Digital photography as a stock assessment tool for Metanephrops challengeri on New Zealand’s continental slope

M. Cryer, K. Downing, B. Hartill, J. Drury, H.J. Armiger, C. Middleton and M.D. Smith
National Institute of Water and Atmospheric Research Ltd.
PO Box 109695, Newmarket, Auckland, New Zealand
<[email protected]>

1. INTRODUCTION

Scampi (Metanephrops challengeri), a burrowing nephropid crustacean, is found on suitably cohesive sediments on New Zealand’s continental slope, predominantly in depths of 200-600m. A trawl fishery, mostly using vessels 20-40m in length that use double or triple trawl rigs of low headline height, it lands about 1000t annually. Catch rates vary by depth, among areas and among years. Stock assessments for this species have been hampered in the past by a lack of a reliable index of stock abundance. Both research trawl and commercial CPUE indices appear compromised by changes in vulnerability to capture, probably a result of changes in emergence behaviour. Following the apparent success of indices of abundance based on underwater video observations for several stocks of European scampi, (Nephrops norvegicus) (e.g. Tuck, Atkinson and Chapman 1994, Tuck et al. 1997, Marrs, Atkinson and Smith 1998), we developed a sturdy, reliable, self-contained digital camera system capable of operating to 1000m depth and deployable from relatively small vessels.

2. NIWA’S DEEPWATER DIGITAL CAMERA SYSTEM

2.1 Development and design rationale

Following successful trials observing scampi burrows with an emulsion-based underwater camera system, we developed a digital system with an emphasis on durability and the ability to be deployed in poor weather and over foul (rocky) ground. We designed our system to be essentially self-contained underwater and to be deployed using a trawl warp rather than a conducting cable or hydrographic wire. All equipment and pressure housings are contained within a protective stainless steel cage surrounded by sprung stainless steel buffers designed to cushion impacts against the ship or seabed and to minimize the chance of snagging rocky seabeds. We adapted off-the-shelf digital cameras (Minolta D’Image EX1500, 1344*1008 pixels, recently upgraded to Nikon Coolpix 5000, 2560*1920 pixels) to operate with a separate deepwater strobe and high capacity NiCad battery storage in purpose-built pressure housings. The camera is triggered using either a bottom contact switch (with a weight on a line of appropriate length attached) or using an interval timer. We estimate and maintain distance off-bottom using acoustic links based on the rugged and dependable Furuno CN22 trawl monitor system widely used in research and commercial fishing. Image files are recorded in 24-bit colour using light JPEG compression, results in files of about 1.3Mb each (although the recent upgrade to 5 megapixels increases this to about 4Mb per image file). Image files are stored on compact flashcards (currently 64 or 128 Mb, to be upgraded soon to 512Mb or more) within the camera. Other equipment carried on the camera cage includes parallel lasers (200mm apart) to allow for image scaling and an acoustic "pinger" to allow for location and retrieval should the camera be lost at sea.

2.2 Deployment and data acquisition

We take still photographs from 3-5m off-bottom, depending on water clarity and sea state (especially swell), resulting in images of mostly 4-12 m2. Typically, surveys consist of 600-1000 images, spread among 20 or more stations (or transects) in 4-6 strata. Strata are defined on the basis of depth and geography (Figure 1). Photographs are taken as the ship drifts with the wind and tide and we try to separate photographs in time such that they are unlikely to overlap spatially. Transects are terminated if they drift outside the boundaries of the stratum. At the end of each transect the flashcard is removed form the camera and all images are downloaded to the hard drive of a dedicated on-board computer and backed up on CD-ROMs.

3. SCREENING AND COUNTING SCAMPI AND THEIR BURROWS

3.1 Image screening

We have developed a rigorous, standardized protocol for screening these images. An image is accepted for analysis if fine seabed detail is discernable and more than 50 percent of the image is visible (i.e. free from disturbed sediment, poor flash coverage, or other features - a good example is shown in Figure 2). The percentage of the frame within which the seabed is clearly and sharply visible is estimated and marked using polygons in "Didger" image analysis software. All emergent scampi and all burrow openings characteristic of M. challengeri are counted by each of three readers (selected at random from a team of six) working blind from one another. Each reader assesses the number of burrow openings using a standardized protocol (Cryer et al. 2003) which defines "major" and "minor" burrow openings separately (respectively, the type of opening at which scampi are usually observed, and the "rear" openings associated with most burrows). We also classify each opening (whether major or minor) as "highly characteristic" or "probable", based on the extent to which each is characteristic of burrows observed to be used by New Zealand scampi. Burrows and holes which could conceivably be used by scampi, but which are not "characteristic", are not counted. Our counts of burrow openings may therefore be conservative (assuming that burrow occupancy is high). All burrow openings and visible scampi scored by each reader are annotated on low quality (highly compressed) digital copies of the original image file to provide an "audit trail" and facilitate comparisons among readers.

The criteria used by readers to judge whether or not a burrow should be scored are, of necessity, partially subjective. We cannot be certain that any particular burrow belongs to M. challengeri and is currently inhabited unless the individual is photographed in the burrow. However, after viewing large numbers of scampi associated with burrows, we have developed a set of descriptors that guide our decisions (Appendix I). Using these descriptors as a guide, each reader assesses each potential burrow opening (paying more attention to attributes with a high ranking such as tracks on the surface, a shallow descent angle, and sediment fans for major openings) and scores it only if it is "probably" (not "maybe") a scampi burrow.

Many assessments of the similar Nephrops norvegicus in ICES areas are conducted using relative abundance indices based on counts of "burrows" (rather than burrow openings) (Tuck, Atkinson and Chapman 1994, Tuck et al. 1997). We count burrow openings rather than assumed burrows because burrows are relatively large compared with the quadrat (photograph) size and accepting all burrows totally or partly within each photograph will result in counts that are positively biased by edge effects (e.g. Marrs, Atkinson and Smith 1998).

Once the images from any particular stratum or survey have been scored by three readers, any image for which the greatest difference between readers in the counts of major openings is more than one is re-examined by all readers (who may or may not change their score). During this process, each reader has access to the score and annotated image files of all other readers. After re-assessing their own interpretation against the original image, all are encouraged to compare their readings with the interpretations of other readers. Thus, the re-reading process is a means of maintaining consistency among readers as well as refining the counts for a given image. The correlation of counts among readers is typically > 0.95 (Figure 3), and "reader bias" (relative to the average of the six readers) is < 10 percent.

3.2 Data analysis

Counts from photographs are analysed using methods analogous to those employed for trawl and other "swept area" surveys. To exclude a possible image size effect (burrows perhaps being more or less likely to be accepted as the number of pixels making up their image decreases) those few (< 5 percent) images with a very small (< 2 m2) or very large (> 16 m2) readable area are excluded. The mean density (of major or minor openings or scampi) at a given station is estimated as the sum of all counts divided by the sum of all readable areas. For a given stratum, the mean density of openings and its associated variance are estimated using standard parametric methods, giving each station an equal weighting. The total number of openings in the stratum is estimated by multiplying the mean density by the estimated area of the stratum. The overall mean density of openings in the survey area is estimated as the weighted average mean density, and the variance for this overall mean was derived using the formula for strata of unequal sizes given by Snedecor and Cochran (1989):

where is the variance of the overall mean density,, of burrow openings in the surveyed area, W_i is the relative size of stratum i, and and n_i are the sample variance and the number of samples respectively from that stratum. The finite correction term , is set to unity because all sampling fractions are less than 0.01. Estimates of abundance for the sampled strata are scaled by the overall mean density by the combined area of all the strata, assuming these to be without error.

The approach we have taken seems capable of generating at least two promising indices of abundance: a minimum estimate of absolute abundance based on the density of visible scampi; a minimum estimate of absolute biomass based on the foregoing and a photographic estimate of length frequency distribution; and an index of relative abundance based on the density of characteristic burrow openings.

Comparable estimates of relative abundance with estimated coefficients of variation (CVs) have been generated for surveys conducted in the Bay of Plenty, New Zealand, in 1998, 2000, 2001, 2002 and 2003. Separate indices have been calculated for major and minor openings, for all visible scampi and for scampi "out" of their burrows (i.e. walking free on the sediment surface). However, only indices for major burrow openings and for visible scampi are presented here because these are currently thought to be the most reliable indices.

