Capture efficiency of a multi-species survey trawl for Snow Crab (Chionoecetes opilio) in the Newfoundland region

Fisheries Research 101 (2010) 70–79

Contents lists available at ScienceDirect

Fisheries Research journal homepage: www.elsevier.com/locate/fishres

Capture efficiency of a multi-species survey trawl for Snow Crab (Chionoecetes opilio) in the Newfoundland region Earl G. Dawe ∗ , Stephen J. Walsh, Elaine M. Hynick Science Branch, Fisheries and Oceans Canada, P.O. Box 5667, St. John’s, NL, Canada A1C 5X1

a r t i c l e

i n f o

Article history: Received 9 July 2009 Received in revised form 15 September 2009 Accepted 17 September 2009 Keywords: Capture efficiency Survey trawl Snow Crab Chionoecetes opilio Logistic regression

a b s t r a c t An experiment was carried out during September of 2001 to estimate the relative catching efficiency of the NAFC’s (Northwest Atlantic Fisheries Centre) standard survey bottom trawl, the Campelen 1800 shrimp trawl, for Snow Crab (Chionoecetes opilio). Secondary trawls were mounted underneath the main trawl directly behind the footgear to estimate escapement of crab passing underneath the main trawl. Overall, the capture efficiency of the survey trawl was found to be much lower than 1.0 and highly variable, with variability being strongly related to substrate type and Snow Crab size. The probability of capture monotonically increased with crab size, and was higher on soft than on hard substrates. Three substratespecific capture efficiency functions were defined; (1) efficiency was highest (about 0.73–0.95), and least dependent on crab size, on the softest mud substrate; (2) efficiency was lower (about 0.39–0.70) and linearly related to crab size throughout a soft mud–sand substrate; and (3) efficiency was lowest (about 0.05–0.33) on a variety of hard substrates within relatively shallow water. Our models over-estimated true efficiency because secondary trawls were not fully efficient, i.e., crab escaped underneath their footgear. Nevertheless, our experiment demonstrated that substrate type can affect capture efficiency and annual changes in trawl configuration or in Snow Crab distribution in relation to substrate type will affect survey biomass estimates. Additional studies using various approaches could provide a basis for standardizing survey catches for varying capture efficiency effects and refining survey-based estimates of biomass and exploitation rates of Snow Crab. Crown Copyright © 2009 Published by Elsevier B.V. All rights reserved.

1. Introduction Bottom trawls are commonly used as a survey tool for monitoring crab resources in the Eastern Bering Sea (Alverson and Pereyra, 1969) and in Canadian Atlantic fishery areas. Bottom trawl surveys of Canadian Atlantic Snow Crab (Chionoecetes opilio) resources are conducted annually in the Southern Gulf of St. Lawrence (Moriyasu et al., 1998; Hébert et al., 2002), in the Northern Gulf of St. Lawrence (Dufour and Dallaire, 1999), on the eastern Nova Scotian Shelf (Biron et al., 2002), and on the Newfoundland-Labrador continental shelf (Dawe et al., 2008; Dawe and Colbourne, 2002). These surveys differ considerably among fishery areas with respect to their history (time series), design, spatial coverage of the resource, and trawls used. Multi-species bottom trawl surveys have been conducted annually on the Newfoundland-southern Labrador shelf since the fall of 1995, using a Campelen 1800 shrimp trawl. These surveys, conducted in fall (September–December), extend from the southern Labrador shelf to the southern Grand Bank, Northwest Atlantic Fish-

∗ Corresponding author. Tel.: +1 709 772 2076; fax: +1 709 772 4106. E-mail address: [email protected] (E.G. Dawe).

ery Organization (NAFO) Divisions 2J3KLNO. In addition, annual surveys during spring (April–June) on the Grand Bank and St. Pierre Bank (NAFO Divisions 3LNOPs) have been carried out since 1996. These surveys follow a stratified random design, with stratification of the survey area based on depth and area (Doubleday, 1981). The fall NAFO Divisions 2J3KLNO surveys have been used extensively in annual Snow Crab assessments (Dawe et al., 2008). Annual changes in fall survey size-specific abundance and biomass indices suggest that the capture efficiency of the Campelen survey trawl for male crabs in fall is considerably lower than 1.0, varies annually, and varies with crab size (Dawe et al., 2006). Capture efficiency, by the Eastern Bering Sea survey trawl, of Snow Crab and Tanner Crab (Chionecetes biardi) (Somerton and Otto, 1999) as well as Red King Crab (Paralithodes camtschaticus) (Weinberg et al., 2004) was shown to be strongly related to crab size. Survey measurement errors, including those due to poor or variable trawl performance, can profoundly affect population abundance and biomass indices. In this paper we describe an experiment carried out using secondary trawls fitted underneath the Campelen survey trawl to directly estimate the relative capture efficiency of the survey trawl for Snow Crab. We also investigate the relationships of capture efficiency with substrate type and with male crab size. Furthermore,

0165-7836/$ – see front matter. Crown Copyright © 2009 Published by Elsevier B.V. All rights reserved. doi:10.1016/j.fishres.2009.09.008

