TY - JOUR
AU - Fernandes, Paul G
AB - Abstract Brown crab (Cancer pagurus) is a widely distributed crustacean that occurs around the British coastline supporting important commercial fisheries. The habitat preferences of brown crab around Scotland are poorly documented and for the purposes of stock assessment, the species is considered data-poor. Based on an analysis of dredge and trawl surveys taking place in the North Sea (2008–2018), we describe the spatial distribution of brown crab and for the first time, develop abundance and recruitment indices for the species. We make use of geostatistical methods and apply generalized additive models to model catch rates in relation to a number of explanatory variables (depth, distance to the coast, sediment type and year). The dredge and trawl abundance indices were correlated showing a similar trend of increasing catch rates in the early years of the time series up to 2016 and a subsequent reduction. The recruitment index showed a gradual increase in captured juvenile crabs up to 2014 followed by a steep decrease with 2018 being the lowest value estimated. The derivation of robust indicators of stock abundance will contribute to the stock assessment of this species and enable the provision of improved fisheries management advice for brown crab around Scotland. Introduction The brown crab (Cancer pagurus) is a decapod crustacean that inhabits sandy and rocky seabeds of northwestern Europe (Edwards, 1979), from the littoral zone to depths of over 100 m (Shelton and Hall, 1981). Like other crustaceans, the growth of brown crab takes place by moulting the shell, with moult frequency decreasing with age. The species reaches the minimum landing size in the Scotland of 150 mm carapace width (CW) at an age of ∼6 years (Mill et al., 2009). The brown crab is considered to be a slow-growing species with a lifespan of ∼20 years, reaching an asymptotic size (L∞) of between 220 and 246 mm in Scottish waters (Bennett, 1974; Tallack, 2002; Mill et al., 2009). It is a commercially valuable species with annual landings into Scotland in excess of 10 000 tonnes and a first sale value of £26.8 million in 2018 (Scottish Sea Fisheries Statistics, 2018). Pot fisheries for brown crabs occur along almost the entire length of the British coastline, and traditionally, these fisheries have been essentially inshore (Eaton et al., 2001). Since the late-1980s, offshore crab fisheries have developed, rapidly becoming the largest pot fisheries both in the North Sea and west coast of Scotland. More recently, the expanded demand for this species from European andemergent Asian markets makes an evaluation of its status more important (Mesquita et al., 2017). Concerns about increased landings and uncertainty over the level of exploitation have highlighted gaps in the current understanding of brown crab stocks (Bennett, 1995; Smith and Addison, 2003). Information on spatial distribution is important for understanding the population structure and dynamics of commercially valuable marine resources (Hunter et al., 2013), but there have been very few studies documenting habitat preferences of brown crab. Juveniles from different crab species prefer nursery habitats close to the shore, presumably to avoid predation before moving to different habitats, potentially representing a major source of recruitment for offshore fisheries (Edwards, 1979; Robinson and Tully, 2000; Beck et al., 2001). Brown crabs are known to be associated with rocky substrates but also inhabit mixed coarse, sand, and soft sediments, particularly offshore where they dig into mud in search of food, or bury themselves to avoid strong currents (Hall et al., 1993). Abundance indices are generally lacking for decapod crustaceans in European waters. Fisheries-dependent data obtained from vessels using pots are difficult to interpret due to problems in standardizing catch rates, which are affected by, for example, pot design, gear saturation, soak time and types of bait (Miller, 1990). The problems with commercial catch data can be avoided by using fisheries-independent research surveys, which provide estimates of abundance and catch composition across the geographic range of a species (Hilborn and Walters, 1992). Given the occurrence of brown crabs in a wide variety of sediment types, the species is often found in the same habitat as other commercial species such as scallops and demersal fish (Jenkins et al., 2004; Öndes et al., 2016). Many nations have long-term scientific research surveys, which contribute to fisheries assessment and management (Jouffre et al., 2010). The brown crab is not currently targeted by any specific fishery-independent survey, and there are no estimates available of abundance or recruitment for populations around Scotland. Despite this, catches of brown crab are routinely recorded during a number of fisheries surveys conducted by Marine Scotland Science (MSS). There are no studies documenting brown crab catch rates from surveys using mobile gears such as dredges or trawls. It is expected that analysing catch rates from these surveys may contain relevant information on habitat preferences of the species, for example in relation to substrate type, depth distribution, or distance to coast. This study provides an analysis of catch rates of brown crab between 2008 and 2018 in two fisheries-independent surveys taking place in the North Sea: a dredge survey directed at the main king scallop fishing grounds (Dobby et al., 2017) and the International Bottom Trawl Survey (IBTS) developed for demersal fish species (ICES, 2018). The aims of this study were to: (i) determine the spatial distribution of brown crab by size and sex; (ii) model crab catch rates in relation to explanatory variables (depth, distance to coast, sediment type, and year); and (iii) derive abundance and recruitment indices for brown crab. The work represents the first study directed at understanding the abundance, distribution and important habitat features of brown crab in the North Sea using fisheries-independent data. Methods Study area and sampling The study area was located in the western North Sea, off the east coast of Scotland. Brown crab data were collected during 11 dredge surveys and 11 trawl surveys (2008–2018, Table 1). The surveys were carried out by MSS in summer (between May and September) on-board the Scottish research vessels “Alba na Mara” (dredge) and “Scotia” (trawl) covering fishing stations around the east coast of Scotland (Figure 1a). Figure 1. Open in new tabDownload slide (a) Study region on the east coast of Scotland with the dredge and trawl survey areas and stations sampled (2008–2018). (b) Bathymetry map of the study area (depth in metres). (c) Distribution of marine sediments in the study area from the British Geological Survey (Cooper et al., 2013). Figure 1. Open in new tabDownload slide (a) Study region on the east coast of Scotland with the dredge and trawl survey areas and stations sampled (2008–2018). (b) Bathymetry map of the study area (depth in metres). (c) Distribution of marine sediments in the study area from the British Geological Survey (Cooper et al., 2013). Table 1. Summary information for dredge and trawl surveys in the present study: vessel used, year, month, numbers of stations and crabs sampled. Survey . Vessel . Year . Month . Number of stationsa . Number of crabsa . Dredge Alba na Mara 2008 June–July 92 241 Trawl Scotia 2008 August 38 7 Dredge Alba na Mara 2009 July 99 176 Trawl Scotia 2009 August 41 11 Dredge Alba na Mara 2010 June–July 114 261 Trawl Scotia 2010 August–September 46 14 Dredge Alba na Mara 2011 June–July 113 272 Trawl Scotia 2011 July–August 43 14 Dredge Alba na Mara 2012 May–June 117 404 Trawl Scotia 2012 July–August 40 8 Dredge Alba na Mara 2013 July 116 324 Trawl Scotia 2013 July–August 38 12 Dredge Alba na Mara 2014 June 107 610 Trawl Scotia 2014 July–August 41 19 Dredge Alba na Mara 2015 May–June 90 327 Trawl Scotia 2015 July–August 51 39 Dredge Alba na Mara 2016 May–June 115 544 Trawl Scotia 2016 August 57 35 Dredge Alba na mara 2017 June 119 345 Trawl Scotia 2017 August 50 42 Dredge Alba na Mara 2018 July 109 284 Trawl Scotia 2018 July–August 56 20 Survey . Vessel . Year . Month . Number of stationsa . Number of crabsa . Dredge Alba na Mara 2008 June–July 92 241 Trawl Scotia 2008 August 38 7 Dredge Alba na Mara 2009 July 99 176 Trawl Scotia 2009 August 41 11 Dredge Alba na Mara 2010 June–July 114 261 Trawl Scotia 2010 August–September 46 14 Dredge Alba na Mara 2011 June–July 113 272 Trawl Scotia 2011 July–August 43 14 Dredge Alba na Mara 2012 May–June 117 404 Trawl Scotia 2012 July–August 40 8 Dredge Alba na Mara 2013 July 116 324 Trawl Scotia 2013 July–August 38 12 Dredge Alba na Mara 2014 June 107 610 Trawl Scotia 2014 July–August 41 19 Dredge Alba na Mara 2015 May–June 90 327 Trawl Scotia 2015 July–August 51 39 Dredge Alba na Mara 2016 May–June 115 544 Trawl Scotia 2016 August 57 35 Dredge Alba na mara 2017 June 119 345 Trawl Scotia 2017 August 50 42 Dredge Alba na Mara 2018 July 109 284 Trawl Scotia 2018 July–August 56 20 a Number of stations and crabs sampled during each survey within the dredge and trawl areas used for this study are shown in Figure 1a. Open in new tab Table 1. Summary information for dredge and trawl surveys in the present study: vessel used, year, month, numbers of stations and crabs sampled. Survey . Vessel . Year . Month . Number of stationsa . Number of crabsa . Dredge Alba na Mara 2008 June–July 92 241 Trawl Scotia 2008 August 38 7 Dredge Alba na Mara 2009 July 99 176 Trawl Scotia 2009 August 41 11 Dredge Alba na Mara 2010 June–July 114 261 Trawl Scotia 2010 August–September 46 14 Dredge Alba na Mara 2011 June–July 113 272 Trawl Scotia 2011 July–August 43 14 Dredge Alba na Mara 2012 May–June 117 404 Trawl Scotia 2012 July–August 40 8 Dredge Alba na Mara 2013 July 116 324 Trawl Scotia 2013 July–August 38 12 Dredge Alba na Mara 2014 June 107 610 Trawl Scotia 2014 July–August 41 19 Dredge Alba na Mara 2015 May–June 90 327 Trawl Scotia 2015 July–August 51 39 Dredge Alba na Mara 2016 May–June 115 544 Trawl Scotia 2016 August 57 35 Dredge Alba na mara 2017 June 119 345 Trawl Scotia 2017 August 50 42 Dredge Alba na Mara 2018 July 109 284 Trawl Scotia 