INTRODUCTION
The correct taxonomic identification of species provides a critical baseline that supports the rest of the biological research (Last, Reference Last2007). Elasmobranchs in general have suffered major taxonomic constraints that have led to identification problems when by-catch and fisheries data are analysed, usually solved by grouping the data in higher (genus or family) taxonomic levels (e.g. Girard et al., Reference Girard, Lorance and Biseau2000; Zeeberg et al., Reference Zeeberg, Corten and Graaf2006).
The identification of the deep water elasmobranchs is even more problematic because of their morphological similarities and the lack of studies on these groups. Among these, thelantern sharks (genus Etmopterus), are particularly problematic because the small size of most species and lack of commercial interest, and even specific elasmobranch studies where most taxa were identified to species level have failed to identify lantern sharks to species level (e.g. McKinnell & Seki, Reference McKinnell and Seki1998). In the FAO official fisheries data (FAO, 2007) there are only two categories to list lantern sharks (Etmopterus spp. and Etmopterus spinax) indicating that identification problems occur in this group.
Elasmobranch fish are vulnerable to exploitation, due to their slow growth rates, late maturity and low fecundities (Cortés, Reference Cortés2000), with the potential for overexploitation even at low levels of fishing mortality (Stevens et al., Reference Stevens, Bonfil, Dulvy and Walker2000). Commercially exploited deep water shark species are amongst elasmobranchs with the highest risks of extinction (Fowler et al., Reference Fowler, Cavanagh, Camhi, Burgess, Cailliet, Fordham, Simpfendorfer and Musick2005), and implementing fisheries management or species conservation programmes is impossible when identification is still an issue.
Two species of lantern sharks of the genus Etmopterus are common in southern Portuguese waters: the velvet belly lantern shark, Etmopterus spinax (Linnaeus, 1758) and the smooth lantern shark, Etmopterus pusillus (Lowe, 1839). A third species, Etmopterus princeps Collett, 1904, has been recorded in these waters but it has not been recorded in any recent studies that have surveyed the area (e.g. Borges et al., Reference Borges, Erzini, Bentes, Costa, Gonçalves, Lino, Pais and Ribeiro2001; Erzini et al., Reference Erzini, Gonçalves, Bentes, Lino and Ribeiro2001; Monteiro et al., Reference Monteiro, Araujo, Erzini and Castro2001; Coelho et al., Reference Coelho, Erzini, Bentes, Correia, Lino, Monteiro, Ribeiro and Gonçalves2005).
Both these species occur in deep waters, on or near the bottoms of the outer continental shelves and slopes and have the same habitats in their overlapping ranges of distribution; E. spinax has been recorded between 70 and 2000 m depth (mostly between 200 and 500 m) and E. pusillus between 274 and 1000 m depth (Compagno et al., Reference Compagno, Dando and Fowler2005). Etmopterus spinax is restricted to the eastern Atlantic Ocean, including the Mediterranean Sea, whereas E. pusillus has a worldwide distribution, having been recorded on both sides of the Atlantic, the western and south-east Pacific, and the western Indian Oceans (Compagno et al., Reference Compagno, Dando and Fowler2005).
Even though both species of lantern shark presented in this study are valid, the morphological differences between them are not very evident and there is a need to determine which morphometric measurements are most useful for researchers to quickly distinguish between them. Even though these species can be separated by studying their exterior marks, there is a need for alternative methods for separating them, especially when identification cannot be done on board the research vessel. The use of multivariate analysis allows the processing of an entire set of morphometric measurements. Furthermore, when each morphometric measurement is contributing very little to the overall differences, multivariate analysis allows for those small differences to be added and analysed as a whole (Clarke & Warwick, Reference Clarke and Warwick2001). The objectives of the present study were to: (1) determine if the two Etmopterus species could be separated using multivariate analysis based on morphological features; (2) determine the most important morphometric measurements for separating the two species; and (3) determine if sexual dimorphism occurs in each species.
MATERIALS AND METHODS
For this work, a total of 104 specimens were sampled, corresponding to 69 Etmopterus spinax and 35 Etmopterus pusillus. Both species had samples from both sexes and covered a wide range of lengths (Table 1). The samples were collected between February and June 2003 by commercial fishing vessels using deep water trawls and longlines and operating off the southern Portuguese coast.
