ICES Journal of Marine Science: Journal du Conseil Advance Access originally published online on March 13, 2007
ICES Journal of Marine Science: Journal du Conseil 2007 64(3):512-518; doi:10.1093/icesjms/fsm017
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Element composition of sardine (Sardina pilchardus) otoliths along the Atlantic Coast of the Iberian Peninsula
Universidad de Vigo, Facultad de Ciencias del Mar, Campus Lagoas-Marcosende, 36310 Vigo, Pontevedra, Spain
tel: +34 986 812623; fax: +34 986 812556; e-mail: bcastro{at}uvigo.es
Castro, B. G. 2007. Element composition of sardine (Sardina pilchardus) otoliths along the Atlantic Coast of the Iberian Peninsula. – ICES Journal of Marine Science, 64: 512–518.The element composition of sardine (Sardina pilchardus) otoliths along the Atlantic Coast of the Iberian Peninsula was determined by bulk analysis, using Inductively Coupled Plasma-Mass Spectrometry. Samples (collected 2003–2005) were classified by area [Bay of Biscay (GV), Northwest (OCN), and Southwest (GC) of the Iberian Peninsula] and size group (
15, 16–18, and 
21 cm total length). The chemical composition was compared statistically between areas within size groups and between size groups within areas by ANOVA and a Support Vector Machine method, using 10-fold cross-validation. The element composition of sardine otoliths was similar to that of other marine fish. In all, eight elements (Li, Ba, Mg, Mn, Na, Sr, K, and Ca) were useful in discriminating between areas and/or size groups. Discrimination between areas within size groups was less effective for the biggest size group (67%) than for the middle (89%) or smallest size groups (84%). Discrimination between size groups within areas was better for GC (76%) and GV (84%) than for OCN (60%). The results are interpreted in terms of a net balance of migration of large sardine from OCN towards GV and GC.
Keywords: Iberian Peninsula, otolith chemistry, sardine, stock identification
Received 24 July 2006; accepted 21 January 2007; advance access publication 13 March 2007.
 |
Introduction |
|---|
 Top Introduction Material and methodsResultsDiscussionReferences |
|---|
The physical and chemical characteristics of water masses influence fish otolith composition, but this influence is not simple and, even in cases where the characteristics are known, it is not yet possible to predict the element composition of otoliths (Campana, 1999; Elsdon and Gillanders, 2004). However, such caveats do not impede the potential use of element otolith composition as an environmental fingerprint, i.e. a biological marker useful for discriminating among groups of fish living, at least temporally, in different waters (Thorrold et al., 2001; Rooker et al., 2003). When the discrimination refers to hatchlings or juveniles of migratory species, it can be helpful for holistic stock identification (Begg and Waldman, 1999; Campana, 2005).
Sardine (Sardina pilchardus) stock structure in the Northeast Atlantic is still an unresolved issue. Studies on meristic and morphometric characters have postulated four stocks in the area (Andreu, 1969; Furnestin and Furnestin, 1970): a septentrional Atlantic stock from the North Sea (57°N) to north of the Iberian Peninsula (43°N), an Iberian stock, from north of the Iberian Peninsula (43°N) to the Straits of Gibraltar (36°N), a Moroccan stock, from Cap Spartel (36°N) to Cap Juby (28°N), and a Saharian stock, from Cape Juby to Levrier Bay (21°N) in Mauritania. Following this classification, the Bay of Biscay has been considered for management purposes the limit between the septentrional and Iberian stocks, based on biological evidence and for administrative reasons (ICES, 1980), and the separation has been retained until now for assessment purposes (ICES, 2005a). However, recent work, including research within and outside the framework of the EU Project SARDYN, has yielded contradictory results about the identity of the stocks. Some results, based on morphometric variables (Silva, 2003) and analysis of catch variability in the area (Carrera and Porteiro, 2003), taken together with management inconsistencies, have challenged the assumption of stock unity for the Atlanto-Iberian sardine (ICES, 2005a). In contrast, based on genetic markers, the existence of an unique stock throughout the Northeast Atlantic, from the Gulf of Cádiz (GC) to the Celtic Sea and English Channel, has been postulated (Kasapidis et al., 2004). Clearly, therefore, there is need to enhance knowledge of the population structure and dynamics of sardine along the Atlantic Coast of the Iberian Peninsula.
