INTRODUCTION
Biological information, such as body length and weight, constitute necessary data for assessing population structure, particularly to estimate the biomass from the length frequency distribution and to convert length-at-age to weight-at-age (Froese, Reference Froese2006). However, conversely to length measurement it is difficult to obtain the weight with good accuracy during sampling at sea or from an underwater stereo-video system. Consequently, the characterization of the Weight-Length Relationship (WLR) allows for establishing the value of the unknown variable from the known variable. Moreover, this relationship is a sustainable proxy for the ‘fatness’ and ‘general well-being’ as the condition factor (Le Cren, Reference Le Cren1951; Tesch, Reference Tesch and Ricker1968; Weatherley & Gill, Reference Weatherley and Gill1987). In fish species, WLR is often defined by an exponential function under conditions of isometric growth (regression follows the cube law; Ricker, Reference Ricker1975). However, in nature, this relationship depends on the environmental conditions – the physiological state of the fish also has to be considered (Le Cren, Reference Le Cren1951; Froese, Reference Froese2006; Pauly, Reference Pauly2010; Mozsar et al., Reference Mozsar, Boros, Sály, Antal and Nagy2015) – and the exponent or growth coefficient (b) can vary between 2.5 and 4 (Hile, Reference Hile1936; Martin, Reference Martin1949; Pauly & Gayanilo, Reference Pauly and Gayanilo1997; Froese, Reference Froese1998, Reference Froese2006). In this study, the influence of factors such as sampling year and quarter, geographic area and sex were evaluated through the WLR which were estimated for 45 species, sampled during five scientific surveys operating from the North Sea to the Bay of Biscay and covering the entire length range from juveniles to adults.
MATERIALS AND METHODS
Sampling was conducted on the research vessels ‘Thalassa’ and ‘Gwen-Drez’ each year from 2013 to 2015, totalling five bottom-trawl surveys (Figure 1):
• IBTS survey (International Bottom Trawl Survey), North Sea and Eastern English Channel, January–February (Vérin, Reference Vérin1992).
• CGFS survey (Channel GroundFish Survey), Eastern English Channel, October (Coppin & Travers-Trolet, Reference Coppin and Travers-Trolet1989).
• CAMANOC survey (CAmpagne MANche Occidentale), Eastern and Western English Channel, September–October (Travers-Trolet & Vérin, Reference Travers-Trolet and Vérin2014).
• EVHOE survey (ÉValuation Halieutique de l'Ouest de l'Europe), the Celtic Sea and the Bay of Biscay, October-November (Mahé, Reference Mahé1987).
• LANGOLF (LANgoustine GOLFe de Gascogne), the Bay of Biscay, May (Garren & Martin, Reference Garren and Martin2013).
For this study, 31,167 marine individuals were individually weighed (total weight, W to the nearest gram) and measured (Total length, L to the nearest centimetre below) on board from all daylight hauls. When possible, the sex of Actinopterygii and Elasmobranchii was determined by macroscopic observation of the gonads (ICES, 2014). A total of 45 species were determined: Actinopterygii (N = 29,083) represented by 37 species (28 roundfishes and nine flatfishes).
Fig. 1. Location of trawling stations from the Bay of Biscay to the North Sea sampled by the five scientific surveys (EVHOE, LANGOLF, CAMANOC, CGFS, IBTS), where the 31,167 individuals used in this study have been sampled.
Elasmobranchii (N = 2084) represented by eight species (Table 1; Anonymous, 2016).
Table 1. Characteristics of the 45 fish species caught from the Bay of Biscay to the North Sea during 2013, 2014 and 2015: number of sampled individuals (N), mean length ± SD (cm), length range (cm), mean weight ± SD (g) and weight range (g).
Within each class, species are listed in alphabetical order of their family.
Before characterization of the WLR took place, all pairs of data for each species were plotted in order to identify and delete obvious outliers. In order to estimate the parameters of the allometric WLR (equation (1)), its base-10 logarithm (equation (2)) was fitted for each species to data using a least squared linear model:
$$W = a\,L^{b}.$$
$${\rm log}\,W = {\rm log}\,a + b\,{\rm log}\,L$$where ‘a’ is the intercept or initial growth coefficient and ‘b’ is the slope i.e. the growth coefficient (Le Cren, Reference Le Cren1951; Ricker, Reference Ricker1975; Froese, Reference Froese2006).