4. RESULTS TO DATE

4.1 Estimates of burrow density

The estimated mean density of scampi burrows (as indexed by their major openings) throughout the Bay of Plenty, in the 300-500 m depth range, varied from 0.08 m-2 in 2000 to 0.13 m-2 in 1998 (with CVs of 8-15 percent of the mean). Scaling to the sampled area (1196km²) results in abundance estimates of 94-154 million burrows or, assuming 100 percent occupancy, an identical number of animals (Figure 4).

4.2 Estimates of scampi density and minimum absolute abundance

The estimated mean density of all visible scampi (i.e. including those in burrows and those walking free on the sediment surface) varied from 0.010 m-2 in 2001 to 0.025 m-2 in 1998 (with CVs of 18-26 percent of the mean). Scaling these counts to the sampled area leads to abundance estimates of 12-28 million animals. Counting only the animals walking free on the sediment surface (i.e. those most susceptible to capture by trawl) greatly reduces the estimates (to 2-11 million animals, Figure 5) and greatly increases their CVs (to 25-62 percent).

4.3 Estimates of scampi biomass

Deriving estimates of relative or absolute biomass from estimates of abundance requires an estimate of the mean weight of individuals. Cryer and Hartill (1998) and Cryer, Hartill and Drury. (2001) estimated the length frequency distribution of visible scampi in 1998 and 2000, respectively and applied length-weight regressions to estimate aver-age weight. They used the average predicted weight from male and female length weight regressions for animals up to 48 mm and the predicted weight from a male length weight regression for all larger animals. Their estimates of average weight for measurable scampi were 35.4 g in 1998 and 38.3 g in 2000. Scaling the abundance estimates for visible scampi in each year by the smaller of these two estimates leads to estimates of absolute biomasses (Table 1). These estimates are probably close to minimum estimates of biomass, although smaller estimates are conceivable if, for instance, the average size were to be considerably smaller in some years for which the mean weight has not yet been estimated.

Making further assumptions (e.g. that each burrow identified as a scampi burrow is occupied by a single scampi of similar average size to those visible), the estimates of major burrow openings can be used to estimate current biomass (Table 2). These estimates may be conservative because we score only those bur-rows that are characteristic of scampi and we know that scampi are sometimes seen in other types of burrows. Or these estimates may be optimistic because not all burrows may be currently occupied or because hidden scampi are, on average, smaller than visible scampi. It is not currently possible to assess whether estimates of biomass made using our estimates of the density of major burrow openings are positively or negatively biased estimates of actual abundance.

TABLE 1
Estimates of the biomass of visible scampi within the Bay of Plenty between 1998 and 2003 made using a mean average weight of 35.4 g

TABLE 2
Estimates of biomass (t) of scampi within the Bay of Plenty between 1998 and 2003

5. COMPARISONS WITH OTHER DATA

Our "minimum" biomass estimates suggest that catch limits and current landings of scampi from the Bay of Plenty (120 t, Annala et al. 2003) could represent a substantial fraction of the biomass, 12.1-27.6 percent, depending on the year it was 23.6 percent for 2003. Conversely, biomass estimates made from burrow counts suggest that fishing takes a relatively small fraction of total biomass, 2.2-3.6 percent with the 2003 estimate suggesting a 3.3 percent removal.

At this stage it is not possible to be certain which of these indices of abundance is the best indicator for scampi. An index based on the density of characteristic burrows should not be affected by changes in emergence behaviour in scampi and can be estimated using photographs taken at any time of day (although it would be badly affected by changes in occupancy rate). Results from photographic surveys before and after fishing on the Mernoo Bank, south-eastern New Zealand (Cryer et al. 2003), however, suggest that there may be seasonal changes in the density or characteristics of burrows as there is for Nephrops norvegicus (e.g. Tuck, Atkinson and Chapman 1994). This would militate against indices based on burrow densities estimated at different times of year. Indices of absolute abundance based on visible scampi are almost certainly conservative and will be affected by the seasonal and diel timing of photography because emergence behaviour is likely to vary daily and seasonally, e.g. Cryer and Oliver 2001.

The decline in our indices of visible scampi, especially between 1998 and 2001, in the Bay of Plenty is consistent with the decline in commercial CPUE observed since about 1995 (see Hartill and Cryer 2003 for unstandardized indices to 2002, and Cryer and Coburn 2000 for fully standardized indices to 1998, although the two are highly correlated). Conversely, our indices of probable scampi burrows have remained relatively steady, a trend that is not consistent with commercial trawl catch rates (Figure 6). This divergence might be expected because the light, "skimming" trawl gear used to catch scampi is most unlikely to be able to catch scampi that are hidden from view in burrows. Critical in this interpretation is the implicit assumption that the proportion of burrows occupied by scampi is constant among years. If burrows last a long time after they are vacated by a scampi, then this assumption may not hold; the density of burrows could remain constant even while the population was declining rapidly. We have no information on burrow longevity and this could be a fruitful area for future research.

6. FUTURE WORK

Since the inception of this fishery in the late 1980s, we have assembled indices of abundance based on photography, research trawl surveys, and commercial CPUE, multiple length frequency distributions based on photography and on measurements by observers on commercial vessels, and estimates of growth increments based on a tagging experiment and aquarium trials. The next step planned is to integrate these data in a stock assessment model, which will probably be length - based.

7. ACKNOWLEDGMENTS

Thanks are due to the New Zealand Ministry of Fisheries for financial support under projects SCI9801, SCI1999/01, SCI2000/02, SCI2001/01, and SCI2002/02, and to Ian Tuck of FRS, Aberdeen for assistance in developing our protocols.

8. LITERATURE CITED

Annala, J.H., K.J. Sullivan, C.J. O’Brien, N.W. 2003. Report from the Fishery Assessment Plenary, May 2003: stock assessments and yield estimates. 616 pp. (Unpublished report held in NIWA library, Wellington.)

Cryer, M. & R. Coburn 2000. Scampi stock assessment for 1999. Fisheries Assessment Report 2000/07. 60 pp.

Cryer, M. & B. Hartill 1998. An experimental comparison of trawl and photographic methods of estimating the biomass of scampi. Final Research Report for Project SCI9701. 26 pp. (Unpublished report held in NIWA library, Wellington).

Cryer, M., B. Hartill & J. Drury 2001. Photographic estimation of the abundance and biomass of scampi, Metanephrops challengeri. Final Research Report for Project SCI1999/02. 49 pp. (Unpublished report held in NIWA library, Wellington).

Cryer, M., B. Hartill, J. Drury, I. Tuck, H.J. Armiger, M.D. Smith & C. Middleton 2003. Indices of relative abundance for scampi, Metanephrops challengeri, based on photographic surveys before and after fishing in QMA 3, 2001. Final Research Report for Project SCI2000/02 (Objective 3). 33 pp. (Unpublished report held in NIWA library, Wellington).

Cryer, M. & M. Oliver 2001. Estimating age and growth in New Zealand scampi, Metanephrops challengeri. Final Research Report for Project SCI9802, Objective 2. 56 pp. (Unpublished report held in NIWA library, Wellington).

Hartill, B. & M. Cryer 2003. Unstandardised CPUE indices for scampi, 1988-2003. Final Research Report for Project SCI2001/02, Objective 2. 35 pp. (Unpublished report held in NIWA library, Wellington).

Marrs, S.J., R.J.A. Atkinson & C.J. Smith 1998. The towed underwater TV technique for use in stock assessment of Nephrops norvegicus. p 88-98 In: I.C.E.S. Report of the study group on the life histories of Nephrops norvegicus, La Coruña, pp.

Snedecor G.W. & W.C. Cochran 1989. Statistical Methods. 8th ed. Iowa State University Press, Ames, Iowa, USA.

Tuck, I.D., R.J.A Atkinson. & C.J. Chapman 1994. The structure and seasonal variability in the spatial distribution of Nephrops norvegicus burrows. Ophelia 40: 13-25.

Tuck, I.D., C.J. Chapman, R.J.A. Atkinson, N. Bailey & R.S.M. Smith 1997. A comparison of methods for stock assessment of the Norway lobster, Nephrops norvegicus, in the Firth of Clyde. Fisheries Research 32: 89-100.

