Efficacy of a remote control closure of cod-end used in a bottom trawl experience *

Bottom trawl surveys are widely used for monitoring demersal stocks when abundance indices are required (Sparre et al., 1989; Fiorentini et al., 1999). Achieving precise quantification is one of the major problems affecting resource assessment and the different dynamic models employed strive to accomplish this both biologically and statistically (Conquest et al., 1996; Pennington and Vølstad, 1991; 1994; Pelletier, 1998). Furthermore, vulnerability of a given species to gear is affected by its behavioural characteristics such as aggregations or vertical distribution patterns (Parrish et al., 1964; Svatimskij, 1985; Laevastu and Favorite, 1988). Technology endeavours to design fishing systems that will allow more exact quantification of catches through precise measurement of the swept area and avoidance of sample contamination by incidental capture of species during shooting and hauling across the water column. Hydroacoustic systems provide information remotely on the horizontal and SCI. MAR., 66 (4): 423-432 SCIENTIA MARINA 2002

vertical opening of the gear, the distance between the wings and otter doors, towing speed, etc.These recent technological innovations represent significant advances in resolving the former problem.The problem of sample contamination has been solved for fixed mouth systems like those used in plankton tows but for the time being it has not been solved for trawling, and remotely operated closure systems are still in the developmental stage.Pennec and Woerther (1994) designed a system for dividing the cod-end of trawls into separately closeable sections or compartments with a view to making it possible to establish sample origin at different points in time during tows.Opening or closing the gear during shooting and hauling enables relatively undesirable species to be avoided.One of the goals for developing this remote closing system was to solve the problems posed by the echo-integration of shoals being able to correlate the echograms from the sounders to the species of fish detected.To this end the sounder data is recorded at the same time as trawling in order to identify different species composition along the transects by closing the system at different time lapses.This would permit various tows to be carried out without hauling the net up from the fishing ground, thereby achieving considerable time savings during research, which is useful when large numbers of samples are needed (Pen-nington and Vølstad, 1991;Folmer and Pennington, 2000).From this starting point, the application of the remote closing system to bottom trawling could be useful for different purposes (replicates, studies on resource patch distribution, avoiding specific biological contamination of samples, etc.).This would be the case, for instance, of trawl surveys using such interpolation methods as kriging when distances between sampling points are short and when we must consider as the sampling point the geographical centre of the tow.A sequence of trawling, hauling, closing the net, shooting at few meters from the bottom, and trawling again would be another application of the system in order to saving time during trawl surveys.
The object of the present study was both to test the technical reliability of the remote control trawl closure system and to ascertain whether catches in the different sections into which the cod-end was divided after closure were similar, in order to determine whether the sub-samples can be regarded as homogeneous replicates.Other aspects linked with sampling and catchability are also discussed jointly with the advantages and disadvantages of applying the closure system to trawl surveys.However, the variability inherent in trawl sampling or catchability due to gear efficiency falls beyond the scope of this paper.

The remotely controlled trawl cod-end closure system
The Genisea Society manufactured in 1996 the remotely controlled trawl cod-end closure system (Fig. 1) used in this study.There are three components.The first is a remotely operated trawl cod-end closure unit.Driven by an acoustic signal sent from the vessel by a transducer, the unit releases a spring that frees a parachute attached at the end of a rope.Water drag on the parachute draws out the rope, tightening a noose around the cod-end at the position where the system has been deployed, thus strangling the net at that point.Depending on the number of units employed, the cod-end may be closed off into various independent sections to select sampling water volumes or depth intervals along the trawl trajectory.The second is a transducer attached to a depressor-wing placed about 20 m below the water surface.This transducer emits the acoustic signal towards the trawl closure unit, which when activated draws the cod-end closed.The third is an acoustic remote control unit on board supplied with electrical power from the vessel.It sends the signal to the transducer and is equipped with different channels for each of the optional trawl closure units.

The gear
The trawl used in this experience is the common commercial trawl net used in the region (Catalan coast), with a headrope of 25 m and a total length of 27 m (see Sardà et al., 1998 for further details).The trawl net was operated with a pair of flat, rectangular, iron otter boards measuring 1.20 x 2 m and weighing 400 kg with bridles of 75 m.Scanmar remote control measures give a mean of 13.5 m of horizontal opening and 1.9 m of vertical opening.The cod-end used was of 40 mm stretch mesh with a lifter inside of 12 mm.

