104
VARIATION IN THE NUTRITIONAL VALUE OF ATLANTIC HERRING (CLUPEA HARENGUS) FROM THE BAY OF FUNDY, CANADA Hillary Anne Lane A Thesis Submitted to the University of North Carolina Wilmington in Partial Fulfillment of the Requirements for the Degree of Master of Science Department of Biology and Marine Biology University of North Carolina Wilmington 2009 Approved by Advisory Committee _______ Fred Scharf___________ Damon Gannon_________ ______Heather N. Koopman______ Chair Accepted by _____________________________ Dean, Graduate School

HAL Thesis post defensedl.uncw.edu/Etd/2009-1/laneh/hillarylane.pdf · 2009. 12. 15. · Title: Microsoft Word - HAL_Thesis_post_defense.docx Author: westgatea Created Date: 4/22/2009

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

  • VARIATION IN THE NUTRITIONAL VALUE OF ATLANTIC HERRING (CLUPEA HARENGUS) FROM THE BAY OF FUNDY, CANADA

    Hillary Anne Lane

    A Thesis Submitted to the University of North Carolina Wilmington in Partial Fulfillment

    of the Requirements for the Degree of Master of Science

    Department of Biology and Marine Biology

    University of North Carolina Wilmington

    2009

    Approved by

    Advisory Committee

    _______ Fred Scharf___________ Damon Gannon_________

    ______Heather N. Koopman______ Chair

    Accepted by

    _____________________________ Dean, Graduate School

  • ii  

    TABLE OF CONTENTS

    ABSTRACT ....................................................................................................................... iv

    ACKNOWLEDGEMENTS ............................................................................................... vi

    LIST OF TABLES ........................................................................................................... viii

    LIST OF FIGURES ........................................................................................................... ix

    INTRODUCTION ...............................................................................................................1

    Atlantic Herring Life History and Population Structure ..........................................3

    Herring Lipid Structure and Metabolism .................................................................4

    Commercial Importance of Herring .........................................................................6

    Ecological Importance of Herring ...........................................................................7

    MATERIALS AND METHODS .......................................................................................10

    Herring Sample Collection and Pre-processing .....................................................10

    Lipid Extraction and Content and Composition Determination ............................14

    Statistical Analyses ................................................................................................15

    RESULTS ..........................................................................................................................23

    Total Lipid Content Results ...................................................................................23

    Fatty Acid Composition Results ............................................................................38

    Fatty Acid Signature Analysis ...............................................................................45

    DISCUSSION ....................................................................................................................62

    Ontogenetic Variation in the Nutritional Value of Herring ...................................62

    Annual Variation in the Nutritional Value of Herring ...........................................66

    Seasonal Variation in the Nutritional Value of Herring ........................................67

  • iii  

    Spatial Consistency in Lipid Content ....................................................................69

    Variability in the Lipid Content of Mature Fish by Gender ..................................69

    Ecological Consequences of Variation in Prey Quality .........................................70

    Bay of Fundy Herring Compared to other Herring Stocks ....................................77

    Future Directions ...................................................................................................81

    CONCLUSIONS................................................................................................................82

    LITERATURE CITED ......................................................................................................85

  • iv  

    ABSTRACT

    Atlantic herring (Clupea harengus) are abundant in the Bay of Fundy (BoF), representing

    important prey for apex predators, but little is known about their nutritional value. Understanding

    spatial and temporal variation in the nutritional quality of prey will provide insight into the

    distributions and foraging strategies of predators reliant on these species. Using total lipid

    content and fatty acid composition as proxies for nutritional value, we evaluated annual, seasonal

    and ontogenetic variation in the fat content and composition of Atlantic herring from the BoF

    from 2005-2008. Linear regression and analysis of covariance (fish fork length (cm) as the

    covariate) were used to evaluate the relationship between total percent lipid and the variable of

    choice in SPSS 16.0. Within each year ontogeny was the most important factor in determining

    lipid content, with small fish yielding significantly less lipid (6.01±0.206wt%) than both medium

    (10.09±0.206wt%) and large fish (9.76±0.203wt%, P

  • v  

    herring changes as fish grow and underscore the fact that considerable inter- and intra-annual

    variability exist in prey quality. The consequences of this variation in prey quality to upper

    predators are vast, possibly impacting both predator ecology and distribution. These data also

    highlight the need for a comprehensive examination of prey species over time and space in

    determining their nutritional value to predators, rather than the snapshot investigations that have

    been common in the past.

  • vi  

    ACKNOWLEDGEMENTS

    I would like to thank the various people and organizations that have helped me

    along the way, for without them this project would not have been possible. I would first

    like to thank my advisor and mentor, Dr. Heather Koopman. Her guidance and

    leadership for the past four years have been invaluable to me and I am forever in her debt.