APPENDIX I

Rankings of criteria (1 being most important) nominated by each of the three readers for identification of major (top) and minor (bottom) openings of burrows of Metanephrops challengeri

The contribution of visual observations to surveying the deep-sea fish community

V.M. Trenkel^[63] and P. Lorance^[64]

1. INTRODUCTION

Since the development of manned deep submersibles (the bathysphere of Beebe in the early 1930s [Beebe 1933]; the F.N.R.S III in 1954, the Alvin in 1964, 1 830 m; and the Cyana in 1969, 3 000 m), scientists have had the ability to observe various components of deep-sea communities. The early scientific studies were primarily interested in natural history, such as collecting depth distribution and community composition data (e.g. Grassle et al. 1975). As any observation was provided information, a whole suite of observations were described and observations on behaviour of deep-water fishes and geological observations were sometimes reported in the same account as descriptions of the seabed and the associated benthic community (Fage 1958, Pérez 1958). Quickly the potential of visual observations for estimating population densities was recognized. However, fishes were much less studied than communities of fixed and mobile macrobenthos. This was mostly because the amount of data collected during dives of limited duration were too small to allow observation of significant numbers of fishes and although seldom mentioned, the high mobility of the fish fauna raised the question of the relevance of the observed number of animals. Nevertheless, some studies did evaluate the visual census approach by comparing density estimates obtained with submersibles with those from trawl or acoustic data, or with photos from camera sledges (Uzmann et al. 1977, Ralston, Gooding and Ludwig 1985, Krieger 1992, Adams et al. 1995, Cailliet et al. 1999). In recent years the use of submersibles in fisheries science has been enlarged to include the study of aspects of fish behaviour and small-scale species dynamics (Lorance, Latrouite and Séret 2000, Yoklavich et al. 2000, Uiblein, Lorance and Latrouite 2002, 2003). Another development has been the application of autonomous landers equipped with bait and camera systems to study the diurnal activities of deep-sea fish (Guennegan and Rannou 1979, Wilson Jr and Smith Jr 1984), population densities (Sainte-Marie and Hargrave 1987, Priede et al. 1994) and behaviour (Armstrong, Bagley and Priede 1992). At times landers and submersibles have been operated in tandem (Mahaut, Geistdoerfer and Sibuet 1990).

In this paper we review visual observation methods that provide quantitative information on species abundance and behaviour and their application for surveying deep-sea communities in support of sustainable management. We distinguish quantities relating to population ecology such as abundance estimates, habitat associations, demographic population structures and behaviour types from information on the interactions of fishing operations with individual fish and the habitat. Quantification of ‘catchability’, or vulnerability to capture falls into the second category. This paper will focus on observation methods and conditions and consider the limitations of visual observations to provide the required information. We suggest where further technological developments seem promising or would merit being investigated. The examples chosen to illustrate the contribution of visual observations to surveying deep-sea fish are taken from two surveys: Observhal 1998 cruise using the manned submersible Nautile and the Vital cruise 2002 using the remotely operated vehicle Victor 6000 (Figure 1).

2. CHOICE OF VEHICLE

Several types of vehicles are available to survey deep-sea and continental slope habitats (Table 1). Each vehicle has its own advantages and drawbacks. Currently there exist worldwide only a handful of manned submersibles and remotely operated vehicles (ROV) that can dive below 1 000 m. Both types of vehicles have been used successfully for fisheries research programmes. Recent technological developments have concentrated on developing autonomous underwater vehicles (AUV) and the application of their use in fisheries, but to date AUVs have not been used for collecting visual observations on fish communities as current projects focus on acoustic data collection (Fernandes et al. 2003). We note that it is doubtful whether AUVs are adequate for visual observations, as videos are demanding in electrical power, which is a limiting factor for AUVs. Camera sleds and towed bodies have also been used in deep-waters to study fish abundances (P. Lorance unpublished data).

Manned submersibles are without any doubt the most flexible of vehicles for surveying deepwater communities. However, while they are appropriate for exploratory studies their efficiency is low as they deliver only a few observations a day at sea due to the limited operation time per dive. In addition while they provide good observation conditions for the scientist on board, obtaining quantitative data is difficult primarily as the field of vision is generally uncalibrated (see Section 3). Although it is possible to overcome this problem by calibrating the field of vision and post-processing registered video records in a quantitative way, ROVs appear to be much more efficient than manned submersibles for surveying fish communities due to the longer observation time possible and consequently larger observed survey area that can be achieved. However, the ability of manned submersibles to survey rugged terrain should not be underevaluated. In contrast, the application of ROVs is limited in these situations as the cable connecting the ROV to the vessel can get caught on rocks and other bottom features. Similarly, telecommunication cables freely spanning canyons at some altitude from the bottom can be a problem for ROVs. Manned submersibles, therefore, might be the only option if steep canyons or areas of rough bottom require surveying. Indeed, being able to assess local relative population densities and species compositions in areas inaccessible to ROVs might be crucial for evaluating the refuge potential of a particular habitat or its suitability to become a marine protected area.

TABLE 1
Characteristics of different vehicles for surveying deep-water fish communities

For both sleds and towed bodies, the fact that their trajectory cannot be altered, as it is dependant on the vessel course and speed, presents an important drawback. Currents are also a factor in affecting towed systems. In addition, they can only be operated close to the bottom in relatively flat environments where large rocks or debris are not present that can foul the vehicle. Thus sleds and towed bodies might be seen as a second option, although depending on the objectives of the study, they could be a cost efficient alternative to manned submersibles or ROVs.

TABLE 2
Types of observations obtainable with different vehicles.
* stereo-video system required

Independent of the type of vehicle chosen, the quality of the optic equipment determines the value of observations and their suitability to study various aspects of deep-sea ecology. Most vehicles can be equipped with photographic cameras, simple videos, or stereo-video systems. In certain cases several systems can be used simultaneously, e.g. simple videos or cameras might be used for surveying vertically down from the vehicle, while stereo-video is used for surveying forwards or sidewards. High-resolution video systems should always be combined with digital video storage to allow optimal post-treatment of videos. Different vehicles allow collection of various types of visual observations suitable for different objectives. Table 2 summarizes what fisheries data can be collected by the variety of vehicles available. A detailed description of the data collection will be provided in the following sections.

3. COLLECTING QUANTITATIVE INFORMATION

The collection of quantitative visual observation from still photos, videos or by an observer, is only possible for near-bottom habitats as it requires accurate knowledge of the size of the observation field. An additional advantage of having the bottom visible is that it can increase contrast in the picture and hence facilitate animal detection. If the size of the observation field is unknown, population densities cannot be estimated, only relative abundances of different species or presence - absence data can be obtained. Various scientists have adopted different techniques to calibrate the observation field. Grassle et al. overlaid a perspective grid (Canadian grid) on their photographs. Auster, Malatest and Donaldson (1997) decomposed video images into successive trapeziums each measuring 0.35 m². Adams et al. and Trenkel, Lorance and Mahévas.(2004) calibrated the width of the survey area at a certain time line, i.e. a virtual line was positioned on the video monitor (Figure 2a).

In order to maintain the calibrated observation field throughout the survey, all camera settings (zoom, pan and tilt angles) and cruising altitude have to be fixed. It is not difficult to fix the angle of a video or still camera. Keeping the cruising speed fixed at a chosen speed will insure bias due to detection problems remains constant (Trenkel, Lorance Mahévas 2004).Obviously, these calibration methods only work if the bottom. these calibration methods only work if the bottom slope is reasonably flat and the vehicle can advance parallel to the slope. Steep slopes or, much worse, large rocks or outcrops alter the observation field and this method of standardisation will fail. Unfortunately, it is not obvious how to measure, or even define, the observation field for rugged habitats (Figure 2b).

4. POPULATION ECOLOGY

4.1 Abundance estimation

Many authors have used visual observations to obtain abundance estimates. To achieve these two methods have been used: strip transects (Adams et al. 1995, Trenkel 2003) and line transects (O’Connell, Carlile and Wakefield 1998). Strip transect methods consist of counting all animals encountered in a given survey strip (Seber 1982). Using quadrates, as shown by Grassle et al. (1975), equates to strip transects. For the line transect method, animals that are encountered are counted and their distance from the virtual transect line is also measured (Buckland et al. 2001). A detection function is fitted to the observations to estimate population abundance or density. In contrast to the strip transect method, it is not necessary to detect all animals for the line transect method, although generally it is assumed that all animals on the transect line have been registered. Measuring the distance of a fish from the transect line is straightforward if manned submersibles are used, as a sonar-gun can be pointed at the fish by the operator (O’Connell et al. 1998). In the case of ROVs more sophisticated equipment would be required and to the best of our knowledge this has not yet been attempted.

