Fishing for Atlantic cod (Gadus morhua) with pots and gillnets: A catch comparison study along the southeast coast of Labrador

Khanh Q. Nguyen, Corey J. Morris

Fisheries and Oceans Canada, Northwest Atlantic Fisheries Centre, St. John’s, Newfoundland, A1C5X1, Canada

Keywords:

Catch rates

Cod pots

Bycatch reduction

Gillnets

Conservation

Atlantic cod fishery

A B S T R A C T

Gillnets are the primary fishing gear used to catch Atlantic cod (Gadus morhua) in coastal areas of Newfoundland and Labrador but are known to catch non-target species and produce lower quality fish than live-catch methods.The purpose of this study was to compare the catch efficiency of collapsible cod pots against gillnets near the Gilbert Bay Marine Protected Area (MPA), Labrador, where a reduction in by-catch is needed. Results showed that one cod pot caught half as much Atlantic cod as a gillnet, and pots had a relatively stable catch rate throughout the sampling period, whereas gillnets caught relatively few fish later in the sampling season.Moreover, nearly all fish sampled with pots were caught alive undamaged, whereas most gillnet-caught fish experienced damage or mortality. As a conservation measure pot- fishing would enable fishing within close proximity to the Gilbert Bay MPA and the live-release of visually distinguishable Gilbert Bay cod, thereby mitigating against the negative effects of commercial fishing on this protected Atlantic cod population.

1.Introduction

The Gilbert Bay Marine Protected Area (MPA) was created in 2005 to protect its genetically distinct and slow growing resident population of Atlantic cod against commercial fishing for northern cod. The Gilbert Bay MPA monitoring program conducted since 1998 indicates that these population differences increase the susceptibility to commercial fishing,and indeed fishing within the home range of Gilbert Bay cod in areas outside the MPA, as described by Morris et al. (2014), has negatively impacted the Gilbert Bay cod population (Morris & Green, 2017). Gillnets have been the primary gear-type used to commercially capture Atlantic cod in the study area in recent decades. Although Gilbert Bay cod is the same species as Atlantic cod, it is distinguishable by colour, in that Gilbert Bay cod appear more reddish to golden-brown in colour.Recent genomic evidence also indicates that Gilbert Bay cod are locally adapted, thereby contributing important intraspecific biodiversity to Atlantic cod (Sinclair-Water et al. 2018) that warrants conservation.Gilbert Bay cod overwinter and spawn inside Gilbert Bay and within the MPA boundaries, but migrate to coastal areas briefy during summer to areas outside MPA boundaries before returning to the MPA in fall(Morris et al., 2014; Morris & Green, 2002). Commercial Atlantic cod fishing, that targets the northern cod stock, takes place on traditional fishing ground adjacent the MPA along the Labrador coast. Commercial landings within the research area, adjacent the MPA but within the home range of Gilbert Bay cod, in 2016 were 28.4 metric tons accounting for CAD $48,000 in landed value (Morris & Green, 2017). The commercial northern cod fishing season in this area starts as early as June and is completed by December, however based on ultrasonic tracking studies to monitor seasonal migration and movement patterns the greatest impact of fishing on Gilbert Bay cod occurs when Gilbert Bay cod have migrated outside MPA boundaries in summer, July and August (Green & Wroblewski, 2000; Morris et al., 2014; Morris & Green,2002, 2014, 2017). During the past two decades the number of active fishing vessels in this small region has varied from 3 to 15 vessels per year (Morris & Green, 2017). While fishing effort and season has varied over time, due to fish availability and market conditions, in recent years fishing effort has increased and the season has lengthened which could affect the populations recovery (Morris & Green, 2017).

Gillnets are a passive stationary fishing gear, that is both easy to use and highly effective, capture depends on fish actively having physical contact with the gear during their natural diel movements or seasonal migration (Rudstam et al., 1984). Gillnets are among the most used fishing gears by commercial and artisanal feets around the world(Gabriel et al., 2005). Currently, 85% of Atlantic cod license-holders in the Newfoundland and Labrador use gillnets (DFO, 2019). Despite the popularity of gillnet fisheries, it is criticized for high bycatch rates and mortality of invertebrates, elasmobranchs, seabirds, and marine mammals (Benjamins et al., 2008, 2010, 2012; Shester & Micheli, 2011;Bayse & Grant, 2020). Lost, abandoned and/or discarded gillnets continue catching target and non-target species, a phenomenon known as “ghost fishing”, and causes negative impacts on the marine environment (Standal et al., 2020).

