthic invertebrate species, the bulk of which are unmanaged, with an average of 4.2 kg of fish discarded for every 1 kg of scampi caught since 2006 (Anderson 2012).Efforts to reduce bycatch through the development of a potting fishery for scampi have so far proven unsuccessful (Major et al. 2017b;Ogilvie et al. 2018).
Relatively low vulnerability of scampi to trawling is likely to further compound the negative impacts of high bycatch retention.As scampi are burrow-dwelling organisms, the ability to catch scampi in trawl nets is largely determined by their patterns of emergence (Tuck 2010;Tuck et al. 2015b).However, the proportion of emerged scampi caught during trawls is not well understood and can only be indirectly approximated by comparing trawl catch records with independent estimates of stock density, which are extremely difficult to procure (Tuck 2010).
Benthic trawling remains the sole method of fishing for scampi in New Zealand waters, with the commercial fleet adopting multiple trawl rigs of two or three lightweight nets of very low headline height (MPI 2013).While most efforts to reduce bycatch and improve efficiency of scampi trawls have focused on improving selectivity once fish have entered the net (Hartill et al. 2006), comparatively little effort has been placed on assessing the efficiency of scampi capture at the front of the trawl.
The aim of this study was to provide insight into scampihabitat associations at the level of the trawl and offer an assessment of scampi trawl efficiency by sampling with video technology attached to scampi trawl nets.Specifically, this study used video recording to; (i) identify any associations between the distribution of scampi burrows and visually apparent topographical features of the seabed, visible epifauna and fish taxa, and (ii) measure the proportion of visible scampi observed to enter trawl nets and determine whether scampi trawl efficiency is influenced by scampi size, positioning or topography.Furthermore, these investigations were conducted at three spatially separated commercial fishing grounds for New Zealand scampi to determine whether results were specific to different locations.

Data collection
Video data were collected from 12 scampi trawls conducted in three commercially important scampi fishing grounds, i.e., Bay of Plenty, Mernoo Bank and north of the Chatham Islands (Fig. 1).Five of the 12 trawls from which video data was analysed were conducted using the research vessel (RV) Kaharoa on the Mernoo Bank (Chatham Rise) (FMA SCI 3) and occurred over two separate scampi voyages; 2010: 2 trawls, 2016: 3 trawls (Fig. 1).The trawl locations were randomly selected within known scampi grounds and used a single rigged bottom trawl net with a 25-m-wide mouth to the body of the trawl net which was made of 80 mm mesh, and with 42 mm mesh for the cod-end.Each trawl was towed at a speed of 5.6 km h −1 (3 knots) for a duration of 1 h (Table 1).
Seven of the 12 trawls were conducted using a commercial trawl vessel in two commercial scampi fishing grounds.Four of the seven trawls occurred off the coast of the Bay of Plenty (FMA SCI 1) in 2016, while the remaining three trawls occurred just north of the Chatham Islands (FMA SCI 4) in 2018 (Fig. 1).All trawl locations for this study were randomly selected within known scampi grounds and used a twin-rigged bottom trawl net with a 60-m-wide mouth with a net body with a mesh size of 120 mm and a cod-end with a mesh size of 55 mm.Trawls were towed at a speed of 4.8 km h −1 (2.6 knots) and ranged from 6 to 7.5 h in duration.
All 12 trawls from both vessels were conducted during daylight hours to coincide with the period of maximum trawl catchability of scampi (Tuck 2010).Sampling at a time when the greatest number of scampi are likely to be out of their burrows is an excellent aid to the identification of certain burrow types as belonging to scampi (Cryer et al. 2001).