    I would also like to thank Dr. Andrew Westgate who has played so many roles in my life

    over the past four years, both educational and personal. I feel privileged to have had the

    opportunity to work with both Dr. Koopman and Dr. Westgate over the past four years.

    Many thanks to Dr. Damon Gannon and Dr. Fred Scharf, who as committee members and

    mentors have helped me to develop my thesis work into a successful project. Lastly, I

    could not have completed my degree without the help of my labmates, past and present:

    Zoey Zahorodny, Zach Swaim, Caitlin McKinstry and Sara McClelland. Special thanks

    to Zach Swaim and Caitlin McKinstry for their friendship through thick and thin both on

    Grand Manan and in Wilmington.

    I am forever in debt to the fishermen of Grand Manan, whose continual

    cooperation has made this project possible. I would like to thank the entire staff of the

    Grand Manan Whale and Seabird Research Station for help and support throughout the

    four years of “The Grind”. The research assistants on Grand Manan provided the

    manpower that allowed for the large sample size obtained in this project. Special thanks

    to Caitlin McKinstry and Jesse Kelly for your many, many hours behind the blender

    during the summer of 2007 and to Rob Ronconi, Sara Wong and Laurie Murison for

    overall project assistance. The staff of the Saint Andrews Biological Station provided me

  • vii  

    with knowledge and support throughout this project, especially Jack Fife, Dr. Gary

    Melvin, Dr. Rob Stephenson and Dr. Mike Powers. I would also like to thank Connors

    Brothers, Inc., especially Tony Hooper, Ken Dougan and Debbie Cooke for their help

    with making sample and data collection smooth during this project. I am most thankful

    to Mike Brant whose love has shown me a new and wonderful way to see the world.

    Finally, I would like to thank Charles, Mom and Dad for their guidance and support in all

    ways possible throughout my entire educational career.

    This project would not have been possible without funding from the UNCW

    Graduate School, GSA, BioGSA, Lindquist Scholarship and the CMS Summer Stipend as

    well as DFO’s Herring Science Council and the Grand Manan Whale and Seabird

    Research Station.

  • viii  

    LIST OF TABLES

    Table Page 1. Distribution of Fish Analyzed for Total Lipid Content .........................................25 2. AIC Analysis Results .............................................................................................27 3. Mean Total Percent Lipid by Year with Yearly Regression Results .....................30 4. Mean Total Percent Lipid by Season with Seasonal Regression Results ..............32 5. Mean Total Percent Lipid by Month for Summer Months ....................................35 6. Mean Total Percent Lipid by Month for Winter Months .......................................36 7. Mean Total Percent Lipid by Location with ANCOVA Results ...........................37 8. Descriptive Statistics and Mean Percent Composition of Fatty Acids that

    Compose at least 5% of each sample .....................................................................41 9. Mean Percent Composition of Omega-3, omega-6 and Essential Fatty Acids ......42 10. Mean Percent Composition of Essential Fatty Acids and other Physiologically

    Important Fatty Acids for all samples and separately for small and large fish ......44 11. Annual and Seasonal Mean Percent Lipid of Size Ranges of Herring that are

    Targeted by Primary Predators in the Bay of Fundy .............................................75 12. Mean Percent Lipid and Percent Composition of 22:1n-11 and 22:6n-3 of Size

    Ranges of Herring that are Targeted by Primary Predators in the Bay of Fundy ..76

  • ix  

    LIST OF FIGURES

    Figure Page 1. Typical Weir off Grand Manan ..............................................................................12 2. Grand Manan and Surrounding Waters .................................................................13 3. Histogram of Transformed Percent Lipid Data .....................................................26 4. Bivariate Scatterplot of Total Percent Lipid by Fork Length for all Samples .......28 5. Bivariate Scatterplots of Total Percent Lipid by Fork Length separately for each

    Year ........................................................................................................................31 6. Bivariate Scatterplot of Percent Composition by Fork Length of Fatty Acids that

    Exhibited a Pattern with Fork Length Upon Visual Inspection .............................43 7. Discriminant Plot of Functions 1 and 2 created from the Fatty Acid Signature

    Analysis of all Fish by Year...................................................................................49 8. Discriminant Plot of Function 1 vs. Fork Length created from the Fatty Acid

    Signature Analysis of 2006-2007 Fish by Season .................................................50 9. Discriminant Plot of Functions 1 and 2 created from the Fatty Acid Signature

    Analysis of Large (≥15.5cm) Fish by Year ...........................................................51 10. Discriminant Plot of Functions 1 and 2 created from the Fatty Acid Signature

    Analysis of Small (

  • INTRODUCTION

    Atlantic herring (Clupea harengus) comprise a major component of the diet of

    many upper trophic level predators in the western North Atlantic (wNA) including seals,

    porpoises, dolphins, whales, predatory fish, sharks, seabirds and humans (Katona et al.