From the Bay of Biscay to southern Portugal and the GC, sardine live in different water masses (OSPAR, 2000). Therefore, some differences in element composition of the otolith would be expected, depending on the location where the sardine have hatched and grown. The differences could be maintained over time if there was no mix of sardine between locations. However, if the hypothesis of a single stock is correct, a mixing effect and no differentiation in otolith composition of groups of fish from different geographical origin would be expected. Consequently, I seek here to evaluate the feasibility of using otolith chemical composition to delineate sardine from different parts of the Iberian Peninsula, to support efforts at clarifying sardine stock identity in the area.
|
Material and methods |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
Sardine samples
Samples for the study were obtained between 2003 and 2005 from research cruises or fishing boats (when the location of hauls was known) in three separate areas of the Iberian Atlantic Coast (Figure 1). These areas were Bay of Biscay (GV), from 2°48'W to 46°00'N, Northwest Iberian Peninsula (OCN), from 7°00'W to 40°54'N, and GC, from 6°20'W to 37°00'N. The sardine were separated into three size groups based on their total length:
15, 16–18 and 21 cm. Table 1 shows the sample sizes by area and size group.
|
|
Otolith extraction, preparation, and chemical analysis
Before microchemical analysis, otoliths were extracted from the fish under a binocular microscope using plastic tools, cleaned with ultra-pure H2O2 (30%) and ultra-pure water, and dried under a laminar flow hood. All plastics used for otolith extraction, manipulation, and storage were cleaned with 2% ultra-pure HNO3 for 14 h and rinsed with ultra-pure water before use.
Element analysis was performed at the SAI (University of Coruña) by a solution-based method using a Magnetic Sector Element 2 Thermo-Finnigan Inductively Coupled Plasma-Mass Spectrometer. Analysis was done separately for each fish except in the case of 48 small sardine ( 15 cm, OCN) belonging to the same haul, which were pooled in pairs, resulting in 24 chemical samples. Nine elements were analysed (7Li, 137Ba, 24Mg, 55Mn, 59Co, 23Na, 88Sr, 39K, and 44Ca), but Co was not employed in discriminating between areas or size groups because its concentration was usually less than the detection limit. The same situation applied in a few cases to Li, Mg, Mn, and Ba (see total number of specimens analysed by element in Figures 2 and 3). Li, Ba, Na, and Sr were measured at low resolution, after checking for the absence of interference at medium resolution. Mg, Mn, and Co were measured at medium resolution, and K and Ca were measured at high resolution. The mean limits of detection, calculated from the concentration of analyte yielding a signal equivalent to 10 x the standard deviation of the blank signal (n = 9), were on a dry weight basis: Li (0.05 µg g–1), Ba (0.05 µg g–1), Mg (1.5 µg g–1), Mn (0.45 µg g–1), Co (0.10 µg g–1), Na (0.02 mg g–1), Sr (0.001 mg g–1), K (0.01 mg g–1), and Ca (0.15 mg g–1). Element quantification was made by external calibration. 115In and 205Bi were added as internal standards for control of possible matrix effects and instrument drift.
|
|
Statistical analysis
Statistical analysis was performed using R.2.2.1 (R Development Core Team, 2005) and SPSS 13.0, and it was directed to test the utility of element composition of sardine otoliths for discriminating between geographical origin and size. The protocol employed was as follows. First, I compared, by chemical element, different areas within size groups or different size groups within areas, using ANOVA and Tukey post hoc tests. Those elements without significant differences between groups were not considered for discrimination purposes. When sizes were compared, I also tested whether there was any significant increase or decrease in the concentration of any element with size, all areas being pooled. Second, I selected non-redundant elements within those showing significant differences between groups. This was done using stepwise linear discriminant analysis. Third, I applied a Support Vector Machine method (SVM) with 10-fold cross-validation to test the discrimination between groups using the non-redundant elements. This procedure of validation was repeated 1000 times for each comparison between groups, obtaining total and partial accuracy means and their variances of correct classification (geographical origin or size). Finally, I compared the significance of correct classifications by a t-test against the results obtained also applying the SVM method, but after random sample assignment to a geographical origin or size group. SVMs were performed employing a radial kernel, and hyperparameters were adjusted after tuning (gamma = 1 and cost = 100), using the tuning routine of R's package e1071. SVM is a non-linear method derived from generalizing linear decision boundaries for classification (Bennett and Campbell, 2000; Hastie et al., 2001; Schölkopf and Smola, 2002).