Camera setup
Underwater images for this study were captured using three camera systems.A pilot study was conducted for the two scampi trawls in 2010 on the Mernoo Bank.This pilot study used a SeaCorder video recording system with built-in white LED lights (minimum output of 120 lm) and battery pack (8.2Ah and operational for 6 h) mounted to a small sledge that was towed immediately behind the footrope of the trawl with the camera field of view directed along the footrope.Recordings from these trawls were continuous (25 fps) with adequate resolution and field of view to observe scampi All subsequent underwater image capture was taken with two near identical camera setups.The camera system aboard the commercial vessel used a single Brinno TLC200 Pro video camera, while the system aboard the RV Kaharoa used a single DeepSea Power & Light Seacam500 camera.Each camera was encased within a custom built aluminium underwater housing.Both camera and a battery pack with a 4.2-Ah capacity were fixed to a custom-built camera stand made of steel and plastic framing (Fig. 2).The camera stand was secured using cable ties to the underside of the top panel of the scampi trawl net, positioned near the middle of the net and about 170 cm back from the headline (Fig. 3).The camera stand was orientated to face in the direction of the trawl so that the camera's field of view included a section of the footrope of the trawl net.Two white Cree XLamp LED lights with a calculated output of 2142 lm were connected to the same battery pack and positioned immediately below the camera and directed toward the camera's field of view (Fig. 2).The camera on RV Kaharoa trawls was set up to take continuous recordings (25 fps) for the duration of the 1-h trawls, whereas the camera on commercial trawls was configured to take an image at 1-s intervals (timelapse recordings) with a synchronised light flash so that battery and data storage would allow recording over the longer 6-to 7.5-h

Scampi-habitat associations
Recordings from nine trawls were used for habitat association analyses (seven from commercial trawls, two from RV Kaharoa trawls) (Table 1), whereby the camera setup, angle of view and spatial scale was consistent across all trawls.
The spatial scale of the image frames from continuous and timelapse recordings ranged from approximately 1.5 to 2 m 2 and were estimated by counting the number of mesh squares at the base of the footrope that were visible in the recordings.Three of the 12 trawls in this study were excluded from habitat-association analysis, including Trawl F (Mernoo Bank 2016), as the angle of view was too shallow relative to other trawls due to imprecise mounting of the camera on the trawl net, and Trawls K and L (Mernoo Bank 2010) which were taken with a different camera setup as a pilot study.

Scampi trawl efficiency
Recordings from all 12 trawls were used for scampi trawl efficiency analyses.These included all trawls derived from the two camera systems outlined above (including Trawl F), as well as two 1-h video trawls conducted on the Mernoo Bank during a scampi research voyage in 2010 (Table 1).