To investigate variations of the relationship between body length and weight for each species a completed Generalized Linear Model was performed according to the following explanatory variables:
• Geographic area (A): North Sea (ICES divisions IVb & IVc); Eastern English Channel (ICES division VIId), Western English Channel (ICES division VIIe), Celtic Sea (ICES divisions VIIg, VIIh & VIIj) and the Bay of Biscay (ICES divisions VIIIa & VIIIb).
• Sex (S): Female and Male.
• Sampling year (Y): 2013, 2014 and 2015.
• Sampling quarter (Q): 1, 2, 3 and 4.
For each species, data were deleted when the data number from explanatory variables was lower than 10. The individual weight of each species was modelled on body length as a continuous effect and geographic area, sex, sampling year and quarter as factors (equation (3)):
$$\eqalign{{\rm log} \,W & \sim {\rm log} \,L + A + S + Y + Q + {\rm log} \,L \times A + {\rm log} \,L \times S \cr & + {\rm log} \,L \times Y + {\rm log} \,L \times Q} $$with the separate influence of factors A (log L × A), S (log L × S), Y (log L × Y) and Q (log L × Q) on the relationship between body length and weight. For each species, the normality of the dataset was tested by a Quantile-Quantile Plot of the residuals (Zuur et al., Reference Zuur, Ieno and Smith2007).
To characterize the difference in the WLR for each species of fish, the condition factor, K, has been employed (Le Cren, Reference Le Cren1951, equation (4)):
$$K = 1000.{\rm W/}{\rm L}^3$$Fish with a high value of K are heavy for their length, while fish with a low value are light for their length.
All statistical analyses were carried out using the ‘CAR’ package (Fox & Weisberg, Reference Fox and Weisberg2011) in the statistical environment R (R Core Team, 2016).
RESULTS
Data relative to each species are presented in Table 1 with the number of measured specimens and the minimum, maximum and mean ± SD of length and weight. For Actinopterygii, measured length (29.0 ± 13.4 cm) and weight (401.3 ± 925.6 g) ranged respectively from 3 cm (Chelidonichthys lucerna) to 220 cm (Conger conger) and from 1 g (several species) to 45,000 g (Conger conger) and for Elasmobranchii, measured length (69.6 ± 21.4 cm) and weight (2173.8 ± 1967.8 g) ranged respectively from 3 cm (Raja clavata) to 150 cm (Galeorhinus galeus) and from 1 g (Raja clavata) to 19,000 g (Mustelus asterias) (Table 1). The samples were distributed by sex, sampling year, sampling quarter and by geographic area (Supplementary Table 1). Among the 45 tested species, all showed a significant correlation (P < 0.05) between body length and weight. The parameters of the WLR are given in Supplementary Table 2. The initial growth coefficient ‘a’ varied from 4.2 × 10−4 ± 1.0 × 10−5 in Conger conger to 6.6 × 10−2 ± 4.8 × 10−2 in Scophthalmus maximus, while the growth coefficient ‘b’ ranged from 2.7 ± 1.2 × 10−2 in Hyperoplus immaculatus to 3.5 ± 8.2 × 10−3 in Conger conger. The coefficients of the WLR are significantly correlated (Figure 2). Among the 45 tested species, the value of b was under 3 for 14 species (31.1%) with 12 roundfishes and two flatfishes (Supplementary Table 2). All Elasmobranchii species presented positive allometric growth (coefficient b higher than 3) (Supplementary Table 2; Figure 2).
Fig. 2. Relationship between the WLR parameters showed by a scatter plot of mean log a over mean b for 45 fish species by distinguishing the Actinopterygii (roundfishes and flatfishes) and the Elasmobranchii (sharks and skates) with body shape information. The regression line was realized from 45 fish species.