The sampling
The experiment was carried out on board the R/V García del Cid (38 m in length, 1200 HP).In all, 26 hauls were performed using the remotely operated trawl closure system, 16 during the "Nerit I" survey (September 1999) and the remaining 10 during the "Nerit II" survey (June 2000).During "Nerit II", 25 other hauls were carried out without the trawl clo-sure system at the same depths and in the technical conditions.In order to avoid depth variability in the samples, a specific station around 107 m depth (106 minimum and 108 m maximum) in a single flat fishing ground off the coast of Catalonia was chosen (40º47'N, 1º12'E; western Mediterranean off the delta Ebro river).The same gear was used during the complete progress of the experiment.Tow duration was 90 min, with closure of the cod-end after 45 min.The remote closure system was placed on the middle of the cod-end (2.5 m from the end of the cod-end).In this way the closure system was perfectly able to strangle the net.During the "Nerit I" cruise, the system was used after the standard instructions provided by Genisea (1996).However, due to its bad functioning (see Table 1 for details), it was necessary to discard these hauls for further data analyses.Only one remotely operated trawl closure unit was used during both surveys, in such a way that the cod-end was only closed off into two sections: a rear section and a front section.The number of individuals per species in each cod-end section was counted.In order to compare the effect of the remote closure system in the trawl, the other 25 hauls without a closing system were used in the data processing.All experiments were performed continuously during four 24 h cycles.
It was assumed that species contamination during shooting or hauling of the net was negligible.There were several factors that corroborated this assumption: 1) null presence of epipelagic species in the samples; 2) the mouth spread of the net during shooting is around 7 m because the trawling speed in this moment is low (measured with Scanmar system it represents 50% of the optimum spread during hauling); 3) the trawl sampling was conducted at 107±1 m; at this depth the gear delayed less than 10 minutes to reach the bottom, a very short time period in comparison with the 45 minutes of effective tow duration over the bottom; and 4) during hauling, the gear ascends in vertical position and the vessel is stopped, which means that the fishery capability is negligible.

Data analysis
The catches were standardised as number of individuals per square mile, calculated on the basis of the swept area estimated from the data recorded by a Scanmar remote sensing system deployed in the mouth of the trawl gear (horizontal spread between wings) and the distance covered by the haul (starting and ending geographical GPS positions).With these data, a matrix by species, haul, and cod-end section was constructed.All the species were considered, except those for which no more than a single individual was recorded per haul or species that exhibit many zeros because to yield high variability (Pelletier, 1998).In order to analyse sampling efficiency, individual species were graphically represented in percentages.If the closing system is efficient, the proportion would have been expected to approach 50% between species from front and rear cod-end sections.In order to test differences of all species from 50% individually, a G-test with William's correction was performed (Sokal and Rohlf, 1981).
To compare hauls carried out with and without the trawl closure system, the complete matrix of catches was employed for processing the seven hauls performed with the closure system and the 25 hauls performed without the closure system on the same cruise under the same working conditions ("Nerit II").To determine the importance of variability between all these catches due to the behaviour of the species, a hierarchical cluster analysis was performed using the linear correlation index and the UPGMA aggregation algorithm after ln (n+1) data transformation (Lleonart and Roel, 1984;Pennington, 1996).

Technical aspects
The operation of the remote trawl closure system encountered certain practical difficulties during the "Nerit I" survey, and the system worked correctly in just one haul.Table 1 gives the details for all hauls and indicates the incidences of closure system operations.The main problem was the failure of the remote trawl closure unit to work properly, for a variety of reasons: a) the trawl closure unit exhibited a tendency to rotate getting clogged by mud due to its re-placement under the net; consequently, it failed to receive the closing signal; b) the system worked but failed to close with sufficient strength to strangle the cod-end, and hence the species frequently passed through the strangled knot and appeared mixed together on the rear section of the cod-end, without any separation between the two sections of the cod-end; and c) the system worked properly, but the rope or the parachute fastener ring broke.As a consequence of these operational difficulties, the "Nerit I" hauls were disregarded for further analysis.
During the "Nerit II" survey, a twin system to the previous one was used.Thus, the following modifications were made: a small float was placed in the trawl closure unit to improve buoyancy (Fig. 1) and prevent blockage by mud; the rope was shortened and the parachute fastener rings were strengthened and attached by a swivel; the original parachute was replaced by a stronger one, and a piece of cork was attached to the cover of the housing of the unit to enhance detachment and buoyancy upon release.With these modifications, the effectiveness of the system operation improved markedly and worked well in 70% (7 of 10) of the experimental trawls.Therefore, only the data collected on the "Nerit II" survey, after these technical improvements had been made, were used in the data analysis.
-Technical success of experimental hauls employing the remotely operated trawl cod-end closure system from Genisea during the cruises "Nerit I" (with standard closure system) and "Nerit II" (with modified closure system).Explanation of the outcomes in Material and Methods. NERIT