2018 July–August 56 20 Survey . Vessel . Year . Month . Number of stationsa . Number of crabsa . Dredge Alba na Mara 2008 June–July 92 241 Trawl Scotia 2008 August 38 7 Dredge Alba na Mara 2009 July 99 176 Trawl Scotia 2009 August 41 11 Dredge Alba na Mara 2010 June–July 114 261 Trawl Scotia 2010 August–September 46 14 Dredge Alba na Mara 2011 June–July 113 272 Trawl Scotia 2011 July–August 43 14 Dredge Alba na Mara 2012 May–June 117 404 Trawl Scotia 2012 July–August 40 8 Dredge Alba na Mara 2013 July 116 324 Trawl Scotia 2013 July–August 38 12 Dredge Alba na Mara 2014 June 107 610 Trawl Scotia 2014 July–August 41 19 Dredge Alba na Mara 2015 May–June 90 327 Trawl Scotia 2015 July–August 51 39 Dredge Alba na Mara 2016 May–June 115 544 Trawl Scotia 2016 August 57 35 Dredge Alba na mara 2017 June 119 345 Trawl Scotia 2017 August 50 42 Dredge Alba na Mara 2018 July 109 284 Trawl Scotia 2018 July–August 56 20 a Number of stations and crabs sampled during each survey within the dredge and trawl areas used for this study are shown in Figure 1a. Open in new tab The scallop dredge survey was directed at king scallop stocks and took place at fixed stations. The dredge gear consisted of a mixture of standard commercial dredges (9 tooth bar, 80 mm internal diameter belly rings) and smaller dredges (11 tooth bar, 60 mm internal diameter belly rings) aimed at capturing smaller scallops (Dobby et al., 2017). The North Sea IBTS quarter 3 (Q3) is an internationally coordinated survey carried out in August each year to collect data on a range of demersal species in ICES Division 3a and Subarea 4 (ICES, 2018). The fishing gear consisted of a GOV (Grand Ouverture Vertical) bottom trawl with standard ground gear as described in ICES (2018). The survey followed a repeat station survey design based on an initial stratified-random allocation, with at least two stations in each ICES statistical rectangle (1 degree longitude × 0.5 degree latitude). At each station, the fishing gear was towed for ∼30 min at a speed of 2.5 knots (dredge) or 4 knots (trawl) and any identified brown crabs were counted, sexed, and measured to the nearest millimetre below. Catch rates were standardized for gear size and tow length by calculating a swept area for each station, i.e. the product of the tow length and the gear width. In the dredge survey, gear size was measured as the total dredge width. For the trawl survey, the horizontal opening of the mouth of the net (wingspread) was used as a measure of the gear width. Catch rates were calculated for each station as numbers of crabs caught per square metre (N m−2). Environmental variables Bathymetry datasets for the areas considered were obtained from the UK Hydrographic Office and supplied to MSS by OceanWise Ltd at a resolution of 6 arc second (OceanWise, 2015) (Figure 1b). The sediment layers were obtained from the British Geological Survey (BGS) (Cooper et al., 2013) based on the Folk sediment classification (Folk, 1954). Four sediment strata covering the survey areas were considered: mud, sand, gravelly sand, and gravel (Figure 1c). At each station, the geodesic distance to the nearest point in the UK coast (including islands) was computed (in km), referenced to the World Geodetic System 1984. Spatial distribution To examine the spatial structure of brown crab distribution, geostatistics (Rivoirard et al., 2000) were applied to analyse catch rates in the dredge and trawl surveys (2008–2018). The analysis focused on all brown crabs caught on both dredge and trawl surveys. For the dredge surveys, where a higher number of stations were completed and larger numbers of crabs were sampled (Table 1), the analysis was performed separately for females, males, and juvenile crabs under 100 mm CW. Due to the skewed nature of crab catch rates in both surveys, with many zeros (no crabs caught during a haul), and to reduce the influence of the largest values (Petitgas et al., 2017), a logarithmic transformation was performed on the data as follows: Cij=logrijr‾j+1, where Cij is the log-transformed catch rate (response variable), rij is the catch rate in station i and year j, and r‾j is the mean catch rate in year j. Given that the analysis combines several years, dividing the catch rate by r‾j ensures the data are mean standardized by year. Variograms can be used to quantify the mean variability (semi-variance) of a regionalized variable as a function of the distance between any two points (Petitgas et al., 2017). Omnidirectional empirical variograms were computed for the transformed catch rates for both dredge and trawl surveys in each year (2008–2018) using the estimator (Matheron, 1963): γh=12Nh∑i=1NhZxi-Zxi+h2, where γ is the calculated semi-variance for each dataset, Z(xi) is the catch rate of brown crabs at sampled station xi, Z(xi + h) is the catch rate value separated from xi by a lag distance h (measured as a straight line, in km) and N(h) is the number of observation pairs separated by h. A distance h of 10 km was used for the dredge survey and 25 km for the trawl survey (number of lags = 15) (Table 2). Table 2. Variogram models fitted to the dredge and trawl surveys including type of model (Sph—spherical, Lin—linear), lag distance information, model parameters (range, sill, nugget, and slope), goodness of fit (RSS—residual sum of squares), and mean predicted variance. . Dredge . Trawl . All . Females . Males . <100 mm . All . Model Sph Sph Sph Sph Lin Lag distance (km) 10 10 10 10 25 Range (km) 71 93 91 59 – Sill 0.82 0.91 0.84 0.86 – Nugget 0.53 0.76 0.54 0.71 0.63 Slope – – – – 4.07e−04 RSS 8.61 4.62 10.29 11.32 1.45 Mean variance 0.62 0.81 0.61 0.78 0.66 . Dredge . Trawl . All . Females . Males . <100 mm . All . Model Sph Sph Sph Sph Lin Lag distance (km) 10 10 10 10 25 Range (km) 71 93 91 59 – Sill 0.82 0.91 0.84 0.86 – Nugget 0.53 0.76 0.54 0.71 0.63 Slope – – – – 4.07e−04 RSS 8.61 4.62 10.29 11.32 1.45 Mean variance 0.62 0.81 0.61 0.78 0.66 Open in new tab Table 2. Variogram models fitted to the dredge and trawl surveys including type of model (Sph—spherical, Lin—linear), lag distance information, model parameters (range, sill, nugget, and slope), goodness of fit (RSS—residual sum of squares), and mean predicted variance. . Dredge . Trawl . All . Females . Males . <100 mm . All . Model Sph Sph Sph Sph Lin Lag distance (km) 10 10 10 10 25 Range (km) 71 93 91 59 – Sill 0.82 0.91 0.84 0.86 – Nugget 0.53 0.76 0.54 0.71 0.63 Slope – – – – 4.07e−04 RSS 8.61 4.62 10.29 11.32 1.45 Mean variance 0.62 0.81 0.61 0.78 0.66 . Dredge . Trawl . All . Females . Males . <100 mm . All . Model Sph Sph Sph Sph Lin Lag distance (km) 10 10 10 10 25 Range (km) 71 93 91 59 – Sill 0.82 0.91 0.84 0.86 – Nugget 0.53 0.76 0.54 0.71 0.63 Slope – – – – 4.07e−04 RSS 8.61 4.62 10.29 11.32 1.45 Mean variance 0.62 0.81 0.61 0.78 0.66 Open in new tab The variograms for each survey/year were combined to obtain an experimental mean variogram standardized by the yearly sample variance following the approach proposed by Rivoirard et al. (2000). Four mean variograms were obtained from the 11 years of data for the dredge survey (all crabs, females, males, juveniles) and one mean variogram for the trawl (all crabs) survey. Visual inspection of the resulting mean variograms (Figure 2a) suggested that spherical models provided a good fit for the dredge data and a linear model was appropriate for the trawl data. Spherical models were fitted to the variograms using a non-linear regression to obtain model parameters, resulting in estimates of a sill, range, and nugget (Bivand et al., 2008). The range is the distance at which the semi-variance (sill) becomes asymptotic. The nugget is the discontinuity from the origin at a distance of zero, due to measurement error and small-scale variability (Rivoirard et al., 2000; Petitgas et al., 2017). For the trawl data, a nested model consisting of a nugget and a linear model, weighted by the number of observation points in each lag, was fitted using ordinary least squares (Pebesma, 2004). The estimated variogram models were used to interpolate the data by kriging (Matheron, 1963; Rivoirard et al., 2000; Bivand et al., 2008). Figure 2. Open in new tabDownload slide (a) Variograms estimating semi-variance (γ) with spherical (parameters: range-rg, sill-si, nugget-ng) and linear (parameters: intercept/nugget-ng, slope-sl) models. (b) Kriged maps showing predicted catch rates (in the transformed log scale). The five analysis scenarios considered are displayed from top to bottom: (1) all crabs (dredge), (2) females (dredge), (3) males (dredge), (4) juveniles (dredge), and (5) all crabs (trawl). Figure 2. Open in new tabDownload slide (a) Variograms estimating semi-variance (γ) with spherical (parameters: range-rg, sill-si, nugget-ng) and linear (parameters: intercept/nugget-ng, slope-sl) models. (b) Kriged maps showing predicted catch rates (in the transformed log scale). The five analysis scenarios considered are displayed from top to bottom: (1) all crabs (dredge), (2) females (dredge), (3) males (dredge), (4) juveniles (dredge), and (5) all crabs (trawl). A number of predictor variables for the study area were available (distance to coast, depth and BGS sediment type), allowing the application of kriging with external drift whereby some of the variance of the data is attributed to the variable in question (Rivoirard et al., 2000; Oliver and Webster, 2014). Data exploration of crab catch rates, in relation to different predictor variables, led to the selection of depth and distance to coast as important variables to explain the spatial distribution of crabs in the dredge surveys. The same variables were used for the trawl surveys with the addition of BGS sediment type, which was found to have an effect in the trawl-caught catch rates. The interpolated catch rate predictions were computed onto an equally spaced grid (3 nautical miles resolution) of the survey area and two-dimensional density maps were generated for each dataset on both surveys. The validity of the variogram models and kriging interpolations was evaluated using a leave-one-out cross-validation procedure (Bivand et al., 2008). A routine was implemented whereby each point was removed from the dataset, using the remaining data to predict