E.G. Dawe et al. / Fisheries Research 101 (2010) 70–79

we apply those relationships to the raw survey data to standardize for effects of substrate type and male crab size and we evaluate the extent to which this standardization adjusts for low capture efficiency. 2. Methodology 2.1. Data collection Experimental trawling was conducted during September 3–7, 2001 in Conception Bay (Fig. 1), aboard the Research Vessel CCGS Wilfred Templeman using the Campelen 1800 shrimp trawl, the standard multi-species survey bottom trawl that has been used by the Northwest Atlantic Fisheries Centre, Fisheries and Oceans Canada, Newfoundland region since 1995 (McCallum and Walsh, 1996; Walsh and McCallum, 1996). Three secondary trawls were attached below the trawl net and directly behind the footgear of the Campelen trawl to collect crabs that had escaped underneath. This experiment is very similar to one carried out by Walsh (1992) to estimate efficiency of a survey trawl for finfishes as well as to

71

that conducted by Somerton and Otto (1999) to estimate efficiency of the Eastern 83/112 survey trawl for Snow Crab and Tanner Crab and by Weinberg et al. (2004) for Red King Crab. The Campelen 1800 survey trawl is a small-meshed (80, 60 and 44 mm) shrimp trawl that is fitted with a 12.7 mm mesh codend liner. It has an average door spread ranging from 41 to 53 m, an average wing spread ranging from 15 to 18 m and an average vertical opening of 4–5 m, depending on depth fished (McCallum and Walsh, 2002). The trawl is spread by 4.2 m2 Morgère cambered, oval, single slot trawl doors weighing 1400 kg each. It has rockhopper footgear featuring 355 mm diameter rubber disks spaced at 200 mm intervals along the 19.6 m footrope. Secondary trawls, of 44 mm mesh size throughout, were specifically constructed for this experiment following the methodology described in detail, and illustrated, by Walsh (1992) for another survey trawl. The three secondary trawls were attached to the Campelen trawl, beneath the port and starboard wings and the bosom, with the 19.5 m headropes of the secondary trawls attached to the 19.5 m fishing line of the Campelen trawl with the eye splices of both lines hammerlocked into the upper hole of the delta plate.

Fig. 1. Location map showing the location of successful tows for estimating trawl capture efficiency for Snow Crab within each of two depth strata in Conception Bay, separated by the 200 m isobath. Depth and substrate type at each tow are identified by symbol shape and colour respectively. The single black square represents a shallow stratum tow on soft substrate.

72

E.G. Dawe et al. / Fisheries Research 101 (2010) 70–79

Table 1 Snow Crab sample sizes and capture efficiency estimates by sex and substrate type. Capture efficiency is defined here as the proportion of catch in the main trawl codend relative to the catch of the codends of the secondary trawl codends plus catch of the main codend. Substrate Type

Mud Mud–sand Coarse Total

Number of tows

Males

Males

Male

Females

Females

Female

Main trawl

Secondary trawl

Capture efficiency

Main trawl

Secondary trawl

Capture efficiency

0.91 0.57 0.15

2 6 5

5,704 2,417 76

549 1811 427

13

8,198

2789

The three secondary trawls shared a common 200 mm diameter rubber disk-encased cable footgear that was attached to both ends of the Campelen rockhopper footrope and loosely in the bosom to prevent damage during shooting and hauling. The secondary trawl footrope followed the Campelen footrope by a maximum distance of 1.0 m, at the centre controlled by 1 m safety chains which connected both ends of the bosum area of each secondary trawl footgear to an adjacent section on the main footgear. These chains took the maximum weight of the secondary trawl footgear off the netting during shooting and hauling. Earlier flume tank testing combined with at sea trials with a similar secondary trawl rigging setup for the Campelen in 1994 (Walsh, unpublished data) indicated that the footrope of the secondary trawls should be about 5% longer than that of the main footgear to minimize damage to the small mesh netting of the secondary trawls. Before attachment of the secondary trawls the Campelen footrope was measured as 19.8 m, 20 cm longer than specification. The common footgear of the secondary trawls was increased to 21 m (6% of the Campelen footrope length) to accommodate this difference. All experimental tows were comparable to standard survey tows with respect to trawling time (15 min tow duration), distance towed (0.75 nm), vessel speed (3 knots), and use of acoustic trawlmounted monitoring instrumentation (SCANMAR). SCANMAR was used to determine trawl touch-down and lift-off and to measure trawl geometry throughout each tow. Two bottom contact sensors (inclinometers) were used during each tow, affixed at various locations along the footrope of either the Campelen trawl or the secondary trawls. The sensors were attached by lopping a 41.5 cm bobbin chain around the iron spacers (lancasters) between the rockhopper disks of the main footgear and connecting to the toggle chain of the secondary trawls. The most consistent placement of the sensors was having one sensor located at the centre of the bosom of the secondary trawl’s footgear while the second sensor was placed at various locations (port, bosom or starboard) along the footgear of the main trawl. These sensors monitored changes in angle throughout each tow (Fig. 2). Data downloaded following each tow provided information on relative differences among tows in the variability of sensor angles, which reflected degree of footrope contact with the substrate. Initially, eight experimental 15 min tows were executed on the northern Grand Bank, about 20 nm outside of Conception Bay (Fig. 1). Three sets of tows were made with the main trawl rigged with and without the secondary trawls (all codends left open) to evaluate the effects of secondary trawls on the geometry and performance of the main trawl. Two additional tows were executed with secondary trawls attached and all codends closed to evaluate the effects of secondary trawls, under load, on main trawl geometry. The vessel then moved to Conception Bay (Fig. 1) where four additional tows were executed to check, again, the rigging and performance of the trawl system including nets, SCANMAR sensors, bottom contact sensors and experimental protocols before beginning the experiment. The experiment to estimate capture efficiency of the survey trawl for crab then began with all codends closed during all subsequent tows. Within Conception Bay two survey strata were sampled (Fig. 1), with eight successful tows in the deep