Analyses of scampi-habitat associations
Video recordings were analysed for scampi-habitat associations for nine of the 12 trawls where the field of view was consistent among tows and suitable for analysis.Seven were from commercial trawls (three from the Bay of Plenty, four from the Chatham Islands) and two were conducted on the RV Kaharoa (Mernoo Bank) (Table 1, Fig. 1).For each of the seven commercial trawls, 1 h of unbroken video recording was randomly selected using a range-defined random number generator in R Studio (version 3.6.3),as a means of standardizing for trawl length.Randomly selected frames from the videos and timelapse recordings of the seafloor (as described below), were then scrutinized for visible features of interest.The following biological and topographical features within each frame were recorded: Scampi: Visible scampi were partitioned into two size categories (small ~ < 10 cm overall body length, large ~ > 10 cm overall body length).Size estimates were made by comparing scampi with trawl net mesh squares within the frame.
Scampi burrows: Scampi burrow openings were identified in accordance with characteristics and descriptors developed  2003), such as surface tracks leading from openings, shallow descent angle, sediment fan and crescent shape (Fig. 4).Burrows were classified as either 'highly characteristic' or 'probable' based on the evaluation of these burrow characteristics.
Conical mounds: The number of conical mounds, likely formed by bioturbators, were counted per frame.These mounds ranged in size up to approximately 30 cm in diameter and were typically greater in height than the 60-mmdiameter of the footrope.
Surface roughness: Benthic topography for each frame was categorized as either predominantly rough or smooth.Rough topography was where the majority of seafloor within the frame undulated more than the height of the footrope, whereas vertical undulations in the seabed smaller in size were considered smooth.
Epifauna and pelagic organisms: Epifauna and pelagic organisms were counted and identified to their lowest taxonomic rank.Smaller macrofaunal organisms (approximately < 3 cm) were often too hard to identify and were left out of analysis.Likewise, occasionally fast-moving finfish species were unidentifiable and were instead categorized according to size (small ~ < 20 cm overall length, large ~ > 20 cm overall length).
A power analysis was run to determine the number of digital image frames that needed to be analysed to reliably detect a difference (at a significance level of 0.05) between the number of scampi burrows per frame as a reliable predictor of scampi abundance (Cryer et al. 2003) and other faunal and benthic features identified from the video imagery.The means and pooled variance per trawl sample of scampi burrows and conical mounds from a random sample of 840 analysed frames across two separate trawls within the Chatham Islands sampling site revealed a minimum sample size of 350 images per sampling location (FMA).Consequently, a sample size of 300 frames for each trawl was chosen, as it resulted in a sample size per sampling location that greatly exceeded the minimum requirement given there were multiple trawls sampled in each FMA.The 300 samples were randomly allocated for each 1 h of the video recordings randomly sampled from each trawl by assigning sequential numbers to each second of continuous recordings, or for each frame of timelapse recordings, and using a range-defined random number generator in R Studio (version 3.6.3)to select 300 images for analysis (i.e., 457 to 1284 frames per FMA) (Table 2).The additional sampling enabled images to be discarded where the image quality was too poor for analysis, due to debris or water clarity, leaving a final total of 2537 images that were analysed for all nine trawl samples (Table 2).In most instances poor image quality resulted from suspended sediment in the water column All image observations were carried out on clear images and by the same researcher to ensure consistency.Near the outset of the study a randomly selected subset of 75 of the initially analysed images was assessed by two benthic ecologists and one scampi expert with extensive experience of scampi image analyses to confirm the correct identification of organisms and scampi burrow characteristics.This quality control resulted in a minor adjustment to one classification which was then retrospectively applied to previously analysed images, and to all subsequent analysis.
Poisson regression analysis was used to investigate the relationships between scampi burrow numbers per frame (response variable) and the explanatory variables: total visible scampi, counts of epifauna taxa, counts of finfish taxa, topography categories and presence or absence of conical mounds.Estimates of the regression coefficients indicate the direction and magnitude of the relationship between burrow numbers and explanatory values, with positive values indicating a positive relationship.In all cases the residual deviance value was smaller than the degrees of freedom and thus the data were not over-dispersed.All statistical analyses were carried out using R statistical software (R Studio version 3.6.3).

Analyses of scampi trawl efficiency
Scampi trawl efficiency was assessed by observing the proportion of visible scampi that were observed to have passed over the footrope and were deemed to have been 'caught' in the following trawl net versus those that passed under the footrope that were 'not caught'.The analysis used 1 h of trawl recordings from each of the 12 trawls, including the nine recordings sampled for habitat-association analysis.
Rather than subsampling, each 1 h of video recording was analysed in its entirety in order to observe all visible scampi encountering the viewed portion of the footrope.The pre-trawl arrival positions of scampi were categorized as follows; surface (visible scampi fully emerged from burrow) (Fig. 4a and b), partially in burrow, fully in burrow.Scampi size was also recorded in accordance with size categories used for habitat association analysis (small vs large).Topography (smooth or rough) and presence/absence of conical mounds for each scampi sighting was also recorded.
A catch assessment could be made for all visible scampi encountering the footrope during RV Kaharoa trawls, due to the use of continuous recordings.In contrast, only a small proportion (9.5%) of visible scampi from timelapse recordings used during commercial trawls were observed encountering the footrope.In those cases, the 1-s interval between frames allowed for the same scampi to be observed in back-to-back frames, approaching and then encountering or passing by the footrope.
Multiple logistic regression analysis was used to test for the effect of variables (scampi position, scampi size, topography type, presence/absence of conical mounds) on the probability of scampi being caught.Chi-square tests were also used to investigate whether the proportions of caught and uncaught scampi were different from expected proportions with respect to scampi position, scampi size, topography type and presence/absence of conical mounds.All statistical analyses were carried out using R statistical software (RStudio version 3.6.3).