The northern cod fishery targets mature Atlantic cod, ≥45 cm total length (TL) (DFO, 2019). Fishery regulations permit a mesh size between 140 and 166 mm (5.5-6.5 inches) to select adult cod and reduce the incidental catch for other species. A maximum of 6 nets of 91 m (50 fathoms) each is permitted per license, and the soak time is less than 48 h (SFPP, 2016). Fishing vessels employed in the inshore cod fishery typically range from 5 to 10 m in length with 2-4 crew members,powered by a range of both gas and diesel engines, and is usually conducted within 12 nautical mile from the coast.

The reduction of bycatch and discards is a priority for fisheries worldwide, by means of measures to improve selectivity and to preserve the environment (FAO, 2011, p. 73). In Canada, the Standing Committee of Fisheries and Ocean, a federal organization, has encouraged the Atlantic cod (NAFO 2J3KL) stewardship fishery to transition to lower impact gear types including cod pots (Simms, 2017). Pots might be more acceptable to fisheries management than gillnets, or other types of fishing gear such as baited long-lines and handlines, because pots capture live high-quality fish (Suuronen et al., 2012) and have limited impact on by-catch (not hooked or damaged) that can be released alive.Ensuring high quality catch and survival of by-catch post release is an important conservation measure for the Gilbert Bay MPA (Morris et al.,2014; Humborstad et al., 2016). To ensure realism representing commercial fishing efforts, our study design considered the sampling ability of typical commercial vessels and crew involved in this inshore Atlantic cod fishery. Since pot technology is still developing, and could likely be improved further with increased usage to modify the gear and increase catch rate and efficiency, this study is a valuable contribution to improve future MPA and fishery management decision making. In addition, ghost fishing by pots can be controlled using biodegradable twine, which is currently industry regulation in the snow crab fishery (Winger et al.,2015). Existing regulations for cod pots stipulate dimensions of no more than 2 m by 2 m by 1 m high, constructed of 5/8 round stock steel as the frame, with the an exterior netting mesh size of 100 mm. Each inshore vessel is allowed to fish with a maximum of 15 pots (SFPP, 2016).

In the process of developing and improving fishing gear technology to support sustainable fishing practices, it is important to evaluate and compare catch rates between novel and existing gear types to demonstrate effectiveness. The purpose of this study is to comparatively assess the performance of Atlantic cod fishing using traditional gillnets and a newly developed collapsible pot design during commercial fishing. The impetus of our study was to evaluate the commercial potential for this new gear type to reduce the fishing mortality of Gilbert Bay cod, in areas adjacent the existing MPA boundaries, through the live capture and release using pots. The two fishing methods were investigated over part of a fishing season near the Gilbert Bay MPA, located along the southeast coast of Labrador, Canada, with regard to species composition and seasonal fuctuation in Atlantic cod catch and demographic structure.

2.Methods

2.1.Field experiment

The field experiment was conducted during the annual Atlantic cod stewardship fishery in 2019 at two well-known commercial fishing locations near the mouth of the Gilbert Bay Marine Protected Area, at Copper Island (Site 1) and West Island (Site 2) (Fig. 1). The study was conducted from August 22 to September 8 (de fined as a summer season)and from September 25 to October 2 (de fined as a fall season). The fishing locations were on traditional commercial fishing grounds,selected by experienced commercial harvesters that fished in this area.The seabed substrate mainly consists of gravel, gravelly-sand, and gravelly sandy mud, which gillnets can be placed using anchors and pots remained stable on the sea bed.

Fig. 1.Map of study site, boundary of the Gilbert Bay Marine Protected Area(MPA) represented by Zone 1a,b, Zone 2, and Zone 3, southeast coast of Labrador. Copper Island is de fined as the sampling Site 1, and West Island area is de fined as the sampling Site 2. Commercial fishing is prohibited in the MPA.