983 667 33

200 648 659

1,683

1,507

0.83 0.51 0.05

Total

7,436 5,543 1,195 14,177

stratum (789; 184–366 m) and five successful tows in the shallow stratum (799; 93–183 m). The sampling procedure at each station involved comparable treatment of the catches from all four codends (main codend and three secondary trawl codends). Each catch was initially sorted by commercial species as well as by sex for Snow Crab. Snow Crab catches were sampled in their entirety or sub-sampled by sex. Individuals of both sexes were measured in carapace width (CW, mm). 2.2. Capture efficiency models 2.2.1. Capture efficiency Capture efficiency of the Campelen survey trawl for crab was estimated by sex and substrate type, for all sizes pooled. Since crabs are not herded by the doors or bridles (Somerton and Otto, 1999; Weinberg et al., 2004) of a trawl we restrict our definition of capture efficiency to the area of the trawl mouth between the wing ends of the trawl. The capture efficiency of the survey trawl is defined here as the proportion of Snow Crab caught by the Campelen main codend to that caught in total by the main codend plus the three secondary trawl codends. This assumes that the capture efficiency (E) of the secondary trawls is 1.0, i.e., all crabs escaping underneath the main trawl footrope are caught and retained in the codends of the secondary trawls. 2.2.2. Size-based capture efficiency models We examined the relationship of capture efficiency with size for male crabs only. Females comprised only 22% of total crabs caught (Table 1), and their sample sizes were considered inadequate for our modelling exercise. Three variants of the logistic regression model commonly used in modelling codend selectivity were fitted hierarchically using maximum likelihood estimation according to the methods of Tokai et al. (1996), Munro and Somerton (2001) and Weinberg et al.

Fig. 2. An example of trends in bottom contact sensor angle of inclination from both the main trawl and the secondary trawl along a tow path. Small angles, of 0 to about 20◦ , reflect near vertical position of the sensors while the trawls were being deployed and retrieved. Maximum angles, of about 90◦ for the secondary trawl footgear and about 50◦ for the main trawl footgear, reflect full contact of trawl footgear with the substrate.

E.G. Dawe et al. / Fisheries Research 101 (2010) 70–79

73

Table 2 Capture probability model selection and estimated parameters for the best model fit based on the lowest Akaike Information Criterion (AIC). Model 1 is a 2-parmater logistic function, model 2 is a three-parameter logistic function and model 3 is a four-parameter logistic function. Bottom type/depth

No. of tows

ALL Muddy-sand/deep Mud/deep Coarse/shallow a

Catch (nos.)

13 6 2 5

AIC Values

10,987 4,227 6,253 503

Estimated parameters

Model 1

Model 2

Model 3

˛

ˇ



ı

11,760.9 278.4 268.6 98.9

11,766.7 280.4 247.2 100.8

11,756.3 281.9 249.2 NCa

−3.061 −0.560 0.753 −3.148

0.028 0.011 0.023 0.022

0.891

0.424

0.975

NC, no convergence.

(2004). Munro and Somerton (2001) estimated footrope selectivity using three parametric models with increasing complexity to define the probability outcomes of fish capture. We adopt their approach here and use the following models in order of complexity:

the RoxAnn system was disabled before the experiment was completed. However, using all available RoxAnn and sensor data we were able to partition tows among three substrate types (Table 2). Accordingly, we fitted the models to three subsets of the data based on substrate type:

(1) two-parameter logistic model; r1 (l) =





1

(1) Mud substrate; two deep tows. (2) Mud–sand substrate; six deep tows. (3) Coarse substrate; five shallow tows.

1 + e−(˛+ˇl)

where the capture probability increases monotonically with length (or carapace width in the case of Snow Crab) to an asymptote of unity (2) three-parameter logistic model; r2 (l) = 



1



1 + e−(˛+ˇl)

where the capture probability reaches an asymptote of less than unity, and (3) four-parameter logistic model;

r3 (l) = 

 1   1 − ı ı  e−ı(˛+ˇl)  1−ı

ı

1 + e−(˛+ˇl)

where the capture probability reaches a maximum then decreases with increasing carapace width.1 Under the assumption that all crabs encountered will either enter the main trawl or be caught by the secondary trawls the catches of the main trawl were conditioned on the total catch of all codends. The selectivity parameters ˛, ˇ, , and ı in the generalized linear model are derived using maximum likelihood estimation with a log-likelihood function based on binominally distributed errors. The best selectivity model was chosen based on the lowest Akaike’s Information Criterion (AIC) (Burnham and Anderson, 1998). 2.2.3. Fitting the data We assumed that capture efficiency of the survey trawl for crabs would be affected by substrate type, and it was known that substrate type varies with depth in Conception Bay. The fishery in this bay is prosecuted primarily on a soft mud substrate at greatest depths throughout the centre of the Bay (Dawe et al., 2006). Therefore we used the existing depth stratification scheme as a proxy for substrate type for sampling purposes. We subsequently examined data from the bottom contact sensors and the vessel’s RoxAnn seabed classification system to determine substrate type for each tow. The RoxAnn system classified substrate type according to roughness and hardness along each tow path. Unfortunately

1

In each case we fitted the data, pooled over all tows, to the three models and used the lowest AIC to select the best model. All model selection was carried out using routines written for SOLVER MS EXCEL (Tokai, 1997). Somerton and Otto (1999) also modelled the probability of Snow Crab and Tanner Crab escaping underneath the footgear of the Eastern Bering Sea survey trawl and entering the main net through the belly mesh because of the large meshes (101 mm) in their trawl. We felt that this was a non-issue given the small mesh size (60 mm) of the netting used in the belly of the Campelen survey trawl.