|
Results |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
Otolith element composition by area within size groups
Figure 2 shows the element composition of sardine otoliths by area for sardine of the three size groups,
15 cm total length, between 16 and 18 cm, and 21 cm. The Figure also shows any statistically significant differences in the concentration of each chemical element analysed between areas within size groups. The number of elements with significant differences in concentration between areas decreased with size (7, 6, and 4, respectively), and there are few cases where an element separates the three areas (Mn for sardine 15 cm, Ba for sardine 16–18 cm, and Li for sardine 21 cm). Also, just one element, Ca, yields significant differences at the same time for the three size groups. After these comparisons, those elements showing significant differences between areas were further depurated to avoid redundancy of variables to be used for area discrimination by means of SVM. Table 2 lists those elements selected, for each size group, after stepwise linear discriminant analysis.
|
SVM was applied after this step to evaluate possible discrimination between areas of sardine of similar size based on otolith element composition. Table 3 shows the results of this procedure as total accuracy (percentage of total sardine correctly classified) and partial accuracy (percentage of sardine correctly classified in each area by size group) after 10-fold validation, compared with similar results obtained after random permutation of sardine geographical origin before SVM. In all cases, the percentages were statistically significantly higher for non-permuted samples (p < 0.001), but in the case of sardine
21 cm from GC, discrimination between size groups was quite limited (36.1%).
|
Otolith element composition by size group within areas
Figure 3 shows the element composition of sardine otoliths by size group for each area, GV, OCN, and GC. The Figure also reveals any statistically significant differences in the concentration of each chemical element analysed between size groups within geographical areas. Differences between sizes are bigger for GV and GC than for OCN, the number of elements with significant differences in concentration being, respectively, 7, 6, and 4. Changes in otolith composition with size do not show any consistent pattern among areas, and in just one case, Li in OCN, there was a clear trend, in this case of a decrease in concentration with growth.
After these comparisons, those elements showing significant differences between size groups were further depurated to avoid redundancy of variables to be used for size discrimination by means of SVM. In this case, all elements were selected after stepwise linear discriminant analysis, whichever size group was being considered. Table 4 shows the elements finally selected for SVM.
|
After this step, an SVM for evaluating possible discrimination between size groups of sardine within area was applied based on otolith element composition. Table 5 shows the results of this procedure as total accuracy (percentage of total sardine correctly classified) and partial accuracy (percentage of sardine correctly classified in each size group by area) after 10-fold validation, compared with similar results obtained after random permutation of sardine size groups before SVM. In all cases, the percentages were statistically significantly higher for non-permuted samples (p < 0.001), but for all size groups from OCN and for the 16–18 cm size group from GC, discrimination between sizes was moderate (<70%).
|
|
Discussion |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
The element composition of Iberian sardine otoliths is similar to that of other marine fish. All the elements analysed had concentration ranges within the limits found for other species (de Pontual and Geffen, 2002), with the exception of Sr, which had a lower concentration in this case, about 700 µg g– 1 compared with >1000 µg g– 1 in other species (de Pontual and Geffen, 2002). However, within normal ranges, the composition changes significantly with geographical area and size group (Figures 2 and 3). The origin of the differences is not clear, but water environment can cause them (Campana, 1999; Elsdon and Gillanders, 2004), because the three areas considered are influenced by different water masses and oceanographic features (OSPAR, 2000; Puillat et al., 2004; Peliz et al., 2005; García-Lafuente et al., 2006). As fish grow, their mobility increases (ICES, 2005b), and they are probably exposed to distinct environmental characteristics that could be reflected in the otolith element composition. Therefore, otolith element composition can be used to discriminate among groups of fish that have experienced different environmental histories. If the differences in otolith element composition are retained through years they can serve as the basis, along with other evidence, of stock discrimination (Begg and Waldman, 1999; Campana, 2005). In this case, sardine captured in different seasons and years were pooled to ensure that the sample size was adequate for statistical testing, so it was not possible to compare different years in terms of stability in the pattern of element composition by area and size group. However, important otolith element composition differences were found between areas and sizes, despite the pooling procedure used, which would not have been expected if the differences were only temporal.