the omitted density value and the residuals were computed and spatially plotted for inspection and validation of the model applied to each dataset. Abundance and recruitment indices To estimate temporal trends and an abundance index of brown crab in the study area, generalized additive models (GAMs) were applied to the survey data using the catch rate per haul (numbers of crabs caught per square metre) as the response variable. Two sets of models were fitted separately, for the dredge and trawl surveys. Large numbers of zero observations were present in both survey datasets, as no crabs were recorded in many of the sampled stations. To deal with zero observations and considering the response variable was markedly right-skewed, a GAM was applied assuming a Tweedie distribution (Tweedie, 1984; Dunn and Smyth, 2005) with a log link. Tweedie models belong to the exponential family of distributions and include a power mean–variance relationship, Vμ=μp ⁠, where V() is the variance function, µ is the mean of the distribution, and p is a parameter (1 < p < 2), where p = 1 implies a Poisson distribution and p = 2 implies a gamma distribution (Dunn and Smyth, 2005). The parameter p was estimated for each survey during the fitting procedure with the tw function in the mgcv R package (Wood, 2011). The size data available for brown crabs caught in the dredge and trawl surveys were compared to develop a recruitment index for the east of Scotland. A size of 100 mm CW was chosen as a cut-off point to separate juvenile crabs from mature adults based on the lowest estimates of gonadal size at maturity in Europe (Haig et al., 2016). A preliminary analysis of the length information showed that very few crabs below 100 mm CW were captured in the trawl surveys. Consequently, the recruitment analysis in the study area was focused only on data from the dredge survey, during which smaller crabs were frequently caught. The response variable was the catch rate of crabs (N m−2) under 100 mm CW. GAMs assuming a Tweedie distribution were fitted to the defined recruitment index but model diagnostics based on the residuals showed generally a poor fit. Therefore, a two-part hurdle model (Mullahy, 1986; Zeileis et al., 2008; Zuur et al., 2009) was used to model the recruitment index of juvenile crabs. In the first part, a binomial GAM (logit link) was used to model the probability that a zero is observed in each sampled station. Positive recruitment index values were plotted and compared with several distributions (Delignette-Muller and Dutang, 2015) and shown to follow a gamma distribution. Therefore, in the second part of the model, non-zero values were modelled with a GAM using a gamma distribution with a log link. The explanatory variables considered for the two abundance models and for the recruitment hurdle model (both parts) were: sediment type (BGS data, categorical variable with four levels), depth, distance to coast, year (2008–2018), and geographical position (latitude and longitude plus an interaction term between these) at each sampling station. A backward stepwise model selection was used to build the GAMs starting with a full model containing all explanatory variables and removing non-significant variables one at a time. The degrees of freedom of each smooth function in the GAMs were not constrained, and the convergence of the smoothness selection optimization was inspected for each model. Testing of the fitted models was based on the analysis of deviance. Marginal F-tests (Wood, 2006) were used to compare nested GAMs, and chi-squared tests (Chambers and Hastie, 1992) applied to compare nested binomial models, selecting the simpler model if no differences were found (at a significance level of α = 0.05). The model selection procedure is summarized in Table 3. Table 3. Variable selection for the abundance (dredge and trawl) and recruitment index models using a backward stepwise approach. Models . Model type (assumed distribution) . Covariates . Dropped . p-Value . Test . Dredge  Model 1 GAM (Tweedie) Year, distance, depth, BGS, Lon × Lat – – –  Model 2a Year, distance, depth, Lon × Lat BGS 0.181 F Trawl  Model 3a GAM (Tweedie) Year, distance, depth, BGS, Lon × Lat – – – Recruitment index  Model 4 Hurdle GAM1 (Binomial) Catch rate (0.1) Year, distance, depth, BGS, Lon × Lat – – –  Model 5a Year, distance, depth, Lon × Lat BGS 0.254 χ2  Model 6 Hurdle GAM2 (gamma) Catch rate > 0 Year, distance, depth, BGS, Lon × Lat – – –  Model 7 Year, distance, BGS, Lon × Lat Depth 0.276 F  Model 8 Year, distance, Lon × Lat BGS 0.277 F  Model 9a Year, Lon × Lat Distance 0.079 F Models . Model type (assumed distribution) . Covariates . Dropped . p-Value . Test . Dredge  Model 1 GAM (Tweedie) Year, distance, depth, BGS, Lon × Lat – – –  Model 2a Year, distance, depth, Lon × Lat BGS 0.181 F Trawl  Model 3a GAM (Tweedie) Year, distance, depth, BGS, Lon × Lat – – – Recruitment index  Model 4 Hurdle GAM1 (Binomial) Catch rate (0.1) Year, distance, depth, BGS, Lon × Lat – – –  Model 5a Year, distance, depth, Lon × Lat BGS 0.254 χ2  Model 6 Hurdle GAM2 (gamma) Catch rate > 0 Year, distance, depth, BGS, Lon × Lat – – –  Model 7 Year, distance, BGS, Lon × Lat Depth 0.276 F  Model 8 Year, distance, Lon × Lat BGS 0.277 F  Model 9a Year, Lon × Lat Distance 0.079 F Hypothesis testing was performed with marginal F-tests (GAMs) and χ2 tests (binomial GAMs). Lon × Lat includes variables longitude, latitude and an interaction term between these. a Selected parsimonious model. Open in new tab Table 3. Variable selection for the abundance (dredge and trawl) and recruitment index models using a backward stepwise approach. Models . Model type (assumed distribution) . Covariates . Dropped . p-Value . Test . Dredge  Model 1 GAM (Tweedie) Year, distance, depth, BGS, Lon × Lat – – –  Model 2a Year, distance, depth, Lon × Lat BGS 0.181 F Trawl  Model 3a GAM (Tweedie) Year, distance, depth, BGS, Lon × Lat – – – Recruitment index  Model 4 Hurdle GAM1 (Binomial) Catch rate (0.1) Year, distance, depth, BGS, Lon × Lat – – –  Model 5a Year, distance, depth, Lon × Lat BGS 0.254 χ2  Model 6 Hurdle GAM2 (gamma) Catch rate > 0 Year, distance, depth, BGS, Lon × Lat – – –  Model 7 Year, distance, BGS, Lon × Lat Depth 0.276 F  Model 8 Year, distance, Lon × Lat BGS 0.277 F  Model 9a Year, Lon × Lat Distance 0.079 F Models . Model type (assumed distribution) . Covariates . Dropped . p-Value . Test . Dredge  Model 1 GAM (Tweedie) Year, distance, depth, BGS, Lon × Lat – – –  Model 2a Year, distance, depth, Lon × Lat BGS 0.181 F Trawl  Model 3a GAM (Tweedie) Year, distance, depth, BGS, Lon × Lat – – – Recruitment index  Model 4 Hurdle GAM1 (Binomial) Catch rate (0.1) Year, distance, depth, BGS, Lon × Lat – – –  Model 5a Year, distance, depth, Lon × Lat BGS 0.254 χ2  Model 6 Hurdle GAM2 (gamma) Catch rate > 0 Year, distance, depth, BGS, Lon × Lat – – –  Model 7 Year, distance, BGS, Lon × Lat Depth 0.276 F  Model 8 Year, distance, Lon × Lat BGS 0.277 F  Model 9a Year, Lon × Lat Distance 0.079 F Hypothesis testing was performed with marginal F-tests (GAMs) and χ2 tests (binomial GAMs). Lon × Lat includes variables longitude, latitude and an interaction term between these. a Selected parsimonious model. Open in new tab Model validation was performed by visually inspecting the residuals obtained for each model that were plotted against each explanatory variable to assess the presence of patterns. The final selected models were used to estimate the abundance and recruitment indices by year. For the hurdle model, estimates of part 1 binomial GAM (probability of a non-zero event) were multiplied by estimates of the zero-truncated part 2 (gamma GAM) to obtain a final combined estimate. Predictions were made for each grid cell within a 3 × 3 nautical mile grid and averaged for each survey year over the scallop survey area (Figure 1a). The relative abundance indices derived from the dredge model (N m−2) were raised to the total dredge area (26 600 km2) to generate an absolute index of abundance and recruitment. The trawl model was used to produce a relative index of abundance (N m−2) estimated inside the dredge area and this was compared with the dredge absolute abundance index for the period 2008–2018. All statistical analyses were implemented in the R statistical programming environment (R Core Team, 2018). The computing of variograms, model fitting, kriging, and cross-validation were performed with the gstat package v. 1.1-6 (Pebesma, 2004). GAMs were fitted with R package mgcv v. 1.8-23 (Wood, 2011). The probability distribution of survey catch rates was evaluated using functions provided by package fitdistrplus v. 1.0-11 (Delignette-Muller and Dutang, 2015). Results The two surveys considered in the study are part of MSS research vessel programme where the east coast of Scotland dredge survey is generally carried out 1 month before the IBTS Q3 trawl survey. The same vessels were involved in the data collection for each survey with consistent gear and sampling procedures. The mean number of stations sampled in each year inside the study area was 108 stations in the dredge surveys and 46 stations in the trawl surveys (Table 1). The dredge survey area where sampling effort is higher (1 station per 248 km2) is ∼22% of the larger trawl area where sampling was sparser (1 station per 2609 km2). Spatial distribution The spherical model fitted to the dredge surveys gave a range estimate between 59 and 93 km and sills ranging from 0.82 to 0.91 (Table 2). The nuggets varied between 0.53 and 0.76, which is indicative of relatively high variance at short ranges compared with the overall variability at the regional scale (sill). The variograms for all individuals and male crabs were similar with comparable sills and lower estimated nuggets than those found in other scenarios (Figure 2a). For the trawl survey, which only used catch rates for all crabs combined, a linear variogram with a 0.63 nugget (intercept) and a 4.07 × 10−4 slope was estimated as the best fit (Table 2). This implies that the variance in the trawl survey increases without bounds unlike the dredge survey where the variance is bounded at the sill. The predicted catch rate for all crabs in the dredge survey was higher in coastal