The study fished using single gillnets that are typical for the commercial cod fishery in Newfoundland and Labrador, with 0.7 mm mono filament, 140 mm diamond stretched mesh opening, 91 m (50 fathoms) in length, 7 m (50 meshes) in stretched height, and 0.5 hanging ratio, (DFO, 2019). Norwegian-style collapsible pots were also used,consisting of round-stock steel frame with 1.2 m in height ×1.5 m in length ×1 m in width, covered with green polyethylene netting with 100 mm mesh opening. This pot type was similar to the pots tested in other cod fisheries for a variety of purposes in the Baltic and the Barents Sea (Furevik et al., 2008; Bryhn et al., 2014; Königson et al., 2015;Jørgensen et al., 2017; Humborstad et al., 2018). The pot was divided by a mesh false bottom that extends midway through the horizontal axis of the pot, creating two equal chambers. A slit in the false bottom mesh allowed cod to enter the upper chamber to help minimize the escape rate. Pots were closed using an overhand knot in the mesh wall of the lower and upper chamber for fish removal and baiting the pot. The two entrances of the pot faced each other from opposite directions within the lower chamber. The entrance dimensions were 0.6 m height ×0.5 m width for outside and 0.15 m height ×0.25 m width for inside. Additional pot con figuration details are described in Meintzer et al. (2018).

To conduct fishing, we deployed gillnets and cod pots at the same time so that each gear type was fished under similar environmental conditions. Pots and gillnets were set at least 200 m apart, which was considered far enough to not affect neighboring fishing gear (i.e. bait effect). We fished 15 pots and two gillnets during each deployment, at each of our two sites, and redeployed fishing gear at the same locations for each set. Fishing gear was set singularly and sank on the seabed, in that all pots were fished in a single-pot con figuration with a single buoy line and each gillnets had a buoy at each end. Pots were baited with ～1 kg (4 fish) of fresh Atlantic herring (Clupea harengus) placed within a mesh bag hanging from the center of the pot (～30 cm off bottom).

For each haul and deployment of gear, we recorded the date, time,soak duration, bottom depth and location (coordinates). For each pot and gillnet retrieved, all individuals captured were separated, counted,and recorded as the catch rate per fishing trial. The total length (mm) of cod caught by both gillnets and pots were measured to compare length frequency distributions sampled by each gear type. Other species were counted and recorded for bycatch analysis, but the lengths were not measured. The gillnet bycatch species were noted with presence or absence in the catch. Only cod ≥45 cm, the minimum legal size (MLS),were retained for commercial purposes. All other individuals including cod under the MLS and bycatch species were immediately returned alive to the water. The number of alive, damaged, or dead fish upon capture was also visually assessed and noted to discuss fish quality obtained from pots and gillnets.

For comparison purposes, the catch rate of pots and gillnets was standardized by the same sweep area and catching duration. For the gillnet, sweep area was determined as the length multiplied by the height of the net, while for pots, it included pot diameter and an expected attractive radius of the bait based on established commercial pot fishing practices (Königson et al., 2015; Petetta et al., 2020; Poirier et al., 2020; Ganias et al., 2021). Specifically, the gillnet catch was standardized for the number of individuals captured by 91 m of net in 24 h, typical for this fishery in Newfoundland and Labrador (Benjamins et al., 2008; Rouxel & Montevecchi, 2018). The commercial pot fishery commonly uses several pots spaced 45.5 m (25 fathom) apart (Bryhn et al., 2014; Jørgensen et al., 2017; Meintzer et al., 2018) and soaked 24 h, which is believed to represent an area over which bait is most effective (Bryhn et al., 2014; Humborstad et al., 2018). The pot catch was standardized by an assumed area of bait attraction distance, 45.5 m radius, based on established spacing used in the commercial fishing industry (Humborstad et al., 2018; Meintzer et al., 2017; See Bryhn et al., 2014). Thus, based on similar catch the effort was standardized at a 2:1 pots to gillnet ratio (identical sweep areas), for a 24 h fishing period. This duration represented the typical commercial fishing time for experimental pots and gillnets fished, although the soak time varied during the sea trials due to the weather and operational logistics.