This model can be simplified to r3 (l) = 

 1−ı ı−1  ı

e−ı(˛+ˇl) 1+e−(˛+ˇl)



.

2.3. Evaluation of capture efficiency estimates We recognized that it is likely that our experiment resulted in inflated estimates of capture efficiency because it is unlikely that our secondary trawl captured all crabs in its path. To evaluate the reliability of our capture efficiency estimates we addressed the assumption that our secondary trawl was fully efficient (E = 1.0). All available RoxAnn data from annual surveys in Conception Bay were utilized to identify two substrate types: a smooth soft bottom type and a rough hard bottom type. The spatial distribution of substrate types in relation to survey strata and sampling stations in Conception Bay was mapped. This technique was then applied to the northern Grand Bank using all RoxAnn data collected from spring and fall multi-species surveys since 1995, to map substrate type in relation to depth across that broad area. Fall survey Snow Crab catches (1995–2006) from the northern Grand Bank were adjusted or ‘standardized’ by applying the appropriate capture efficiency function to the raw survey size-specific catch data based on substrate type. For those tows and strata for which there were no RoxAnn data and so no substrate type could be assigned the lowest capture efficiency function was applied. The effect of the standardizing, or adjusting catches, for low capture efficiency on estimates of exploitation rate was examined to determine whether this adjustment fully accounted for low capture efficiency. Annual (1996–2007) exploitation rates were estimated, as the ratio of the commercial catch to the exploitable biomass estimate projected from the fall survey of the previous year using standardized exploitable biomass estimates.

74

E.G. Dawe et al. / Fisheries Research 101 (2010) 70–79

Table 3 Comparison of SCANMAR trawl geometry with different riggings. The main trawl codend was open in riggings 1 and 2 whereas it was closed in rigging 3. ‘Opening’ represents the vertical trawl opening. ‘ND’ means no data. Rigging

Tow Number

Stats.

Door (m)

Wing (m)

1. No secondary trawls

1 2 3

Mean (CV, %) Mean (CV, %) Mean (CV, %)

44.89 (4.78) 45.42 (3.75) 45.56 (0.69)

16.14 (4.50) 16.21 (2.38) ND

4.44 (8.39) 4.40 (9.43) 4.31 (7.74)

Rigging # 1

Overall mean

45.29

16.18

4.38

2. Secondary trawls on with codends open

4 5 6

40.06 (12.35) 44.40 (3.74) 45.05 (2.16)

14.91 (7.86) 15.86 (4.48) 16.07 (3.62)

4.69 (9.13) 4.40 (10.99) 4.42 (10.00)

Rigging #2

Overall mean

43.17

15.61

4.50

3. Secondary trawls on with codends closed

7 8

44.58 (2.60) 41.67 (4.68)

15.94 (6.12) 15.21 (3.70)

4.43 (7.51) 4.38 (6.31)

Rigging #3

Overall mean

43.13

15.58

4.41

Mean (CV, %) Mean (CV, %) Mean (CV, %)

Mean (CV, %) Mean (CV, %)

Effect on main trawl geometry: No load: mean of rigging # 1 − mean of rigging #2 Under load: mean of rigging # 1 − mean of rigging #3

3. Results

2.12 2.17

0.56 0.60

Opening (m)

Depth (m) 189.90 189.80 189.70

189.10 190.06 189.91

190.18 190.69

−0.12 −0.02

trawls had minimal effect on the overall geometry of the survey trawl.

3.1. Trawl geometry 3.2. Stratification by substrate type The geometry of the trawl with and without the secondary trawls attached underneath was measured during six tows over the same ground with the codends left untied (Table 3). The use of the secondary trawls had a minimum effect on the means of door spread (decreased by 2 m), wingspread (decreased by 0.6 m) and trawl opening (increased by 0.1 m). Two additional tows were made with the codend tied and a similar effect on the geometry was noted (Table 3). During the annual surveys in the 100–200 m depth range door spread averaged 44.3 m (CV = 5.2%) and wing spread averaged 15.8 m (CV = 3.8%) and these correspond well to the average values measured during fishing with secondary trawls attached, i.e., 43.1 and 15.6 m, respectively. Our conclusion is that the secondary

The RoxAnn seabed classification data showed that substrate type was closely associated with depth (Fig. 1). The deeper stratum corresponded with a soft mud or mud–sand substrate throughout the centre of the Bay, whereas the shallower substrate corresponded to a harder or coarse substrate around the perimeter of the Bay. The tow-specific RoxAnn seabed classification data and the bottom contact sensor data confirmed that all tows in the deeper stratum (Fig. 1) corresponded to a soft substrate. This was indicated by low variability in angle of inclination of bottom contact sensors along tow paths (Fig. 3 for an example) reflecting almost constant footgear contact with bottom typical of soft substrate tows. This

Fig. 3. Trends along trawl tow paths from RoxAnn data showing substrate smoothness and roughness (above), and, from the bottom contact sensor, angle of inclinations (below) for tows typical of the deep water soft substrate (left) versus the shallow water hard substrate (centre), as well as an atypically soft substrate within the shallow stratum (right).