Scampi-habitat associations
A total of 2537 images were analysed for the investigation of scampi-habitat associations (Table 2).Trawls with lower sample numbers were a result of the exclusion of frames with poor image quality due to debris and poor water clarity.The total number of visible scampi varied across trawls, ranging from 5 to 55 per trawl.Of these, large scampi (> 10 cm overall body length) accounted for 58% of all observed scampi.Visible scampi as a proportion of scampi burrow openings ranged from 0.05 to 0.41 among trawls.These numbers are comparable to findings from photographic surveys conducted as a component of scampi stock assessment (Tuck et al. 2015a(Tuck et al. , 2016)).Conical mounds were present across all trawls, with the highest rate of occurrences on the Mernoo Bank (54% of all samples) and lowest in the Bay of Plenty (26%).Rough topography was most prevalent for trawls conducted on the Mernoo Bank (74% of all Mernoo Bank samples) and Chatham Islands (80% of all samples), while two of the three Bay of Plenty trawls recorded predominantly smooth topography (Table 2).

Epifauna and pelagic organisms
A total of 33 taxa were identified from the images, comprising 11 epifauna and 22 fish taxa (Table 3).Total visible epifauna counts were relatively low per trawl with scampi comprising a sizeable proportion of all epifauna numbers, accounting for between 13 and 76% of the epifauna per trawl (Table 3).For seven of the nine trawls, scampi comprised greater than 56% of total visible epifauna.A variety of other crustaceans including antlered crabs Dagnaudus petterdi, squat lobsters Munida spp., krill, decapod shrimp and other (unidentifiable) decapod crabs were identified as minor proportions of total epifauna in other trawls.Sea pens were prevalent during Mernoo Bank (SCI 3) trawls and a large grouping of sea urchins comprised 77% of total epifauna numbers during trawl I (Chatham Islands).
The five most commonly identified fish taxa included rattails from the family Bathygadidae, hoki Macruronus novaezelandiae, hagfish Eptatretus cirrhatus, sea perch Helicolenus spp., and anguilliform eels: conger eels.Rattails were prevalent in Mernoo Bank and Chatham Island trawls but were only observed 3 times in the Bay of Plenty.Twenty of the 24 observed ling Genypterus blacodes, an active predator of scampi, were observed during Chatham Island trawls.

Associations per individual trawl
A Poisson regression analysis identified a statistically significant positive relationship between the number of burrows (response variable) and total scampi numbers in four of nine trawls.Total scampi numbers rather than scampi size classes were analysed due to low scampi counts in three trawls.A positive relationship was also observed between burrow numbers and number of conical mounds in eight of the nine trawls (Table 4).Consequently, smooth benthic topography was also found to be less likely to have scampi burrows present for six of the nine trawls (Table 4).As individual taxa counts were low, taxa were pooled by total epifauna and total fish per trawl to investigate broad associations between burrows and benthic or pelagic species.A negative relationship was found for burrow numbers and total epifauna counts for a single Chatham Island trawl, whereas no relationships were observed for burrow numbers and total fish counts per trawl (Table 4).

Associations by fisheries area
When grouping data by location, a positive relationship was observed between burrows and total visible scampi for Bay of Plenty and Chatham Island trawls.This relationship was also observed for small scampi in the Bay of Plenty and for small and large scampi in the Chatham Islands.A positive relationship between numbers of burrows and conical mounds was observed for all three locations, while there was a corresponding negative relationship between burrows and smooth topography at all locations.The relatively high standard errors of total scampi relative to the coefficients suggest high variability and patchiness in the distribution of visible scampi relative to burrows at all three sampling locations (Table 5).
For regression analyses of individual taxa, it was necessary to pool data for trawls within sampling location due to low counts per trawl.No significant relationship between numbers of scampi burrows and the numbers of individual taxa was observed for trawls occurring in the Bay of Plenty; however, for Mernoo Bank a positive relationship was found with ghost-sharks (estimate = 0.93, P = 0.025), and for the Chatham Island location a positive relationship was found with antlered crabs (estimate = 0.65, P = 0.016), other crabs (estimate = 1.01,P = 0.042), eels (estimate = 0.44, P = 0.010), small finfish (estimate = 0.27, P = 0.014) and lookdown dories (estimate = 0.76, P = 0.002).