Table 1. Size (N) and total length (TL, in mm) range of the sample used in this study, for each sex and each species.
Min, minimum; Max, maximum.
Once in the laboratory, specimens were sexed and measured for a total of 27 different morphometric characteristics. These measurements covered all the general body areas of these sharks (head, trunk and tail), including general length measurements, body heights and widths and measurements on each of the fins (Table 2). Most measurements were taken with a digital calliper with 0.01 mm precision, except the largest body measurements that were taken with a 1 mm precision. All measurements throughout this study refer to millimetres and names and abbreviations were defined according to Compagno (Reference Compagno2001). All measurements were expressed in % of total length (TL).
Table 2. Morphometric measurements used in this study, with the corresponding abbreviations (Compagno, Reference Compagno2001). All measurements were taken with a precision of 0.01 mm except for measurements marked with an asterisk (*), that were taken with 1 mm precision.
For the paired structures, such as the pectoral and pelvic fins, both the left and the right side structures were measured and compared with analysis of covariance (ANCOVA), using TL as the covariate. Once it was determined that no differences occurred between the two sides, only the left side structure was used for the multivariate analysis.
Multivariate analysis was carried out with the PRIMER 6 software (Clarke & Gorley, Reference Clarke and Gorley2006). A matrix of the morphometric data as percentages of TL was created, and the variables species (E. spinax or E. pusillus) and sex (male or female) added as factors to test. This matrix was then used to build a similarity matrix, using Euclidean distances. Non-metric multidimensional scaling (MDS) of distances in this similarity matrix was used to visually evaluate differences between the two species and between sexes in each species. Analysis of similarities (ANOSIM) statistical tests were used to assess if the differences observed in the MDS plots were significant, both between species and between sexes. This statistic tests the null hypothesis that no differences occur between groups. In the cases where differences were found, a SIMPER (similarities of percentages) analysis was carried out, in order to determine the percentage of contribution of each morphological measurement to the overall differences. This analysis was carried out until the cumulative differences were higher than 50%.
In order to create a model capable of correctly separating these two species, a discriminant analysis was carried out with the STATISTICA 6 software (Statsoft, 2004). All the morphometric measurements (as percentages of TL) were used and the model was constructed using a backward stepwise approach, where all variables are included in the initial model and variables that contribute least to the prediction of group membership are sequentially removed. Thus, only the most important variables (those that contribute the most to the discrimination of the groups) are kept (Statsoft, 2004).
The coefficients of the discriminant analysis were calculated by determining optimal variable combinations in different functions (canonical analysis). Given that in this case there were only two species (groups), only one canonical function was calculated, given by:
where a is the constant and b 1 through to b m are the canonical coefficients of the morphometric variables x 1 through to x m. The correct identification to the species level of an Etmopterus specimen can be obtained by applying this formula.
RESULTS
The complete morphometric characterization of the two Etmopterus species is presented in Table 3. Some measurements such as pre caudal-fin length (PCL), fork length (FL) or dorsal caudal-fin margin (CDM) were fairly different between species, with a very slight overlapping of the ranges whilst others, such as the pectoral-fin anterior margin (P1A) or the pelvic-fin anterior margin (P2A) were very similar, with the ranges of values mostly overlapped. For some measurements such as abdomen width (ABW), there were large differences between males and females, probably due to the fact that the ABW in females increases proportionally more than in males once maturity is achieved.
Table 3. Morphometric characterization of males and females of Etmopterus spinax and E. pusillus. All values are presented as percentage of total length (TL), except TL that is given in mm. The mean values and ranges (in parentheses) are given.
For definition of abbreviations see Table 2.
In both species, no differences were detected for the measurements taken for the paired structures, namely the left and right side P1A (ANCOVA E. spinax: F = 1.01; P = 0.32; ANCOVA E. pusillus: F = 0.08; P = 0.78) and the left and right side P2A (ANCOVA E. spinax: F = 1.36; P = 0.25; ANCOVA E. pusillus: F = 0.64; P = 0.43). Therefore, for the remaining multivariate analysis only the left side structures were considered.