2.2.Statistical analysis

Generalized linear mixed-effect models (GLMMs) based on the Poisson regression was applied to identify differences in standardized CPUE between gillnets and pots, following procedures outlined in Zuur et al. (2016). Our catch data violated many of the assumptions needed for parametric approaches (e.g. ANOVA), which were count data and not normally distributed, as well as gear deployment nested within sampling day. Statistical techniques such as GLMMs enable us to measure the effect of gear types on CPUE, while accounting for the non-normal distribution and data with nested structure such as ours. Initially, we modelled using thelme4 package(Bates et al., 2013), with R statistical software (R Development Core Team, 2020). The dependent variable was standardized CPUE, the independent variables included gear types(gillnet, pot), deployment depth (m), sampling sites (Copper Island and West Island at the northern and southern entrances to the Gilbert Bay MPA), fishing season (Summer and Fall), and we added sampling data(day of year) as the random effect on the intercept to account for the variation in catch rates over time. A preliminary analysis indicated that zero-infation was not significant. No significant interaction between fixed factors were detected using thecoplotfunction, and it was therefore dropped from the models. We fit the following model:

whereyis the standardized CPUE, α is the intercept,β1,2,3,4is regression parameters,GearTypeis the treatment factors (gillnet versus pot),Siteis the deployment site (Northern and Southern sites),Depthis the deployment depth,Seasonis deployment time (Summer and Fall), andbis the random variable representing the variability among fishing dates(whereb～ N [0, σ2]), andεis the error term. Since the soak time was standardized by 24 h, it was not included in the model, although the effect of soak time on the catch rate using original count data (raw data)was separately estimated.

First, our models were analyzed using a Poisson distributed model with theglmerfunction. The zero infation and dispersion were determined via simulations with theDHARMa package(Hartig, 2017). If over-dispersion was determined (dispersion＞1.0), then a negative binomial model was applied using theglmerfunction. The analysis was conducted with the Poisson distributed model when equi-dispersion(dispersion ～ 1.0) was observed, and fit with a quasipossion model using the functionglmmPQLin the packageMASSwhen under-dispersion (dispersion＜1.0) was determined. Evidence of zero infation was considered using GLMMs in theglmmTMB package(Bolker,2016). GLMMs were also used to undertake a pairwise post hoc comparisons where there were differences in CPUE between gear types,fishing locations, and fishing seasons.

In order to evaluate the catch efficiency between two gear types in terms of the size of the individuals caught, mean length, and length frequency distribution of cod caught by pots and gillnets were investigated. Comparison of the mean cod length caught by different treatments was conducted using ANOVA. Fish length was the dependent variable and experimental treatment was the independent variable. The model structure was built as follows:

whereαis the intercept,βis the regression parameter (slope),GearTypeis the treatment factors (gillnet versus pot), and ε is the error. The 95% CI was calculated using aTukeyHSD functionin R. We validated the model and found that all assumptions needed for a parametric tests were met with regard to homogeneity of variance, normal distribution of errors,and independence of errors. Fish length frequency distributions were compared between pots and gillnets using Kolmogorov-Smirnov twosample Z test. ANOVA and Kolmogorov-Smirnov two-sample Z test were also used to examine the mean length and length frequency distribution of cod caught in different seasons and sites, respectively.

3.Results

3.1.Catch and body length comparison

Between August and October of 2019, 262 successful cod pot deployments and 22 gillnet deployments were conducted, which were analyzed in this study (Table 1). An additional 14 pots and one gillnet were deployed but became disabled or broken and were excluded from analysis. Pots were soaked between 13.8 h and 73.2 h with a mean of 27.6 ±0.71 standard error (SE). Gillnets were soaked from 13.8 h to 25.3 h (mean of 22.8 ±0.64 SE). Fishing depths for both gear types varied between 8 and 37.1 m (mean 19.5 ±0.24 SE).

Table 1Summary details for the comparative fishing experiment.

We caught 1538 Atlantic cod in pots and 196 in gillnets (Table 2).The number of cod caught per set by pot and gillnet varied from 0 to 25,and from 0 to 38, respectively. Pots captured 99.7% of fish alive,compared to 44.8% for the gillnets of which all fish were affected by entanglement in the mesh. Rarely were sublegal-sized cod caught by gillnets, thus, we compared the standardized CPUE between gillnets and pots using data for legal-sized fish and all sizes combined; data for undersized fish were not suf ficient for statistical analysis.

Table 2Summary of catch and bycatch species caught during the experiment. Y and N denote the species presence or absence in the catch, respectively.

