ICES Journal of Marine Science: Journal du Conseil Advance Access originally published online on June 27, 2007
ICES Journal of Marine Science: Journal du Conseil 2007 64(6):1220-1234; doi:10.1093/icesjms/fsm084
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
In situ measurements of orange roughy (Hoplostethus atlanticus) target strength
1 Innovative Solutions Limited, PO Box 22-235, Wellington, New Zealand
2 Blue Mountains Road, Whitemans Valley, Upper Hutt, New Zealand
Correspondence to R. F. Coombs: tel: +64 4479 0151; e-mail: roger.coombs{at}isl-solutions.co.nz
Coombs, R. F., and Barr, R. 2007. In situ measurements of orange roughy (Hoplostethus atlanticus) target strength. – ICES Journal of Marine Science, 64: 1220–1234.Orange roughy (Hoplostethus atlanticus) support one of New Zealand's most valuable commercial fisheries, and its assessment poses many problems. Acoustic estimation using echo integration has become one of the main sources of biomass information, and for this an estimate of orange roughy target strength (TS) is needed. Its schooling characteristics together with patterns in the rate of change of phase vs. TS plots are used to identify ensembles of orange roughy targets from in situ TS data collected from a wide range of fishing areas off eastern New Zealand. The results suggest a TS of –49.3 dB for an orange roughy of 35 cm standard length.
Keywords: echo classification, hydroacoustics, orange roughy, target identification, target strength
Received 25 September 2006; accepted 10 April 2007; advance access publication 27 June 2007.
 |
Introduction |
|---|
 Top Introduction MethodsResultsDiscussionReferences |
|---|
Acoustic techniques, and in particular echo integration, have been widely used to assess the stocks of orange roughy (Hoplostethus atlanticus) off New Zealand, so an estimate of target strength (TS) is essential. However, orange roughy (hereafter referred to as roughy) have a low TS and live in depths of around 1000 m either in dense aggregations or dispersed among other species with similarly low TSs. Therefore, the well-known difficulties (e.g. Foote, 1991) of TS estimation are exacerbated. A particular problem for in situ TS estimation is identification of targets, and herein we use the "rate-of-change-of-phase" method of Barr and Coombs (2005) to classify targets and to estimate roughy TS.
Roughy are widely distributed around New Zealand in depths of 900–1200 m and support one of the country's most valuable deep-water fisheries. Most trawls in that depth range yield at least a few roughy, but the commercial fishery is based mainly on large aggregations which are found primarily on the North Chatham Rise (Anderson et al., 1998). These typically form on underwater hills during the spawning season in midwinter, but may also be present at other times of the year and over substrata without pronounced hills or slopes. Assessment of roughy populations posed a problem from the start of the fishery in the late 1970s. Trawl surveys were used initially, but were ineffective because of the large aggregations (Clark, 1996). Acoustic techniques were used from 1986, and these were more appropriate for aggregations, but had their own difficulties because of the great ranges involved, the often poor weather and the low TS of roughy (Do and Coombs, 1989).
The work described here was carried out as part of the assessment and management of New Zealand roughy stocks and was used to set quotas. For this, absolute biomass estimates were required by the New Zealand Ministry of Fisheries, with a target coefficient of variation of 20–30% (Doonan et al., 1999).
Target strength
Roughy TS is low relative to that of other teleosts of similar size, because its swimbladder is filled with wax esters rather than gas (Sargent et al., 1983). Wax esters are present in other parts of the body cavity too, and the bones are poorly ossified. Roughy also exhibit marked avoidance behaviour (Koslow et al., 1995; McClatchie et al., 1999), and they are generally mixed with other species with a wide range of TSs.
The first roughy TS estimates came from dead fish in a tank (Do and Coombs, 1989). Subsequent estimates were made by Elliott and Kloser (1993), in a "fishing-down" experiment, and Kloser et al. (1997), who made measurements on dead roughy in the sea and live roughy in situ. They also made TS estimates based on a simple model. Together, these yielded a wide range of values, and McClatchie et al. (1999) set out to narrow this by making measurements of live roughy in a tank. They produced a length vs. TS relationship for fish at surface temperatures and pressures with a tilt-angle distribution estimated from roughy in the ocean, using a still, stereo-camera system. This relationship was subsequently adjusted to roughy depths based on the properties of roughy oils (McClatchie and Ye, 2000).
The first in situ data were collected in the main roughy fishery on the Chatham Rise in 1998 in independent surveys commissioned by the New Zealand Orange Roughy Management Company (Kloser et al., 2000) and the New Zealand Ministry of Fisheries (Doonan et al., 1999). Kloser et al. (2000) used multifrequency capability to classify targets and selected echoes on the basis of types of echogram mark and implied fish behaviour. Doonan et al. (1999) used a conventional mode-matching approach, but concluded that small swimbladdered fish such as myctophids produced the main mode. Subsequently, Barr et al. (2000) proposed the use of phase information for classifying targets, particularly for differentiating between fish with gas-filled swimbladders and those without, such as roughy. Barr and Coombs (2001) applied this method to data collected between 1998 and 2000 on the Chatham Rise, and the results were used to adjust the McClatchie et al. (1999) relationship.