Scampi trawl efficiency
Over all of the sampling, regardless of location, only 30% of observed scampi encountering the footrope were caught, with the majority passing under the footrope of the trawl (Table 6).
Scampi size had little effect on their chance of capture with 29% of small and 31% of large scampi observed entering the net.Scampi trawl efficiency was lower on rough topography (27%) versus smooth topography (38%) and with the presence versus the absence of conical mounds, (23% vs 35% respectively).Scampi position greatly affected their chance of capture with scampi observed partially in their burrows escaping capture 90% of the time compared with 62% of scampi observed on the surface (Table 7).
Multiple logistic regression analysis found scampi position to be the only variable with a statistically significant effect on scampi catchability (estimate = − 1.38, P < 0.001).Likewise, chi-square tests revealed that the proportion of caught and uncaught scampi were significantly different from expected proportions given scampi positions (χ 2 = 10.63,P = 0.001), but not for scampi size (χ 2 = 0.06, P = 0.81), topography (χ 2 = 1.73,P = 0.19) or presence of conical mounds (χ 2 = 2.08, P = 0.15).Scampi positioned fully and partially in their burrows were combined for chi-square tests to improve approximations, as only two of the observed 186 scampi were fully in burrows.

Habitat associations
The present study offers a first account of the patterns of habitat association in New Zealand scampi at the trawl scale.Scampi numbers were negatively associated with benthic smoothness and positively associated with conical mounds that were prevalent across sampling locations within the three sampled FMAs.These results augment our current understanding of scampi-habitat preferences and the factors influencing scampi distribution patterns.Few acoustic surveys, allowing for the classification of seabed type of New Zealand scampi habitats, have been conducted, with only a small proportion of those being ground-truthed by photographic surveys of scampi occupying the seafloor (Cryer et al. 2001).Nonetheless, groundtruthed acoustic surveys indicate that scampi distribution is likely tied to the availability of soft sediments suitable for burrow formation by scampi (Cryer et al. 2001).Furthermore, broad sediment preferences of scampi may be inferred by overlaying data collected during acoustic seabed classification surveys conducted along New Zealand plateaus (Nodder et al. 2011;Harris and Baker 2012) with the distribution of scampi catches from trawl-catch records (Tuck 2009).It is likely that conical mound-forming bioturbators and scampi have similar sediment requirements and their distributions overlap due to this shared habitat preference.There is currently no research directly relating the presence of New Zealand scampi burrows to sediment grain size composition.Investigations of habitat associations and sediment requirements for the related Norway lobster Nephrops norvegicus (henceforth referred to as Nephrops) suggests that the availability of suitable sediment for burrow building is a pre-requisite for Nephrops distribution (Johnson et al. 2013).Examinations of burrow density with sediment composition has found 'mud' substrates (i.e., substrates with higher silt-clay ratios and lower coarse sand ratios) correspond with higher burrow densities (Afonso-Dias 1998; Campbell et al. 2009;Power et al. 2019).An understanding of sediment preferences and associations with other environmental factors including benthic topography has allowed for habitat suitability models to predict Nephrops distributions in existing and unexplored areas (Lauria et al. 2015).A similar understanding of scampi-habitat associations in New Zealand waters could potentially allow for more discriminating and targeted fishing efforts.
In the present study it is unclear what organism is responsible for conical mound formation, due to a lack of infaunal analysis of scampi habitats.Burrowing polychaete worms are a likely candidate, having produced similar looking mounds and of a similar size (up to approximately 30 cm in diameter) in other soft sediment habitats (Ray and Ciampi 1956).Also, the presence of scampi burrows in areas without mounds suggest that scampi are unlikely to be responsible and observations of burrowing behaviour of scampi in captivity indicate that they are unable to construct mounds (Heasman and Jeffs 2019).As rough seabed topography was typically accentuated by the presence of conical mounds, the negative association between burrows and smooth topography may also be partially explained by shared sediment preferences between scampi and conical mound-forming organisms.It is possible that rough topography and conical mounds may aid in scampi predator avoidance by offering a more three-dimensional habitat in which to build their burrows and provide some refuge from predators.While increased habitat structure has been shown to reduce detection from predators in other soft sediment crustacea (Main 1987), these are typically associated with increased vegetation coverage.Investigations of predator avoidance strategies in Nephrops have revealed the use of olfaction to sense chemicals released by predators, swimming by tail flipping to rapidly avoid approaching predatory fish, and burrow submergence (Katoh et al. 2013).