For the legal-sized cod comparison, the fitted model for the Poisson distribution showed indications of overdispersion (dispersion =4.8,P ＜0.001), therefore the negative binomial distribution was used. Dispersion of the negative binomial model was calculated at the 0.9 level, and no patterns were observed in the plot of residuals versus fitted values,predictor, and random variable. The validated model for legal sized cod failed to detect differences between pots and gillnets (P=0.569,Table 3). The mean standardized CPUE of legal-sized cod was 8 for pots(CI: 6.37-9.84) and 7.38 for gillnets (CI: 5.76-8.8). The fishing depth had a significant effect on catch rate, with more fish being caught in the shallower depths (P ＜0.001, Table 3). The catch rate at Site 1 was significantly higher than Site 2 (P ＜0.001, Table 3) and significantly higher catch rates were observed during Summer compared to the Fall sampling period (P=0.014, Table 3)

Table 3Parameter estimates, fit statistics, and variation from the random effect of a GLMM model for cod with pooled data using retrieving date as a random factor(n =25). SE is the standard error of the estimate and SD is the standard deviation.

For all sizes combined, the Poisson model results indicated that the catch data were over-dispersed (dispersion =1.24,P ＜0.001), thus, it was fit using a negative binomial distribution. Based on the mean standardized CPUE, which was 2 pots being equal to 1 gillnet, the mean standardized CPUE was 9.23 for pots (CI: 7.57-11.11) and 7.55 for gillnets (CI: 4.2-13.75) (Fig. 2), the two fishing sites were pooled for comparative analysis. While pots caught 5.8% (CI: -4.4% - 24%) more fish than gillnets, statistical models failed to detect significant differences in standardized CPUE between the pot and gillnet (P=0.754,Table 3) over the entire fishing season.

Fig. 2.Boxplots of standardized CPUE of cod classified by sampling sites for the different fishing gear types.

Catch rates were significantly different between sites, but similar among pots and gillnets within each site (Table 3, Fig. 2). The standardized CPUE for pots was 11.92 (CI: 9.45-15.15) at Site 1, and captured 40.64% (CI: 25%-53%) more cod than at Site 2 (mean CPUE of 7.08 (CI: 4.45-11.31)), which was a statistically significant difference(P ＜0.001; Fig. 2). Gillnets at Site 1 also caught significantly (71.22%;CI: 58%-81%) more cod than at Site 2 (P ＜0.001; Fig. 2), corresponding with a mean of 11.21 (CI: 5.23-22.36) and 3.23 (CI: 1.01-9.34),respectively (Fig. 2). Fishing depth also affected the CPUE for both pot and gillnet (P=0.002), such that more fish were caught in shallower water (Table 3).

Fishing season had a significant effect on the standardized CPUE for gillnet data (Table 3, Fig. 3), decreasing from 17.75 in summer (CI:4.5-65.76) to 2.8 in Fall (CI: 1.49-4.8), a 534% reduction (CI: 202%-1270%) (Fig. 3). There was no significant differences in the standardized CPUE for pots between summer (mean of 10.38, CI: 5.34-20.03) and fall(mean of 7.57, CI: 5.53-10.1) (P=0.073; Fig. 3).

Fig. 3.The effects of fishing season on standardized daily catch of cod caught by different fishing gear types for two sampling sites combined. A daily weighted smoother and associated 95% con fidence interval envelope are superimposed on the data.

There was a positive relationship between the catch rate and soak time for the pot using original count data; every additional hour of soak time increased the average catch rate by 2.12% (Table 4). The catch rate did not increase with time for the gillnet (Table 4). The average catch for pots soaked 24 h was 4.56 fish per pot, and 6.74 fish per pot when fished greater than 24 h, indicating higher rates in the first 24 h compared to the additional day of fishing. The gillnets soaked less than 26 h, thus the average catch of gillnet was not calculated for soak times greater 24 h.

Table 4Parameter estimates, fit statistics, and variation from the random effect of a GLMM model for gear type deployed at different soak times. SE is the standard error of the estimate.