The use of phase was further developed by Barr and Coombs (2005), who proposed simplified models of fish to explain the sorts of features seen in TS vs. phase plots. In their analysis, Barr and Coombs (2001) used in situ data only from collection runs on the largest aggregations. They relied exclusively on TS vs. phase plots to classify echoes, and processed all data collected in each of the runs they analysed. Here, we use TS vs. phase plot together with signal characteristics, types of echogram mark, and differences between areas to classify targets and hence to estimate roughy TS.
|
Methods |
|---|
|
TopIntroductionMethodsResultsDiscussionReferences |
|---|
Rate of change of phase
Our main classification method characterizes targets in terms of both amplitude and phase (Barr and Coombs, 2005). The acoustic signal backscattered from a fish in water, Lbs(f), at a frequency f, differs from the incident signal in both magnitude, |Lbs(f)|, and phase, arg[Lbs(f)]. The TS is derived from the magnitude and is commonly defined (e.g. Medwin and Clay, 1998) as
|
| (1) |
From the phase component we define a descriptor, the "target rate of change of phase" (RCP), which is a measure of the rate at which the phase changes (in degrees per metre) across an echo relative to the transmitted pulse (Barr and Coombs, 2005). Plots of TS against RCP show distinctive patterns that result from the way the sound wave is reflected by the body of the fish. In particular, phase responses differ considerably depending on whether the fish has a gas-filled, oil-filled, or no swimbladder (Barr et al., 2000; Barr, 2001; Barr and Coombs, 2005). Barr and Coombs (2005) investigated these patterns using simple theoretical models of fish, and showed that roughy produce an arc-shaped feature (shown later).
TS–RCP or "complex" plots presented here are in the form of scatter diagrams if the number of points is small, or as contour plots of log10 numbers of targets if there are many points. The contour plots are either low resolution based on 1 dB TS and 10° m–1 RCP bins, or high resolution with 0.1 dB TS and 4° m–1 RCP bins, again depending on the overall number of targets. The plots therefore show the relative distribution of targets in complex TS space. Grouping the data in the TS plane will produce a conventional TS distribution.
Barr and Coombs (2001) classified targets using a template consisting of two concentric arcs delimiting an area containing 
95% of the roughy-like signatures in the Spawning Box 2000 data. The same template was used for all transects and for the Northwest (NW) Hills as well as the Spawning Box. Coombs (2004) fitted quadratic relationships to points within the arc region, but this gave variable results. The simulations in Barr and Coombs (2005) showed the response to be more a segment of a circle, with a well-defined upper arc and a less well-defined chord, and this pattern is used here. It was fitted by manually tracing the upper limit to obtain a set of defining points, then fitting an arc of a circle to them. A lower limit was then chosen which was typically either well below the echoes or close to the upper limit of any readily apparent, swimbladdered fish with a lower TS that were present. A horizontal line was then fitted passing through this point, limited in the phase plane by the limits over which the arc was traced. An example is given in the Results section. The mean TS for all echoes falling in this region was estimated by averaging their scattering cross-sections, |Lbs(f)|2, and all other mean TSs are averaged in this way.
Single target selection
As roughy data were all collected from in and around aggregations, filtering them to eliminate echoes from more than one target is essential. This is particularly so for targets of the same amplitude, but it is also desirable not to reject more than is necessary. Echoes were initially identified by locating peaks in the combined beam signal. The measurements from individual quadrants of the transducer were combined in pairs to calculate the position of the selected echoes in the beam (Ehrenberg, 1979), as well as the amplitude corrected for the position in the beam. The maximum amplitude of each echo was then estimated as described in the section on acoustic data below. Selection of echoes from individual fish used the approach detailed in Barr et al. (2000) and Barr and Coombs (2001), and was based on echo width, position in the beam, and variability of angle of arrival between different samples in the echo. Position in the beam was set to the transducer 3 dB beam width (±3.5°), to avoid applying too much correction to the amplitudes. The echo width for each echo identified was calculated at points on the received pulse 3 dB down on the peak amplitude. The acceptable range was set to reflect the length of the transmitted pulse, and the effects of using different widths are given later. Variability in the angle of arrival is intended to reject echoes that do not emanate from a consistent direction, and indicates that an echo is not from a single localized scatterer. Arrival angles in both the alongship and athwartship directions were estimated for different samples in the echo, and the standard deviation of these was used as the measure of variability.
Fish behaviour
There have been no formal studies of roughy behaviour, so the descriptions that follow are derived from the observations of fishers and fisheries scientists reported informally through RV reports and the New Zealand Ministry of Fisheries' deep-water working group meetings.
As noted earlier, roughy spawn in midwinter when they may form large characteristic aggregations, usually over underwater hills. However, aggregations may also form over gently sloping or undulating ground ("flat") and on hills at times outside the spawning season. Most of these hills are of volcanic origin, and most have a pronounced cone shape (Figure 1).
|
From a range of 500 m or more, large roughy aggregations on hills appear in echograms as a "cap" covering the top of the hill, and the difference in depth between the top and the bottom of the aggregation may be 100 m or more. Aggregations usually have a dense core, which at the fringes thins into striking horizontal or gently sloping "finger" structures (Figure 1). In these fingers, contiguous chains of echoes may extend for up to 600 m or more. Smaller aggregations may form a "lump" on the top of, or down the side of, the hill, often appearing as a "halo" partially or wholly surrounding the top. On flat ground, aggregations form "plumes", often with horizontal elements similar to the halo structures on hills (Figure 2). Fingers are present here as well, but they are usually not as prominent as those in large hill aggregations. These features were used to locate appropriate roughy aggregations for in situ data collection, and for this the transducer was positioned 50–100 m from the fish such that individual echoes could be resolved. However, this usually affects the behaviour of the fish so that schools become denser and their shape changes. Nevertheless, separate echoes that show finger structures usually surround dense marks.