areas particularly in the south of the Moray Firth and along the south east coast towards the Firth of Forth (Figure 2b). In the southern area off the Firths of Forth and Tay, the higher predicted catch rates extended further offshore to the east. Catch rates from the trawl survey had a similar distribution pattern to that found for the dredge survey, although the prediction area of the former was larger. Trawl catch rates showed that the high density area in the south found in the dredge survey may have extended further to the southeast and, in addition, showed a region of medium to high densities estimated to the north of the Orkney Islands corresponding mostly to gravel grounds (Figure 1c). The male and female distribution (derived from the dredge survey) was found to be relatively similar, the main difference being the fact that females were more often found in the inshore waters south of Orkney and to the east side of the survey area. Juvenile crabs <100 mm CW showed a very clear inshore distribution along the shoreline up to ∼20 km from the coast, except in the north of the survey area where catch rates were low both inshore and offshore. The kriging prediction variances on the five scenarios presented were lowest at and around the observation points and highest in areas where the number of stations was low. Higher predicted mean variances were estimated for the female (0.81) and juvenile (0.78) models in relation to the males and all crab models (0.61 and 0.62, respectively) (Table 2). Cross-validation analysis performed to verify the validity of spherical and linear variogram fits and kriging predictions supported the suitability of the models used. Spatial plots of the residuals for each scenario showed no spatial patterns in the data. Abundance indices Selected GAMs to model catch rates of all crabs (abundance index) and juvenile crabs (recruitment index) are summarized in Table 4. As part of the model validation, visual inspection indicated no obvious patterns in the residuals when plotted against each explanatory variable. The deviance explained for the dredge and trawl models was 45.7 and 61.6%, respectively. The Tweedie parameter p was estimated as 1.14 for the dredge survey and 1.02 for the trawl survey. Explanatory variables “year”, “distance to coast”, “depth” and “longitude × latitude” were selected for the dredge and trawl models in the abundance index. The “sediment” (BGS) covariate was found to be not significant in the dredge model (p = 0.181) based on the analysis of deviances, while in the trawl model, this variable had a significant effect (p = 0.04) (Table 3). The trends of the smooth functions and the BGS parametric term in the trawl model fitted to the catch rates are illustrated in Figure 3. The response variable for both the dredge and trawl surveys showed an approximate gamma distribution (positive values) but also contained a large proportion of zeros (21% in the dredge survey; 78% in the trawl survey) supporting the use of a Tweedie distribution to model the data (Table 4 and Figure 3a). The estimated crab catch rates from the dredge survey (1300–3200 animals/km2) were 1–2 orders of magnitude greater than those from the trawl survey (16–65 animals/km2) as recorded catches of crabs were much higher in the former (Table 1). The smooth functions fitted for both models showed mostly non-linear responses to the different covariates used. The exception was the covariate “distance to coast” in the trawl model for which a linear (d.f. = 1) decrease in catch rates was estimated over the range of distances, which were recorded up to 160 km from the coast. The model applied to the dredge data also showed a decrease in catch rates with distance to coast (although non-linear), in particular within 30 km from the coast where most stations were located with data becoming sparser and larger confidence intervals estimated for distances above 40 km up to the maximum value of 67 km (Figure 3c). In both surveys, the abundance of crabs seemed to be affected by depth with lower catch rates found generally at greater depths. Crab catch rates showed a non-linear decrease with depth in the dredge survey where the depth range was 23–84 m. A similar pattern was observed in the trawl surveys (depth range 42–192 m) but only for depths shallower than ∼150 m (Figure 3d). Smoothed plots for year effects were also clearly non-linear (Figure 3b). The plot of the effect of BGS sediment (trawl model only) shows that crab catches in the trawl survey increased gradually from softer to harder grounds with catch rates peaking in the “gravel” sediment type (Table 4 and Figure 3e). An absolute abundance index was derived from the dredge model and compared with the relative index from the trawl model. These were compared inside the scallop dredge area where the two surveys overlap (Figure 4). The total abundance as estimated by the dredge model varied between 34 and 86 million of crabs in the 2008–2018 period (trawl catch rates ranged between 16 and 65 crabs per square kilometre). The dredge and trawl indices resemble the respective year effects (Figure 3b) and seem to be correlated showing a very similar trend of increasing catch rates in the early years of the time series up to 2016 and a subsequent reduction. Figure 3. Open in new tabDownload slide Results for the abundance and recruitment indices derived from the dredge model (top panel), trawl model (second panel), and hurdle models (third and bottom panels) applied to brown crab catch rates. (a) Distribution of the response variable; (b) partial effect of “year”; (c) partial effect of “distance to coast”; (d) partial effect of “depth”; and (e) parametric effect of BGS sediment type (trawl model only; gravel is used as the reference type). The dashed lines give the standard errors around the parametric and smooth effects (shown in the scale of the linear predictor). Figure 3. Open in new tabDownload slide Results for the abundance and recruitment indices derived from the dredge model (top panel), trawl model (second panel), and hurdle models (third and bottom panels) applied to brown crab catch rates. (a) Distribution of the response variable; (b) partial effect of “year”; (c) partial effect of “distance to coast”; (d) partial effect of “depth”; and (e) parametric effect of BGS sediment type (trawl model only; gravel is used as the reference type). The dashed lines give the standard errors around the parametric and smooth effects (shown in the scale of the linear predictor). Figure 4. Open in new tabDownload slide Brown crab abundance indices by year estimated from GAMs applied to dredge (solid line, left vertical axis) and trawl (dashed line, right vertical axis) surveys (2008–2018) with 95% confidence intervals. The abundance index for the dredge model is an absolute estimate (millions of individuals) while the trawl index is relative (N m−2). Predictions are averaged for grid cells within the dredge survey area (Figure 1a). Figure 4. Open in new tabDownload slide Brown crab abundance indices by year estimated from GAMs applied to dredge (solid line, left vertical axis) and trawl (dashed line, right vertical axis) surveys (2008–2018) with 95% confidence intervals. The abundance index for the dredge model is an absolute estimate (millions of individuals) while the trawl index is relative (N m−2). Predictions are averaged for grid cells within the dredge survey area (Figure 1a). Table 4. Selected models for estimating abundance (dredge and trawl GAMs) and recruitment (hurdle). Model . Abundance index . Recruitment index . Dredge . Trawl . Dredge . GAM . GAM . Hurdle . GAM1 . GAM2 . N 1 191 501 1 191 155 Perc. zeros (%) 20.9 77.8 87.0 0 Family Tweedie Tweedie Binomial Gamma Link function Log Log Logit Log P 1.14 1.02 – – Dev. explained (%) 45.7 61.6 32.5 41.6 GCV/UBRE score 3.606e−04 1.957e−05 −0.421a 0.323 Model terms Est. d.f. p-Value Est. d.f. p-Value Est. d.f. p-Value Est. d.f. p-Value Smoothers  Year – 8.7 <0.001 – 3.6 <0.001 – 7.7 <0.001 – 7.5 0.005  Distance – 6.9 <0.001 – 1 <0.001 – 5.2 <0.001 – – –  Depth – 7.4 <0.001 – 7 <0.001 – 5.3 0.054 – – –  Lon × Lat – 25.3 <0.001 – 28.1 <0.001 – 14.7 0.001 – 13.1 0.012 Parametric terms  Intercept −9.084 1 <0.001 −11.78 1 <0.001 −2.859 1 <0.001 −9.487 1 <0.001  BGS   Gravelly sand – – – −1.472 1 0.009 – – – – – –   Sand – – – −1.729 1 0.005 – – – – – –   Mud – – – −1.901 1 0.022 – – – – – – Model . Abundance index . Recruitment index . Dredge . Trawl . Dredge . GAM . GAM . Hurdle . GAM1 . GAM2 . N 1 191 501 1 191 155 Perc. zeros (%) 20.9 77.8 87.0 0 Family Tweedie Tweedie Binomial Gamma Link function Log Log Logit Log P 1.14 1.02 – – Dev. explained (%) 45.7 61.6 32.5 41.6 GCV/UBRE score 3.606e−04 1.957e−05 −0.421a 0.323 Model terms Est. d.f. p-Value Est. d.f. p-Value Est. d.f. p-Value Est. d.f. p-Value Smoothers  Year – 8.7 <0.001 – 3.6 <0.001 – 7.7 <0.001 – 7.5 0.005  Distance – 6.9 <0.001 – 1 <0.001 – 5.2 <0.001 – – –  Depth – 7.4 <0.001 – 7 <0.001 – 5.3 0.054 – – –  Lon × Lat – 25.3 <0.001 – 28.1 <0.001 – 14.7 0.001 – 13.1 0.012 Parametric terms  Intercept −9.084 1 <0.001 −11.78 1 <0.001 −2.859 1 <0.001 −9.487 1 <0.001  BGS   Gravelly sand – – – −1.472 1 0.009 – – – – – –   Sand – – – −1.729 1 0.005 – – – – – –   Mud – – – −1.901 1 0.022 – – – – – – For the BGS sediment effect (trawl model only), gravel is used as the reference type. The significance of parametric model terms was tested with t-tests. The p-values associated with the GAM smothers for each model term were derived from Wald tests (Wood, 2013). GCV, Generalized Cross Validation; UBRE, Un-Biased Risk Estimator. a GCV was used as the prediction error criteria for all models except the binomial GAM where the scale parameter is known and an UBRE is used instead (Wood, 2011). Open in new tab Table 4. Selected models for estimating abundance (dredge and trawl GAMs) and recruitment (hurdle). Model . Abundance index . Recruitment index . Dredge . Trawl . Dredge . GAM . GAM . Hurdle . GAM1 . GAM2 . N 1 191 501 1 191 155 Perc. zeros (%) 20.9 77.8 87.0 0 Family Tweedie Tweedie Binomial Gamma Link function Log Log Logit Log P 1.14 1.02 – – Dev. explained (%) 45.7 61.6 32.5 41.6 GCV/UBRE score 3.606e−04 1.957e−05 −0.421a 0.323 Model terms Est. d.f. p-Value