Catch analysis
Tables 2 and 3 show the species density on hauls with and without remote closure system.The comparison between species in the hauls using remote closure system revealed clear differences between the two cod-end sections (Gtest, p<0.05), and very few species occurred in the same proportion between the rear and front cod-end sections (Table 2; indicated in bold).The line separating the species caught in the front and rear sections of the cod-end can be observed to vary from 50% in a completely irregular manner (Fig. 2).Percentage shares by species were also observed to vary from haul to haul, for example, for species nos. 3 (Boops boops), 10 (Helicolenus dactylopterus), 27 (Pontocaris lacazei) and 28 (Solenocera membranacea), revealing erratic patterns for species occurrence in the front and rear sections of the codend, not only in the same haul but also among the different hauls.That is, both inter and intra-haul variability exists.Even the hauls conducted at the same hour from correlative days had significantly different results between species of the haul pairs: 40-49; 40-57, 49-57 and 41-50 (G-test; p < 0.05).In some cases certain species were observed to occur only in the front or the rear cod-end compartment.This was true for several benthopelagic species, such as Boops boops and Trachurus spp., but it was also an unexpected finding for such typically benthic species (Table 2) as Nephrops norvegicus (trawl P40), Symphurus nigrescens (P40), Pontocaris lacazei (P41), and Eledone cirrhosa (P42).Some species tended to show great variability, which could be indicative of non-homogeneous spatial distributions of the individuals of these species.Helicolenus dactylopterus, Ophidion barbatum, and Solenocera membranacea are three examples of this latter case.
The cluster analysis (Fig. 3) showed an initial segregation into two major groupings, i.e., day-time and night-time hauls, with different similarity values.Within these groupings, the hauls carried out using the remotely operated closure system were clustered separately from the other hauls.In the daytime group, hauls P40, P49, P41 and P57, formed a sub-group separated from the other day catches.In the night-time group, hauls P42 and P50 also  appeared in the same group, though with a lower level of similarity.Only haul P56 was in the group with the rest of the hauls.Then, it is demonstrated that day-night variability due to the benthopelagic behaviour of some species influences catches more than general catch variability or the effect that the remote closure system can produce on the working characteristics of the gear during the haul, in the conditions in which the experiment was performed.