|
Study areas and data sources
The main fisheries for roughy are on the north Chatham Rise and most acoustic-survey effort has been directed there. Fisheries on the east coast of the North Island are much smaller, but locally important, and several acoustic surveys have been carried out there too. Attempts were made to collect in situ TS data on all roughy surveys from 1997 on, but with varying degrees of success. The analysis here considered all in situ TS datasets collected, including some gathered during surveys for oreo on the south Chatham Rise in the roughy depth range. The areas from which data were collected are shown in Figure 3 and listed in Table 1 together with references describing the surveys.
|
|
Historically, the Spawning Box was the main fishing ground for roughy, containing many large plumes. The area has a largely undulating seabed dropping from a depth of about 650 m at the southern boundary to 1200 m in the north. During the surveys there, the plumes occupied a band between 800 and 900 m deep. The NW Hills is a complex of 28 underwater hills with summit depths of 750–1263 m and bases in depths of 1000–1600 m. Roughy occur in significant quantities on just nine of the hills, all within 6 km of each other. By far, the largest catches came from Graveyard (750 m) and Scroll (870 m). The 2002 NW Hills survey included trawling to map the density of roughy as they left Graveyard after spawning (Doonan et al., 2003a, "out-migration") and some in situ data were also collected then in marks on the western flanks of Graveyard.
There are six Northeast (NE) Hills, of which Smiths (895 m) and Camerons (784 m) have significant quantities of roughy, usually mixed with other species that have a much higher TS. The Eastern Hills are about 100 km to the south with 16 identified underwater hills. Roughy catches are much lower here. There is a chain of underwater hills along the eastern part of the south Chatham Rise, which in the early years of the roughy fishery, yielded large catches in the spawning season but are now only fished for smooth oreo (Pseudocyttus maculates). Hegerville (730 m) is one of these, and a large catch of roughy was made there on an aggregation showing characteristics typical of roughy during the oreo survey of 2001 (TAN0117). In the final area, the mid-East Coast, the fishing was initially based on hilltops, but is now on ridges and flatter areas away from the peaks.
Acoustic data and equipment
The in situ data were collected using a split-beam version of the "Computerised Echo Sounder Technology" (Coombs et al., 2003) developed by the New Zealand Institute of Water and Atmospheric Research (NIWA). In the system, signals received by the transducer are amplified, digitized, then fed to a digital-signal processor (DSP56002), where they are complex-demodulated, filtered, decimated, corrected for spherical spreading and sound absorption with a 40 log10R + 2
R time-varied gain, and stored. All four transducer quadrants (beams) are energized simultaneously from the transmitter, and each quadrant has its own receiver. The four receivers are synchronized to preserve the phase relationships between quadrants.
The output voltage, A, from the receiver for any fish is, from Equation (1), proportional to |Lbs(f)|. We have characterized this using the maximum amplitude, Amax, of a received pulse, estimated by locating the highest amplitude sample in a pulse envelope and fitting a parabola to it and to the points on either side, three points in all. The TS, in decibels, is derived from the square of the amplitude, with calibrations referenced to a tungsten-carbide sphere. Similarly, the signal phase, arg[Lbs(f)], changes across a received pulse as a consequence of the range to the target and, more important, the acoustic properties of the target. We fit phase change, 
, to a quadratic relationship with time, t, where
|
| (2) |
|
| (3) |
For a sequence of samples at t = –1, 0, +1, where t = 0 is the time of the maximum amplitude, Amax, the "rate of change of phase" descriptor, 'max, is simply b from Equation (3). Together, Amax and 'max describe a target in complex target space. The velocity of sound in seawater is used to convert the unit of RCP from degrees per second to degrees per metre.
The acoustic data were all collected from NIWA's RV "Tangaroa". From 1998 to mid-2000 (surveys A–C in Table 1), a flat-nosed, heavy-weight, deep-towed body fitted with a Simrad ES38DD split-beam transducer and electronics package intended primarily for biomass surveying (Coombs et al., 2003) was used. After mid-2000 (surveys D–G in Table 1), data were collected with a deep-towed "frame" system consisting of an open stainless-steel structure supporting an ES38DD transducer and electronics package, designed primarily for in situ TS work (Barr et al., 2002; Coombs et al., 2003). In all systems, short, single-frequency pulses of 0.32 ms (12 cycles) at 38.156 kHz were transmitted at intervals of 1.4 s.
During surveys A–C, the towed-body transducer signals were sampled at 100 kHz, digitally filtered (3 dB bandwidth 4.9 kHz) and decimated, to give a final sample rate of 10 kHz. The frame-system transducer signals were sampled at 125 kHz, and a 3.1 kHz bandwidth digital filter was applied. A decimated sample rate of 12.5 kHz was used on survey D, and 15.625 kHz on surveys E and F.
Data were collected while drifting or steaming at very slow speeds, with the towed body (or frame) deployed on a tow cable 2 km long. The height of the towed body above the seafloor varied with location between 50 and 150 m, and the speed of the vessel between 0.2 and 1.0 m s–1. Each deployment typically lasted 1 h, and the resulting acoustic data were stored for later analysis in the New Zealand Ministry of Fisheries' fisheries-acoustics database. Entries in this database are characterized by a voyage identifier (e.g. "TAN0008"), whether the data were collected by a towed vehicle ("towbody") or other means, and by a numerical file identifier.
