Efficiency of Bottom Sampling Trawlsin Deriving Survey Abundance Indic - PDF document

Efficiency of Bottom Sampling Trawlsin Deriving Survey Abundance IndicesStephen J. WalshNorthwest Atlantic Fisheries Centre, P. O. Box 5667IntroductionAnnual bottom trawl surveys are commonly used to measure temporal variation in stock size, mortalityand recruitment, along with other biological characteristics, of various groundfish stocks under managementregulation. In the NAFO area, annual groundfish surveys generally employ a stratified-random design toestimate population abundance indices which are often expressed as mean numbers (or biomass) perstandard tow distance or area swept. These survey indices are also used in calibrating fishery dependentmodels (Sequential Population Analysis) to increase the precision of these abundance estimates. Surveyindices are more advantageous because of the rigorous standardization of protocols used to collect dataand with the use of small mesh trawls are better for estimating the strength of recruiting year-classes. Inthe current assessment of many NAFO groundfish stocks, trawl surveys provide the only source of stocksize estimates available to fishery managers. Consequently, errors and unexplained variability in surveyGenerally, researchers have directed their efforts towards increasing precision and accuracy of surveyestimates by improving survey designs and analyses to deal with the spatial variability of the target species.Changes in sampling trawl geometry and performance can affect the catching efficiency and also contributeexamine the role and validity of various assumptions made about the sampling tools used to derive surveyyellowtail flounder, (Pleuronectes ferrugineaSampling trawlsselection and subsequent estimates of abundance and recruitment. For example, in the Canadian surveysof the Grand Bank, the estimates of stock size and age composition of yellowtail flounder derived fromsmall and large mesh survey trawls often differ by an order of magnitude even though the timing of thesurveys are one month apart. The ideal sampling gear would catch, with equal efficiency, all demersal lifestages of many target species. However, bottom trawls are flexible structures and do not catch all fish inthe area sampled during a fishing tow because of changes in bottom trawl geometry, performance andfish behaviour. Thus all sampling trawls are size and species selective to varying degrees. In many countries,it is common to choose a common commercial trawl used in the local fishery as the standard survey trawland modify it by inserting a small mesh liner into the codend to reduce mesh selection.Fish capture processMuch of our knowledge about how a trawl catches fish comes from daylight studies of round fish, inparticular cod, haddock (), mackerel (harengus). The vessel, trawl doors, sweeplines (here defined as bridles plus ground warps), sand clouds,footgear and net panels present a combination of visual and auditory stimuli to herd the fish into the trawlmust be sufficient to exhaust the fish. As the fish get closer to the trawl mouth they swim orientated in thedirection of the tow (an optomotor response) and when exhausted either turn to swim into the net or over orunder the trawl. Swimming speed and endurance of the target species plays a key role in herding and The reader is directed to major reviews by Gunderson (1993), EngŒs (1994), God¿ (1994) and Walsh (MS 1996)which I have relied upon as sources of information for this synthesis. Fig. 1.Drawing of a bottom trawl in action, showing the herding effect of the sand (ormud) cloud generated by the trawl doors (Adopted from Gunderson 1993).capture success. At night, the visual herding response is reduced and the orientation of roundfish to thetow direction is generally more random. Generally, gear avoidance during the day time is high and at nightit is reduced.In contrast to roundfish, most flatfishes such as American plaice and yellowtail flounder, are non-schooling, bottom dwelling fish and do not react to doors, sweeplines, footgear until these componentsare very close (0.5 to 1 metre) or collide with them. Flatfishes generally do not show orientation andmovement along the direction of the tow and there is very little difference in day and night behaviours.Although flatfishes can avoid capture by escaping underneath the trawl they are never seen to rise upwardto escape over the headline nor are they affected by the noise of the survey vessel.The process of catching fish in bottom trawls is a complex interaction between fish and the trawl inthree zones and is illustrated in Fig. 2. It begins ahead