Once abundance estimates have been obtained, the question of what proportion of a stock or population has been surveyed remains. It is well known that many deep-sea species spend much time in the water column. For example, roundnose grenadier (Coryphaenoides rupestris), black scabbardfish (Aphanopus carbo) and orange roughy (Hoplostethus atlanticus) are found high off the bottom (Nakamura and Parin 1993, Atkinson 1995, Koslow, Kloser and Stanley 1995, McClatchie et al. 2000). Some species also carry out diel migrations (Atkinson 1995). Hence density estimates for the same area can vary on a diel cycle. Trenkel, Lorance & Mahévas (2004) found significant day-night effects for population density estimates of orange roughy (Hoplostethus atlanticus) and roundnose grenadier.

Avoidance of, and attraction to, the underwater vehicles is a further problem. Many researchers have noted reactions to approaching manned submersibles (Uiblein et al. 2002) and ROVs (Adams et al. 1995, Trenkel et al. (2004). Reactions at considerable distances have been documented (Koslow et al. 1995), however, it is difficult to assess how many individuals avoid the approaching vehicle at a great distance and hence are not detected. Directional swimming can also lead to biased abundance estimates due to the generally slow surveying speed; however if fish swimming velocity and direction is known, a correction for this bias is possible (Trenkel 2003).

4.2 Demographic structure

The deep-sea fish community includes a large number of small-sized species that may be numerically dominant. These are poorly sampled by commercial type trawls with large mesh sizes, that are used to sample commercial fish populations (Gordon 1986, Merrett et al. 1991, Gordon and Bergstad 1992). Even for larger species, trawl samples do not always provide a representative sample from the population length distribution due to mesh selectivity (escape of small or young individuals) and escapement of larger more mobile individuals (Godø and Walsh 1992, Walsh 1992). The impact of these factors might be reduced to acceptable levels by the use of small mesh sizes and appropriate survey speeds. As in other marine sampling applications, the real size distribution of local deep-water populations is generally unknown; hence it is impossible to validate the length distribution obtained from trawl samples. This is where visual observations can provide crucial insights.

The body size of fish or attached benthic species can be measured in situ in several ways. For simple camera systems, parallel lasers and measurements based on the focal point of the camera have been used (Rochet, Cadiou & Trenkel, inpress). These methods require the positioning of the fish to be measured at a right angle to the focal axis. Flexed tails and continuous movements of the fish will bias these measurements.

Surprisingly, in contrast to the differences in population densities for North Atlantic codling (Lepidion eques) estimated by both the visual census data and trawl samples during the Vital cruise, the length distributions derived from the two sampling methods were rather similar (Figure 3). However, given the small number of individuals caught and measured visually, this result might not be significant.

From our survey data results we note that estimating the size distribution of large commercial fish from ROV-recorded videos may be difficult for several reasons. First, commercial species tend to be observed in small numbers due to the low sampling intensity of the method. Second, several of these species are fast moving and are often seen passing quickly at the edge of the observation field and hence are not measurable. For large mobile individuals the use of stereo-video methods (Harvey et al. 2003) might be more promising. To date these have not been used for moving vehicles but only in static set ups, such as in connection with tuna aggregating devices.

Thus, visual methods can provide size distributions of small species, which will allow an assessment of the effects of fishing on these non-target species. Average length data of individual species as well as the fish community as a whole are important indicators for describing the state of an exploited community and for detecting changes that might be caused by fishing (Rochet and Trenkel 2003). These indicators may be especially relevant in the context of deep-sea communities as some length distribution data are available prior to the start of commercial exploitation. These data can provide an estimate of the "pristine" state of such communities, which will allow the setting of target reference points. This is a unique situation not often encountered for shelf populations, where the lack of knowledge of pristine states prevents the definition of such reference points.

4.3 Behaviour

Currently little is known about deep-sea fish behaviour. Most knowledge has been inferred from specimens taken from trawl samples where the morphology of fish has allowed inferences of their possible behaviour. For example, stomach content data shed light on fish locations in the water column and their foraging strategy (e.g Mauchline and Gordon 1984a,b). In contrast to this indirect information, underwater vehicles allow the direct in situ observations of natural fish behaviour. In addition, they enable various types of reaction behaviours to be studied by observing their reactions towards the approaching vehicle. Along with the studies using solely underwater vehicles, fish reactions at trawl opening have also been observed directly by attaching cameras at various parts of the trawl (Hemmings 1973) and by using ROVs working in parallel with the trawl. An example of this based on ROV observations is given by Bublitz (1996) who categorized the reaction types of flatfish as that of direction change, including turning on their side or back and avoidance behaviour by rising upwards.

Individual behaviour has been studied by post-processing video records collected during the Observhal 1998 cruise (Lorance, Uiblein and Latrouite 2002, Uiblein et al. 2002, 2003). These authors characterized behaviour according to several variables: position in the water column, locomotion and activity mode. The study showed that in addition to having diverse shapes, adaptations, bioenergetics and diets, deep-water fishes have different natural behaviours. They adjust their behaviour to small-scale variations in habitat and environmental conditions.

Investigating how reaction behaviour might explain vulnerability to fishing is of major interest for both management and survey-based stock assessments. Some species, e.g. Trachyscorpia c. echinata, display almost no reaction to an approaching submersible. If this species behaved in the same way in front of a trawl, only individuals within the path of the net would be caught. In contrast, some individuals of roundnose grenadier and orange roughy have been observed slowly swimming in front of the manned submersible Nautile and the ROV Victor (Figure 4). Such behaviour might result in herding by a bottom trawl. If this is true, then these two species that strongly differ biologically and ecologically (Koslow, 1996) may have similar vulnerabilities to capture by active fishing gears. Similarly, the high catch rates of smoothheads (Alepocephalus bairdii was by far most noteable during the Vital cruise) and trichiurids (here Aphanopus carbo) probably result from their flight movement towards the bottom in reaction to the noise and motion of trawls. This flight movement was observed during the Vital 2002 cruise. It may be that the bottom is a refuge for benthopelagic fishes from predation. However, this flight strategy will increase such species’ chances of being caught by bottom trawls. In contrast, large chondrichthyans such as chimaeras and squalid sharks that remain close to the bottom react strongly to manned submersible and ROVs, and some escape laterally (Lorance et al. 2000, Lorance, Trenkel &Uiblein 2005). It might be that similar behaviour in front of a trawl would result in reduced vulnerability of that species. In summary, the behaviour of deep-water species can have major implications for their vulnerability by bottom trawls and visual observations can provide unique information for studying this important component of their behaviour (see also Section 5.1 Catchability).

4.4 Habitat association

Individual spatial distributions and the definition of a species habitat may be considered at different scales. Catch records properly address the biogeographical scale and the large-scale depth and area distributions; as the area swept by a survey trawl will cover several hectares. For example, with a large orange roughy trawl of 30 m wingspread towed at 2 knots, an one-hour tow sweeps an area of more than 11 hectares. In contrast, fish habitat preferences might imply much smaller scales. It is well known that orange roughy form dense aggregations with low densities (Koslow 1997, Clark 1999, Clark et al. 2000). Observations from a manned submersible during the Observhal cruise have shown that these aggregations can be quite small (Lorance et al., 2002) so that a single trawl tow may cover different habitats and fish densities. This phenomenon also seems to apply to other deep-sea species. The density of several species has been observed to vary at the macro-habitat scale spanning a few hectares (Uiblein et al. 2003). Individual fishes may select preferred locations at an even smaller scale of a few metres. This seems to be the case for deep-sea scorpionfish (Trachyscorpia c. echinata) or Lepidion eques, which were often seen associated with small bottom features such as stones or benthic fauna colonies during the Vital cruise (unpublished data, Figure 5), while Neocyttus helgae was mainly associated with vertical cliffs. Flat, sandy and unchanging seabed bottom types would provide less shelter than similar habitats with additional dispersed features. Small-scale bottom structures are likely to be a major factor for explaining local population densities of deep-sea fish species and might be crucial for community diversity. The importance of small-scale habitat features has been observed elsewhere. Auster et al. (1997) using an ROV found that juvenile silver hake densities were greater on bottoms with high amphipod tube cover compared to featureless grounds.