The systems were calibrated following the approach described in Coombs et al. (2003), using a tungsten-carbide sphere of diameter of 38.1 mm ± 2.5 µm and a nominal TS of –42.4 dB as calibration standard. Calibrations were carried out in a seawater-filled tank at NIWA's Greta Point Laboratory before and after each survey, as well as at sea during surveys. The frame was usually deployed with a tungsten-carbide calibration sphere suspended beneath it, and several "deep-drop" calibrations were carried out during the various surveys. For the latter, the towed body or frame, with a calibration sphere hanging 10 m below it, was lowered in steps down the water column.
Catch composition and roughy length data
After each in situ acoustic-data-collection run, the marks seen on echograms were trawled to identify the fish and to evaluate their size composition. Trawling was carried out either by an accompanying commercial fishing vessel or by "Tangaroa". All trawling was with bottom trawls tailored to fish for roughy (Bull et al., 2000). The commercial vessels used a 100 mm mesh in the codend, and "Tangaroa" used both 100 and 40 mm meshes, the former typically on hills and the latter over flat ground. The relatively large mesh size used for sampling meant that few small fish such as myctophids were caught, generally entangled in the meshes; they might have been caught anywhere in the water column.
As many fish as possible were measured and, where practical, the whole catch. Fish were measured to the millimetre below using NIWA's electronic fish-measuring system. A typical roughy length frequency distribution, from the Spawning Box (Survey C) is shown in Figure 4, and mean lengths from all areas are listed in Table 2.
|
|
For most of the echogram marks used here, large catches of roughy up to 70 t or more were taken, and the proportion of roughy in the catch exceeded 90%. The main bycatch was smooth oreo in all areas around the Chatham Rise, plus Baxter's lantern dogfish (Etmopterus baxteri) on the NW Hills and Johnson's cod (Halargyreus johnsonii) on the NE Hills. On the mid-East Coast, there were also significant quantities of large swimbladdered fish such as Johnson's cod, cardinalfish (Epigonus telescopus), and alfonsino (Beryx splendens).
|
Results |
|---|
|
TopIntroductionMethodsResultsDiscussionReferences |
|---|
Single fish selection
The selection criteria were set using data from an extended transect from the 1999 Spawning Box survey over the bottom depth range 830–850 m, which had many large plumes of roughy. With no arrival-angle selection and with a wide range for echo width,
60% of all echoes detected were three samples long (
0.3 and <0.4 ms). When arrival-angle filtering was used, more echoes were three samples long, the proportion increasing rapidly as the arrival-angle standard deviation threshold was reduced. With heavy filtering, the widths of the calibration-sphere echoes were mainly three samples long with a width of 0.335 ms. Therefore, the data consisted largely of distinct pulses consistent with the transmitted-pulse length and, provided the range used included at least three samples, the upper limit was not critical. The echo-width range was therefore set to 0.3–0.6 ms. The variability in angle of arrival was similar in the alongship and athwartship directions. The standard deviation when only the three central samples in an echo were used was about twice that when all samples were included. Also, when only three samples were used the proportions in the frequency distribution dropped in an exponential fashion, whereas when all samples were used, they dropped linearly, as shown in Figure 5. The greater variation in the latter case we ascribe to the inclusion of low-level returns at the skirts of the pulse, where interfering echoes were more likely. On this basis, and for consistency with the peak and RCP fitting procedures, we used the standard deviation of the three central samples. The threshold level used was based on the response from a calibration sphere, and the effects on the complex target response. Calibration-sphere arrival angles showed a skewed response, with a mode at a standard deviation of 0.05° and a long tail out to higher values. Fewer than 1% of the arrival angles were >2.5°, and those returns all had complex values smaller than the rest. The effect of different arrival-angle thresholds is shown in the high-resolution complex plot of Figure 6, from which it can be seen that the overall pattern remained much the same, independent of the threshold chosen. The effect on TS distribution is shown in Figure 7; the main effect was to reduce the number selected and to define the mode in the region of the "roughy arc" more clearly. The overall pattern remained the same, and the positions of the modes did not change. As the choice of arrival-angle standard deviation did not seem to be critical, a value of 0.25° was used, based on the target-sphere response.
|
|
|
Vertical distribution
Figure 8 shows TS plotted against height above the seabed for two arrival-angle variability thresholds: the one on the left is with minimal selection, and that on the right 0.25°. The plots are contour maps of log10 relative numbers of targets based on 1-m height, and 0.5-dB TS bins. The numbers were compensated for the increase in beam volume with distance from the transducer. The left panel of Figure 8 highlights a change in distribution of scatterers at
50 m above the seabed, a feature also evident in echograms (Figure 2). At 50 m, there is a band spanning a wide range of TSs, and at 90 m another that is less evident. The higher level of filtering in the right panel tends to remove more echoes at longer distance from the transducer, because there are likely to be more overlapping echoes and the numbers of echoes generally reduce towards the seabed. The general pattern remains the same, but the two vertical bands of targets are more pronounced, as is the band of echoes close to the seabed, which contains the scatterers of highest TS. The vertical band centred on a TS of –48 dB corresponds to the roughy arc (Figure 6), and the other, centred on a TS of –54 dB, corresponds to the cluster of echoes below the arc in Figure 6, which Barr and Coombs (2005) ascribed to fish with small swimbladders, such as myctophids.
|
Figure 9 shows high-resolution complex plots for echoes <70 m and
70 m above the seabed. Most of the low TS scatterers are shallower than 70 m and have a positive RCP displacement, forming a tilted ellipse, as in Feature 4 of Figure 15 (later). These patterns are not evident in the lower layer, and the roughy-arc echoes are more diffuse.
|
All files selected show some equivalence to the bottom layer of Figures 2 and 9, although the height above the seabed varies, and because this contains most of the –50 dB targets, a height cut-off based on this layer structure was used for all subsequent analyses.