DISCUSSION
After suitable modifications to the remotely operated closure unit had been made, the operational success rate was 70%.We expect that, with additional modifications, higher values of technical success can be achieved.However, the system still requires too careful handling bearing in mind that fishing manoeuvres call for systems capable of withstanding the heavy jolts that come with hard, fastpaced work.
Two hauls, P48 and P51, had no catches in the front section of the cod-end.The conclusion drawn from these results is that the system was sometimes not strong enough to fully strangle the net or that the parachute did not open completely and hence failed to prevent displacement of the catch back into the rear of the cod-end.If system operation is still problematic for a small gear like the one used in our experimental hauls, effectiveness can perhaps be expected to be even lower in a larger gear with a full cod-end towed at higher speed.
Fishery stock assessment is often based on data of abundance indices found from trawl surveys aimed at specific target species.These indices are taken from different kind of vessels with different fishery capabilities.This made it necessary to conduct studies on trawl and vessels inter-calibrations (Svatimskij, 1985;Gordon and Berstsd, 1992;Sangter and Breen, 1998;Pellitier, 1998;Fiorentini et al., 1999).All these methods are usually based on the assumption that: 1) the target species are present in a great percentage of the hauls; 2) the number of species captured is sufficient to obtain significant and conclusive results; and 3) the species distribution is spatially homogeneous.
Although the study of the general variability in the catches was not a goal of this paper, the results presented here demonstrate that the catch variability was very high.This aspect has been studied widely to estimate densities on trawl surveys (Godø et al. 1990;Pennington, 1996;Pennington andVølstad, 1991, 1994;Folmer and Pennington, 2000).Despite this variability inherent to trawl surveys to assess target species, the cited authors indicate that short hauls are at least as efficient as long ones in catching fish of any size and that the efficiency of trawl surveys can be increased by reducing haul duration and increasing the total number of hauls.On the other hand, two hauls will always be different if the density of a given resource changes with the daynight time.Therefore, Pennington (1996) supported that the spatial distribution of marine organisms is highly patchy.Because of this patchy distribution, data from marine abundance surveys are highly skewed and have a large variance, and the sample mean has a low level of precision even for relatively intense surveys (Pennington and Godø, 1995).Also, Maynou et al. (1996Maynou et al. ( , 1998) demonstrated a highly patched distribution for crustaceans in the Mediterranean.In this sea a multiespecific fishery exists (Farrugio et al., 1993), and the differences observed between hauls may be due to the aggregation pattern of the species or to their day-night behaviour along the water column.Svatimskij (1985) found that diurnal migrations of cod resulted in changes in the size and sex composition of catches, and Laevastu and Favorite (1988) indicate that the catches were also affected by the direction of trawling with respect to the current.The results also showed that differences between day-time and night-time catches were the major differences observed between hauls at the same depth.Working at different depths would introduce noise masking the experimental results because some species occur at different depths.This finding draws attention to the need to conduct research into day-night variations in resource distribution and assessment.It should also be noted that these results are valid only for the gear and sampling protocols used in the present experiment and that the results could well differ for other trawl gear configurations or sampling strategies covering a broader swept area.
In several hauls carried out using the remotely operated closure system, certain species appeared only in the front or rear section of the cod-end.It can therefore be concluded that the species were not distributed homogeneously in the study area.In other words, certain not very mobile benthic species may be present at high densities in aggregations while other, more highly mobile species are not uniformly represented.These factors are intrinsic to all trawl sampling of marine species.If sample design had taken into account a certain minimum area as being representative of the community and longer tows had been used, the similarity values between the hauls and between the cod-end sections would probably have increased.Conversely, there would be a corresponding loss in the precision of the information on the small and medium-scale spatial-distribution of the resources and on day-time and night-time abundance, two aspects which are of great importance in determining the exact resources locally and making a precise assessment, as pointed out by Pennington and Vøldtad (1991), Pennington (1996) and Folmer and Pennington (2000).Though the variability in trawl surveys is an important matter to discuss, this aspect has never been studied in Mediterranean waters.
In conclusion, the remotely operated trawl codend closure system considered here is still not fully reliable technically and thus is still not suitable for regular deployment.Therefore, we believe that in the future, after further technical improvement, the system's quantitative effectiveness in both bottom and pelagic trawls will increase considerably.The remote closing system shows a low level of efficiency in taking homogenous samples from rear and front sections of the cod-end.However, we do not have any evidence that the remote closing system affected the species composition of the catches in any way, so the representativeness of the catch in each section of the cod-end and the high variability observed must be an actual reflection of the species community distribution.From this point of view, our results may suggest that remote closing systems can be useful for replicate samples assuming a high variability of the resource on the western Mediterranean shelf.In a monospecific context the remote closure system would probably increase its efficiency and take more homogeneous samples, but the use of this system in Mediterranean waters is highly doubtful.Further studies aimed at improving design in trawl surveys should be encouraged in the Mediterranean, bearing in mind the high spatial, seasonal and daynight variability of some species.
FIG. 3. -Dendrogram analysis showing similarity among hauls carried out with and without the remotely operated trawl cod-end closure system.*, hauls employing the remotely operated trawl cod-end closure system; D, day, N, night.

TABLE 2 .
-Species abundance matrix by haul ("Nerit II").The number of the cod-end section is given in brackets.(1), rear section and (2), front section.In bold, species not significantly different from the expected proportion 1:1.