The four explanatory variables presented a significant effect on the WLR (Table 2), but only for whiting (Merlangius merlangus) and striped red mullet (Mullus surmuletus), were all four effectively significant at the same time. The influence of sex was estimated on the 39 species for which macroscopic observation was sufficient to determine sex identification. Slopes of WLR were significantly different between males and females for only 18 species (46.1%) of which 14 were Actinopterygii (Family Pleuronectidae: Pleuronectes platessa, Limanda limanda, Microstomus kitt, Platichthys flesus; Family Soleidae: Solea solea; Family Scophthalmidae: Scophthalmus maximus; Family Moronidae: Dicentrarchus labrax; Family Merlucciidae: Merluccius merluccius; Family Gadidae: Merlangius merlangus, Trisopterus esmarkii; Family Mullidae: Mullus surmuletus, Family Trachinidae: Trachinus draco; Family Phycidae: Phycis blennoides; Family Triglidae: Chelidonichthys cuculus) and four were Elasmobranchii (Family Trakidae: Mustelus asterias; Family Scyliorhinidae: Scyliorhinus canicula; Family Rajidae: Raja clavata, Raja montagui) (Table 1). The effect of the sex factor is more often observed in Elasmobranchii (50%) than in Actinopterygii (35.1%). Nevertheless, in Actinopterygii, this result fluctuated according to the fish shape (66.6% of flatfishes vs 21.9% of roundfishes). The geographic factor of dividing the results into five sampling ecoregions from the Bay of Biscay to the North Sea, was significant on WLR of only 18 species among 38 tested species (where species occur in sufficient number in these areas) (47.4%). These species were composed of 17 Actinopterygii (Hyperoplus immaculatus, Limanda limanda, Merlangius merlangus, Chelidonichthys cuculus, Lophius piscatorius, Sardina pilchardus, Gadus morhua, Lophius budegassa, Microstomus kitt, Phycis blennoides, Merluccius merluccius, Melanogrammus aeglefinus, Dicentrarchus labrax, Solea solea, Mullus surmuletus, Pollachius pollachius, Pleuronectes platessa) and only one Elasmobranchii (Raja undulata) (Table 2). Contrary to the sexual dimorphism, the spatial effect on the WLR was measured essentially for the Actinopterygii. The temporal effect on the WLR must be divided at two observation scales with the variations inter-years and intra-year (seasonality effect represented by the quarters). Among the 29 tested species with both temporal effects, only five (17.2%, Gadus morhua, Merlangius merlangus, Lophius piscatorius, Mullus surmuletus, Mustelus asterias) presented both significant variations inter-years and intra-year. Additionally, the year effect and the seasonality effect were significant at the level of 32.4 and 35.3% respectively. In Elasmobranchii the seasonality effect (42.8%) was more significant than between years (11.1%; Table 2).
Table 2. P-value for the relationship between weight and body length (W-L) and for the influence of sex.
Geographic area, Sampling year and Quarter on the WLR (P < 0.05 in grey cell) of the 45 fish species caught from the Bay of Biscay to the North Sea during 2013, 2014 and 2015. No value in the cell (–) indicates that the factor was not tested because there was only one modality.
To compare the fatness of each fish species according to geographic area, sex, sampling year and quarter, the condition factor (K) was estimated (Table 3). In the event of significant sexual dimorphism, all condition factors (K) of females were higher than those of males (Table 3). For the other tested factors, the highest values of K were distributed between all sampled years, areas and quarters; there was no observable trend (Table 3).
Table 3. Mean value of condition factor (K) of the 45 fish species according to each modality of the explanatory factors (Geographic area, Sex, Sampling year and Quarter) on the WLR. Grey cells indicate that a factor appears to have a significant effect (P <0.05) on the WLR (see Table 2 for P-values).
Only the individuals where the sex was determined were tested (F: female; M: male; -1: no sex information available).
DISCUSSION
The large sample data (N = 31,167) used in this study allows exploration of the possible effects of factors influencing the allometric WLR. According to Hile (Reference Hile1936); Martin (Reference Martin1949); Pauly & Gayanilo (Reference Pauly and Gayanilo1997) and Froese (Reference Froese1998, Reference Froese2006), ‘b’ values may range from 2.5 to 4 for fish, which is the case for the values estimated in our study. Moreover, the study showed that the coefficients of the WLR were significantly correlated. The growth coefficient (b) reflected firstly the shape and the fatness of the fish species. Consequently, the Elasmobranchii (sharks and skates) and the flatfishes presented only one body shape, known as depressiform, and consequently the weight growth was higher than the length growth (b > 3; Figure 2). This result corroborated the results obtained for Elasmobranchii (Pallaoro et al., Reference Pallaoro, Jardas and Santic2005; Yeldan & Avsar, Reference Yeldan and Avsar2007; Yığın & Ismen, Reference Yığın and Ismen2009) and for Soleidae (Torres et al., Reference Torres, Ramos and Sobrino2012). Among 28 roundfish species, the b values were within the range of 2.5–3.5 and there was no observed trend in body shape due to its large range of shapes as fusiform (i.e. Gadus morhua), arrow-like (i.e. Hyperoplus immaculatus), ribbon-like (Conger conger) or laterally flattened (i.e. Trachurus trachurus). The difference of shapes could be characterized by the ‘form factor’ equation of the log a–b relationship (Froese, Reference Froese2006; Verreycken et al., Reference Verreycken, Van Thuyne and Belpaire2011).