A positive relationship was observed between the numbers of burrows and the numbers of visible scampi for two of the three FMAs that were sampled.As scampi are thought to spend a considerable amount of their time in, or near, burrows (Tuck 2010), it is unsurprising that a relationship was detected.However, the lack of a statistically significant relationship between numbers of burrows and visible scampi for Mernoo Bank trawls may be related to seasonal patterns of burrow emergence associated with scampi moulting cycles (MPI 2021).These sampling dates overlapped with predicted female scampi moulting periods, which peak in October in SCI 3, and are associated with substantially reduced burrow emergence (Tuck 2013).This would potentially reduce the proportion of visible scampi for Mernoo Bank samples in this study.Bay of Plenty sampling dates (February/March) did not coincide with predicted moulting periods (October/November) (Tuck 2010).While an analysis of scampi emergence patterns has not been undertaken for SCI 4, Chatham Island trawls were conducted during months that aligned with historically higher catch rates for the fishery and are therefore less likely to be a moulting season.
For the binomial regression analyses of individual taxa, it was necessary to pool data for trawls within sampling location due to low counts per trawl.However, pooling data by location assumes relative uniformity within each scampi FMA despite the vast size of FMAs in comparison to individual trawls.Further video trawls are required to determine the level of variation in the distribution of taxa within sampling locations.
Scampi burrow numbers were only positively associated with four of the 33 identified taxa, and none consistently across all three FMAs.Trawls were characterised by low total epifauna numbers but with a relatively high proportion of those observed being scampi (13-76% within trawls), suggesting that scampi are the dominant benthic epifaunal species.Detecting associations between sedentary burrowdwelling species, such as scampi, and highly mobile pelagic species can be difficult when assessing at spatial scales of only a few metres (i.e., scale of video imagery).Pelagic and demersal fish species can move and migrate over large areas and their distribution and abundance are typically assessed via acoustic surveys (Simmonds and MacLennan 2008), catch records and mark-recapture methods over much larger spatial scales.Scampi trawl catch records can offer insights into species associations over much broader scales, although they do not account for uncaught species due to gear selectivity parameters.An analysis of annual bycatch estimates from 1990 to 2010 across the entire scampi fishery revealed that more than half of the total bycatch biomass was comprised of five fish taxa: javelin fish Lepidorhynchus denticulatus, unspecified rattail species of the family Macrouridae, sea perch Helicolenus spp., ling and hoki (Anderson 2013).All five taxa were observed in the present study with 'rattails' (grouping of javelin fish and unspecified rattails) making up the largest number of identifiable fish taxa.
While comparatively time consuming, visual assessments of the seafloor potentially offer more accurate assessments of scampi-species associations than catch records as it includes scampi burrows, a more reliable indicator of scampi presence than trawl catches (Cryer et al. 2001(Cryer et al. , 2003)), and may account for uncaught species.In the present study a number of epifaunal species, including crustaceans, sea urchins and sea-pens were observed passing under the net and evading capture.Likewise, strongly swimming fish were occasionally viewed actively avoiding capture using bursts of speed to swim upward and out of view of the video.The criteria used to assess the presence of burrows as belonging to scampi were, of necessity, partially subjective, but were based on a set of descriptors developed for ongoing photographic stock assessments (Cryer et al. 2003).Further replication may enable the use of scampi data to provide further information on the distribution of other species underrepresented in scampi trawls.