In both Etmopterus species, no sexual dimorphism was detected (Figure 1). Statistically, the visual analyses made with the MDS were corroborated by ANOSIM tests, with no differences detected between sexes: ANOSIM E. spinax: R = 0.055; P = 0.078 and ANOSIM E. pusillus: R = 0.057, P = 0.220.
Fig. 1. Non-metric multidimensional scaling of the morphometric differences between males (black dots) and females (white dots) of (A) Etmopterus spinax and (B) Etmopterus pusillus. The stress value refers to the error that is created when multidimensional data are plotted in two dimensions.
Given that no sexual dimorphism was detected, the comparison between the two species was made using the sexes combined. The multivariate visualization of the data with a MDS plot produced a clear separation between the species (Figure 2). Statistically, these differences were significant (ANOSIM: R = 0.491, P < 0.01).
Fig. 2. Non-metric multidimensional scaling of the morphometric differences between the two Etmopterus species. White dots refer to E. spinax and black dots refer to E. pusillus. The stress value refers to the error that is created when multidimensional data are plotted in two dimensions.
The SIMPER analysis evidenced the morphometrics that contributed most to the species separation (Table 4). A >50% cumulative difference was achieved when 11 morphometric traits were considered: the measurement that individually contributed more for the species separation was the FL, and accounted for 5.94% of the differences, followed by the CDM with 5.84% and PCL with 5.64%.
Table 4. Morphometric measurements that contributed most to the differences found between the two species. Only the morphometrics up to a cumulative difference of >50% are indicated.
For definition of abbreviations see Table 2.
With the backward stepwise discriminant analysis a model using only 6 variables (morphometrics) was created that adequately explains the variability between the two species (Wilks' Lambda = 0.077; approximately F = 178.69; P < 0.001). Given that in this case only two groups (species) were being separated, only one canonical function was needed to separate the data. With these canonical coefficients, an equation for identifying the species was created and defined as:
The centroids of this model were respectively −2.176 for Etmopterus spinax and 5.362 for Etmopterus pusillus when applied to the original data (Figure 3). The best cutting point is half way between these values (1.593) and in this way it is possible to determine the category (species) of a new observation (specimen) depending on the value obtained. For values greater than 1.593 the specimen is probably E. pusillus and if it is lower it is probably E. spinax. When a post hoc prediction was run on all specimens observed during this study, 100% accuracy was obtained between the model estimated and the actual observed species.
Fig. 3. Histogram of the values obtained when applying the model to the specimens used in this study. The expected normal curve, if only 1 group occurred, is also given.
DISCUSSION
This work showed both the importance and difficulty of separating some morphologically similar deep water shark species, in this case from the Etmopterus genus. Difficulties in separating elasmobranch catches to species level are common in fisheries and discards studies, and many authors have had to analyse the data considering a higher taxonomic level such as genus or even family. Examples of such cases are the deep water squaloid fishery in northern France where commercialized squaloid sharks are all identified under the same common name (Girard et al., Reference Girard, Lorance and Biseau2000) and the analysis of by-catch from pelagic trawlers in western Africa (Zeeberg et al., Reference Zeeberg, Corten and Graaf2006). In Portugal, the official fisheries statistics have most of the elasmobranch catch listed under a general ‘unidentified sharks and rays’ group (DGPA, 2006).
The particular case of the genus Etmopterus seems to be even more problematic and there are some studies, where most elasmobranch catches have been identified to species level except for the genus Etmopterus where specimens were only identified to genus level. Examples of such cases are the studies by Anderson & Clark (Reference Anderson and Clark2003) on the by-catch of the orange roughy, Hoplostethus atlanticus, in New Zealand or the work of McKinnell & Seki (Reference McKinnell and Seki1998) on shark by-catch in a Japanese squid fishery.