A total of 1683 cod (161 cod for gillnets and 1522 cod for pots) were measured for total length (TL) during the experiment. Legal sized fish(≥45 cm) dominated the catch in both pots and gillnets. The TL ranged from 37.2 to 91.5 cm for pots and 41.5 to 91.5 for gillnets (Fig. 4). The proportion of legal-sized and sublegal-sized cod accounted for 89.5% and 10.5% for pot, and 99.4% and 0.6% for gillnet, respectively. Pairwise comparisons indicated fishing gear types had a significant effect on cod length distribution (D =0.51,P ＜0.001) using Kolmogorov-Smirnov test (Fig. 4). Mean body length was 55.48 cm (CI: 55.1-55.9)for pot and 63.22 cm (CI: 61.4-65) for gillnet, which was significantly different using ANOVA (deviance: 7.74, F =122,P ＜0.001).

Fig. 4.Length frequency distribution of cod recorded in the different fishing gear types. Vertical dashed lines represent mean length of cod caught by pots and gillnets.

Fishing season affected the length distribution of cod caught between summer and fall for both pots (D =0.18,P ＜0.001) and gillnets (D =0.53,P ＜0.001). Mean body length of cod caught by pot in summer (n=1173) was 56 (CI: 55.5-56.5) cm, significantly larger than in fall (n=349), with mean of 53.84 (CI: 52.3-55.4) cm (deviance: 2.13, F =16.57,P ＜0.001; Fig. 5). By contrast, the gillnet captured larger cod in fall than in summer (deviance: 6.32, F =15.83,P ＜0.001). Mean body length of cod caught by the gill net was 62.48 (CI: 61.4-63.6) cm and 68.8 (CI: 64.6-73) cm in Summer (n=142) and Fall (n=19), respectively (Fig. 5).

Fig. 5.Length frequency distribution of cod recorded in the different fishing season for the pot (A) and the gillnet (B). Vertical dashed lines represent mean length of cod caught in Summer and Fall, respectively.

3.2.Bycatch

The length distribution of cod was different between sampling sites(D =0.12,P ＜0.001; Fig. 6). Mean body length of cod sampled at Site 1(Copper Island) was 57.13 cm (56.46 cm for pot and 64.58 cm for gillnet) (CI: 56.6-57.7), significant larger (deviance: 2.28, F =27.92,P＜0.001) than fish sampled at Site 2 (West Island) which was 54.85 cm(54 cm for pot and 61.75 cm for gillnet) (CI: 53.5-56.2, Fig. 6).

Fig. 6.Length frequency distribution of cod recorded in the different sampling sites for the pot (A) and the gillnet (B). Vertical darkblue and darkgreen dashed lines represent mean length of cod caught in the Site 1 and Site 2, respectively.

Pots caught eight species and gillnets caught five species of by-catch(Table 2). No rock cod (Gadus ogac) were caught by gillnets, compared to 580 individuals for pots. Toad crab (Hyas araneus) dominated the bycatch in pots, followed by rock cod and sculpin (Myoxocephalussp.)(Table 2). Currently, none of these species have commercial value in this area.

4.Discussion

This study compared the commercial catch of Atlantic cod from pots and gillnets, on traditional inshore fishing grounds adjacent a Marine Protected Area, and demonstrated the effectiveness of pots as conservation-based alternative to gillnets enabling live-release of bycatch. Results showed that over the course of a season, pots could capture cod in comparable amounts to traditional gillnets using a 2:1 ratio of pots to nets, while soak time, location, and season explains some of the variation in catch per unit effort. The catch rate of gillnets was highly dependent on fishing season, significantly higher during the early season and reduced during the late season, while pots maintained a more consistent catch rate. Although the proportion of large fish (≥45 cm)caught by gillnets was higher than pots, nearly all fish retained from pots(99.7%) were alive, therefore smaller cod can be returned to the water with a low mortality (Suuronen et al., 2012; Morris et al., 2014; Humborstad et al., 2016). By contrast, more than 55% of Atlantic cod caught by gillnets died or were injured by entangling in gillnet meshes during～23 h soak durations. Minimal fishing mortality was observed for cod caught using pots, therefore, non-target species including Gilbert Bay cod can be easily released alive.