Submersible observations revealed that rockfish were mainly associated with rugged habitats (Krieger 1992).

Strong habitat associations have implications for survey-based stock abundance estimation. In particular in order to increase the precision of estimates, it can be useful to use habitat characteristics to define homogeneous areas. Nasby-Lucas et al. (2002) used a high-resolution multibeam sonar for obtaining habitat information and submersible visual counts to estimate population densities by habitat type. Of course this approach requires detailed habitat knowledge for the whole population area.

5. EXPLOITATION

5.1 Catchability^[65]

Catchability is an important factor for both abundance indices derived from scientific trawl surveys and catch based stock assessments (Fréon, Gerlotto and Misund 1993). Visual observations from submersibles and ROVs have been used in two ways to obtain information on trawl catchability:

i. direct comparison of visual based density estimates with trawl swept-area based estimates (Krieger and Sigler 1995, Somerton et al. 1999) and

ii. estimation of the different components of catchability, such as (a) habitat preference, (b) differences in diel activity, (c) body position in water column, (d) body size, (e) patterns of spatial distribution (spatial randomness) and (f) reaction to the approaching vehicle (Trenkel et al. 2004).

Habitat preference can be considered on several scales. On a large scale the comparison of population densities on seamounts, terraces, canyons and other types of habitats determines preferences. On a finer scale, population densities in different types of habitats within these broad categories might also be considered. In terms of trawl catchability, it is the larger scale that is relevant, in particular the relationship between population densities in trawlable (terraces and sea mounts) and non-trawlable (canyon) areas. Note however, that ROVs in particular might encounter difficulties with surveying rugged canyons (see Section 3, Collecting quantitative information).

Diel activity patterns can be detected by surveying a given area around the clock. In order to separate diel patterns from other effects, ideally all other parameters such as depth and tidal condition should be kept fixed. The impact of diel variations in population densities can lead to complex patterns in catchability.

Body position in the water column affects what part of the population is accessible to a trawl and hence has an impact on catchability but also determines the accessibility to visual observations. If the vertical visual observation field is smaller than the vertical opening of a given trawl of interest, the obtainable information is incomplete. Nevertheless, the form of the vertical distribution gives an indication for the extent of this mismatch (Figure 6).

The type of spatial distribution affects catchability in as much as schooling species might have increased vulnerability to capture once detected. It also affects the variance of abundance estimates derived from trawl and visual observations. For visual observations the exact absolute location of each fish encountered is known provided accurate positioning systems are used. From these the distance between individuals can be derived (Figure 7). If individuals are randomly distributed in space, these histograms should follow an exponential distribution. This is the case for Coryphaenoides rupestris but not for Synaphobranchus kaupi, which shows signs of non-randomness (overdispersion). Alternatively, counts per transect line can be used to analyse the spatial distribution (Trenkel et al. 2004). When analysing spatial distributions, fixed effects due to, for example, depth need to be taken account of for both approaches. Trenkel et al. (2004) found that the type of spatial distribution (represented by the dispersion factor of the Poisson model for counts) was the most important explanatory factor across all species considered for modelling the ratio between visual census based density estimates and trawl based estimates.

Reaction to an approaching trawl has clear implications for catchability. The observation of reaction behaviour from videos has already been discussed in Section 4.3 Behaviour). The question remains of how relevant these reactions are towards an approaching ROV or any other vehicle used for determining reactions towards a trawl. Obviously the stimuli will differ, as trawls generally do not have lights. However, both make noise, which has been found to trigger reactions by orange roughy (Koslow et al. 1995). Thus for all intents and purposes, differences in reactions between species observed by underwater vehicles should be informative at least on a relative scale.

5.2 Physical impact

To assess the physical impact of fishing gears on the seabed it is necessary to survey large areas of fished seabed in order to catalogue gear marks, lost gears or parts of gear. It is likely that ROVs and benthic sleds are the most efficient vehicles for this task due to their capacity to sample large areas. Indeed, in shallow areas towed cameras have provided satisfactory results (Collie, Escanero and Valentine 2000). Unfortunately, the longevity of trawl marks on deep-water bottom is poorly known and may vary with depth, substrate, natural disturbance or activity of burying fauna. For example, it is likely that in areas with many ripple marks, trawl marks are quickly erased when tidal currents regenerate ripple marks. On areas without ripple marks, the processes that erase trawl marks are sedimentation (expected to be slow) and bioturbation whose efficiency at erasing trawl marks is unknown but should be proportional to the density of bioturbated marks.

We categorized trawl marks recorded by videos during the Vital cruise according to their size and aspect (recent or not) together with the substrate type (sand, mud, bioturbated or not, presence of ripple marks) and the density of large benthic macrofauna. The results clearly show that deepwater trawling strongly alters the first centimetres of sediments and reduces the density of fixed macrofauna (Figure 8). However, quantification of the degree of damage done by trawling is difficult as comparable images prior to exploitation are not available so that the changes in benthic density and species composition cannot be assessed. Finding comparable exploited and unexploited areas becomes difficult as fisheries have worked on all suitable grounds. During the Vital cruise we sampled two small terraces one being exploited and the other not. It was obvious from the results that the two terraces differed in terms of substrate and hence their natural fauna. As a consequence, the untrawled terrace could not serve as reference for the trawled terrace as initially hoped.

Although the absolute level of destruction caused by trawling cannot be evaluated, indices of the visible impact of trawling (e.g. number of trawl marks per unit of distance travelled) can be determined and related to macrofauna density, diversity and species composition. One important component of the affected macrofauna are the cold-water corals. Their reefs are clearly sensitive to trawling as one single tow destroys almost all corals within the path swept by the doors. In Norwegian waters, trawling was estimated to have affected or destroyed 30 to 50 % of Lophelia reefs......(Fossa et al., 2002). Trawling for orange roughy in deeper waters has also been reported to destroy cold-water corals (Koslow et al. 2000, Clark 1999). ROVs and manned submersible are the only vehicles that can undertake non-destructive monitoring of cold-water corals. They are suitable for assessing the proportion of the reefs that is impacted and they might also allow observation of the regeneration/recolonisation processes.

6. SOME SOLUTIONS TO CURRENT LIMITATIONS OF VISUAL OBSERVATIONS

This paper has concentrated on the contribution of visual observations to surveying the deep-water fish community. The alternative approach, in addition to trawling, are acoustic methods. However, acoustic and visual observations provide complementary observations on two different spatial scales for the same phenomena. Acoustic methods can detect fish reactions at large distances to an approaching video camera (Koslow et al., 1995) and a trawl (McClatchie et al. 2000) without necessarily providing species identification. Fishery research echo-sounders provide observations within a range of 50-200 m while visual observations are limited to about 10 m around the vehicle depending on turbidity and light conditions. Combining both systems would enable use of visual observations at smaller scales and acoustic information at larger scales. Visual observations would provide information on species composition and size classes while acoustic information would allow spatial extrapolation. In addition, acoustic methods can provide crucial information on vertical distributions above the range that is visually discernable.

A second area where further developments are important is that of absolute population density estimates. This could involve operating simultaneously underwater vehicles for visual observations, towed bodies or AUVs for acoustic information and scientific trawls. Currently there might be logistic limitations to carry out all three operations from the same research vessel. Hence, several vessels would be required which would make this a costly operation. In the short to medium term, it seems appropriate to consider using a fishery research vessel to operated an ROV or manned submersible for collecting visual observations together with a hull-mounted echo-sounder and to have trawl tows carried out from a commercial vessel. In the longer term, concomitant use of visual (e.g. ROV) and acoustic (e.g. AUV) vehicles is an attractive option. It should also be considered whether a sonar suitable for fish detection could be added to the ROV or manned submersible. In terms of improving visual observations, most advances are to be expected from the use of stereo-video systems. Such developments are currently underway at our institute.

7. LITERATURE CITED

Adams, P.B., J.L. Butler, T.E. Laiding, K.A. Dahlin & W.W. Wakefield 1995. Population estimates of Pacific coast groundfishes from video transects and swept-area trawls. Fishery Bulletin. 93: 446-455.