Est. d.f. p-Value Est. d.f. p-Value Est. d.f. p-Value Smoothers  Year – 8.7 <0.001 – 3.6 <0.001 – 7.7 <0.001 – 7.5 0.005  Distance – 6.9 <0.001 – 1 <0.001 – 5.2 <0.001 – – –  Depth – 7.4 <0.001 – 7 <0.001 – 5.3 0.054 – – –  Lon × Lat – 25.3 <0.001 – 28.1 <0.001 – 14.7 0.001 – 13.1 0.012 Parametric terms  Intercept −9.084 1 <0.001 −11.78 1 <0.001 −2.859 1 <0.001 −9.487 1 <0.001  BGS   Gravelly sand – – – −1.472 1 0.009 – – – – – –   Sand – – – −1.729 1 0.005 – – – – – –   Mud – – – −1.901 1 0.022 – – – – – – Model . Abundance index . Recruitment index . Dredge . Trawl . Dredge . GAM . GAM . Hurdle . GAM1 . GAM2 . N 1 191 501 1 191 155 Perc. zeros (%) 20.9 77.8 87.0 0 Family Tweedie Tweedie Binomial Gamma Link function Log Log Logit Log P 1.14 1.02 – – Dev. explained (%) 45.7 61.6 32.5 41.6 GCV/UBRE score 3.606e−04 1.957e−05 −0.421a 0.323 Model terms Est. d.f. p-Value Est. d.f. p-Value Est. d.f. p-Value Est. d.f. p-Value Smoothers  Year – 8.7 <0.001 – 3.6 <0.001 – 7.7 <0.001 – 7.5 0.005  Distance – 6.9 <0.001 – 1 <0.001 – 5.2 <0.001 – – –  Depth – 7.4 <0.001 – 7 <0.001 – 5.3 0.054 – – –  Lon × Lat – 25.3 <0.001 – 28.1 <0.001 – 14.7 0.001 – 13.1 0.012 Parametric terms  Intercept −9.084 1 <0.001 −11.78 1 <0.001 −2.859 1 <0.001 −9.487 1 <0.001  BGS   Gravelly sand – – – −1.472 1 0.009 – – – – – –   Sand – – – −1.729 1 0.005 – – – – – –   Mud – – – −1.901 1 0.022 – – – – – – For the BGS sediment effect (trawl model only), gravel is used as the reference type. The significance of parametric model terms was tested with t-tests. The p-values associated with the GAM smothers for each model term were derived from Wald tests (Wood, 2013). GCV, Generalized Cross Validation; UBRE, Un-Biased Risk Estimator. a GCV was used as the prediction error criteria for all models except the binomial GAM where the scale parameter is known and an UBRE is used instead (Wood, 2011). Open in new tab Recruitment index The dredge survey captured smaller crabs (median = 134 mm, mean = 135.9 mm) than the trawl survey (median = 160 mm, mean = 154.5). The median size per year was found to be significantly correlated (Pearson correlation = 0.83, p = 0.002) in the two surveys but fluctuated with no trend throughout the study period (Figure 5). Because animals below 100 mm were seldom caught in trawl surveys, the recruitment index was based on dredge data and was modelled with a two-part hurdle model. The results from the first part of the model (binomial GAM) showed that the variables year, distance to coast, and geographical position were significant (p < 0.05) for explaining the probability of juvenile crabs being captured in the dredge survey (Table 4). The depth smoother was borderline insignificant (p = 0.054, also included in the model) implying a slightly lower likelihood of finding juvenile crabs away from the coast and in depths >50 m. A χ2 test showed that including the type of sediment (p = 0.254) as an explanatory variable did not improve the model significantly and this was not included in the final model (Table 3). Figure 5. Open in new tabDownload slide Length frequency distribution of brown crabs captured in the dredge and trawl surveys (2008–2018). The medians and means in each year are represented by the full and dotted vertical lines, respectively. Figure 5. Open in new tabDownload slide Length frequency distribution of brown crabs captured in the dredge and trawl surveys (2008–2018). The medians and means in each year are represented by the full and dotted vertical lines, respectively. The second part of the hurdle model used a GAM assuming a gamma distribution of the response variable, i.e. the catch rate of juvenile crabs caught in each dredge station where juveniles were recorded (positive catch rates). The results from this model showed that only variables “year” and geographical position were significant (p < 0.05) (Table 4). Juvenile crabs in the dredge survey were mostly found in stations close to the coast in shallow waters, as shown by the geostatistical analysis (Figure 2a). The variables depth (p = 0.276), sediment (p = 0.277), and distance to coast (p = 0.079) did not significantly improve the second part of the model and were excluded during the stepwise selection (Table 3). A predicted recruitment index obtained from combining the two model parts in the dredge survey area is shown in Figure 6. The recruitment index shows a gradual increase in juvenile crabs up to 2014 (0.58 million crabs), followed by a recent steep decrease with the lowest value estimated (0.15 million) observed in 2018. Figure 6. Open in new tabDownload slide Brown crab recruitment index by year with 95% confidence intervals. The recruitment index is an absolute estimate calculated as the number (millions) of juvenile crabs <100 mm CW. Recruitment estimates were derived from a hurdle model applied to dredge surveys (2008–2018). Predictions are averaged for all grid cells within the dredge survey area (Figure 1a). Figure 6. Open in new tabDownload slide Brown crab recruitment index by year with 95% confidence intervals. The recruitment index is an absolute estimate calculated as the number (millions) of juvenile crabs <100 mm CW. Recruitment estimates were derived from a hurdle model applied to dredge surveys (2008–2018). Predictions are averaged for all grid cells within the dredge survey area (Figure 1a). Discussion There is an increasing demand from fisheries managers and stakeholders to obtain information on stock status of exploited marine species. Detailed information on the distribution patterns of species is often missing in fisheries science. Mapping is an important tool to identify habitats and the distribution range of important marine resources. Fishery-independent surveys at sea provide high-quality data, as sampling and collection are scientifically designed and standardized (Pennino et al., 2016). Despite the fact that the two surveys analysed in this study were aimed at other species such as scallops (dredge) and demersal fish (trawl), data collected on the distribution of brown crab were informative and useful abundance estimates were derived. Hilborn and Walters (1992) list a number of limitations of fisheries surveys that may lead to unrepresentative sampling, including the location of sampling stations in relation to the target species and the timing of the survey and its relation to the seasonal cycle. The dredge survey analysed is mostly carried out around the coast, following the scallop fishing distribution in the East of Scotland, which often overlaps with the inshore crab and lobster fishery (Öndes et al., 2016). The trawl survey includes the area of the dredge survey but also extends further offshore to the east and northern North Sea. The dredge survey has a much higher sampling effort resulting, not only in more stations per area, but also in much higher catch rates than the trawl surveys, given that brown crabs are mostly found in the inshore grounds of the North Sea. The timing of the dredge and trawl surveys is similar as both take place in summer, making the two surveys comparable. Brown crabs have previously been reported to be caught more frequently by scallop dredges than trawls (Kaiser et al., 1996). The different catchabilities of the gears may be related to the different types of operation, including the way these gears interact with the sediment and the towing speed (higher in the trawl survey). The teeth on scallop dredges are able to go deeper in the sediment, capturing crabs that are buried, whereas trawls are more likely to catch only those crabs that are active on the surface of the seabed (Öndes et al., 2016). For example, dredges have been shown to be more efficient than trawls to capture blue crabs (Callinectes sapidus) in the western Atlantic given that these species remain inactive and partially buried in the sediment during winter months (Vølstad et al., 2000). Brown crabs are also known to lie in the sediment partially buried (Edwards, 1979; Howard, 1982) making them less available to trawl gears. Spatial distribution Geostatistics were used to describe the patterns of spatial distribution of brown crab in the North Sea around the East coast of Scotland. Kriging with external drift is a common way to handle a trend in non-stationary geostatistics, taking advantage of auxiliary variables that are known across the study area including the sampling stations (Petitgas et al., 2017). In this study, the predictors used in the kriging analysis included depth and distance to coast (dredge and trawl) and BGS sediment type (trawl model only). Results of variographic analyses of the dredge survey suggest that, in the four scenarios (all crabs, females, males, and juveniles), the underlying covariate function is represented by a spherical model, showing that there is spatial autocorrelation in crab density estimates at stations located within small distances of each other. The spherical model has been widely used in fisheries geostatistics given its properties of linear behaviour at the origin stabilizing on a sill, implying the model is bounded and, therefore, associated with stationary covariance (Rivoirard et al., 2000; Petitgas et al., 2017). For the trawl data (which uses all crabs), the nugget effect is higher than for the dredge survey (all crabs scenario) and the semi-variance increases linearly with distance, corresponding to non-stationary covariance, which implies a trend in the distribution. The spatial distributions of males and female crabs in the dredge survey were found to be quite similar but there is some evidence of higher female catch rates in the north of the survey area. Several studies based on tagging provide evidence that female brown crabs move more frequently and over longer distances than male crabs, e.g. in the North Sea (Edwards, 1979), English Channel (Bennett and Brown, 1983; Hunter et al., 2013), Bay of Biscay (Latrouite and Le Foll, 1989), Sweden (Ungfors et al., 2007) and north of Scotland (Jones et al., 2010; Coleman and Rodrigues, 2016). Edwards (1979) describes an inshore movement of crabs in spring for hatching, moulting and mating and an offshore movement in the autumn (in particular females) for spawning. The sediment type may also play a role during