While gillnets selected for nearly all commercial sized fish (≥45 cm),fish smaller than marketable size accounted for 10.5% of fish caught using pots, which is in part dependent on the demographic structure of the stock and mesh sizes used. In the Barents Sea and Baltic, pots tested using 45-50 mm mesh consisted of up to 64% undersized fish (Bryhn et al., 2014; Jørgensen et al., 2017), while another study conducted near Fogo Island of Newfoundland reported＜7% undersized fish using 100 mm mesh opening (Meintzer et al., 2018). These variations could also be related to differences in soak times and gear con figurations among the studies (Bryhn et al., 2014; Königson et al., 2015). We measured higher catch rates at Site 1 compared to Site 2 using both gear types. Other studies also measured differences in catch among fishing sites, which is not surprising for coastal fisheries that are located in areas with highly variable habitats and environmental conditions (Bryhn et al., 2014;Major et al., 2017). Although the number of sublegal-sized cod caught by pots was relatively low, and considering that this gear type is new with opportunities for improvement, changes in mesh sizes or the addition of escape mechanisms could improve the selectivity of pots(Winger & Walsh, 2007; Boutson et al., 2009; Ovegård et al., 2011;Bilkovic et al., 2012; Broadhurst et al., 2020; Zhang et al., 2020).Moreover, natural exclusion of small individuals from the pots helps reduce the time required to sort the catch.

The small-boat inshore fishery in Newfoundland and Labrador is limited by windy conditions that often prevents harvesters from checking nets on a daily basis (Nguyen & Winger, 2019a). Fishing gear can soak several days during wind events which significantly reduces the quality of gillnet-cod (Standal et al., 2020), however, fish caught in pots are retained alive for several days (Meintzer et al., 2018) and catch rates continued to increase after 24 h as shown in this study. For example,compared to gillnets over a ～24 h soak period could yield 58% grade A catch and 23% grade C catch, based on assessment of the fillet quality which considers fillet colour, texture, bruising, staining, blood clots,ripping, temperature, parasites, and odour (SFPP, 2016; Meintzer et al.,2018). By contrast, using cod pots, 94% of the landings produced grade A fillets (Meintzer et al., 2018). Harvesters using gillnets often reduce soak times, often hauling nets twice during a single day to increase fish quality when catch rates are high to ensure high quality. Commonly,there are marketing incentives, such as pricing bonuses, paid to harvesters that supply a consistently high proportion of high-quality catch(AFPP 2016), to deter poor quality landings from gillnets which has been a marketing concern for the Newfoundland and Labrador Atlantic cod fishery for some time (Rouxel & Montevecchi, 2018; Simms, 2017).

Compared to gillnets pots are considered environmentally friendly however a transition from the gillnet fishery to cod pots has not occurred, largely because catch volume has been lower than gillnets(Suuronen et al., 2012; Meintzer et al., 2018). Moreover, fishing with pots requires added bait costs (Winger et al., 2016) and could require vessel or pot modifications to use the new gear more efficiently. Cod pots are a relatively new type of fishing gear for cod in our study area, and there is both uncertainty in its effectiveness and increased startup costs for harvesters to switch gear-types, which also contributes to limited usage of cod pots commercially. Gillnets are low cost, and harvesters have become accustomed to their usage. Fisheries are however changing, with respect to their management and regulations globally as conservation requirements increase in general. Although harvesters are not in favor of closing this area to fishing the abundance of Gilbert Bay cod has decreased, therefore in order to improve MPA effectiveness it is important for managers to consider adaptive management options based on available information (Morris & Green, 2014). This study provides information for both harvesters and managers to evaluate cod pots as a potential mitigation measure to address concerns regarding effectiveness of the Gilbert Bay MPA and the surrounding cod fishery.