E.G. Dawe et al. / Fisheries Research 101 (2010) 70–79

75

On further investigation, it appeared that two tows at the innermost extreme of the deeper stratum (Fig. 1) were executed on an unusually soft mud substrate resulting in the entire catch of both tows being heavily fouled with mud. In addition, these two tows produced the largest Snow Crab catches, representing 60% of the overall total catch of male crabs from all eight tows provisionally assigned to the deep water/soft substrate group. We expect that these two tows would be highly influential in any modelling exercise for the deep water/soft substrate group and decided to model the data from those two tows separately from the data derived from the other six tows assigned to that group. The end result was the creation of three substrate-specific datasets for our modelling exercise (Table 1). Fig. 4. Estimates of capture efficiency for each of male and female Snow Crab and unsexed American Plaice across all sizes for each experimental tow within the two depth strata.

agreed with the low and stable values of substrate softness and roughness from the available RoxAnn seabed classification analysis. By contrast, tows conducted in the shallower stratum (Fig. 1) typically showed severe fluctuations in angle of inclination of bottom contact sensors (Fig. 3), reflecting bouncing of rockhopper footgear on hard rough substrates. This agreed with the high and variable values of substrate hardness or roughness from the available RoxAnn seabed classification analysis. A single exception to this association of substrate types with depth strata was the deepest tow executed in the shallower stratum which showed low variability in the bottom contact sensor data and low stable values of substrate hardness and roughness along the tow path (Fig. 3). By including this tow with those from the deeper stratum we created two clear groups of tows based on substrate type (Fig. 1), a deep water/soft substrate group (eight tows), and a shallow water/coarse substrate group (five tows).

3.3. Catches and capture efficiency Eighteen tows were completed during the experiment but only 13 tows were deemed successful because of damage to one or more of the secondary trawls. A total of 14,177 crabs was caught; 10,987 (78%) males and 3190 (22%) females (Table 1). Overall capture efficiency in numbers of Snow Crab across all sizes showed similar trends among tows between the sexes (Fig. 4). Capture efficiency (E) was generally higher in the deeper stratum (range 0.24–0.96) than in the shallower stratum (range 0–0.45). Capture efficiency of Snow Crab was compared with that of a bottom feeding flounder, American Plaice (Hippoglossoides platessoides), the most frequently caught fish species in the experiment. American Plaice showed a much higher capture efficiency than Snow Crab (about 0.70–0.98) with no apparent difference between depth strata, consistent with earlier trawl capture efficiency studies of plaice (Walsh, 1996). The data summarized by substrate group showed clear effects of substrate type on capture efficiency (Table 1). Capture efficiency, for both males and females, was highest on the mud substrate (0.91

Fig. 5. Size frequency of male Snow Crabs caught in the survey trawl (Main) and those that escaped underneath the trawl into the secondary trawls (Secondary), for all experimental tows combined, and by substrate group.

76

E.G. Dawe et al. / Fisheries Research 101 (2010) 70–79

Fig. 6. Selectivity curves using the capture efficiency estimates by size for male Snow Crab derived using logistic models for all experimental tows combined, and by substrate group.

and 0.83 respectively) and lowest on the coarse substrate (0.15 and 0.05 respectively). It was consistently higher for males than females on all three substrate types.

tows) and lowest on hard rough bottoms (five shallow water tows) (Fig. 6).

3.5. Evaluation of catch efficiency estimates 3.4. Effect of substrate and size on male capture efficiency In all sub-grouping of the data sample sizes were adequate overall, but the smallest male crabs were not well represented in the catches (Fig. 5). This was most evident for the softest (mud) substrate and least evident for the coarse substrate. Based on the lowest AIC, the two parameter logistic model (model 1) gave the best fit to the data for two of the four datasets (although all capture efficiency values for large crab were less than unity). A three parameter logistic (model 2) gave the best fit for the two tows on mud substrate, whereas a four-parameter logistic model gave the best fit for all data combined (Table 2). Capture probability of male crabs for all 13 tows shows a monotonic increase with carapace width to an asymptote of 0.89 (Fig. 6), characterized by the four-parameter logistic model, r3 (l), which had the lowest AIC (Table 2). The survey trawl has a low efficiency for small crab with zero catches in the size range of 12–18 mm. A plot of the original values overlaid on the predicted curve showed highly variable efficiency for small crab, 21–48 mm carapace width (Fig. 6), likely due to low sample sizes at those carapace widths (Fig. 5). Separation of the carapace width data for crab by substrate types (Fig. 6) showed that capture efficiency was highest and least variable in the two deep water tows on soft mud substrate, increasing from 0.73 to 0.98 (Fig. 6) across the size range sampled. Capture efficiency was lower for the six deep water tows on muddy-sand substrate, increasing with size from about 0.39 to 0.70. Size-specific capture efficiency for the five tows on shallow coarse substrates was the lowest, ranging from 0.05 to 0.33 (Fig. 6). In summary, the capture efficiency of crab at various sizes by the survey trawl was highest on softest muddy bottoms (two deep water tows), moderate on mud–sand bottoms (six deep water