spawning as female crabs seem to prefer softer substratum (more likely found offshore) to scoop out a hollow to rest the lower abdomen and ensure attachment of the eggs to the pleopods (Brown and Bennett, 1980). This study clearly demonstrates that juveniles have a strong preference for inshore grounds as catch rates peak at small distances from land and most crabs below 100 mm were recorded within 20 km from the coast. This agrees with previous descriptions of juvenile brown crabs, which have been found to live mainly in shallow inshore waters (Bennett, 1974; Bennett, 1995; Robinson and Tully, 2000) indicating selective habitat preferences possibly related to favourable environmental and ecological conditions (Pallas et al., 2006; Pardo et al., 2007; Dickens, 2012). Heterogeneous habitats such as those found inshore may be selected by crabs as a response to the selective pressure of predation. Structurally complex habitats may result in reduced predation rates by the provision of refugia/shelter to juvenile animals or by decreasing the chances of encounters between prey and predators (Harrison and Crespi, 1999). Abundance indices Abundance indices were estimated from the survey data using GAMs assuming a Tweedie distribution. Tweedie models are natural candidates for modelling continuous non-negative variables where many zeros are present in the data (Dunn and Smyth, 2008). Tweedie distributions have been used as a modelling approach in a diverse range of fields including fisheries (Candy, 2004; Shono, 2008; Arcuti et al., 2013). In this study, a value of between 1 and 2 was used for the p parameter as the Tweedie model was represented by a mix of distributions with a positive mass around zero, given the high number of stations where no crabs were captured. Results from the GAMs applied to the dredge and trawl survey data showed a significant effect of covariates depth and distance to coast in predicting brown crab catch rates. These effects (with the exception of distance to coast in the trawl survey) were non-linear but reflected a general decrease in catch rates at greater depths and away from the coast. This study represents a first insight into the distributional pattern and potential effects of depth and distance to coast on the distribution of brown crab in the North Sea during the summer. However, these results may not applicable to other times of year or other stocks because the distribution of brown crab varies with season and locality, and the trawl surveys in the present study included few hauls at depths >150 m (while the dredge survey was restricted to areas of <100 m depth). For example, female brown crabs support a large fishery in offshore grounds to the north and west of Scotland at depths between 100 and 200 m (Mesquita et al., 2017). More recently, numerous individuals of adult brown crab were observed at >400 m depth (well over the depths normally reported for the species) on the coast of Norway using video transects (Bakke et al., 2019). In the English Channel, larger adult crabs tend to be caught in the fishery at greater depths while juveniles and smaller crabs were mostly reported inshore (as described in this study) but no standardized catch rates were available (Brown and Bennett, 1980). A study in the Isle of Man found that the catch per unit of effort of brown crab caught in traps increased with depth but the study was limited to a small inshore area with depths up to 65 m (Öndes et al., 2019). Depth and location preferences of brown crabs may also be associated with substrata preference. An important difference between the dredge and trawl models was that the latter included sediment type as an explanatory variable. GAMs applied to the trawl data showed that sediment type was a significant variable for explaining crab catch rates, with higher rates found in the harder sediment type (gravel) in comparison with the softer sandy and muddy sediments. This explains the relatively high catch rates predicted in the trawl geostatistical analysis for the areas to the north of the Orkney Islands where gravel sediment is predominant. This contrasts with the results obtained for the dredge survey where sediment was not found to be significant, although the amount of gravel inside the dredge area is restricted to some small patches mostly found in the south of Moray Firth and to the south of Orkney. The limited amount of stations within the gravel sediment in the dredge survey in relation to the other sediment types was perhaps not sufficient for the model to find a significant effect. It is also possible that the physical interaction of dredges and trawls with each substrate is distinct, implying potential differences in catchability by sediment type. Crabs are known to live in a variety of different sediment types (Shelton and Hall, 1981) and habitat preferences may depend on several factors including the migratory and reproductive cycle, which is thought to differ between males and females. The GAMs for both dredge and trawl catch rate included a significant year effect, which was used to estimate abundance indices for both gears inside the scallop dredge area. It was clear that the dredge survey is more efficient at catching brown crabs than the trawl survey, and for this reason, we propose a (minimum) absolute abundance estimate for the dredge survey and a relative abundance index for the trawl survey (used for comparison). Different biases may arise from the proposed catch rates in the two surveys. The dredge survey only covers an inshore area off Scotland, although its station density is relatively high. The trawl survey covers a larger area than the dredge survey but has a lower station density and also a catch rate bias (given its lower catchability). An absolute abundance estimate for the whole area could potentially be extrapolated from the dredge survey but caution is needed when interpreting the results, given the absence of dredge data in the offshore areas. Estimation of absolute abundance from surveys is difficult because it typically requires assumptions about gear efficiency (Miller et al., 2019). An absolute abundance estimate implies, for example, assumptions that crabs are not able to swim over the dredge/nets and that no herding takes place. Crabs of both sexes are known to converge for the reproductive season to facilitate mating encounters (Edwards, 1979; Orensanz and Gallucci, 1988), but there is no evidence of large mating aggregations (Hartnoll, 1969), implying that the assumption of no herding is reasonable. Catch rates estimated from the dredge survey in this study ranged between 1300 and 3200 animals/km2. These estimates are comparable to the catch rate for the same species, of 2100 animals ≥70 mm/km2, estimated from a mark-recapture model applied to a population of crabs in the east coast of England (Bell et al., 2003). There is a very clear similarity in the crab abundance signal estimated from the two surveys in the present study. The fact that the surveys show the same signal in terms of crab abundance trends indicates that active gear surveys can be used to estimate the distribution of brown crabs, even though dredging/trawling are not the main methods employed by the fishery to capture the species. Recruitment index The median size of crabs caught during the study period was shown to be correlated in the two surveys. No crabs below 35 mm were caught by any of the surveys. This is not surprising, given the habitat preferences of smaller crabs, which are mainly found very close to shore and outside the areas covered by the surveys. The dredge survey generally captured smaller individuals than the trawl survey, including juvenile animals below 100 mm CW, suggesting that small crabs are also more likely to escape trawling and the dredge survey is a better choice to detect recruitment pulses. Brown crab recruitment estimates were produced for the dredge study area using a two-part hurdle model. The difficulties associated with fitting a Tweedie model to the recruitment data may be explained by the high percentage of zeros (87%) in the dredge data, i.e. individuals below 100 mm CW were only caught in a small percentage of sampled stations. Some authors have described a poor performance of Tweedie models in handling extremely unbalanced zero data (∼90% or more zeros), in which case, hurdle models are considered reliable alternatives to analyse the data (Shono, 2008; Zhou et al., 2018). Two-part GAMs (first part: presence and absence; second part: catch rates given presence) have been used in fisheries and ecology studies to describe trends in abundance and geographic distribution in fisheries (Maravelias, 1997; Murase et al., 2009; Goetz et al., 2012). The hurdle model applied to the recruitment data showed that several covariates (year, depth, distance to coast) were important in determining juvenile crab presence (part 1) while only year was significant for juvenile crab catch rates (given presence) (part 2), although this may be at least in part an artefact of the much smaller sample size for presence records. From the first part of the model, there was a greater probability of juvenile crabs being present close to the coast between 40 and 50 m depth, as previously shown by the geostatistical analysis. When considering only the stations where juveniles were present in the second part of the model (shallow areas closer to the coast), there was not enough contrast in the data to detect an effect of depth or distance to coast. The combination of the two parts of the hurdle model results in a recruitment index peaking in 2014. The highest recruitment estimate was detected 2 years before the abundance peak of 2016 on both the dredge and trawl surveys suggesting that the higher recruitment signal detected in 2014 may have contributed to the higher estimates 2 years later. The low recruitment estimated in 2018 suggests that the abundance decline recently observed is likely to continue in the coming years. Growth studies based on tagging in the North Sea suggest that it takes ∼2 years for a 70-mm crab to reach 115 mm (Edwards, 1979), although growth may vary with the frequency of moult increments, which is likely to be related with temperature and may be subject