Differences in otolith element composition between areas decreased with size (Figure 2). In the case of sardine 15 cm, the concentration of seven out of eight chemical elements changed with the origin of sardine, whereas in the largest size group ( 21 cm), just four out of eight elements showed a difference. An increase in the variability of element composition with size could explain this result. However, the change was not found when making comparisons of variation coefficients within size group of those chemical elements (Ba, Mn, Na, and Sr) not showing differences between areas for sardine 21 cm (Figure 3, Table 4). In those instances, the absence of a difference cannot be attributed to variability in element composition with size. A second explanation of the differences could be allometric change of composition with growth; a common pattern of variation would be expected in the three areas (GV, OCN, and GC), but with different slope. Again, however, if the elements Ba, Mn, Na, and Sr are compared, there is no common pattern in the three areas (Figure 3, Table 4). A third explanation, based on increased migration between areas with size, could be more plausible. As the composition of an otolith seems to depend, at least partially, on the characteristics of the water where the fish live (Campana, 1999; Buckel et al., 2004; Elsdon and Gillanders, 2004), if the extent of migration between areas increases with size, large sardine should have an otolith element composition that is more similar overall than would small less-migrant fish. Additional support for this homogenization in composition between areas with size is provided by the reduction in total accuracy of classification by SVM of large sardine (Table 3). Application of this classification was based exclusively on non-redundant elements, with significant differences in concentration between areas. The number of these elements was small (3) for sardine 21 cm (Table 2), but an equal number was used for size group 16–18 cm and the total accuracy obtained was the best (89.4%) of all three size groups analysed (Table 3). Therefore, the decrease in accuracy with size does not seem to be a problem of the number of variables employed in the classification, but of the homogeneity of composition. If increased migration by size is the reason for this homogeneity, the migration should not be significant in fish < 18 cm, because the medium size group and the smallest both gave good accuracies between areas, determined by SVM (Table 3).
Growth alone does not seem to be the cause of the differences in element composition of sardine otoliths within areas, when expressed with respect to otolith weight (Figure 3). If growth were the single cause of the differences, a similar pattern of changes with size in different areas would be expected, but this was not found (Figure 3). Differences with size were more or less pronounced, or absent, depending on area (Figure 3). Moreover, there was just one clear trend of element composition change with size, i.e. for Li in OCN (Figure 3). Therefore, it was considered unnecessary to analyse differences in the element composition of sardine otoliths within areas by MANCOVA or similar techniques, including size as covariate. Again, migration could explain the changes observed with size within areas. If migration between areas, or between waters with different characteristics inside areas, is related to size, and its importance changes depending on the origin of the fish, these results would be plausible. Therefore, based on SVM accuracies for size group within area (Table 5), sardine in OCN will remain in homogeneous water longer than would sardine in GV or GC.
In summary, differences in the element composition of otoliths are smaller within areas for larger sardine ( 21 cm) than for smaller fish ( 15 and 16–18 cm), and greater between sizes for the Bay of Biscay and the GC than for northern Portugal and Galicia. A migratory pattern where the net balance of movement of large sardine is from northern Portugal and Galicia to the Bay of Biscay and the GC could explain these results, and is compatible with other characteristics of the Iberian sardine (Silva, 2003). If sardine movements are significant in terms of population numbers, the existence of a single sardine stock in Atlanto-Iberian waters would be supported.
|
Acknowledgements |
|---|
The study was developed within the framework of the EU Project SARDYN (Contract Q5RS–2002–000818). I thank IEO (Spain), IPIMAR (Portugal) and the AZTI Foundation (Basque Country) for providing me the sardine samples, and Alicia Cantarero (SAI–Coruña University) for conducting the analytical work and resolving my doubts about ICPMS analysis. Christoph Stransky and an anonymous referee provided appreciated comments on an early draft of the manuscript.
|
References |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
-
Andreu B. (1969) Las branquiespinas en la caracterización de las poblaciones de Sardina pilchardus. Investigación Pesquera 33:425–607.
Begg G. A. and Waldman J. R. (1999) An holistic approach to fish stock identification. Fisheries Research 43:35–44.[CrossRef][Web of Science]
Bennett K. P. and Campbell C. (2000) Support vector machines: hype or hallelujah? SIGKDD Explorations 2:1–13.
Buckel J. A., Sharack B. L., Zdanowicz V. S. (2004) Effect of diet on otolith composition in Pomatomus saltatrix, an estuarine piscivore. Journal of Fish Biology 64:1469–1484.[CrossRef][Web of Science]
Campana S. E. (1999) Chemistry and composition of fish otoliths: pathways, mechanisms and applications. Marine Ecology Progress Series 188:263–297.[Web of Science]
Campana S. E. (2005) Otolith elemental composition as a natural marker of fish stocks. In Cadrin S. X., Friedland K. D., Waldman J. R. (Eds.). Stock Identification Methods(Elsevier Academic, Burlington, USA) pp. 227–245.