The relationship of substrate type with depth throughout Div. 3L (Fig. 7) indicates that the softest substrate is associated with greatest depths whereas harder substrates are associated with shallower depths. Based on available substrate classification data we applied the mud–sand–deepwater capture efficiency function (six tows); r1 (CW) =

1 1 + e−(−0.506+0.011(CW))

(Fig. 6) to those tows on the soft substrate, whereas we applied the hard substrate-shallow water capture efficiency function (five tows); r1 (CW) =

1 1 + e−(−3.148+0.022(CW))

(Fig. 6) to those tows on harder substrate types. This hard substrate function was also applied to those tows for which no substrate type data were available. The portion of the total area for which substrate type could be assigned ranged 79–99% across years (Fig. 8). The annual standardized exploitation rates for the survey area of the northern Grand Bank, adjusted for low capture efficiency according to substrate type, ranged 0.10–1.87 (Fig. 8) and remained above 0.60 since 2000. The rates exceeded 1.0 in 2001 (1.13) and 2007 (1.87) when 88 and 89%, respectively, of the total survey area could be assigned substrate types based upon the survey RoxAnn data. Exploitation rates were estimated as 0.66 for 2002 and 0.70 for 2006, two years of virtually complete substrate type assignment (98 and 99% respectively, Fig. 8).

E.G. Dawe et al. / Fisheries Research 101 (2010) 70–79

77

Fig. 7. Distribution of bottom substrate types in relation to depth across the northern Grand Bank, based on research vessel RoxAnn data collected from 1995 to 2007.

4. Discussion Our study showed that escapement of crabs underneath the footgear of the Campelen survey trawl was a function of substrate type and crab size. Overall trawl capture efficiency for males, was highest on a soft mud substrate, lowest on coarse hard substrates and intermediate on a mud–sand substrate. For the combined dataset overall capture efficiency increased with size but did not reach unity at the largest crab size caught, i.e., asymptote equivalent was 0.89. Capture efficiency of male crabs increased with increasing size on all substrates. This explains the lower capture efficiency of the survey trawl for the females than males on all three substrates, since female crabs are smaller on average than males. Although trawl capture efficiency for male crabs increased with increasing size, it failed to reach unity regardless of which subset of data was modelled. In most cases, there was little difference in the AIC among models 1 (asymptote of unity), 2 (less than unity) and 3 (where the capture probability reaches a maximum then decreases with increasing carapace width). These results agree with results of modelling of capture efficiency data for Alaskan Snow and Tanner Crab (Somerton and Otto, 1999), Red King Crab (Weinberg et al., 2004) and flatfish (Munro and Somerton, 2000). Failure of the modelled capture efficiency estimates to reach an asymptote of unity was not unexpected, given the dynamic nature of the Campelen survey trawl’s rockhopper footgear, which hops when it encounters any obstruction in its path, creating large escapement openings beneath the footgear. Data from the RoxAnn substrate classification system and the bottom contact sensors confirmed the close association of substrate type with depth and facilitated our grouping of the data by sub-

Fig. 8. Exploitation rate estimates based on the ratio of the commercial catch to the exploitable biomass index projected from the annual fall survey of the previous year, using standardized survey data. A plot of the annual percentage of the total area trawled categorized by substrate type is overlain.

strate type. Particularly important, this substrate data allowed us to clearly distinguish a shallow hard coarse substrate from the softer substrates. We acknowledge that the further partitioning of our soft substrate deep water tows between two soft substrates (mud versus sandy-mud) was somewhat subjective. However, as we noted earlier, there were field observations to support this distinction, in that mud was retained in all trawl codends for the two tows we assigned to the mud substrate. Furthermore, these two mud substrate tows would have been unduly influential in any modelling of the total data from all eight deep water soft substrate tows because of their large catches. We believe that the soft mud substrate at these two stations, located at the innermost reaches of the deep stratum, is an atypically compact substrate, with higher silt content, than that found throughout the remainder of the deep stratum. This explains why it was retained within the small mesh codends. Regardless of the distinction between the two soft substrate types, the effects of hard, coarse substrates versus soft substrates on capture probability were clear. Model-predicted capture efficiency was highest (0.70 and 0.95) for the largest males (about 126 mm CW) on the softer substrates and was lowest (0.06) for smallest males (about 21 mm CW) on the hardest substrate. The higher trawl capture efficiency of American Plaice, when compared to Snow Crab, on the coarse substrate is a reflection of the difference in reaction behaviour between species to the main footgear. While Snow Crabs can escape beneath a bouncing footgear, plaice tend to rise up in front of the footgear when disturbed (Walsh, 1996; Godø et al., 1999) and so are more efficiently captured. Trawl capture efficiency for crab was greater on the soft mud substrate than on the mud–sand substrate because the rockhopper disks of the Campelen would have sunk deeper into the mud than into the mud–sand substrate, bringing the footrope closer to the bottom on mud and decreasing the escapement area. We believe the footgear on the secondary trawls was more frequently in contact with the substrate than was the rockhopper gear of the main trawl. However, the bottom contact sensor data showed that the secondary trawl footgear was not consistently in contact with the bottom, especially on hard substrates. Our size-capture efficiency relationships for crab, excluding that for the two deep stations on mud substrate, were similar to relationships derived for both Snow Crab and Tanner Crab in the eastern Bering Sea (Somerton and Otto, 1999) in that capture efficiency was highest for largest crabs. The size-specific capture efficiency for Snow and Tanner Crab larger than about 40–50 mm CW in that earlier study, were higher than our estimates on soft mud–sand and coarse substrates but not as high as our estimates for the mud substrate. These differences in size selection of larger crabs probably reflect differences in substrates between the eastern Bering Sea and our experimental area. Higher efficiency of the