Scampi trawl efficiency
The detrimental effects of benthic trawling on the marine benthic environment are widely acknowledged (Kaiser et al. 2002;Thrush and Dayton 2002).In New Zealand, the degree of benthic modification from scampi trawls is poorly understood due the relatively great depths at which they operate that hinders scientific research (200-600 m).Adequate and direct observations of the seabed following scampi trawls is both difficult and costly at such depths.However, analysis of catch composition from overlapping demersal trawls over selected sites in north-eastern New Zealand suggests that trawling alters benthic community structure and reduces biodiversity over broad spatial scales (Cryer et al. 2002).
Benthic trawling for several crustacean fisheries, including shrimp and scampi, are ranked as some of the most energy intensive fisheries globally (Tyedmers 2004).Improving scampi trawl efficiency could reduce fishing times, while also potentially reducing energy consumption and seabed damage.The results of the current study highlight the apparent relative inefficiency of capture of the available scampi encountered by trawl nets, which has implications both for commercial fishing and the use of trawl catch data for fisheries stock assessment.
Commercial trawl catch records are already known to severely underestimate scampi density, as a large proportion of scampi remain uncaught whilst in their burrows.For example, visible scampi counts from stock assessment photos have typically been 10-20% of burrow counts (Tuck 2010), with estimates from the present study falling within the margins of error (5-40%).However, the proportion of emergent scampi that are captured and the factors influencing scampi trawl efficiency have not been investigated until now.These preliminary findings suggest that less than a third of visible scampi are captured during trawls.While scampi size had seemingly no effect on scampi trawl efficiency, burrow submergence had a significant effect on catch success.The presence of conical mounds tended to further reduce catch success.Rough topography and conical mounds reduced footrope contact with the seabed, allowing scampi to pass under the net.Likewise, partially emergent scampi were only observed to be caught when reacting to approaching trawl gear by flicking their tails and swimming upwards.Swimming, as a predator avoidance response, has been found to play a role in the capture of Nephrops in trawl nets with a proportion of those that are stimulated to swim being captured in the net (Newland et al. 1992).
Other factors influencing Nephrops catch success have included swimming orientation (Newland and Chapman 1989) and the number of trawl rigs used, with one study finding Nephrops catch rates improved 340% for twin rigged trawls over single rigged trawls (Sangster and Breen 1998).
The study suggested that differences in ground gear shape could imply different ground gear contact affecting fishing efficiencies.In the present study, a higher percentage of scampi were caught in twin-rigged commercial trawls (47%) than in single-rigged trawls (24%).However, these results may be confounded by other variables, and by low numbers of observed scampi encountering the net for commercial trawls (17) relative to RV Kaharoa trawls (169).Of the 17 scampi assessed from commercial trawls, eight that were caught were all observed on the surface, while the nine uncaught were all partially observed in burrows, suggesting that catch success was predominantly influenced by scampi positioning.
It is possible that the presence of the artificial lighting used to facilitate the video image capture in this current study could have affected the behaviour of scampi and some bycatch species, which could have affected the resulting observations and catches.Artificial lighting can both attract and deter the movement of marine species, including in relation to a moving trawl net (Nguyen and Winger 2019).For example, some deepwater crustacean trawl fisheries use artificial lighting on trawl gear to enhance catches (Geraci et al. 2021).The possible influence of camera light on catches in scampi trawls could be controlled in future research by using only strobe lighting to illuminate timelapse image capture, given that the current study has demonstrated from one of the camera systems that was used in the pilot study, that timelapse image capture on a scampi trawl is effective in producing useful observations for analyses.
Estimates of scampi abundance and biomass are vital components of scampi stock assessments and help determine the total allowable commercial catch for each FMA.Uncertainty over trawl catchability associated with scampi emergence patterns has resulted in the development of photographic survey counts of scampi burrows to help provide an independent method for estimating scampi abundance and density (Froglia et al. 1997;Tuck et al. 1997;Cryer et al. 2003;Smith et al. 2003).Scampi biomass is typically estimated based on catches from trawl surveys which are then scaled to survey strata (e.g., Tuck et al. 2019).It is acknowledged that trawl survey estimates of biomass are likely to considerably underestimate absolute stock biomass as only scampi that have emerged from their burrows are available for capture (e.g., Tuck et al. 2019).The discrepancy between photographic burrow counts and trawl catches informs an estimate of scampi catchability which is currently built into the Bayesian model used for scampi stock assessments (e.g., Tuck 2009).These loosely defined 'priors' (estimated probability distributions) for the parameters of scampi catchability determine how best to scale-out burrow counts and trawl catches to stock population abundance and biomass estimates.However, the 'prior' distribution for catchability initially assumed that all emerged scampi were caught and did not account for the fact that a significant proportion of emerged scampi may avoid capture, even over smooth terrain, as indicated by this study.More recent priors have been informed by this research (e.g., Tuck 2019).Failure to incorporate low scampi trawl efficiency rates may underestimate stock biomass.Furthermore, a better understanding of how scampi trawl efficiency changes with different terrain and habitat types could help provide more spatially refined estimates of stock biomass and help assess the relative vulnerability of stocks.The use of video technology on trawl nets in this study proved to be an effective tool for assessing the performance of trawls for catching emergent scampi.While analysis of the recorded video images is timeconsuming, there is opportunity to include trawl-video sampling on a random subset of future research trawls to provide more robust estimates of scampi trawl efficiency over a range of locations and habitat types.The emergence of artificial intelligence for analysing recorded underwater observations has the potential to greatly reduce the time-consuming analyses of recordings, and also improve sampling resolution by being able to include all recordings (Allken et al. 2021), rather than the significant subsampling as required in this current study to reduce researcher time for image analyses.

