Sampling Methods

Herring will be collected by purse seine from the focal spawning areas of seven major spawning aggregations throughout Alaska.  At collection, the lengths and sexes of the sampled fish will be recorded and scale samples will be removed for aging. In contrast to Otis and Heintz (2003), this study will target all age classes, maturity stages and sexes. Herring hearts will be removed and placed immediately in liquid nitrogen for shipment to Juneau (NMFS-Auke Bay Lab), where they will be stored at –80º C until analyzed.

Fatty acid analysis will be performed on the lipids of whole hearts, extracted using the Folch method. Before extraction, 50 μL of extraction surrogate (C23:0) and BHT will be added to the homogenized cardiac tissue. The homogenates will be extracted with a Dionex 200 Accelerated Solvent Extractor (ASE), followed by rotary evaporation and volumetric dilution to a final volume of 1 ml. A 30 mg sample of the purified lipid will be suspended in toluene, mixed with 50 μL of transesterification surrogate (C21:0) and transesterified in methanol and sulfuric acid for two hours at 80°C. The resulting solution of fatty acid methyl esters will be extracted into hexane and dried by passing through a column packed with Na2SO4. The eluant will be spiked with 50 μL of internal standard (C19:0).

Fatty acids will be identified by injecting 1.0 μL of each transesterified sample into a Varian CP3800 gas chromatograph equipped with a Saturn model 2200 mass-detector. Fatty acids will be separated with a 30 m Omegawax 250 fused silica column operating under a ramped temperature program. Mass detection will be in single ion mode and 39 fatty acid peaks will be identified. These will be quantified relative to five- point calibration curves for each of the fatty acids and normalized to the recovery of the internal standard. Calibration curves will be developed for each batch of 15-20 tissue samples. Analytical accuracy for each batch of samples will be determined by examining the fatty acid composition of an in-house standard reference material (SRM). The composition of the SRM was initially calibrated against the National Institute for Standards and Technology SRM-1946, whose fatty acid composition has been certified. Precision of the estimated fatty acid concentrations will be evaluated by examining the variation observed in a duplicated sample. Sample purity will be examined by processing blank samples.

OTOLITH CHEMISTRY

Sagittal otoliths will be extracted from Pacific herring using standard techniques (Campana et al. 1995; Campana 1999; Bickford et al. 2003).  Otoliths will be removed from the fish and placed in centrifuge tubes to dry until processing. We will thin section the otoliths, using a Beuhler isomet low speed saw and the exposed otolith core will be used for aging and chemical analysis.  Thin sectioned otoliths will be polished with fine grit and velvet polishing pads (Beuhler) until the core is distinctly visible.  Otoliths will be randomly analyzed with dry laser ablation (LA; New Wave 213 nm Nd:YAG) - inductively coupled plasma – mass spectrometry (ICP-MS; 7500c Agilent).  ICP-MS instruments can also assay inter-element ratios, such as Sr/Ca, with precision (0.05% relative standard deviation [RSD]) approaching that of thermal ionization mass spectrometry. We envision that Mg/Ca, Mn/Ca, Sr/Ca and Ba/Ca ratios will be assayed in the otoliths along with other trace elements that provide a distinguishable signature.

The center of the otolith, the core, represents larval otolith deposition.  We will use core otolith chemistries to compare fish from various collections sites.  This will identify the number of population sub-units that otolith chemistry can identify.  The chemistry of the core region of the otolith records the chemistry of the natal habitat.  In spawning fish the chemistry of the outer edge can be matched to core chemistries to assess whether the fish returned to their natal grounds for spawning.  Finally, a relational database will be constructed and the data statistically parameterized so that individual fish can be classified to their nursery area based on these geochemical signatures.

Data Analysis and Statistical Methods

Differences in the heart tissue fatty acid compositions and otolith elemental compositions collected from spawning aggregates at a given site will be determined by multivariate analysis of variance (MANOVA). Evaluation of our first two hypotheses will employ a one-way nested MANOVA with stock as the main factor and sampling period nested within stock. Response variables will be the percentages of analytes transformed following the method of Aitchison (1992). Wilk’s lambda with a set to 0.05 will be used to test the hypothesis that sampling period has no effect on fatty acid composition and that stock has no effect on fatty acid composition. If the first hypothesis is accepted and the second rejected, the same approach will be repeated for samples collected in the second year.

Assuming no effect of spawn timing is found, then comparisons across years will be made to test the third hypothesis. Data described by Otis and Heintz (2003) will be used for one of the years. Data from the different sampling times will be pooled into a common stock in a given year, and a two-way MANOVA with years and stocks as the main factors will be evaluated by calculating Wilk’s lambda with a set to 0.05. If significant differences are found between spawning events within regions, sample events will not be pooled and the above analysis will be conducted for each unique spawning stock identified during evaluation of hypothesis 2.

Differences detected among groups under each hypothesis will be further examined by descriptive discriminant analysis (DA) to identify which groups differ. DA resolves differences among groups by identifying a series of canonical functions, each of which is a linear combination of the response variables. These functions progressively reduce the error in the data set. The number of functions that account for the error represents the dimensionality of the data set, which is determined by iteratively fitting a function and testing the hypothesis that the residual error is equal to zero (Huberty 1994). Bi-plots for each function will be constructed to examine how the functions separate the data. We will also examine the pooled within group canonical structure to identify which fatty acids and elements exert the most influence on the separating functions. The results of the MANOVAs will be further examined by predictive discriminant analysis to examine the robustness of the conclusions.  The analysis will employ the leave-one-out method to determine how frequently the discriminant functions accurately identify “unknown” samples. Results of these tests will be expressed as the probability of correctly identifying members of a test group to the appropriate stock or aggregate. The MANOVAs and discriminant analyses will be performed in SAS release number 6.12 using the non-parametric DISCRIM procedure.  Prior to analysis, the homogeneity of the covariance matrices will be examined.  If they are found to be not homogenous then correlation matrices will be used.