Armstrong, J.D., P.M. Bagley & I.G. Priede 1992. Photographic and acoustic tracking observations of the behaviour of the grenadier Coryphaenoides (Nematonurus) armatus, the eel Synobranchus bathybius, and other abyssal demersal fish in the North Atlantic Ocean. Marine Biology. 112: 535-544.

Atkinson, D.B. 1995. The biology and fishery of roundnose grenadier (Coryphaenoides rupestris Gunnerus, 1765) in the north west Atlantic. A. Hopper. Dordrecht, Kluwer Academic Publishers, 51-111 pp.

Auster, P., R. Malatest & C. Donaldson 1997. Distributional responses to small-scale habitat variability by early juvenile silver hake, Merluccius bilinearis. Environmental Biology of Fishes. 56: 195-200.

Beebe, W. 1933. Preliminary account of deep sea dives in the Bathysphere with special reference to one at 2 200 feet. Proc. Natl. Acad. Sci. 19: 178-188.

Bublitz C.G., 1996. Quantitative evaluation of flatfish behavior during capture by trawl gear. Fisheries research, 25, 2-3: 293-304.

Buckland, S., D. Anderson, K. Burnham, J. Laake, D. Borchers & L. Thomas 2001. Introduction to distance sampling. Estimating abundance of biological populations. Oxford University Press, Oxford.

Cailliet, G., A. Andrews, W. Wakefield, G. Moreno & K. Rhodes 1999. Fish faunal and habitat analyses using trawls, camera sleds and submersibles in benthic deep-sea habitats off central California. Oceanologica Acta. 22: 579-592.

Clark, M. 1999. Fisheries for orange roughy (Hoplostethus atlanticus) on seamounts in New Zealand. Oceanologica Acta. 22: 593-602.

Clark, M.R., O.F. Anderson, R.I.C.C. Francis & D.M. Tracey 2000. The effects of commercial exploitation on orange roughy (Hoplosthetus atlanticus) from the continental slope of the Chatham Rise, New Zealand, from 1979 to 1997. Fisheries Research. 45: 217-238.

Collie, J.S., G.A. Escanero & P.C. Valentine 2000. Photographic evaluation of the impacts of bottom fishing on benthic epifauna. ICES Journal of Marine Science. 57: 987-1001

Fage, L. 1958. Les campagnes scientifiques du bathyscaphe F.N.R.S. III. Annales de l’Institut Océanographique de Monaco. 35:237-242.35: 242.

Fernandes, P.G., P. Stevenson, A.S. Brierley, F. Armstrong & E.J. Simmonds 2003. Autonomous underwater vehicles: future platforms for fisheries acoustics. ICES Journal of Marine Science. 60: 684-691.

Fossa J.H., Mortensen P.B., Furevik D.M., 2002. The deep-water coral Lophelia pertusa in Norwegian waters: distribution and fishery impacts. Hydrobiologia, 471, 1-3: -12.

Fréon, P., F. Gerlotto & O.A. Misund 1993. Consequences of fish behaviour for stock assessment. ICES Journal of Marine Science. 196: 190-195.

Godø, O.R. & S.J. Wals 1992. Escapement of fish during bottom trawl sampling - implications for resource assessment. Fisheries Research. 13: 281-292.

Gordon, J.D.M. 1986. The fish populations of the Rockall Trough. Proceedings of the Royal Society of Edinburgh. 88B: 191-204.

Gordon, J.D.M. & O.A. Bergstad 1992. Species composition of demersal fish in the Rockall Trough, North-Eastern Atlantic, as determined by different trawls. J. mar. biol. Ass. U. K. 72: 213-230.

Grassle, J., H. Sanders, R. Hessler, G. Rowe &T. McLellan 1975. Pattern and zonation: a study of the bathyal megafauna using the research submersible Alvin. Deep-Sea Research. 22: 457-481.

Guennegan, Y. &M. Rannou 1979. Semi-diurnal rhymthmic activity in deep-sea benthic fishes in the Bay of Biscay. Sarsia. 64: 113-116.

Harvey, E., M. Cappo, M. Shortis, S. Robson, J. Buchanan &P. Speare 2003. The accuracy and precision of underwater measurments of length and maximum body depth of southern bluefin tune (Thunnus maccoyii) with a stereo-video camera system. Fisheries Research. 63: 315-326.

Hemmings, C.C. 1973. Direct observation of the behaviour of fish in relation to fishing gear. Helgoländer wiss. Meersunters. 24:348-360.

Koslow J.A., 1996. Energetic and life-history patterns of deep-sea benthic, benthopelagic and seamount-associated fish. Journal of Fish biology, 49 (Suppl. A), 54-74.

Koslow, J., R. Kloser & C. Stanley 1995. Avoidance of a camera system by a deepwater fish, the orange roughy (Hoplostethus atlanticus). Deep-Sea Research. 42: 233-244.

Koslow, J.A. 1997. Seamounts and the ecology of deep-sea fisheries. American Scientist. 85: 168-176.

Krieger, K.J. 1992. Distribution and abundance of rockfish determined from a submersible and by bottom trawling. Fishery Bulletin. 91: 87-96.

Krieger, K.J. & M.F. Sigler 1995. Catchability coefficient for rockfish estimated from trawl and submersible surveys. Fishery Bulletin. 94: 282-288.

Lorance, P., D. Latrouite & B. Séret 2000. Observations of chondrichthyan fishes (rays and chimaeras) in the bay of Biscay (North-Eastern Atlantic) from submersibles. B. Séret and J.-Y. Sire. Boulogne-sur-Mer, Societé Francaise d’Ichtyologie &IRD, 29-45 pp.

Lorance, P., F. Uiblein & D. Latrouite 2002. Habitat, behaviour and colour patterns of orange roughy Hoplostethus atlanticus (Pisces: Trachichthyidae) in the Bay of Biscay. Journal of the Marine Biological Association U.K. 82: 321-331.

Lorance, P., V.M. Trenkel, V.M. & F. Uiblein. Characterizing Natural and Reaction Behaviour of Large Mid-slope Fish Species: Consequences for Catchability. In R. Shotton (Ed). Deep Sea 2003: Conference on the Governance and Management of Deep-sea Fisheries. Conference Poster Papers (1-5 December 2003, Queenstown) and Workshop Papers (27-29 November 2003, Dunedin) New Zealand. FAO Fisheries Proceedings. No. 3/2. FAO, Rome. Proceedings of the Deep Sea Conference (in preparation).

Mahaut, M.L., P. Geistdoerfer & M. Sibuet 1990. Trophic strategies in carnivorous fishes: their significance in energy transfer in the deep-sea benthic ecosystem (Meriadzek Terrace - Bay of Biscay). Progress in Oceanography. 24: 223-237.

Mauchline, J. &J.D.M. Gordon 1984a. Diets and bathymetric distributions of the macrourid fish of the Rockall Trough, northeastern Atlantic Ocean. Marine Biology. 81: 107-121.

Mauchline, J. &J.D.M. Gordon 1984b. Feeding and bathymetric distribution of the gadoid and morid fish of the Rockall Trough. J. mar. biol. Ass. U. K. 64: 657-665.

McClatchie, S., R. Thorne, P. Grimes &S. Hanchet 2000. Ground truth and target identification for fisheries acoustics. Fisheries Research. 47: 173-191.

Merrett, N.R., J.D. M. Gordon, M. Stehmann & R.L. Haedrich 1991. Deep demersal fish assemblage structure in the Porcupine seabight (Eastern North Atlantic): Slope sampling by three different trawls compared. Journal of the Marine Biological Association U.K. 71: 329-358.

Nakamura, I. & N. Parin 1993. FAO species catalogue. Snake mackerels and cutlassfishes of the world (families Gempylidae and Trichiuridae). An annotated and illustrated catalogue of the snake mackerels, snoeks, escolar, gemfishes, sackfishes, domine, oilfish, cutlassfishes, scabbardfishes, hairtails and frostfishes known to date.

Nasby-Lucas, N., B. Embley, M. Hixon, S. Merle, B. Tissot &D. Wright 2002. Integration of submersible transect data and high-resolution sonar imagery for a habitat-based groundfish assessment of Heceta Bank, Oregon. Fishery Bulletin. 100: 739-751.

O’Connell, V.M., D.C. Carlile &W.W. Wakefield 1998. Using line transects and habitat-based assessments techniques to estimate the density of yelloweye rockfish (Scorpaenidea: Sebastes) in the Eastern Gulf of Alaska. pp.