78

E.G. Dawe et al. / Fisheries Research 101 (2010) 70–79

83–112 Eastern Trawl used in the Bering Sea survey, may also be due to the much smaller and compact design of the footrope (i.e., a rubber-encased cable (Somerton and Otto, 1999)) which should maintain better contact with the substrate when compared with the larger Campelen 1800 shrimp trawl rockhopper footgear. The assumption underlying our capture efficiency estimates, that the secondary trawls were fully efficient in capturing all crabs that passed underneath the main trawl footgear (E = 1.0) was not valid. Bottom contact of the secondary trawl footgear was better on soft substrates than on hard substrates. However it was evident that the footgear of the secondary trawls frequently lifted off bottom on coarse substrates, thereby permitting an unknown amount of escapement of crabs underneath both footgears. Therefore, we conclude that our estimates of capture efficiency over-estimate the true values. Our standardized annual estimates of exploitation rate for the northern Grand Bank crab fishery over-estimated the true exploitation rates, as reflected by estimates exceeding 1.0 in 2001 and 2007. This overestimation is a function of underestimation of swept area biomass due to low capture efficiency of the survey trawl. Those standardized exploitation rates, exceeding 0.6 since 2000, are also inflated because the lowest capture probability function was applied to areas for which a substrate type could not be assigned. Our exploitation rates for recent years, even for survey areas with near complete substrate classification, are unrealistically high considering that the fishery does not target the entire area but is concentrated within a limited depth range (Dawe et al., 2006, 2008). The particularly high estimates, exceeding 1.0 in two years, suggest a change in crab distribution and/or a change in rigging of the trawl which would affect capture efficiency in those years. Overestimation of true capture efficiency of Snow Crab in our experiment was due to low efficiency of our secondary trawls, which were attached to a 200 mm diameter rubber footrope. It is noteworthy that the Bering Sea 83/112 Eastern trawl, which used a 52 mm diameter rubber wrapped cable as a footgear (designed to remain in contact with the bottom), showed low capture efficiency for small and medium size snow and Tanner crabs (Somerton and Otto, 1999). Future experiments to estimate trawl capture efficiency should consider using smaller or alternate footgear on the secondary trawls to increase efficiency of the secondary trawls. Our experiment showed that the capture efficiency of Snow Crab by our multi-species Campelen survey trawl was strongly related to substrate type and crab size, and that capture efficiency was low, even for largest crabs. However our experiment failed to quantify absolute capture efficiency for Snow Crab due to bias associated with poor and variable contact of secondary trawl footgear with the substrate. Such large measurement errors confound any attempts to estimate absolute abundance or biomass. Furthermore, survey measurement errors are expected to be especially large, and variable, if there is a change in the distribution of crabs and/or a change in gear performance due to a change in rigging of the trawl. Such effects would confound interpretation of indices of abundance or biomass. Further experiments are needed to fully elaborate the relationship of capture efficiency with crab size and substrate type, and to reliably estimate capture efficiency of the survey trawl. Comparison of trawl-based biomass or density estimates with those from trawl-independent methodologies may offer the most promise for quantifying capture efficiency. Direct observation of trawl and Snow Crab encounters using video technology may prove to be useful. This would provide a basis for standardizing survey catches for efficiency effects and refining survey-based estimates of biomass and exploitation rates. We recognize that the number of tows in