Temporal and spatial variation
In situ data were collected on Graveyard in 1999 and 2002 and on the Spawning Box in 1998 and 2000, allowing temporal comparisons to be made. Figure 10 shows data for Graveyard for the two years, selected to be comparable, and the patterns are obviously almost identical. For the Spawning Box, the areas and depths covered differed slightly between the two years, and the towed-body depth was more variable in 1998, although the patterns were still similar and had the same principal features.
|
Within the Spawning Box, the large roughy aggregations were predominantly in the bottom depth range 850–870 m. Most transects in 2000 ran along these contours, but one transect ran at right angles up the slope from about 910 to 820 m. This had heavy plumes around 860 m and lighter traces running in to the shallower water, but few roughy marks at the deep end. Figure 11 shows low-resolution, complex plots for the deep, plume, and shallow sections of the transect, and it is clear that the roughy arc was most pronounced in the plume section, whereas the deep section had the most prominent myctophid (TS –54 dB) and large-swimbladdered fish (TS –45 to –38 dB) patterns.
|
In the NW Hills area, in situ data were collected on four of the hills with differing summit depths, and Figure 12 shows low-resolution, complex plots for these, with data selected to make the plots comparable. Graveyard and Morgue were similar, except that Morgue, which is deeper, had many more "high target-strength" scatterers (TS –45 to –35 dB). Scroll and Zombie both showed myctophid patterns at around –54 dB and larger fish in the TS range TS –45 to –40 dB. These hills have the shallowest bases, but tops which are 100 m deeper than Graveyard.
|
The data collected on the flanks of Graveyard during the out-migration show quite a different picture. The transect crossed a marked change in slope of the seabed, on the high side of which was an aggregation with roughy characteristics. Figure 13 shows a low-resolution complex plot and a TS histogram for the aggregation. Most scatterers had a TS close to –50 dB, but there was a distinctly separate group of much lower TS scatterers centred on –65 dB, so producing a bimodal TS distribution.
|
Figure 14 shows low-resolution complex plots for representative transects from each of the areas considered, plus an area on the south Chatham Rise ("Flat" in Figure 3) where roughy catches were small and there was no roughy arc. The data were selected to be <70 m above the seabed. The three panels on the left are from hills and the three on the right from flat areas. The non-roughy area (Flat) stands out as very different from the rest, whereas the two top panels, which represent the sites of largest roughy catches (Graveyard and the Spawning Box), have the most distinct roughy arcs. However, Hegerville, with fewer data points, was quite similar to the Spawning Box. As noted above, some of the NW Hills had myctophid patterns (Figure 12) and were similar to the NE Hills (Camerons). Camerons also has some very high TS scatterers (TS around –30 dB), which are probably Johnson's cod, whereas both Camerons and Rock Garden (mid-East Coast) had significant numbers of very low TS echoes (–80 to –70 dB). There were no consistent differences between the hills and the flat roughy areas. The most important difference in these datasets in terms of separating roughy from other scatterers is the absence of the myctophid mark on Graveyard.
|
TS-RCP patterns
Most of the plots in Figures 10–15 feature a roughy arc spanning an RCP range of –100 to +100° m–1 and a TS range of –52 to –46 dB. The number of echoes needed for an arc to form varies depending on the number of non-arc scatterers present, but generally at least 500 are required (Table 3). The end of the arc on the right usually shows some sign of a superimposed ellipse (see below). Most show a strong myctophid signal. However, by far the most clear-cut arc patterns were in the Spawning Box plumes and on Graveyard among the NW Hills, where there were fewer other scatterers. Figure 15 is a high-resolution complex plot of the unselected plume data with superimposed ellipses highlighting the main non-arc features. At least one feature is present in all plots, although the positions vary slightly. Feature 3 is better defined in some areas than others, and clearly overlaps with the roughy arc. Barr et al. (2000) thought that this mark might also be roughy, and it is not obviously present in the non-roughy area.
|
|
TS estimates
Data used for estimating TS were taken from 70 m or less above the seabed, and only from those echogram marks showing roughy characteristics and where the associated trawls caught a high proportion of roughy. The effects of the possibility of the presence of non-roughy targets in the roughy arc, particularly in the right half (Figure 15), were investigated by estimating TS using data from the whole arc, as compared with only the left part. If a type-3 feature (Figure 15) was strongly present, the arc was stopped at its boundary. Figure 16 shows an example of a selected roughy arc for the latter case. The data are from Graveyard (survey B).
|
When displayed with the resolution in, for example, Figure 15, the upper arc is quite clear, but the precision in locating it is not high. Higher resolution plots such as Figure 16 allow greater precision in placing the arc, but it is less clear where it should go. For datasets with a clear arc, provided the point density was such that the arc was not obscured by non-arc echoes, tracing was repeatable, particularly from scatter diagrams. Ten repeat roughy-arc classifications were carried out on the long Spawning Box transect and also on one from Graveyard. These gave the results presented in Table 4, which shows mean TSs and associated standard deviations for the whole arc, the left half, and the high point in turn. There is only a small difference (0.15 dB) between the whole and half arcs for the Spawning Box, and even less (0.01 dB) for Graveyard. However, the latter omitted the far right of the arc (Figure 16). On the basis of this result, the whole arc was used, because it yielded more points, but type-3 features were not included.