of the trawl doors () where selection can beinfluenced by ship/trawl avoidance reactions with fish moving from the pelagic zone, in response to thevessel noise, into the bottom zone. A proportion of the fish in front of the trawl doors will enter the trawl) and of these fish a proportion will be herded by the sweeplines and thesand clouds created by the trawl doors towards the net (can occur in any of the three zones, then catchability is affected by size selection, horizontal and verticalspatial distribution of fish in the trawling area and the behaviour interactions of the fish with various physicalin survey trawls is often assumed to be negligible, however,in large mesh sampling trawls mesh selection can occur ahead of the small mesh codend liner. For the restof the text we will be concerned mainly with , the area from the trawl doors to the mouth of the net.Underlying MathematicsThe main objective of a bottom trawl survey is to derive an index of abundance which is proportional tothe true abundance and tracks the relative changes in the population through time. In order to achieve thisobjective trawl surveys require: 1) complete coverage of the distribution of target species, and 2) a measureof the catchability or fishing power of the sampling gear. In stratified-random surveys average catch-per-tow is extrapolated to the total survey area and it is assumed that catch (C) and absolute abundance (N)are related. In fishery dependent models this relationship is defined by the following familiar catch equation: C=qfNwhere q is the catchability coefficient and f some measure of fishing effort. Catchability is defined here asthe proportion of the stock caught by a defined unit of effort. Although q and f are difficult to preciselyestimate in fishery dependent models, in survey trawl data f is standardized by using the same tow duration. WALSH: Bottom Sampling TrawlsFig. 2.The three catching zones of influence in the fish capture proc-ess. (Adapted from God¿ 1994).The technological changes associated with changing catchability in commercial catch per unit effort (CPUE)data are reduced in scientific surveys by the process of standardization of vessel, fishing gear and survey CPUE=qHere survey catchability q is the proportionality constant between C or CPUE and the true abundance(or biomass) (D). CPUE generally represents catch (numbers or weight) per standard tow length or perunit area. In the NAFO area, the primary sampling unit is the area swept by the trawl (AS) and is generallyestimated by the product of the tow distance (t) and wing spread (WS). The true estimate of swept area isprobably best represented by trawl door spread (DS), instead of wing spread (see Fig. 2) and will bediscussed later. of the survey trawl can be further decomposed into availability and trawl or catchingefficiency as seen in Fig. 3 and Fig. 4. The above equation is re-written as: D=CPUE/(qHere trawl efficiency, is the proportion of fish in the trawl path retained by the net. Fig. 3.Availability of fish to the trawl can be a function of fish above and below the head-line of the sampling trawl in species such as cod. (Adopted from God¿ 1994).Fig. 4.Trawl efficiency is the ratio of the number of fish which is caught and retained by the net tothe number of fish in the trawl path. (Adopted from Fridman 1986). WALSH: Bottom Sampling TrawlsIn trawl surveys within the NAFO area it is commonly assumed that = 1, i.e. entire population oftarget species is in the survey area and accessible to the trawl, and escapement from the gear. Since both are unknown and assumed to be constant coefficients(although both are probably less than 1), the stock size estimates become relative abundance indices. If and could be precisely measured then absolute estimates could be derived. Assuming q B=Dwhere A is the area of the survey region.Although recent studies have focused on interpreting survey catchability by examining changes in in response to environmental changes and stock reduction, it is, nevertheless, importantthat we try to understand how these and other changes affect trawl efficiency Survey catchabilityMany factors can affect survey catchability, chief among them are:¥the horizontal and vertical distribution of the target species in relation to the trawl, i.e. natural behaviour,¥behaviour of fish ahead of the trawl and in the vicinity of the trawl, i.e. vessel/trawl induced behaviour,¥selectivity of the trawl.Survey catch rates, i.e. average catch-per-tow, are considered proportional to the true stock densityonly if factors such as vertical and horizontal fish distribution, fish behaviour reactions to the