to interannual variability. Future work This study demonstrated that spatial modelling may be used to create brown crab distribution maps and abundance indices from fisheries survey data, allowing for the detection of trends in the crab fishery. The deviance explained by the GAMs (33–62%) could be improved by the inclusion of other potentially influential variables not considered in this study, e.g. bottom water temperature or current strength. These abiotic factors related to environmental and oceanographic conditions are strongly associated with the spatial distribution of crustacean assemblages (Basford et al., 1989; Fariña et al., 1997). One advantage of using dredge and/or trawl gears instead of passive gears such as traps for obtaining standardized catch rates is that the catchability of the former is not affected by factors such as season, bait, current strength/direction (affecting the spread of the bait plume) and soak time. In addition, catch rates calculated from mobile gears eliminate issues related with animals, which avoid traps, such as berried females (Howard, 1982), and are not affected by saturation effects related to the behaviour of the first animals entering the traps. The main limitation associated with the use of active gear surveys applied to benthic crustaceans is that towed nets or dredges are limited to trawlable seabed, which may introduce some biases considering that crustaceans may also live in rocky habitats (Smith and Tremblay, 2003). For this reason, the abundance estimates provided in this study should be interpreted with caution given the assumptions (no herding and crabs unable to escape by swimming over the nets/dredge) and should be regarded as a minimum estimate. Abundance data may be used in stock assessment models and when combined with other sources of data such as effort, landings and catch composition (e.g. length) may be able to provide information on the dynamics of the stock (Maunder and Piner, 2015). There have previously been no abundance indices available for the Scottish brown crab fisheries. The methodology developed here was applied to the North Sea but could potentially be extended to other areas in Europe covered by scallop and dredge surveys. This approach has the potential to be used in stock assessment to determine stock status and provide future advice on sustainable levels of catches, contributing to the management of the brown crab fisheries around Scotland and elsewhere. This paper was presented (oral presentation) at the International Council for the Exploration of the Sea (ICES) Shellfish Symposium (under the session “Assessment and population dynamics of shellfish”), which took place in November 2019 in Tromsø, Norway. Acknowledgements We would like to thank the MSS staff who participated in the east coast dredge surveys and IBTS quarter 3 surveys for the data sampling work that has been used in this study. Thanks to Andrzej Jaworski for providing advice on the use of geostatistical methods. Finally, weuld also like to thank the editor and two anonymous reviewers for their helpful comments. Funding This study was funded by the Marine Collaboration Research Forum (MarCRF), with Marine Scotland Science and the University of Aberdeen. Data availability statement The data underlying this article will be shared on reasonable request to the corresponding author. References Arcuti S. , Calculli C. , Pollice A. , D’Onghia G. , Maiorano P. , Tursi A. 2013 . Spatio-temporal modelling of zero-inflated deep-sea shrimp data by Tweedie generalized additive . Statistica , 73 : 87 – 101 . Google Scholar OpenURL Placeholder Text WorldCat Bakke S. , Buhl-Mortensen L. , Buhl-Mortensen P. 2019 . Some observations of Cancer pagurus Linnaeus, 1758 (Decapoda, Brachyura) in deep water . Crustaceana , 92 : 95 – 105 . Google Scholar Crossref Search ADS WorldCat Basford D. , Eleftheriou A. , Raffaelli D. 1989 . The epifauna of the northern North Sea (56–61 N) . Journal of the Marine Biological Association of the United Kingdom , 69 : 387 – 407 . Google Scholar Crossref Search ADS WorldCat Beck M. W. , Heck K. L. , Able K. W. , Childers D. L. , Eggleston D. B. , Gillanders B. M. , Halpern B. , et al. 2001 . The identification, conservation, and management of estuarine and marine nurseries for fish and invertebrates: a better understanding of the habitats that serve as nurseries for marine species and the factors that create site-specific variability in nursery quality will improve conservation and management of these areas . Bioscience , 51 : 633 – 641 . Google Scholar Crossref Search ADS WorldCat Bell M. , Eaton D. , Bannister R. , Addison J. 2003 . A mark-recapture approach to estimating population density from continuous trapping data: application to edible crabs, Cancer pagurus, on the east coast of England . Fisheries Research , 65 : 361 – 378 . Google Scholar Crossref Search ADS WorldCat Bennett D. B. 1974 . Growth of the edible crab (Cancer pagurus L.) off south-west England . Journal of the Marine Biological Association of the United Kingdom , 54 : 803 – 823 . Google Scholar Crossref Search ADS WorldCat Bennett D. B. 1995 . Factors in the life history of the edible crab (Cancer pagurus L.) that influence modelling and management . ICES Marine Science Symposia , 199 : 89 – 98 . Google Scholar OpenURL Placeholder Text WorldCat Bennett D. B. , Brown C. 1983 . Crab (Cancer pagurus) migrations in the English Channel . Journal of the Marine Biological Association of the United Kingdom , 63 : 371 – 398 . Google Scholar Crossref Search ADS WorldCat Bivand R. S. , Pebesma E. J. , Gómez-Rubio V. , Pebesma E. J. 2008 . Applied Spatial Data Analysis with R. Springer, New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Brown C. , Bennett D. 1980 . Population and catch structure of the edible crab (Cancer pagurus) in the English Channel . ICES Journal of Marine Science , 39 : 88 – 100 . Google Scholar Crossref Search ADS WorldCat Candy S. 2004 . Modelling catch and effort data using generalised linear models, the Tweedie distribution, random vessel effects and random stratum-by-year effects . CCAMLR Science , 11 : 59 – 80 . Google Scholar OpenURL Placeholder Text WorldCat Chambers J. M. , Hastie T. J. 1992 . Statistical Models in S . Wadsworth & Brooks/Cole Advanced Books & Software , Pacific Grove, CA . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Coleman M. , Rodrigues E. 2016 . Orkney Shellfish Project End of Year Report: January–December 2015 . Orkney Sustainable Fisheries Ltd . Report No 13: 86 pp. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Cooper R. , Green S. , Long D. 2013 . User guide for the British Geological Survey DiGSBS250K dataset. British Geological Survey Internal Report, IR/11/026. 17 pp. Delignette-Muller M. L. , Dutang C. 2015 . fitdistrplus: an R package for fitting distributions . Journal of Statistical Software , 64 : 1 – 34 . Google Scholar Crossref Search ADS WorldCat Dickens S. 2012 . Surveying the abundance and distribution of juvenile Cancer pagurus (L.) in littoral areas of Anglesey and the Llŷn Peninsula, North Wales. MSc thesis, School of Ocean Sciences, University of Wales, Bangor. 108 pp. Dobby H. , Fryer R. , Gibson T. , Kinnear S. , Turriff J. , Mclay A. 2017 . Scottish Scallop Stocks: Results of 2016 Stock Assessments. Scottish Marine and Freshwater Science, 8 : 178 . Dunn P. K. , Smyth G. K. 2005 . Series evaluation of Tweedie exponential dispersion model densities . Statistics and Computing , 15 : 267 – 280 . Google Scholar Crossref Search ADS WorldCat Dunn P. K. , Smyth G. K. 2008 . Evaluation of Tweedie exponential dispersion model densities by Fourier inversion . Statistics and Computing , 18 : 73 – 86 . Google Scholar Crossref Search ADS WorldCat Eaton D. , Brown J. , Addison J. , Milligan S. , Fernand L. 2001 . Larvae surveys of edible crab (Cancer pagurus) off the east coast of England: implications for stock structure and management. ICES CM 2001/J:14. 10 pp. Edwards E. 1979 . The Edible Crab and Its Fishery in British Waters. Buckland Foundation Books/Fishing News Books Ltd., Bath, Great Britain/Farnham, Surrey, England. Fariña A. , Freire J. , González-Gurriarán E. 1997 . Megabenthic decapod crustacean assemblages on the Galician continental shelf and upper slope (north-west Spain) . Marine Biology , 127 : 419 – 434 . Google Scholar Crossref Search ADS WorldCat Folk R. L. 1954 . The distinction between grain size and mineral composition in sedimentary-rock nomenclature . The Journal of Geology , 62 : 344 – 359 . Google Scholar Crossref Search ADS WorldCat Goetz K. T. , Montgomery R. A. , Ver Hoef J. M. , Hobbs R. C. , Johnson D. S. 2012 . Identifying essential summer habitat of the endangered beluga whale Delphinapterus leucas in Cook Inlet . Alaska. Endangered Species Research , 16 : 135 – 147 . Google Scholar Crossref Search ADS WorldCat Haig J. A. , Bakke S. , Bell M. C. , Bloor I. S. M. , Cohen M. , Coleman M. , Dignan S. , et al. 2016 . Reproductive traits and factors affecting the size at maturity of Cancer pagurus across Northern Europe . ICES Journal of Marine Science , 73 : 2572 – 2585 . Google Scholar Crossref Search ADS WorldCat Hall S. , Robertson M. , Basford D. , Fryer R. 1993 . Pit-digging by the crab Cancer pagurus: a test for long-term, large-scale effects on infaunal community structure . Journal of Animal Ecology , 62 : 59 – 66 . Google Scholar Crossref Search ADS WorldCat Harrison M. K. , Crespi B. J. 1999 . A phylogenetic test of ecomorphological adaptation in Cancer crabs . Evolution , 53 : 961 – 965 . Google Scholar Crossref Search ADS PubMed WorldCat Hartnoll R. 1969 . Mating in the Brachyura . Crustaceana , 16 : 161 – 181 . Google Scholar Crossref Search ADS WorldCat Hilborn R. , Walters C. J. 1992 . Quantitative fisheries stock assessment: choice, dynamics and uncertainty . Reviews in Fish Biology and Fisheries , 2 : 177 – 178 . Google Scholar Crossref Search ADS WorldCat Howard A. E. 1982 . The distribution and behaviour of ovigerous edible crabs (Cancer pagurus), and consequent sampling bias . ICES Journal of Marine Science , 40 : 259 – 261 . Google Scholar Crossref Search ADS WorldCat Hunter E. , Eaton D. , Stewart C. , Lawler A. , Smith M. T. 2013 . Edible crabs “Go West”: migrations and incubation cycle of Cancer pagurus revealed by electronic tags . PLoS One , 8 : e63991 . Google Scholar Crossref Search ADS PubMed WorldCat ICES. 2018 . Report of the International Bottom Trawl Survey Working Group (IBTSWG), 19–23 March 2018, Oranmore, Ireland. ICES Document CM 2018/EOSG: 01. 