|
Summarized results for the locations meeting the selection criteria are listed in Table 3. Although the mean lengths of the roughy do not vary much, the TS estimates cover a range of 2 dB. For the two best areas in terms of matching the criteria, the Spawning Box plumes and Graveyard, the TS estimates differ by >1 dB. No obvious relationships stand out in Table 3, and although TSs tend to be greater in deeper locations (Table 3, and Figures 12 and 14), the trend is not statistically significant.
|
Discussion |
|---|
|
TopIntroductionMethodsResultsDiscussionReferences |
|---|
A weakness of much in situ TS work is that if the length and the species of the individual fish is not known independently, there is always a substantial element of subjectivity or arbitrariness in deciding which echoes to use. This is true for all roughy in situ estimates. Barr and Coombs (2001) used a fixed template of two concentric arcs, as described earlier. A pair of arcs was chosen subjectively and then applied, without change, to all datasets. Working with TS only, Kloser et al. (2000) chose essentially arbitrary limits over which to average TS, and Kloser and Horne (2003) examined a range of limits directly from their data and by modelling, but their final choice of limits was inevitably somewhat arbitrary. With this in mind, we attempted to make our approach as objective as possible, but although we think that our approach is not arbitrary, it is undoubtedly subjective, because fitting the roughy arc to TS–RCP plots is a pattern-recognition problem, and it is still the case that the human brain is much better at picking out patterns than any automated process.
The TS result for the Spawning Box plumes is similar (0.2 dB higher) to that of Barr and Coombs (2001), averaged over all the data. For NW Hills, their average was 50.1 dB, and the average for the subset of the hills in Table 3 is also 50.1 dB, so although Graveyard alone is lower at –50.54 dB, the present results are also consistent with the earlier ones. The estimates from the other areas are all new.
The rationale in the approach taken here is that there "must" be echoes from roughy among those collected from in and around roughy aggregations and, in this respect, the Spawning Box data are best. The catches were almost entirely roughy, the aggregations extended for up to 6 km, and the towed body was far enough away from the fish for there not to be any obvious diving reaction. However, the negative side of this is that overlapping echoes are more likely, and after single-target selection, very few echoes remain from the aggregations themselves, most coming from finger traces at the fringes. Although the finger traces are characteristic of roughy schools, there is no guarantee that they are actually roughy, and could be an associated species. The individual echoes making up these traces vary considerably in TS, but so potentially do those of roughy, as shown in Figure 5, and by the tilt-dependence figures in McClatchie et al. (1999). In contrast, on the hills, there was either only a small contact zone with the fish, perhaps a few hundred metres, or the fish were driven down the hillside, exhibiting a marked avoidance reaction. McClatchie et al. (1999) simulated this situation and showed that a reduction in TS of up to 3 dB was possible (Kloser et al., 2000).
As noted earlier, the distinctly separate myctophid pattern is missing from Graveyard (including the out-migration), Morgue (Figure 12), and Hegerville, but strongly present in the Spawning Box and to some degree in all other datasets. The mean TS of the myctophids ranges from –54.0 to –52.7 dB. This placed a lower limit on the range over which echoes were averaged, which was higher than the limit placed when there were no myctophids, as in Figure 16. Both Hegerville and the out-migration show strongly concentrated marks in the roughy arc, and perhaps these are small fish with swimbladders. However, the pattern is also consistent with them being roughy; see the simulations in Barr and Coombs (2005).
The out-migration transect close to Graveyard showed many roughy features, including fingers, although these were not as well formed as on Graveyard summit. However, the aggregation was less compressed than a typical roughy mark, which allowed many individual echoes to be resolved. It has a much more limited range of scatterers than any of the other transects, and the TS distribution shows a narrow, clearly defined, roughy-arc mode, together with a lower, broader mode from low-TS scatterers (Figure 13). The mark has similarities with Hegerville, and again could possibly be small fish with swimbladders. The TS estimated is 1.7 dB higher than for the same fish on the summit of Graveyard. However, as shown in Figure 13, there are two distinct parts to the mark, a lower dense part and a right-tending upper part. If the arc is positioned below this, the TS drops to –49.42 dB. Ignoring echoes closer to the bottom than 20 m has a similar effect. This TS is still a lot higher than for the summit, but now about the same as for the Spawning Box, suggesting that the flight reaction depresses the TS estimates for the summit of Graveyard.
It seems certain that the roughy in all these data are mixed to a greater or lesser extent with other fish and scatterers of similar TS. In addition to fish with swimbladders of various size, there are likely to be non-fish targets, such as siphonophores with air bubbles (Barr and Coombs, 2005) and other fish without swimbladders, such as Baxter's lantern dogfish, which featured in most catches. The fish and other scatterers that fall outside the roughy arc in the TS–RCP plots vary significantly between the different areas, and it is likely that these differences are also reflected within the arc. This may explain some of the variation in our estimates of roughy TS from the different locations.
Kloser and Horne (2003) discuss uncertainties in measuring roughy TS and present in situ results suggestive of mean TSs lower than those estimated here. Their results were based on tracking individual fish at the periphery of large aggregations, identified as roughy by comparing responses at 38 and 120 kHz (Kloser et al., 2002). Their measurements and modelling suggest a maximum TS of –46 dB for a 35-cm roughy, and a range for the mean of –52.9 to –51.0 dB. The maximum is reasonably consistent with the upper limits of the arcs fitted here (Table 3), and the higher of the mean values is also consistent with our results. As with "rate of change of phase", the frequency comparison method does not allow individual fish to be identified, and the mean values estimated for the tracked fish were constrained by a subjectively chosen upper limit.