Pérez, J.-M. 1958. Remarques générales sur un ensemble de quinze plongées effectuées avec le bathyscaphe F.N.R.S. III. Annales de l’Institut Océanographique de Monaco. 35: 259-285.

Priede, I.G., P.M. Bagley, A. Smith, S. Creasey &N.R. Merrett 1994. Scavenging deep demersal fishes of the porcupine seabight, North-East Atlantic: Observations by baited camera, trap and trawl. Journal of the Marine Biological Association U.K. 74: 481-498.

Ralston, M., R. Gooding &G. Ludwig 1985. An ecological suvey and comparison of bottom fish resource assessments (submersible versus handline fishing) at Johnston Atoll. Fishery Bulletin. 84: 141-155.

Rochet, M.-J. & V.M. Trenkel 2003. Which community indicators can measure the impact of fishing? A review and proposals. Canadian Journal of Fisheries and Aquatic Sciences. 60: 86-99.

Rochet, M.-J., J.F. Cadiou &V.M. Trenkel submitted. Precision and accuracy of fish length measurements obtained with two visual underwater methods. Fishery Bulletin.

Sainte-Marie, B. & B.T. Hargrave 1987. Estimation of scavanger abundance and distance of attraction to bait. Marine Biology. 94: 431-443.

Seber, G. 1982. The estimation of animal abundance and related parameters. Macmillan, New York.

Somerton, D., J. Ianelli, S. Walsh, S. Smith, O. Godø &D. Ramm 1999. Incorporating experimentally derived estimates of survey trawl efficiency into the stock assessment process: a discussion. ICES Journal of Marine Sciences. 56: 299-302.

Trenkel, V.M. 2003. Density estimators for strip transects when animals show directional movement and observation speed is slow. Computational Statistics and Data Analysis. 44: 305-312.

Trenkel V.M., Francis R., Lorance P., Mahevas S., Rochet M.J., Tracey D.M., 2004. Availability of deep-water fish to trawling and visual observation from a remotely operated vehicle (ROV). Mar. Ecol. Prog. Ser., 284,293-303.284, 293-303

Trenkel, V.M., P. Lorance, D. Tracey, S. Mahévas &M.-J. Rochet. submitted. Estimating components of catchability using visual observations from an ROV.

Trenkel, V.M., P. Lorance & S. Mahévas 2004. Do visual transects provide true population density estimates for deep-water fish? ICES Journal of Marine Science, 61(7):1050- 1056.

Uiblein, F., P. Lorance & D. Latrouite 2002. Variation in locomotion behaviour in northern cutthroat eel (Synaphobranchus kaupi) on the Bay of Biscay continental slope. Deep-Sea Research I. 49: 1689-1703.

Uiblein, F., P. Lorance & D. Latrouite 2003. Behaviour and habitat utilization of seven demersal fish species on the Bay of Biscay continental slope, NE Atlantic. Marine Ecology Progress Series. 257: 223-232.

Uzmann, J.R., R.A. Cooper, R.B. Theroux &R.L. Wigley 1977. Synoptic comparison of three sampling techniques for estimating abundance and distribution of selected megafauna: submersible vs camera sled vs otter trawl. Marine Fisheries Review. 39: 11- 19.

Walsh, S.J. 1992. Size-dependent selection at the footgear of a groundfish survey trawl. North American Journal of Fisheries Management. 12: 625-633.

Wilson Jr, R.R. & K.L. Smith Jr 1984. Effect of near-bottom currents on detection of bait by the abyssal grenadier fishes Coryphaenoides spp., recorded in situ with a video camera on a free vehicle. Marine Biology. 84: 83-91.

Yoklavich, M., H. Grenne, G. Cailliet, D. Sullivan, R. Lea &M. Love 2000. Habitat associations of deep-water rockfishes in a submarine canyon: an example of a natural refuge. Fishery Bulletin. 98: 625-641.

Technical requirements and prerequisites for deepwater trawling

W. Thiele^[66] and G. Niedzwiedz^[67]

1. INTRODUCTION

With the establishment of EEZs (Exclusive Economical Zones) in the late 1970s, many of the far distant fishing fleets in Europe lost their traditional fishing grounds. The reaction of the fishing nations concerned were different. Some countries, such as Germany, reduced their fleets drastically while other countries looked for opportunities to fish under licence in the EEZs of coastal states; yet others tried to explore and exploit resources outside the EEZs in greater depths.

The last mentioned option was a real challenge, both for fishermen and scientists, stock assessment scientists as well as gear technologists as vessels, winches and some gear elements were inadequate for working in depths greater than 800 m.

Our research work as gear technologists focussed on three fields,

i. elements of ground gears
ii. otter boards and
iii. warps

because these three trawl gear elements are the most important one for successful deepwater trawling.

2. TECHNICAL RESEARCH WORK FOR DEEPWATER TRAWLING

2.1 Ground gears

Because of the fact that the traditional used steel bobbins of different sizes starts to implode in depths greater than 800 m, rubber bobbins of different sizes were used in Eastern Germany for experimental fisheries. Figure 1 shows the arrangement of a ground gear for a 90’ bottom trawl.

One section of the ground rope consists of 11 big rubber bobbins, 44 small rubber bobbins and 10 lancasters (chains). The whole ground rope consisted of four such sections. The weight in air was around 400 kg and weight in water, around 150 kg. Of course, today the few remaining German trawlers also used rockhopper gears and several other ground fishing gear elements.

In the Polish distant water fleet similar investigations were made with bobbins similar to those shown in Figure 2. Those bobbins consisted of wood, with a heavy iron cover. In the middle section of the ground rope 12 bobbins were used and another 22 were inserted in each trawl wing.

The two different ground gear elements mentioned are so called static elements because they create the pressure on the sea bed only by their weight in water. Therefore it should be mentioned that a hydrodynamic bobbin was developed by the Russian scientist Karpenko. This bobbin creates downwards directed hydrodynamic forces by rolling over the bottom.

This bobbin consist of two half spheres, fixed on a shaft. Between the half spheres four vanes are fixed, shaped as a cylindrical segment. Openings go through the cover of the half spheres to add strength and avoid diverting the current.

During towing the current turns the bobbins and because of the "Magnus effect" a downwards-directed force is created. The size of this force is proportional to the towing speed. If the bobbin touches an obstacle (boulders, stones, etc.) it stops turning and the hydrodynamic forces become zero and the ground gear can rise more easily over the obstacle (Figure 3).

2.2 Otter boards

Otter boards used in deepwater trawling must fit specific requirements. To speed up the shooting process and for reaching maximum depths with a given warp length, they must sink quickly. Therefore otter boards for deepwater trawls should be heavy. And, because of the fact that the reachable depths depends on resistance of the trawl and otter boards, the drag of the otter boards used should be as low as possible.

And finally, taking into account the relatively long shooting process, the otter boards must have a high static stability to avoid unclear or twisted gears. Static stability is defined as the ability of the otter board to turn back to the original position in case of disturbance or perturbations. Not all otter boards used in bottom trawling fulfil such requirements and research has been undertaken to investigate these effects.

Comparative investigations were made with two types of otter boards on board the same vessel and using the same trawl. The first trial was made with oval-shaped three-nozzle otter boards (Figure4) with an area of 5.5 m2, weight in air of between 1350 and 1500 kg and a weight in water of between 1150 and 1280 kg.

The otter boards were adjusted to the smallest possible angle of attack (˜ 35°). Figure 5 shows the hydrodynamic coefficients Cx (drag coefficient), Cy (lifting coefficient) and the quality factor k, defined as the quotient Cy/Cx. According to the abovementioned angle of attack of around 35°, the coefficients are Cy = 1.0; Cx= 0.67 and the quality factor Cy/Cx = 1.5.

This otterboard has a good hydrodynamic efficiency, but the main disadvantage is the low static stability resulting from using long warps. Several times during the trials the trawl became snarled. Therefore, it is not recommended to use oval shaped otter boards for deepwater trawling. Much better results were achieved with v-shaped rectangular otter boards of area, 5.5 m2 and a weight between 1350 and 1550 kg (weight in water 1150-1320 kg accordingly) (Figure 6). With the smallest angle of attack of around 30°, the hydrodynamic coefficients of this type of otter boards were CY = 0.8; Cx = 0.5 with a quality factor, k = 1.6 (Figure 7).