trawl and theperformance of the trawl are constant over time. Thus, within and between annual surveys, it is assumed¥size and species selection is constant under various conditions,¥trawl performance remains constant under various conditions, and¥swept area of the trawl is known and is constant under various conditions.Consequently, if any of these assumptions are invalid then the estimates are biased. Hence, the primeobjective during the survey is to minimize bias and sampling variability by standardizing all operationsand maintaining a constant catchability. Since we generally estimate relative abundance indices, aproportional bias is acceptable, provided that it is constant.Trawl selectionTypically, in bottom trawl catches, the length distribution is biased towards larger fish due to sizeand differences in swimming speeds. Of primary importance in this mechanical sorting is the size of thefootgear has been investigated for many survey trawls by rigging mini-trawl bags underneath the maintrawl to catch escaping fish. Figure 5 shows the escapement of cod under the footgear of both Canadiansurvey trawls when equipped with similar 36 cm diameter rubber rockhopper gear. Here trawl efficiency ) is calculated using the following equation: /(Mwhere q represents the efficiency of the trawl defined as the area between the wing ends and the centreof the footgear, Mc= catch in the main trawl and = total catch in the bag trawls. We assume here thatall fish entering the trawl mouth area are caught either by the main trawl or the bag trawls. Size selectioncan also differ considerably between species e.g. more yellowtail flounder escape underneath the footgearthan American plaice and more cod than haddock.Age selectionIn age-based assessment models, age data is derived from length of fish in the catch. Becauseescapement is strongly length dependent, estimates of age specific selection values will vary in accuracyfrom year to year if changes in growth, i.e. mean length-at-age, occur. For example, using the selection Fig. 5Escapement of cod underneath the footgear of the Campelenthe Canadian bottom trawl surveys off Newfoundland. (Walsh andcurve for the Engel survey trawl in Fig. 5, if the mean length of age 2 cod decreased from 21 cm (efficiency= 20%) to 18 cm (efficiency = 1%) then subsequent errors would occur in estimates of abundance-at-ageand prediction of year-class strength coming into the fishery. Periodic changes in growth will affect ageselection and if large enough could invalidate the reliability of the time series unless conversion factorsare derived. This has been demonstrated in the northeast Arctic cod time series. Variations in the efficiencyof survey trawls can reduce the accuracy and reliability of trawl data when these data are used in acousticmodels. The quality of survey trawl data, as representative of the population abundance, size andcomposition, should be critically examined in relation to the demands of these various models.Diel variation in light intensity will also influence selectivity. For example, in Fig. 6, the efficiency of theCanadian Engel survey trawl in catching yellowtail flounder is higher at night than during the day, becauseof a reduction in gear avoidance at night. A similar pattern was evident for American plaice but the oppositepattern was seen in cod. This bias in efficiency could seriously invalidate a time series if not accountedfor in the station allocation scheme used in the stratified random design. For example, in yellowtail flounder,if there is a higher proportion of night tows allocated in one survey year and a higher proportion of daytows in another, then the abundance and recruitment indices would be biased in opposite directions. Thiscan be minimized by temporal stratification of number of fishing sets, i.e. a 50:50 proportional allocation ofday and night sets within each strata, before the start of the survey. For tuning VPA's one could also time-partition the survey data into day and night indices and choose the one which has less sampling variability.Trawl geometry and performanceGenerally the variance around the average catch-per-tow is assumed to reflect changes in abundanceand does not account for changes in catchability resulting from size and species specific fish behaviour.Fish behaviour, both natural and trawl induced, can confound the interpretation of catchability data. A WALSH: Bottom Sampling TrawlsFig. 