However, even though these species are morphologically very similar, they have different ecological life cycles and in order to have accurate fisheries management and conservation programmes, there is a need to correctly discriminate them to species level. In fact, deep water elasmobranch fish are amongst the most vulnerable marine organisms (Fowler et al., Reference Fowler, Cavanagh, Camhi, Burgess, Cailliet, Fordham, Simpfendorfer and Musick2005) and there is an urgent need for the implementation of efficient management and conservation programmes in the short term that will require fisheries and by-catch data specified to species level.
Even though lantern sharks are mainly discarded species, part of the catch may sometimes be landed and should therefore be recorded by the countries' fisheries statistics. The FAO fisheries statistics (FAO, 2007) have two categories to list lantern sharks: the general ‘lantern sharks’ group that refers to Etmopterus spp. and the ‘velvet belly (ETX)’ group that refers specifically to Etmopterus spinax. Up until 1994 no recordings appeared for either of these groups, meaning that before this date these species were being identified only to family level. Starting in 1995, some catches started to be recorded but most of the problems remained. In the case of Europe, from 1999 to 2005, the Etmopterus spp. group recorded 871 tonnes while the E. spinax group only recorded 75 tonnes, and this evidences clear limitations of these data sets at two levels: (1) most E. spinax are not being correctly identified to species level and therefore are being placed in the general Etmopterus spp. group; and (2) the relatively low biomass of these groups indicates that most data are not even being identified to genus level and are probably being placed in the Squalidae family group.
In this work the two Etmopterus species studied were well separated by multivariate analysis. Each morphometric measurement contributed relatively little for the overall differences, but when all these small differences were considered together it was possible to separate the species.
The discriminant analysis carried out is typically used to determine which variables discriminate between two or more naturally occurring groups (Statsoft, 2004). In this study, it was possible to create a model that could effectively separate between the two species, in this case using a backward stepwise approach that used only 6 of the 27 possible morphometric measurements. The post hoc tests, used to compare what was observed to what was being predicted by the model showed an accuracy of 100%, meaning that the model created accurately identified all the specimens used for this study. Four morphometric measurements were common both to the SIMPER and to the discriminant analysis and two of them relate the proportion between the specimen length and the caudal-fin length. This fact is important since these measurements are relatively easy to obtain: even in photographed specimens, these measurements can be estimated with relative ease and eventually allow for a posterior identification or confirmation of the species.
Some measurements, such as the abdomen width were clearly not adequate for separating these species. For these particular measurements, there is large intra-specific variability that is mostly related with sex and maturity stage and not so much with species characteristics. Once maturity is achieved, the abdomen width of females increases proportionally more than that of males due to the presence of large oocytes or embryos, and this intra-specific variability prevents these measurements from reflecting specific species characteristics.
No sexual dimorphism was detected in these species. Sexual dimorphism in terms of maximum sizes of Squalidae sharks is common with females usually attaining larger sizes than males and heavier weights for the same size (e.g. Ford, Reference Ford1921; Wetherbee, Reference Wetherbee1996; Girard & Du Buit, Reference Girard and Du Buit1999; Coelho & Erzini, Reference Coelho and Erzini2005; McLaughlin & Morrissey, Reference McLaughlin and Morrissey2005). However, the results presented here evidence that even though females might reach larger sizes than males, the body proportions are maintained and therefore no sexual dimorphism is observed, once the effect of growth is removed.
The two Etmopterus species studied have overlapped distributions, both in terms of depth and geographical areas, so most fisheries and by-catch studies carried out in the outer shelves and slopes of the eastern Atlantic Ocean are likely to catch both. With this work, a relatively easy way to separate both species based on morphometric traits that can inclusively be applied after the biological sampling procedures, using detailed photographs of the specimens, was demonstrated.
ACKNOWLEDGEMENTS
This study was funded by POCI 2010 (Programa Operacional Ciência e Inovação 2010) and FSE (Fundo Social Europeu) through a FCT (Fundação para a Ciência e Tecnologia) PhD grant (Ref. SFRH/BD/10357/2002) attributed to the first author. The authors are grateful to all fishermen who collaborated in collecting specimens for this study: special thanks go to Ilídio Diogo and Francisco Diogo, skippers of the longliner ‘Branca de Sagres’ and Helder Cavaco, skipper of the bottom trawler ‘Gamba’.