each substrate category may lower the statistical power of the analysis but feel confident that the high catches in most substrates add some statistical weight to our conclusion. Our experiment demonstrated that substrate has an important effect on trawl and capture efficiency and that, with low trawl capture efficiency, slight annual changes in trawl configuration or in Snow Crab distribution in relation to substrate type will strongly affect biomass estimates. Therefore it would be prudent to consider such survey estimates as relative indices, and to exercise caution in interpreting their long-term trends. Acknowledgements We thank Barry McCallum and Paul Beck of the Northwest Atlantic Fisheries Centre for their expert assistance in conducting this work. We wish to acknowledge Scott McEntire of NOAA (National Oceanic and Atmospheric Administration) for his technical support and kind loan of the bottom contact sensors. We also thank Geoff Evans and Don Stansbury for reviewing an earlier version of this manuscript. The authors are indebted to Prof. Tadashi Tokai at Tokyo University of Marine Science and Technology for consultation on model selection and use of his SOLVER in EXCEL routines for estimating capture probability. References Alverson, D.L., Pereyra, W.T., 1969. Demersal fish explorations in the Northeastern Pacific Ocean-an evaluation of exploratory fishing methods and analytical approaches to stock size and yield forecasts. J. Fish. Res. Board Canada 26, 1985–2001. Biron, M.L., Savoie, R., Campbell, R., Wade, E., Moriyasu, M., DeGrace, P., 2002. Assessment of the 2001 Snow Crab (Chionoecetes opilio) fishery off eastern Nova Scotia (CFAs 20-24). DFO Can. Sci. Adv. Sec. Res. Doc. 2002/011, 93 p. Burnham, K.P., Anderson, D.R., 1998. Model Selection and Inference: A Practical Information-theoretic Approach. Springer, New York (p. 353). Dawe, E.G., Colbourne, E.B., 2002. Distribution and demography of Snow Crab (Chionoecetes opilio) on the Newfoundland and Labrador shelf. In: Paul, A.J., Dawe, E.G., Elner, R., Jamieson, G.S., Kruse, G.H., Otto, R.S., Sainte-Marie, B., Shirley, T.C., Woodby, D. (Eds.), Crabs in Cold Water Regions: Biology, Management, and Economics. University of Alaska Sea Grant, AK-SG-02-01, Fairbanks, pp. 577–594 (p. 876). Dawe, E.G., Mullowney, D., Stansbury, D., Parsons, D.G., Taylor, D.M., Drew, H.J., Veitch, P.J., Hynick, E., Skanes, K., O’Keefe, P.G., Beck, P.C., 2006. An assessment of Newfoundland and Labrador Snow Crab in 2005. DFO Can. Sci. Advis. Sec. Res. Doc. 2006/031, 129 p. Dawe, E., Mullowney, D., Stansbury, D., Taylor, D., Colbourne, E., Hynick, E., Veitch, P., Drew, J., O’Keefe, P., Fiander, D., Stead, R., Maddock-Parsons, D., Higdon, P., Paddle, T., Noseworthy, B., Kellend, S., 2008. An assessment of Newfoundland and Labrador Snow Crab in 2006. DFO Can. Sci. Advis. Sec. Res. Doc. 2008/009, 138 p. Doubleday, W.G., 1981. Manual on groundfish surveys in the Northwest Atlantic. NAFO Sci. Coun. Studies 2, 7–55. Dufour, R., Dallaire, J.-P., 1999. Le crabe des neiges de l’estuaire et du nord du golfe du Saint-Laurent: Etat des populations de 1995 a 1998. DFO Can. Sci. Advis. Sec. Res. Doc. 99/19, 46 p. Godø, O.R., Walsh, S.J., Engås, A., 1999. Investigating density dependent catchability in bottom trawl surveys. ICES J. Mar. Sci. 56, 292–298. Hébert, M., Wade, E., Surette, T., Moriyasu, M., 2002. The 2001 assessment of Snow Crab stock in the southern Gulf of St. Lawrence (Areas 12, E and F). DFO Can. Sci. Adv. Sec. Res. Doc. 2002/013, 65 p. McCallum, B., Walsh, S.J., 2002. An update on the performance of the Campelen 1800 during the bottom trawl survey in NAFO subareas 2 and 3 in 2001. NAFO SCR Doc. 02/32, 16 p. McCallum, B., Walsh, S.J., 1996. Groundfish survey trawls used at the Northwest Atlantic Fisheries Centre, 1971-present. NAFO Sci. Coun. Studies 29, 93–103. Moriyasu, M., Wade, E., Sinclair, A., Chaisson, Y., 1998. Snow Crab, Chionecetes opilio, stock assessment in the southwestern Gulf of St. Lawrence by bottom trawl survey. In: Jamieson, G.S., Campbell, A. (Eds.), Proceedings of the North Pacific Symposium on Invertebrate Stock Assessment and Management, vol. 125. Can. Spec. Publ. Fish Aquat. Sci., NRC Research Press, pp. 29–40. Munro, P.T., Somerton, D.A., 2000. Maximum Likelihood and non-parametric methods for estimating trawl footgear selectivity. ICES J. Mar. Sci. 58, 220–229. Somerton, D.A., Otto, R.S., 1999. Net efficiency of a survey trawl for Snow Crab, Chionecetes opilio, and Tanner Crab, C. bairdi. Fish. Bull. 97, 717-625. Tokai, T., Omoto, S., Sato, R., Matuda, K., 1996. A method of determining selectivity curve of a separator grid. Fish. Res. 27, 51–60.

E.G. Dawe et al. / Fisheries Research 101 (2010) 70–79 Tokai, T., 1997. Maximum likelihood parameter estimates of a mesh selectivity logistic model through SOLVER on MS EXCEL. Bull. Jpn. Fish. Oceanogra. 61, 288–298. Walsh, S.J., 1992. Size dependent selection at the footgear of a groundfish survey trawl. N. Am. J. Fish. Manag. 12, 625–633. Walsh, S.J., McCallum, B.R., 1996. Performance of the Campelen 1800 shrimp trawl during the 1995 Northwest Atlantic Fisheries Centre autumn groundfish survey. NAFO Sci. Coun. Studies 29, 105–116. Walsh, S.J., 1996. Ecology, resource surveys and management of Long Rough Dab (American Plaice), Hippoglossoides platessoides, (Fabricius) in the Barents Sea and the Newfoundland-Labrador shelf. Ph.D.Thesis. Department of Mathematics and Natural Science, University of Bergen, Norway. ISBN 8277440340, 185 p.

79

Weinberg, K.L., Otto, R.S., Somerton, D.A., 2004. Capture probability of a survey trawl for Red King Crab (Paralithodes camtschaticus). Fish. Bull. 102, 740–749.

Further reading Dawe, E.G., Drew, H.J., Veitch, P.J., Turpin, R., Seward, E., Beck, P.C., 2003. An assessment of Newfoundland and Labrador Snow Crab in 2002. DFO Can. Sci. Advis. Sec. Res. Doc. 2003/025, 67 p.