Another source of variability in roughy TS comes from the way that sound is reflected by the fish. The response of roughy depends very much on the interaction of the reflections from the dorsal and ventral surfaces of the fish and from the oil-filled swimbladder, and these in turn depend on the size of the fish, its tilt angle, and the offset angle of the swimbladder. The result of this is that the relationship between length and TS is of an oscillatory nature (Barr, 2001). However, the ripples tend to be smoothed out by variations in tilt angle, as shown in Figure 17. The solid oscillating line in that figure is the response of fish with fixed tilt angles of –7° and swimbladder offset angles of –18°, comparable with values from McClatchie et al. (1999). The dots show the results of varying the tilt angle sampled from a normal distribution with mean –7° and s.d. –19° (McClatchie et al., 1999). The crosses are the scattering cross-section means of 1-cm length groups of the dot data. A TS–RCP plot of the dot data shows a diffuse arc, such as that in Figure 16. To duplicate the more tightly clustered arcs in the Spawning Box data (e.g. Figure 6), Barr and Coombs (2005) used a tilt angle of 0° with a 2° standard deviation and a swimbladder offset of 12°. With less variability, more of the oscillating structure is retained (Figure 18).
|
|
The mean TS values from Table 3 are plotted against length in Figure 18, together with the results for live roughy held in a tank from McClatchie et al. (1999), adjusted as in McClatchie and Ye (2000), and the two mean values defining the range from Kloser and Horne (2003). Also shown are the two relationships currently used in New Zealand roughy stock assessments. The upper solid line is based on the live roughy measurements adjusted by the in situ results of Barr and Coombs (2001), where TS = 16.15(log10L) – 74.27 and L is standard length. The lower, lighter, solid line has the same slope as the upper, but the constant term is based on the lower mean value from Kloser and Horne (2003). Two modelled relationships are also included: the dot–dash line is the mean TS from Figure 17 based on observed tilt-angle distributions from McClatchie et al. (1999), and the dotted line is based on tilt angles from Barr and Coombs (2005). The modelled and live-roughy results are averaged with respect to tilt angle only and are directly comparable; the in situ results are, in addition, averaged with respect to length and will tend to further smooth the underlying oscillations. From Figure 18, it is clear that despite the underlying oscillatory nature of the TS response, the conventional log-linear relationship does approximate the likely response in the sea. However, it also suggests that wide variations in in situ measurements are to be expected, and that large numbers of measurements are desirable.
What value should be used for TS in stock assessment? All estimates, except that for the out-migration, suffer from the shortcoming that the fish are mostly from the fringes of the aggregations surveyed by echo integration. Roughy schools are structured, and the sex ratio can vary greatly between the catches from different trawls, so fish at the fringes of the school may well be atypical. In this respect, stirring them up by driving them down a hillside may yield better results. As argued above, the best estimate is that from the Spawning Box, i.e. a TS of –49.3 dB for a fish of 35.1 cm standard length.
The oscillatory nature of the underlying response means that the TS is highly dependent on fish behaviour, and both live-fish measurements and models make explicit, and in situ results tacit, assumptions about this in taking the behaviour of the fish involved to be typical. This underlines the uncertainty in all published estimates of roughy TS. The length frequencies in all the roughy areas in which acoustic biomass surveys have been carried out are similar (Table 2, Figure 18), so rather than using differing values for different areas, a reasonable approach might be to use –49.3 dB for them all.
As a postscript, the acoustic system used in this study was explicitly designed to preserve phase information, because it had been found of great value in similar areas of research, such as radiophysics and geophysics. The work described here shows that it has comparable value in fisheries acoustics. However, there are currently no commercially available fisheries-research echosounders that preserve phase, and we feel that this would be a valuable addition to the design of the next generation.
|
Acknowledgements |
|---|
The fieldwork described here was carried out while the authors were with the New Zealand Institute of Water and Atmospheric Research and was funded by the New Zealand Ministry of Fisheries under many roughy and oreo projects. Many people were involved in the data collection, and we thank them all. We are also grateful to various referees for their valued comments. The analysis was funded by the Ministry of Fisheries under a general services contract with Innovative Solutions Limited.
|
References |
|---|
|
TopIntroductionMethodsResultsDiscussionReferences |
|---|
-
Anderson O. F., Bagley N. W., Hurst R. J., Francis M. P., Clark M. R., McMillan P. J. Atlas of New Zealand fish and squid distributions from research bottom trawls. (1998) 42:303. NIWA Technical Report.
Barr R. A design study of an acoustic system suitable for differentiating between orange roughy and other New Zealand deep-water species. Journal of the Acoustical Society of America (2001) 109:164–178.[CrossRef][Web of Science]
Barr R., Coombs R. F. In situ target-strength measurements and chirp responses of orange roughy (Hoplostethus atlanticus). (2001) Final Research Report for Ministry of Fisheries Project ORH1999/01A (Unpublished report held by the Ministry of Fisheries, Wellington).