233 pp. Jenkins S. , Mullen C. , Brand A. 2004 . Predator and scavenger aggregation to discarded by-catch from dredge fisheries: importance of damage level . Journal of Sea Research , 51 : 69 – 76 . Google Scholar Crossref Search ADS WorldCat Jones G. , Gibson P. , Dobby H. , McLay A. 2010 . Brown crab (Cancer pagurus) migrations off the Northern Scottish Coast. Scottish Industry Science Partnership Report 02/10. 16 pp. Jouffre D. , Borges M. d. F. , Bundy A. , Coll M. , Diallo I. , Fulton E. A. , Guitton J. , et al. 2010 . Estimating EAF indicators from scientific trawl surveys: theoretical and practical concerns . ICES Journal of Marine Science , 67 : 796 – 806 . Google Scholar Crossref Search ADS WorldCat Kaiser M. , Hill A. , Ramsay K. , Spencer B. , Brand A. , Veale L. , Prudden K. , et al. 1996 . Benthic disturbance by fishing gear in the Irish Sea: a comparison of beam trawling and scallop dredging . Aquatic Conservation: Marine and Freshwater Ecosystems , 6 : 269 – 285 . Google Scholar Crossref Search ADS WorldCat Latrouite D. , Le Foll D. 1989 . Données sur les migrations des crabes tourteau Cancer pagurus et les araignées de mer Maja squinado . Océanis , 15 : 133 – 142 . Google Scholar OpenURL Placeholder Text WorldCat Maravelias C. D. 1997 . Trends in abundance and geographic distribution of North Sea herring in relation to environmental factors . Marine Ecology Progress Series , 159 : 151 – 164 . Google Scholar Crossref Search ADS WorldCat Matheron G. 1963 . Principles of geostatistics . Economic Geology , 58 : 1246 – 1266 . Google Scholar Crossref Search ADS WorldCat Maunder M. N. , Piner K. R. 2015 . Contemporary fisheries stock assessment: many issues still remain . ICES Journal of Marine Science , 72 : 7 – 18 . Google Scholar Crossref Search ADS WorldCat Mesquita C. , Miethe T. , Dobby H. , Mclay A. 2017 . Crab and lobster fisheries in Scotland: results of stock assessments 2013–2015 . Scottish Marine and Freshwater Science , 8 : 87 . Google Scholar OpenURL Placeholder Text WorldCat Mill A. , Dobby H. , McLay A. , Mesquita C. 2009 . Crab and Lobster fisheries in Scotland: an overview and results of stock assessments. Marine Scotland Science Internal Report 16/09: 71. Miller R. J. 1990 . Effectiveness of crab and lobster traps . Canadian Journal of Fisheries and Aquatic Sciences , 47 : 1228 – 1251 . Google Scholar Crossref Search ADS WorldCat Miller T. J. , Hart D. R. , Hopkins K. , Vine N. H. , Taylor R. , York A. D. , Gallager S. M. 2019 . Estimation of the capture efficiency and abundance of Atlantic sea scallops (Placopecten magellanicus) from paired photographic–dredge tows using hierarchical models . Canadian Journal of Fisheries and Aquatic Sciences , 76 : 847 – 855 . Google Scholar Crossref Search ADS WorldCat Mullahy J. 1986 . Specification and testing of some modified count data models . Journal of Econometrics , 33 : 341 – 365 . Google Scholar Crossref Search ADS WorldCat Murase H. , Nagashima H. , Yonezaki S. , Matsukura R. , Kitakado T. 2009 . Application of a generalized additive model (GAM) to reveal relationships between environmental factors and distributions of pelagic fish and krill: a case study in Sendai Bay, Japan . ICES Journal of Marine Science , 66 : 1417 – 1424 . Google Scholar Crossref Search ADS WorldCat OceanWise. 2015 . Marine and Coastal Data Products User Guide. https://www.oceanwise.eu/wp-content/uploads/2014/10/Marine-and-Coastal-Data-Products-User-Guide-v1.6.pdf(last accessed 07 March 2019). Oliver M. , Webster R. 2014 . A tutorial guide to geostatistics: computing and modelling variograms and kriging . Catena , 113 : 56 – 69 . Google Scholar Crossref Search ADS WorldCat Öndes F. , Emmerson J. A. , Kaiser M. J. , Murray L. G. , Kennington K. 2019 . The catch characteristics and population structure of the brown crab (Cancer pagurus) fishery in the Isle of Man, Irish Sea . Journal of the Marine Biological Association of the United Kingdom , 99 : 119 – 133 . Google Scholar Crossref Search ADS WorldCat Öndes F. , Kaiser M. J. , Murray L. G. 2016 . Quantification of the indirect effects of scallop dredge fisheries on a brown crab fishery . Marine Environmental Research , 119 : 136 – 143 . Google Scholar Crossref Search ADS PubMed WorldCat Orensanz J. M. , Gallucci V. F. 1988 . Comparative study of postlarval life-history schedules in four sympatric species of Cancer (Decapoda: Brachyura: Cancridae) . Journal of Crustacean Biology , 8 : 187 – 220 . Google Scholar Crossref Search ADS WorldCat Pallas A. , Garcia-Calvo B. , Corgos A. , Bernardez C. , Freire J. 2006 . Distribution and habitat use patterns of benthic decapod crustaceans in shallow waters: a comparative approach . Marine Ecology Progress Series , 324 : 173 – 184 . Google Scholar Crossref Search ADS WorldCat Pardo L. M. , Palma A. T. , Prieto C. , Sepulveda P. , Valdivia I. , Ojeda F. P. 2007 . Processes regulating early post-settlement habitat use in a subtidal assemblage of brachyuran decapods . Journal of Experimental Marine Biology and Ecology , 344 : 10 – 22 . Google Scholar Crossref Search ADS WorldCat Pebesma E. J. 2004 . Multivariable geostatistics in S: the gstat package . Computers & Geosciences , 30 : 683 – 691 . Google Scholar Crossref Search ADS WorldCat Pennino M. G. , Conesa D. , López-Quílez A. , Munoz F. , Fernández A. , Bellido J. M. 2016 . Fishery-dependent and-independent data lead to consistent estimations of essential habitats . ICES Journal of Marine Science , 73 : 2302 – 2310 . Google Scholar Crossref Search ADS WorldCat Petitgas P. , Woillez M. , Rivoirard J. , Renard D. , Bez N. 2017 . Handbook of geo-statistics in R for fisheries and marine ecology. ICES Cooperative Research Report No. 338. 177 pp. R Core Team. 2018 . R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/. Rivoirard J. , Simmonds J. , Foote K. , Fernandes P. , Bez N. 2000 . Geostatistics for Estimating Fish Abundance . Blackwell Science , Oxford . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Robinson M. , Tully O. 2000 . Seasonal variation in community structure and recruitment of benthic decapods in a sub-tidal cobble habitat . Marine Ecology Progress Series , 206 : 181 – 191 . Google Scholar Crossref Search ADS WorldCat Scottish Sea Fisheries Statistics. 2018 . Scottish Sea Fisheries Statistics 2018. The Scottish Government, Edinburgh. 116 pp. Shelton R. G. J. , Hall W. B. 1981 . A comparison of the efficiency of the Scottish creel and the inkwell pot in the capture of crabs and lobsters . Fisheries Research , 1 : 45 – 53 . Google Scholar Crossref Search ADS WorldCat Shono H. 2008 . Application of the Tweedie distribution to zero-catch data in CPUE analysis . Fisheries Research , 93 : 154 – 162 . Google Scholar Crossref Search ADS WorldCat Smith M. T. , Addison J. T. 2003 . Methods for stock assessment of crustacean fisheries . Fisheries Research , 65 : 231 – 256 . Google Scholar Crossref Search ADS WorldCat Smith S. J. , Tremblay M. J. 2003 . Fishery-independent trap surveys of lobsters (Homarus americanus): design considerations . Fisheries Research , 62 : 65 – 75 . Google Scholar Crossref Search ADS WorldCat Tallack S. M. L. 2002 . The Biology and Exploitation of Three Crab Species in the Shetland Islands, Scotland: Cancer pagurus, Necora puber and Carcinus maenas. PhD thesis, The North Atlantic Fisheries College, Shetland, and the University of the Highlands and Islands, Inverness. Tweedie M. 1984 . An index which distinguishes between some important exponential families. In Statistics: Applications and New Directions: Proceedings of the Indian Statistical Institute Golden Jubilee International Conference. Indian Statistical Institute, Calcutta. pp. 579 – 604 . Ungfors A. , Hallbäck H. , Nilsson P. G. 2007 . Movement of adult edible crab (Cancer pagurus L.) at the Swedish West Coast by mark-recapture and acoustic tracking . Fisheries Research , 84 : 345 – 357 . Google Scholar Crossref Search ADS WorldCat Vølstad J. , Sharov A. , Davis G. , Davis B. 2000 . A method for estimating dredge catching efficiency for blue crabs, Callinectes sapidus, in Chesapeake Bay . Fishery Bulletin , 98 : 410 – 420 . Google Scholar OpenURL Placeholder Text WorldCat Wood S. N. 2006 . Generalized Additive Models: An Introduction with R. Chapman Hall/CRC, Florida, USA . 392 pp. Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Wood S. N. 2011 . Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models . Journal of the Royal Statistical Society Series B-Statistical Methodology , 73 : 3 – 36 . Google Scholar Crossref Search ADS WorldCat Wood S. N. 2013 . On p-values for smooth components of an extended generalized additive model . Biometrika , 100 : 221 – 228 . Google Scholar Crossref Search ADS WorldCat Zeileis A. , Kleiber C. , Jackman S. 2008 . Regression models for count data in R . Journal of Statistical Software , 27 : 1 – 25 . Google Scholar OpenURL Placeholder Text WorldCat Zhou H. , Yang Y. , Qian W. 2018 . Tweedie gradient boosting for extremely unbalanced zero-inflated data. arXiv preprint arXiv: 1811.10192. Zuur A. F. , Ieno E. N. , Walker N. J. , Saveliev A. A. , Smith G. M. 2009 . Zero-truncated and zero-inflated models for count data. In Mixed Effects Models and Extensions in Ecology with R , pp. 261 – 293 . Springer, New York . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC © Crown copyright 2020. This article contains public sector information licensed under the Open Government Licence v3.0 (http://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/)
TI - Abundance and spatial distribution of brown crab (Cancer pagurus) from fishery-independent dredge and trawl surveys in the North Sea
JO - ICES Journal of Marine Science: Journal du Conseil
DO - 10.1093/icesjms/fsaa105
DA - 2020-08-05
UR - https://www.deepdyve.com/lp/oxford-university-press/abundance-and-spatial-distribution-of-brown-crab-cancer-pagurus-from-3hqEebuFBq
SP - 1
EP - 1
VL - Advance Article
IS - 
DP - DeepDyve
ER -