For all 45 species, the body length-weight relationship was significant. Our analyses confirmed those observed in the North-eastern Atlantic Ocean (Dorel, Reference Dorel1986; Coull et al., Reference Coull, Jermyn, Newton, Henderson and Hall1989; Silva et al., Reference Silva, Ellis and Ayers2013; Wilhelms, Reference Wilhelms2013), in Greek waters (Petrakis & Stergiou, Reference Petrakis and Stergiou1995), in the Persian Gulf (Naderi et al., Reference Naderi, Zare and Azvar2013) and in the Aegean Sea (Moutopoulos & Stergiou, Reference Moutopoulos and Stergiou2002). Consequently, it is possible for these marine species to use WLR to estimate weight from length or vice versa. For each species, significant differences could nevertheless be observed according to sex, sampled year, seasonality and geographic area. The first tested factor is the sex. The sexual dimorphism influenced significantly the WLR of a few species as observed in the Azores Islands (Morato et al., Reference Morato, Afonso, Lourinho, Barreiros, Santos and Nash2001). The difference observed between males and females for striped red mullet (Mullus surmuletus) corroborated the previous study on this species during 2004 in the Eastern English Channel (Mahé et al., Reference Mahé, Coppin, Vaz and Carpentier2013). The results of sexual dimorphism effect on the WLR were similar in the Eastern Adriatic Sea, except for Mustelus asterias, but the low number of data in the Mediterranean Sea for one species could be one explanation (Pallaoro et al., Reference Pallaoro, Jardas and Santic2005). According to the value of K, sexual dimorphism manifests as females being heavier than the males at the same length. This trend was observed both in the Actinopterygii and Elasmobranchii. The current study was realized using five surveys covering all ecoregions, from the Bay of Biscay to the North Sea. Consequently, significant differences in their WLR were observed for many widely distributed species across their distribution area. These differences were a result of many morphotypes within a species or a family. For striped red mullet (Mullus surmuletus), there were two morphotypes according to the head shape between South and North populations (Bay of Biscay/Eastern English Channel; Mahé et al., Reference Mahé, Villanueva, Vaz, Coppin, Koubbi and Carpentier2014), which could explain the observed difference of condition factors. The head morphological variation, for one species between two geographic areas or habitats, is influenced by feeding behaviour (Hyndes et al., Reference Hyndes, Platell and Potter1997; Janhunen et al., Reference Janhunen, Peuhkuri and Piironen2009). Within a family, values or the trend of condition factors between two similar species could be opposite. This has been observed between Lophius budegassa and Lophius piscastorius and between Scophthalmus maximus and Scophthalmus rhombus during the same sampling years and quarters. Seasonal or annual differences in WLR and therefore in condition factor may be generally related to reproduction (gonad development and spawning period) or feeding activities (food availability and feeding rate) (Bagenal & Tesch, Reference Bagenal, Tesch and Begenal1978; Weatherley & Gill, Reference Weatherley and Gill1987; Wootton, Reference Wootton1990) but also attributed to differences in sampling, particularly length ranges. Throughout a year, significant difference of the condition factor according to the spawning period for each species (Supplementary Table 3), showed that the specimens were heaviest just before and during the spawning period. This seasonal oscillation of the WLR and the condition factor could be explained by environmental factors such as temperature but also by the availability of food and the physiological state of the fish (i.e. degree of gonad development) (Le Cren, Reference Le Cren1951; Froese, Reference Froese2006; Pauly, Reference Pauly2010; Mozsar et al., Reference Mozsar, Boros, Sály, Antal and Nagy2015).
SUPPLEMENTARY MATERIAL
The supplementary material for this article can be found at https://doi.org/10.1017/S0025315416001752
ACKNOWLEDGEMENTS
We would like to express our gratitude to all people involved in the collection of samples required in this study. Thanks are expressed to all scientists and the crew of RV ‘Thalassa’ and RV ‘Gwen-Drez’ » during the IBTS, CAMANOC, CGFS, EVHOE and LANGOLF surveys for their help with sample collection. We acknowledge the help of Mark Etherton for improvement of the language. We are grateful for the comments provided by the reviewers.
FINANCIAL SUPPORT
This study was supported by the Data Collection Framework (DCF; EC Reg. 199/2008, 665/2008; Decisions 2008/949/EC and 2010/93/EU) and the French FFP project DYSTRETE.