6.Diel variation in trawl (catching) efficiency for yellowtail flounder derived fromcatches using an Engel 145 High Lift otter trawl. (Adopted from Walsh 1991).major source of uncertainty in trawl surveys is the effect of changes in catchability on estimates of abundancedue to changes in trawl geometry and performance. Variations during construction, repairs, deployment,retrieval and actual fishing practices can affect trawl geometry and performance efficiency. These can beminimized by rigorous standardization of protocols. The recent use of acoustic trawl monitoring sensorshas shown that both trawl geometry and performance are mainly affected by towing speeds, water currents,bottom type, poor bottom contact, insufficient warp (scope) ratios, sea state and weather. Fish in front ofthe trawl will react to the instability of the trawl and thus catchability will vary. Biases created by any ofthese factors can affect the accuracy of the survey estimates. For example in Fig. 7, both door spreadsand wing spreads are below their respective average spreads when the trawl is towed against the current(north) and above the average spread in tows which are with the current (south). Swept area will also beaffected in a similar manner.Survey abundance indices are primarily calculated using a swept area model, i.e. a constant wingspread is used and multiplied by constant tow distance. This assumes that swept area within and betweensurveys is constant. In reality, this is not the case because trawls are flexible structures which changeseen in Fig. 8, the shape of the trawl in deep water is more collapse than in shallower water due to differencesin trawl door spreads.If a fixed wing spread is used in the model, then the swept area will be underestimated in deeperwaters and, correspondingly, the abundance will be overestimated, with the opposite trend in shallowwaters. Since we are dealing with relative indices of abundance, this proportional bias in catches and age(length) composition may be acceptable as long as there is no change in depth distribution of the stock orage classes. In recent years, some stocks in the NAFO area, e.g. American plaice, have shown a shift todeeper water, and depending on the magnitude of this change and the temporal and spatial scales, thefixed wing spread model would lead to an overestimate of the abundance in deep water and reduce thereliability of the time series. A similar change in size (age) depth distribution could confound interpretationof recruitment and mortality estimates of various year-classes based on swept area. Fig. 7.Changes in door spread and wing spread of the Engel 145 surveytrawl aboard the FRV during tows north against thecurrent and the reciprocal tow at the same station heading south withthe current. (Adopted from Walsh and McCallum 1995).Fig. 8.Difference in trawl geometry of the Campelen 1800 shrimp trawlused in Norwegian bottom trawl surveys of the Barents Sea atdepths of 50 m and 450 m. (Adopted from God¿ and EngŒsIf trawl geometry data are available, the survey indices can be re-calculated based on a varyingswept area model. Table 1 shows a comparison of the results of a constant wing spread model with amodel that incorporates average wing spreads at 100 m depth intervals. Here the constant swept areamodel underestimates the younger ages of northeast Arctic cod (1Ð3 years) in the survey area because of WALSH: Bottom Sampling TrawlsTABLE 1.Comparison of standard swept area indices of Barents Sea cod derived from afixed wing spread model (I) with those indices corrected for trawl width variationby using a varying wing spread model (II) at different reference depth zones.Division (%) represents deviation from stanard indices. (Adopted from God¿ and Age 123456�7T(A) Model I: Standard indices0Ð10012.979.954.914.11.80.80.1164.5100Ð20010.431.312.86.51.51.50.564.5200Ð3003.417.64.03.31.31.90.532.0300Ð4000.32.31.62.61.21.90.510.4�4000.11.01.01.50.81.50.46.3Total27.1132.174.328.05.67.62.0277.7(B) Model II: Corrected indicesReference depth zone 0Ð100 m WS0Ð10012.979.954.914.11.80.80.1164.5100Ð2008.525.510.45.31.21.20.452.5200Ð3002.513.03.02.41.01.40.423.7300Ð4000.21.61.11.80.81.30.37.3�4000.10.70.71.00.51.00.34.2Total24.2120.770.024.65.35.71.5252.0Division (%)-11-9-6-12-19-24-25-9Reference depth zone 200Ð300 m WS0Ð10017.4108.174.319.12.41.10.1222.5100Ð20011.434.414.17.21.71.70.671.0200Ð3003.417.64.03.31.31.90.532.0300Ð4000.32.21.52.51.11.80.59.8�4000.10.90.91.30.71.30.45.6Total32.7163.294.733.37.27.82.0340.9Division (%)21242819102123Reference depth zone 400Ð600 m WS0Ð10019.5120.582.821.32.71.20.2248.1100Ð20012.838.415.78.01.81.80.679.1200Ð3003.819.64.53.71.42.10.635.7300Ð4000.32.41.72.71.32.00.510.9�4000.11.01.01.50.81.50.46.3Total36.4181.9105.637.28.18.72.2380.1 Division (%)3438423322141237 Although wing spread is commonly used in the swept