Carrera P. and Porteiro C. (2003) Stock dynamic of the Iberian sardine (Sardina pilchardus, W.) and its implication on the fishery off Galicia (NW Spain). Scientia Marina 67:Suppl. 1, 245–258.[Web of Science]
de Pontual H. and Geffen A. J. (2002) Otolith microchemistry. In Panfili J., [de] Pontual H., Troadec H., Wright P. J. (Eds.). Manual of Fish Sclerochronology(Ifremer-IRD coedition, Brest, France) pp. 243–302.
Elsdon T. S. and Gillanders B. M. (2004) Fish otolith chemistry influenced by exposure to multiple environmental variables. Journal of Experimental Marine Biology and Ecology 313:269–284.[CrossRef][Web of Science]
Furnestin J. and Furnestin M. L. (1970) La sardine marocaine et sa pêche. Migrations trophiques et génétiques en relation avec l'hydrologie et le plancton. Rapports et Procès-Verbaux des Réunions du Conseil International pour l'Exploration de la Mer 159:165–175.
García-Lafuente J., Delgado J., Criado-Aldeanueva F., Bruno M., Río J., Vargas J. M. (2006) Water mass circulation on the continental shelf of the Gulf of Cádiz. Deep-Sea Research II 53:1182–1197.[CrossRef]
Hastie T., Tibshirani R., Friedman J. (2001) The Elements of Statistical Learning. Data Mining, Inference, and Prediction(Springer, New York).
ICES. (1980) Rapport du Groupe du Travail pour l'évaluation des stocks de sardines dans les Divisions VIIIc et IXa du CIEM. pp. 41 ICES Document CM 1980/H: 53.
ICES. (2005a) Report of the Working Group on the Assessment of Mackerel, Horse Mackerel, Sardine and Anchovy. pp. 615 ICES Document CM 2005/ACFM: 08.
ICES. (2005b) Report of the study group on regional scale ecology of small pelagic fish (SGRESP). pp. 45.
Kasapidis P., Planes S., Laurent V., Quinta R., Teia A., Lenfant P., Darivianakis S., et al. (2004) Stock discrimination and temporal and spatial genetic variation of sardine (Sardina pilchardus) in northeastern Atlantic, with a combined analysis of nuclear (microsatellites and allozymes) and mitochondrial DNA markers. pp. 22 ICES Document CM 2004/Q: 21.
OSPAR. (2000) Geography, hydrography and climate. Quality Status Report 2000: Region IV—Bay of Biscay and Iberian Coast(OSPAR Commission, London) pp. 5–31.
Peliz A., Dubert J., Santos A. M. P., Oliveira P. B., Le Cann B. (2005) Winter upper ocean circulation in the western Iberian Basin—fronts, eddies and poleward flows: an overview. Deep-Sea Research I 5:621–646.
Puillat I., Lazure P., Jégou A. M., Lampert L., Miller P. I. (2004) Hydrographical variability on the French continental shelf in the Bay of Biscay, during 1990s. Continental Shelf Research 24:1143–1163.[CrossRef][Web of Science]
R Development Core Team. (2005) R: A language and environment for statistical computing. R Foundation for Statistical Computing. , Vienna, Austria ISBN 3 900051 07 0.
Rooker J. R., Secor D. H., Zdanowicz V. S., Metrio G., Relini L. O. (2003) Identification of Atlantic bluefin tuna (Thunnus thynnus) stocks from putative nurseries using otolith chemistry. Fisheries Oceanography 12:75–84.[CrossRef][Web of Science]
Silva A. (2003) Morphometric variation among sardine (Sardina pilchardus) populations from the northeastern Atlantic and the western Mediterranean. ICES Journal of Marine Science 60:1352–1360.
Schölkopf B. and Smola A. J. (2002) Learning with Kernels. Support Vector Machines, Regularization, Optimization, and Beyond. (The MIT Press, Cambridge, USA).
Thorrold S. R., Latkoczy C., Swart P. K., Jones C. M. (2001) Natal homing in a marine fish metapopulation. Science 291:297–299.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A. Silva, D. W. Skagen, A. Uriarte, J. Masse, M. B. Santos, V. Marques, P. Carrera, P. Beillois, G. Pestana, C. Porteiro, et al. Geographic variability of sardine dynamics in the Iberian Biscay region ICES J. Mar. Sci., April 1, 2009; 66(3): 495 - 508. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||