V-shaped otter boards have lower lifting forces, but also lower hydrodynamic resistance. The main advantage is the high static stability. This allows high running speeds of the warps during shooting, which reduces the unproductive time of the vessel.

Otter boards used for deepwater trawling must have a high static stability, which is influenced by the following main factors:

• hydrodynamic characteristics

• length and tension of warps and

• fixation coordinates of the doors.

Poorly adjusted trawl doors can be very sensitive to hydrodynamic forces created by the warps and this can led to an unstable run of the gear. The lower lifting force of v-shaped otterboards can be partly compensated through the use of differently twisted warps on the starboard and port side. This is discussed in more detail in Section 2.4.

Based on the results from these trials with different otter boards it was recommended that v-shaped otter boards be used for trawling in great depths. At present, most of the otter boards used for deepwater trawls are v-shaped, so our research results achieved several years ago have been confirmed.

2.3 Warps

2.3.1 One-warp systems

Trawlers not specially designed for deepwater trawling are equipped with winches of relatively small warp capacity. Towing speed and working depths are therefore limited. By use of a single warp system is it possible to nearly double the working depths for such vessels. Of course, the forces in the remaining warp will be higher and the shape of the warp during towing will be different. Figure 8 shows the results of practical trial with such a system

As shown in Figure 8 the warps run nearly parallel, the one warp system needs only around 50 m more warp length to reach the same depths. But the length of available warp is nearly double and operators can reach greater depths.

Handling of such system is easy and can be done according to Figures 9 and 10. Using such a system for stock assessment surveys is strongly recommended, in order to keep a constant trawl opening independent of warp lengths.

2.3.2 Some specific mechanical aspects

In the past, models for the calculation of warp shape and tension were strongly simplified. Most of the calculations were made under the assumption that

• warps are inelastic

• warps can only transmit traction, not bending forces, torsion moments and gross forces

• the shape of the warps is a straight line

• only drag and lifting forces are important for shape and tension, and

• hydrodynamic gross forces are not taken into account (Hahlbeck 1976).

Such simplified assumptions are valid for short warp lengths.

With warp lengths over 1000 m, the impact of the warps on the stability of the gear becomes more important. The weight of the warps also becomes an important factor. The ratio weight of warps/tension of warps increases, which leads to a change in the ratio warp length to trawl depth.

Additional hydrodynamic forces (gross forces) created by the warps must be taken into account. If the impact of such forces is ignored, it will result in

• asymmetric gear by unequally twisted warps

• instable otter boards, because of incidences of doors approaching critical angles of attack

• limited control over important gear performance parameters and

• loss of catch due to fouled gear (Niedzwiedz 1991, Scheel 1979).

Therefore, it is justifiable to install more gear control devices to be sure that the gear is working properly. Installation of symmetry sensors, bottom contact sensors, and a tension sensor in deepwater trawls are strongly recommended. For successful and efficient deepwater trawling high performance trawl and cable winches with higher drum capacity are needed. The otter boards need specific adjustment, and shooting and hauling of the trawl is more time consuming.

2.4 Estimation of gross forces in warps

The influence of the measurable hydrodynamic cross forces by using long warps are often underestimated. Such forces originated in overlying water currents due to horizontal movement of the rope through water with a circulating water layer, which is caused by the spiral like surface of the rope. A lot of systematic experimental investigations on that subject have been made to estimate the hydrodynamic coefficients for all kind of warp constructions. Figure 11 shows, as an example, coefficients for a six-strand rope. The size of the hydrodynamic gross forces can be estimated by using those coefficients.

The direction of the forces depends on the twist direction of the rope. By creating a right-lay (or Z-lay) rope, means the strands are twisted in a clockwise direction producing a force directed to the right. A left-lay (or S-lay) rope will create a force in the opposite direction. Such phenomenon can be used to support the efficiency of otter boards. The use of warps, twisted in the same direction on starboard and port side will also result in the gear being off center relative of the vessel (Niedzwiedz 1988, Paschen et al. 1995. The following graph (Figure 12) shows this effect.

The phenomenon demonstrated above makes it clear that the use of warps twisted in different directions (ZorS) will support the horizontal opening of the gear. It can be seen from the graph (C), that the distance of the otter boards will be 128 m; by using warps twisted in opposite directions and the trawl will run in line with the vessel (No. 1-4). The use of equal twisted warps gives an horizontal opening of only 41m and the trawl will tow 44 m off the course of the vessel (No. 1-2, 3-4) (SeeFigure 12).

Based on these calculations and practical experience it is recommended to use on the starboard side, Z-lay warps and on the port side S-lay warps in order to support the spreading forces of the trawl doors with the gross forces created by the warps. So there is a possibility to choose smaller doors of lower hydrodynamic resistance.

Torque from the warps may create another negative effect. As a twisted rope comes under strain, it tries to "unlay" and creates twisting forces, which may affect the tilt angles of the otter boards because this force attempts to rotate the otter boards. If both warps are right-lay, both otter boards will be tilted in the same direction relative to the boat, with the result that the trawl is running not in the same line as the vessel. But this force can be eliminated by the use of swivels and, or, use of warps twisted in different directions.

Trawls are often connected by cables for gear control and measurement instruments such as net sounders etc., which can also influence the symmetry of the gear. Therefore it is of advantage to know the shape and forces created by such equipment. An example that shows how the shape and the direction of forces can change is shown in Figure 13 where the shape of a cable fixed to a trawl is calculated at different towing speeds.

3. CONCLUSIONS

• The essential elements for deepwater trawling are the otter boards and warps.

• Otter boards must have a high static stability, which is influenced by the following main factors

  - hydrodynamic characteristics
  - length and tension of warps and
  - fixation coordinates of the doors.

• The reaction of trawl doors not exactly adjusted is sensitive to hydrodynamic forces created by the warps and this can led to an unstable run of the gear.

• The use of calculation and simulation programmes for the estimation of shape and tension of warps are useful tools to optimize deepwater fishing operations.

• Simplified rope calculations are not appropriate for deepwater trawling as more hydro-dynamic aspects arising from towing the year must be taken into account.

• It is strongly recommended to use different twisted warps on star board and port side of the vessel.

4. LITERATURE CITED

Hahlbeck, W.-H. 1976. Die Bewegungsgleichung eines ideal biegsamen, undehnbaren, strömungsbelasteten Fadens bei räumlicher ungleichförmiger Bewegung eines pelagischen 1-Schiff-Schleppnetzes Dissertation A, Universität Rostock.

Niedzwiedz, G. 1991. Neue theoretische und experimentelle Aspekte beider Ermittlung von Form und Belastung angeströmter, ideal biegsamer Fäden, Seile, Trossen und Rohre Vortrag und Script für das Internationale Schiffstechnische Symposium 1991 an der Universität Rostock, Fachbereich Maschinenbau und Schiffstechnik, Sektion F.

Niedzwiedz, G. 2000. Das EDV-Programm KABKURR - Berechnungsprogramm für Form, Belastung und Beanspruchung strömungsbelastete Trossen, Leinen und Kabel ggf. unter Berücksichtigung der Reibkräfte am Meeresboden Unveröffentlicht, Universität Rostock. 1991-2000.

Niedzwiedz, G. & M. Hopp 1998. Rope and net calculations applied to problems in marine engineering and fishery research Archive of Fishery and Marine Research 46 (2), 125-138,.

Niedzwiedz, G.1988. Model specification and further application of the calculation of rope systems Vortrag und Paper auf dem Internationalen Workshop DEMAT’01, 7-10.11.2001 in Rostock, Proceedings: Contribution on the Theory of Fishing Gears and Related Marine Systems, Vol. 2, ISBN: 3-935319-88-6.

Paschen, M., W. Köpnick, G. Niedzwiedz, U. Richter & H.-J. Winkel 1995. Contributions on the Theory of Fishing Gears and Related Marine Systems Proceedings of the 4th International Workshop on Methods for the Development and Evaluation of Maritime Technologies, Neuer Hochschulschriften verlag Rostock, ISBN3-929544-95-4.

Scheel, D. 1979. Die Berechnung der Durchhangsform und Zugkraft schwerer, strömungsbelasteter, fadenförmiger Elemente bei gleichförmiger Bewegung unter Berücksichtigung ihrer Dehnung und Steifigkeit Dissertation A, Universität Rostock.