Barr R., Coombs R. F. Target rate of change of phase: an extra dimension for fish and plankton target identification. Journal of the Acoustical Society of America (2005) 118:1358–1371.[CrossRef][Web of Science][Medline]
Barr R., Coombs R. F., Macaulay G. Can we differentiate between different deep-water fishes using a standard acoustics target-strength ping. Bradley S., ed. (2000) 27 11–1 12. Proceedings of the 10th International Symposium on Acoustic Remote Sensing of the Atmosphere and Oceans and Associated Techniques. Auckland, New Zealand: The University of Auckland. 2000.
Bull B., Doonan I., Tracey D., Coombs R. F. An acoustic estimate of orange roughy abundance on the Northwest Hills, Chatham Rise, June–July 1999. (2000) 36. New Zealand Fisheries Assessment Report 2000/20.
Clark M. R. Biomass estimation of orange roughy: a summary and evaluation of techniques for measuring stock size of a deep-water fish species in New Zealand. Journal of Fish Biology (1996) 49((Suppl. A)):114–131.[CrossRef][Web of Science]
Coombs R. F. Target strength and counts of orange roughy in the background layer, Northwest Hills area, Chatham Rise, June–July 2002. (2004) Final Research Report for Ministry of Fisheries Research Project ORH2001/01, Objective 1, January 2004 (Unpublished report held by Ministry of Fisheries, Wellington).
Coombs R. F., Macaulay G. J., Knol W., Porritt G. Configurations and calibrations of 38 kHz fishery acoustic-survey systems, 1991–2000. (2003) 24. New Zealand Fisheries Assessment Report 2003/49.
Do M. A., Coombs R. F. Acoustic measurements of the population of orange roughy (Hoplostethus atlanticus) on the North Chatham Rise, New Zealand, in winter 1986. New Zealand Journal of Marine and Freshwater Research (1989) 23:225–237.[Web of Science]
Doonan I. J., Coombs R. F., Barr R., McClatchie S., Grimes P., Hart A., Tracey D., et al. Estimation of the absolute abundance of orange roughy on the Chatham Rise. (1999) Wellington): Unpublished report held by Ministry of Fisheries. Final Research Report for Ministry of Fisheries Research Project ORH9701, October 1999.
Doonan I. J., Coombs R. F., Hart A. Acoustic estimates of the abundance of orange roughy on the Northwest Chatham Rise, ORH 3B, June–July 2002. (2003a) 23. New Zealand Fisheries Assessment Report 2003/58.
Doonan I. J., Hicks A. C., Coombs R. F., Hart A. C., Tracey D. M. Acoustic estimates of the abundance of orange roughy in the Mid-East Coast fishery, June–July 2001. (2003b) 22. New Zealand Fisheries Assessment Report 2003/4.
Doonan I. J., McMillan P. J., Hart A., Coombs R. F. Smooth oreo abundance estimates from the October–November 2001 acoustic survey of the south Chatham Rise (OEO4). (2003c) 21. New Zealand Fisheries Assessment Report 2003/26.
Ehrenberg J. E. A comparative analysis of in situ methods for directly measuring the acoustic target strength of individual fish. IEEE Journal of Oceanographic Engineering (1979) OE-4:141–152.[CrossRef]
Elliott N. G., Kloser R. J. Use of acoustics to assess a small aggregation of orange roughy, Hoplostethus atlanticus (Collett), off the eastern coast of Tasmania. Australian Journal of Marine and Freshwater Research (1993) 44:473–482.[CrossRef]
Foote K. G. Summary of methods for determining fish target strength at ultrasonic frequencies. ICES Journal of Marine Science (1991) 48:211–217.
Kloser R. J., Horne J. K. Characterizing uncertainty in target-strength measurements of a deep-water fish: orange roughy (Hoplostethus atlanticus). ICES Journal of Marine Science (2003) 60:516–523.
Kloser R. J., Ryan T. E., Sakov P., Williams A., Koslow J. A. Species identification in deep water using multiple acoustic frequencies. Canadian Journal of Fisheries and Aquatic Sciences (2002) 59:1065–1077.
Kloser R. J., Ryan T. E., Williams A., Soule M. Development and implementation of an acoustic survey of orange roughy in the Chatham Rise spawning box from a commercial factory trawler, FV "Amaltal Explorer". (2000) Deep Water Working Group Document 00/34 (Unpublished report held by Ministry of Fisheries, Wellington).
Kloser R. J., Williams A., Koslow J. A. Problems with acoustic target-strength measurements of a deep-water fish, orange roughy (Hoplostethus atlanticus, Collett). ICES Journal of Marine Science (1997) 54:60–71.
Koslow J. A., Kloser R. J., Stanley C. Avoidance of a camera system by a deep-water fish, the orange roughy (Hoplostethus atlanticus). Deep-Sea Research I (1995) 42:233–244.[CrossRef]
McClatchie S., Macaulay G., Coombs R. F., Grimes P., Hart A. Target strength of the deep-water fish, orange roughy (Hoplostethus atlanticus). 1. Experiments. Journal of the Acoustical Society of America (1999) 106:131–142.[CrossRef][Web of Science]
McClatchie S., Ye Z. Target strength of the deep-water fish orange roughy (Hoplostethus atlanticus). 2. Modelling. Journal of the Acoustical Society of America (2000) 107:1280–1285.[CrossRef][Web of Science][Medline]
Medwin H., Clay S. C. Fundamentals of Acoustical Oceanography. (1998) London: Academic Press. 712.
Sargent J. R., Gatten R. R., Merrett N. R. Lipids of Hoplostethus atlanticus and H. mediterraneus (Beryciformes: Trachichthyidae) from deep water to the west of Britain. Marine Biology (1983) 74:281–286.[CrossRef]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||