area model, door spread is probably morerepresentative of effective swept area than wing spread because of the auditory and visual herding stimuliof doors, sweeplines and sand clouds which herd fish towards the net. For example, if sweeplines arelengthened, causing the door spread to increase, higher catches of cod and haddock can result. Table 2illustrates a comparison of Scotian Shelf cod indices estimated using a fixed wing spread in the sweptarea model (CI) and re-calculated indices derived from a varying door spread model (VI). In this casethere is a greater reduction in the estimates of older cod using door spread. This is due to the fact thatolder fish are distributed in deeper water, i.e. increased door spread with depth in the fixed model wouldoverestimate the abundance indices. It is noteworthy that during the two time periods (1970Ð82 and 1983Ð92) there appears to have been a shift in depth distribution even though the average survey depths wereIt is obvious that trawl width variability can affect survey indices, however, in the few attempts thathave been made in comparing adjusted and standard swept area indices, the CVs still remain high becauseof the typical patchy distribution of many groundfish species. However, this does not imply that modellingvariation in swept area should not be used. Intuitively, the more relevant data that can be incorporated intothe assessment, the more confident we should be about advising managers of our best estimate of thestock size. For example, in Table 3 the vessel survey indices of northern cod, NAFO Div. 2J and 3KL,derived using two vessels are overestimated (combined overestimate is 37%) because the standard wingspread used in the swept area model is unrealistically low when compared to that measured during thesurveys. Re-adjustment of the time series with the new wing spread would not change the overall timeseries trend, however, this bias in the estimates of abundance (and recruitment) could be critical if amanagement decision to close or re-open a particular fishery depended solely on survey trawl data.TABLE 2.Ratio of constant swept area indices(CI) to varying swept area indices (VI) AgeVessel 1Vessel 2 1983Ð921970Ð8211.111.1621.131.1831.141.2141.191.2251.221.2361.221.2371.211.2581.191.2391.221.27 101.181.30(Adopted from Clarke, 1993).TABLE 3.Differences in assumed standard meanwing spread and calculated mean wingspread of the Canadian Engel 145 Hi-Lift survey trawl aboard the two surveynorthern cod, NAFO Div. 2J3KL. Wing% Overestimated spreadWidth (m)indicesStandard13.7ÐVessel 122.239 Vessel 219.831Adopted from Walsh and McCallum, MS 1995. WALSH: Bottom Sampling TrawlsEffect of tow durationThe distance trawled is the second variable in the swept area model and is also generally assigned anautical miles during a 30 min tow using a towing speed of 3.5 knots. Generally, the start of a tow isrecorded when the brakes on the trawl winches are applied, even though there is evidence to suggest thatdistance is far shorter than the target value and the swept area estimate is biased. In Fig. 9, the variability(Trips H037 and 038; A306 and 307) in distance trawled from the target of 1.75 nm is common within andbetween Scotian Shelf trawl surveys and can be controlled (Trips N123 and 124) by introducing standardizedvessel speed protocols.Typically, tow duration in groundfish surveys are 30 or 60 minutes in length. Recently, in the Canadiansurveys of NAFO stocks off the coast of Newfoundland the tow duration time was reduced to 15 minutesfrom 30 minutes. Figures 10 and 11 show that shorter tows are just as efficient as longer tows for manyspecies of gadoids and flatfish, and similar results were obtained in Norway and the United States. Whenswitching to a 15 min tow, it becomes more critical to accurately determine when the trawl reaches bottom.Here the active use of acoustic trawl instrumentation determines trawl bottom touch down, start of tow,end of tow and gear malfunctions. This will minimize variation in tow duration and reduce the number ofnavigational aids such as DGPS exact tow distances can be computed. Together with estimates of trawlwidth, the exact area sampled by the gear can be quantified on a tow by tow basis thus reducing biases inthe area swept by the trawl.The effect of fish densityIn commercial CPUE data, the catchability coefficient q is inversely related to population abundance.Recent studies have suggested that in scientific surveys, trawl efficiency is directly related to abundanceof fish in the trawl path, i.e. smaller numbers of fish led to an increase in gear avoidance. For example, inFig. 