While the gillnet has become part of the social license to catch inshore Atlantic cod in Newfoundland and Labrador, change is possible.For example, since May 2015 the majority of the upper Gulf of California has been declared a no gillnet zone, pots were then quickly chosen as an alternative solution (Königson et al., 2019). Increasing the catching performance of fishing gear to reduce environmental impacts while improving pro fitability is important for fisheries sustainability (Nguyen& Winger, 2019b; Tran et al., 2020). For example, artificial light can improve cod pot catch rates, perhaps as much as 80% (Bryhn et al.,2014; Humborstad et al., 2018; Utne-Palm et al., 2018; Nguyen &Winger, 2019b). Floated pots with one entrance captured 52% more cod and twice as many haddock, compared to two entrances like those used in this study (Jørgensen et al., 2017), and would likely reduce by-catch of benthic invertebrates such as the toad crab caught in this study. Bait type and amount also impacts the efficiency of pots (Winger et al.,2016). Like other aquatic species, cod are attracted to baited pots by olfaction and vision, and will move toward the bait following a counter-current route (Ljungberg et al., 2016; Meintzer et al., 2017).Behavioural observations of cod using underwater cameras showed that while cod often approach pots many do not enter them, and among those that actually enter the pot, up to 25%, were able to escape (Ljungberg et al., 2016; Meintzer et al., 2017; Humborstad et al., 2018). While we demonstrate that cod pots are a useful fishing gear for Atlantic cod,improvements can further increase the catch efficiency of existing pots,i.e. one entrance pot, pot shape and size, fotation, attraction with artificial light, and bait types (Bryhn et al., 2014; Jørgensen et al., 2017;Meintzer et al., 2018), which has not yet been thoroughly explored by the fishing industry, unlike gillnet fishing. Additionally, the catch ef ficiency of pots largely depends on animal density, level of satiation, and food availability (Hubert et al., 2012), suggesting further studies conducted in different conditions are recommended.

In this study a relatively large number of pots were used to sample Atlantic cod, compared to gillnets, largely because of the uncertainty in their catch rates compared to more commonly used gillnets. We also used a level of fishing effort per day that reasonably refected commercial fishing efforts. For comparison of the gear types, however, both were fished at the same time and on the same commercial fishing grounds where harvesters regularly fish using gillnets. While a smaller number of gillnets were used, the data collected indicated that the sample size was suf ficiently large enough (i.e. 22 gillnets vs. 262 pots,Table 1) to statistically represent the different treatments for comparison, and importantly the methodology does not require a balanced design (Schnute, 1994; Kain et al., 2015). Deviance residuals from the GLMM outputs were randomly distributed (Fig. S1, Fig. S2), while the variance for the deviance residuals vs. the fitted plot was equal (P=0.982 for legal-sized cod, andP=0.547 for all sizes combined), indicating that our applied model met the assumptions of linearity (Fig. S1A,Fig. S2A) (Fox, 2008), and homoscedasticity (Fig. S1B, Fig. S2B) (Bates et al., 2015), respectively. However, there are exceptions at high values,so the deviance residuals are larger than predicted by the theoretical quantiles (Fig. S1C, Fig. S2C). Variance of the random effect was 0.18 and 0.17 for legal-sized cod and all sizes combined (Table 3), respectively, indicating that the differences between fishing day were small.Therefore the study design enabled adequate comparison of the fishing gear, while minimizing the number of gillnets used to reduce potential fishing mortality on Gilbert Bay cod during the study.

In conclusion, this study shows that the pots tested in this study provide a promising alternative to the traditional gillnets used along the coast of Labrador. Using pots near the Gilbert Bay MPA would bene fit existing conservation efforts to protect a local population of Atlantic cod in Gilbert Bay, helping to meet MPA objectives. Moreover, pot fishing could extend the cod fishing season to include periods when gillnet catch rates are lower than that of pots, while ensuring a high quality product during periods when poor weather could further reduce the poor quality of gillnet caught cod.

CRediT authorship contribution statement

Khanh Q. Nguyen: Conceptualization, Methodology, Investigation,Formal analysis, Writing - review & editing. Corey J. Morris:Conceptualization, Funding acquisition, Methodology, Investigation,Formal analysis, Writing - original draft.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to infuence the work reported in this paper.

Acknowledgements

The study was funded by the Fisheries and Oceans Canada. This work was done in collaboration with the Nunatukavut Community Council(NCC), Fisheries and Ocean Canada, and Fisheries and Marine Institute.We are grateful to Stan Oliver, George Morris, and Patrick Davis for identifying suitable experiments and assisting with field logistics, to Philip Walsh for assisting with field logistics. We thank Gillian Forbes and Laura Park for help with data collection and support throughout the study. We also extend many thanks to Elizabeth Moore for providing means of efforts and contributions during the paper development. The field experiments could not have been accomplished without kind help and co-operation from commercial fish harvesters of Southeast Labrador. Finally, we would like to acknowledge Dr. Brett Favaro for assisting in the development of this project, consulting on the statistical analyses,and providing valuable comments on an earlier version of the manuscript.

Appendix A.Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.aaf.2021.05.006.