Further application of study findings
While the main scampi fishing grounds of New Zealand are well established, exploratory trawls still occur on the outskirts of plateaus and along the west coast of New Zealand (Tuck 2009).Locating areas of localized higher scampi abundance outside of established grounds is an energy intensive exercise given current methods.Exploratory trawls are typically based on water depth and proximity to established grounds and may yield little to no scampi.The present study suggests that the detection of conical mounds may be used as a way of locating scampi patches for more targeted fishing.A better understanding of scampi-habitat associations with respect to sediment preferences, species interactions and other environmental factors could allow for more accurate predictions of scampi distribution and further reduce the need for 'blind' trawling.Given that the catch of available scampi in the present study was so low, further video assessments of the factors affecting scampi trawl efficiency are needed to potentially improve our understanding of fishing efficiency and inform more robust scampi stock assessments.Overall, these results confirm the effectiveness of video sampling from commercial trawls to improve our understanding of deep-sea habitats and fishing efficiency, whilst having the potential to be applied to improve environmental and economic outcomes for this fishery.Rangatahi: Revolutionary Potting Technologies and Aquaculture for Scampi' (CAWX1316) funded by the New Zealand Ministry of Business, Innovation and Employment.The authors extend thanks to the Waikawa Fishing Company Ltd skippers and crew for carrying out video deployment during commercial trawls.We are grateful to the National Institute of Water and Atmospheric Research Ltd, and the crew of the RV Kaharoa for help with deployment of video during the 2016 voyage and for allowing the use of video recordings captured during their 2010 research voyage.Our thanks go to John Radford of Zebra-Tech Ltd for camera frame construction.The authors also kindly acknowledge the support of the Scampi Team at the Cawthron Institute (Rob Major, Chris Batstone, Kevin Heasman and Sara Jamieson) and Waikawa Fishing Company Ltd (Steve Connor, Geoff Connor, Lance Connor, Glenice Paine and Richard Paine).
Funding Open Access funding enabled and organized by CAUL and its Member Institutions.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material.If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Fig. 3
Fig. 3 Scampi trawl net components with camera positioning and field of view (mockup).The RV Kaharoa vessel used a single rigged net, while the commercial vessel used a twin-rigged net

Fig. 4
Fig. 4 Four examples of sampled images from video trawls; a 1 large scampi away from burrow; b 2 large scampi partially in burrow; 3 conical mound; c 4 scampi burrow, 5-7 conical mounds; d 8 lookdown dory; 9 small scampi on the surface.The bottom two images

Table 1
Video trawl locations [Bay of Plenty (BoP), Mernoo Bank, Chatham Islands], dates, depth, speed, video type (timelapse or continuous) and analyses conducted [scampi association analysis and/or scampi trawl efficiency analysis (STE)]

Table 2
Summary data from trawl video images (samples) examined for scampi-habitat associations

Table 3
Counts of visible epifauna and fish species classified to their lowest possible taxonomic level *Unidentifiable fish smaller than approx.200 mm.**Unidentifiable fish larger than approx.200 mm.

Table 4
Relationships between scampi burrows and other variablesP values and regression estimates (indicating direction and magnitude of the relationship) from Poisson regression analyses are provided.

Table 5
Relationship between number of scampi burrows and other habitat variables for results of analyses of video recordings pooled by location P-values, regression estimates (indicating direction and magnitude of the relationship) and standard errors from Poisson regression analyses are provided.