12, derived from the bag trawl escapement experiments, the trawl efficiency increases with density ofcod in the trawl mouth. Table 4 shows the catch-tow frequency of northern cod at low stock size in the1993 fall survey. Here it appears that the stock is comprised of small numbers of individual cod distributedFig. 9.Frequency of tow distances during 30 min tows before (H037/038; A306/307) andafter (N123/124) introduction of standardized vessel speed control protocols.(Adopted from Koeller 1991). Fig. 10.Effect of tow duration on of American plaice usingthe Engel 145 survey trawl. (Adopted from Walsh 1991).Fig. 11.Effect of tow duration on length composition of cod in catchesof the Engel 145 survey trawl. (Walsh, unpublished data).throughout the survey area. The density estimates can then be expected to be biased downward becauseof this decrease in trawl efficiency. If this trend in stock behaviour continues over several years, it willaffect the accuracy and reliability of the time series when compared to other years when the stock sizewas larger and distribution was more patchy. WALSH: Bottom Sampling TrawlsFig. 12.Effect of density on trawl efficiency of cod derived from bag trawl escapement experiments.(Walsh and McCallum, unpublished data).TABLE 4.Frequency of the number of tows with catches of cod in eachcategory. Area51Ð100 cod101Ð300 cod&#x = 5;� co; -19;.7; = 300 cod2J1050003K1600003L149131 Total414131Source: 1993 Canadian autumn survey of Northern cod.Controlling trawl geometry, performance and swept areaAs seen above, the accuracy of measuring the two components of the swept area model, tow distanceand wing spread, are possible with the correct instrumentation which is widely available. Although modellingvariation in trawl width has only increased the level of precision of survey estimates by 5%, usingmeasurements of trawl width and tow distance variations into the calculation of swept area estimates shoulddecrease some of the bias in swept area estimates.There are also other approaches to improving swept area estimates. Figure 13 and Table 5 illustratesa Norwegian method of standardizing trawl width variation by physically restricting door spread with arope placed ahead of the trawl doors to minimize the depth dependent trawl width variation. The methodof rope attachment is quick and easy. Paired tows shown in Fig. 14 illustrates that there is little differencein length composition of trawl catches of yellowtail flounder in a restricted trawl and unrestricted trawl.Similar results were obtained for cod, American plaice, skate, Greenland halibut and redfish. By minimizing Fig. 13.Controlling trawl door spread us-ing a rope strung 150 m aheadof the trawl doors. (Adopted fromTABLE 5.with the door spreads restricted and unrestricted from alternate tows ina depth of 43 to 1 244 m. CV = coefficient of variation expressed as apercentage. UnrestrictedRestricted CVHauls Doorspread (m)4151.9134145.17 Wingspread (m)4015.574114.39Walsh and McCallum: unpubl. data.the depth-dependent variation in swept area, the variance around average catch-per-tow (or biomassindices) should reflect primarily changes in abundance. One other major advantage of the restrictor ropeis that it may be used to standardize fishing power when two or more vessels are used in the annualsurvey.Absolute abundance indicesIn order to estimate absolute abundance indices accurate measurements of availability efficiency are needed, i.e. catchability. Acoustic instruments can be used to provide the vertical WALSH: Bottom Sampling TrawlsFig. 14.A comparison of the length frequency of yellowtail flounder in catchesof the Campelen 1800 shrimp trawl with restricted unrestricteddoor spread. (Walsh and McCallum, unpublished data). but may not be able to accurately resolve the horizontal component. There has beensome success at measuring trawl efficiency using various techniques such as acoustic tags, submersiblesof trawl-induced behaviour, most are costly not readily adaptable or applicable to other survey trawls orRecently, the focus of some studies in Norway, Australia and United States has been on mathematicallymodelling the behaviour of fish in the three zones of trawl influence in an effort to improve on swept areaSummary and Conclusions¥Estimates of abundance from trawl surveys generally have large variances generated by fluctuationsin spatial distribution, catchability, environmental conditions in the survey area or ecosystem and throughthe interaction of any one of these variables with the other.¥Although a survey is generally designed to minimize the effects of catchability through allocation ofeffort and survey coverage, less attention is directed towards sampling trawl variability. Standardizationof fishing protocols are assumed to minimize differences in trawl performance and the survey designis expected to average out the effects of vessel noise, size, species, light intensity, tow duration,depth, and density on catchability and hence swept area estimates. The validity of this assumption is¥Many of the assumptions we make about our sampling tools are invalid and our survey estimates arebiased as a result. A proportional bias in our estimates, however, is acceptable providing that it isconstant. When the bias of the survey estimate is not consistent from year to year due to changes, e.g.in fish growth or catchability brought on by environmental changes, the quantification of the accuracyof an abundance index is difficult and this can reduce the validity of the time series.¥This review does not conclude that existing survey data are useless but underlines the necessity ofincorporating studies to examine and account for potential errors in estimation of stock size andrecruitment that are contributed by the sampling trawl. Generally, variances around abundance indicesare high so that many biases affecting their accuracy in estimating stock size and recruitment may not ¥Coinciding with the recent severe reductions in stock size of many species in the NAFO area, there isevidence of changes in stock behaviour. These changes could affect availability and trawl efficiencyand result in misleading trends in the time series and serious problems in assessment of the resource.¥Survey indices are still treated as relative abundance estimates because the shortcomings of surveydesign and estimation of sampling trawl efficiency have not been resolved. Given the enormous amountof resources and costs to conduct annual surveys and the fact that those survey indices may be theonly source of information on which to base management advice then greater precision and accuracyin estimating stock size and year-class strength are highly desirable. This should be a continuingprocess.References and Other Suggested ReadingsCLARKE, D. S. 1993. The influence of depth and bottom type on swept area by ground trawl and consequences forDICKSON, W. 1993a. Estimation of the capture efficiency of trawl gear. I. Development of a theoretical model. DICKSON, W. 1993b. Estimation of the capture efficiency of trawl gear. II Testing a theoretical model. DOUBLEDAY, W. G., and D. RIVARD. 1981. Bottom trawl surveys. ENGS, A. 1994. The effects of trawl performance and fish behaviour on the catching efficiency of demersal sampling: Marine Fish Behaviour in Capture and Abundance Estimation. A. Fernš and S. Olsen [eds.]. FishingNews Books. Oxford: p. 45Ð68.FRIDMAN, A. L. 1986. Calculations for fishing gear design (FAO Fishing Manuals). Fishing News Books Ltd.: 241 p.GOD¯, O. R. 1994. Factors affecting reliability of groundfish abundance estimates from bottom trawl surveys. InFish Behaviour in Capture and Abundance Estimation. Fishing News Books. A. Fernš and S. Olsen [eds.]. Oxford:GOD¯, O. R., and A. ENGS. 1989. Swept area variation with depth and its influence on abundance indices of groundfishfrom trawl surveys. J. Northw. Atl. Fish. Sci.GOD¯, O. R., and S. J. WALSH. 1992. Escapement of fish during bottom trawl sampling Ð implications for resourceGUNDERSON, D. F. 1993. Surveys of Fisheries resources. Wiley, Toronto, New York.KOELLER, P. A. 1991. Approaches to improving groundfish survey abundance estimates by controlling variability ofsurvey gear geometry and performance. J. Northw. Atl. Fish. SciMcCALLUM, B., and S. J. WALSH. MS 1995. Survey trawl standardization used in groundfish surveys. ROSE, C. S., and G. E. WALTERS. 1990. Trawl width variation during bottom trawl surveys: Cause and consequences.Int. North. Pac. Fish. Comm. BullWALSH, S. J. 1991. Diel variation in availability and vulnerably of fish to a survey trawl. J Appl. IchthyolMS 1991. The effect of tow duration on gear selectivity. 1996. Ecology, resource surveys and management of long rough dab (American plaice), (Fabricius), in the Barents Sea and the NewfoundlandÐLabrador Shelf. Dr. Philos. Thesis. UniversityWALSH, S. J., and B. R. McCALLUM. MS 1995. Survey trawl mensuration using acoustic trawl instrumentation. MS 1996. Performance of the Campelen 1800 shrimp trawl during the Northwest Atlantic Fisheries Centre1995 fall groundfish surveys. WALSH, S. J., P. A. KOELLER, and W. D. McKONE. 1993. Proceedings of the international workshop on survey trawlmensuration, Northwest Atlantic Fisheries Centre, St. John's, Newfoundland, March 18Ð19, 1991. Can Tech. Rep.