    1993). Herring are also a main consumer of zooplankton in the wNA (De Silva 1973).

    Because of the intimate interactions that herring have with both the upper and lower

    levels of the food chain, the nutrition of these fish reflects the health of the ecosystem

    they inhabit and directly impacts the predators that consume them. The Bay of Fundy

    (BoF) is one of the most ecologically diverse bodies of water in the world (Shacknell and

    Frank 2003) and is known to be a “mixing pot” of different stocks of many fish species,

    including Atlantic herring (McPherson et al. 2001). The heavy reliance on herring by

    upper level predators in the BoF illustrates the important role that this predictable,

    nutritionally valuable species plays in this ecosystem (e.g. Haug et al. 1995, Gannon et al.

    1998, Chase 2002, and Eastman 2003). Prey is chosen by predators based on its quality,

    availability and the physical constraints of the predator (Emlen 1966, Heithaus et al.

    2007). Variability in the quality of forage can result in dramatic effects up the food chain

    (e.g. Flinkman et al. 1998) including reductions in survival (Moore et al. 2003), fecundity

    (Rodenhouse and Holmes 1992) and other negative physiological responses (Thompson

    et al. 1997). Understanding the nutritional dynamics of herring in the BoF is important to

    our overall understanding of their impact on the ecosystem but also is useful for modeling

    the effects of variability in prey quality on predators in general.

  • 2

    In addition to their ecological significance, herring represent an important

    commercial fishery with landings of 166, 965 tons in Canadian waters in 2007

    [Department of Fisheries Canada (DFO) 2008] and 73,781 tons in US waters in 2007

    [National Marine Fisheries Service (NMFS) 2008]. This catch is mainly processed for

    direct human consumption in the form of canned fish, but a portion is also used as feed

    for the salmon aquaculture industry and as lobster bait. Apex predators eat fatty,

    schooling fish such as herring because of their abundance and high fat content (8-20%

    wet weight; data from this study). The abundance of herring, the fact they feed on a

    lower trophic level (zooplankton), and that they comprise a significant portion of the diet

    of middle- and upper-level trophic predators, all make them ideal for examining health of

    ecosystem resources.

    Considering the ecological importance of herring, it is surprising that so little is

    currently known about the nutritional value of this species. The goal of this study was to

    determine the nutritional value of the herring population in the BoF and evaluate the

    differential influence that herring of varying sizes and from different seasons and years

    might have in the diets of upper-level predators. Nutritional value can be comprised of

    many things, including amounts of fat, protein, carbohydrates and vitamins. This study

    used the total lipid content and fatty acid composition of individual fish as proxies for

    nutritional value because lipid transfer is important in marine ecosystems and likely the

    most variable compared to the other proxies listed above. Quantifying ontogenetic

    variation in lipid content and composition will allow for determination of the nutritional

    quality of herring in different seasons and years. This study will also help to understand

  • 3

    how changes in lipid (and therefore energy) content and composition affect the quality of

    food predators are exploiting in the BoF in different seasons and years. These results will

    provide a better understanding of the role that variability in forage quality plays in the

    potential diets of predators in the BoF and other herring-dominated ecosystems.

    Atlantic Herring Life History and Population Structure

    Atlantic herring spawn demersal eggs during the late summer through early fall

    on the shoals of Georges Bank and other areas around the perimeter of the Gulf of Maine

    (Bigelow and Schroeder 1953, Boyer et al. 1973, Lough and Bolz 1979). Eggs lay on

    bottom for 8-9 days at about 10°C (Cooper et al. 1975) and the resulting larvae are

    dispersed throughout the water column by the large tides that occur in the region

    (Bumpus 1976). Larval herring grow quite quickly until metamorphosis into adults at

    about 5-5.5cm in the spring (Blaxter and Staines 1971, Ehrlich et al. 1976). Spawning

    occurs when adults are 1-2 years old (or about 20cm in length, Blaxter and Hunter 1982).

    Atlantic herring travel in schools during both the day and night that are composed

    of fish of similar sizes (Breder 1976). Tightly knit, large schools may be better for

    predator avoidance while small, less cohesive schools may be better for optimum

    foraging due to less competition for food items and less overlap of individuals (Blaxter

    and Hunter 1982). Herring travel large distances in schools to spawning and feeding

    grounds around the Gulf of Maine. Schools tend to stay near the bottom during the day

    and migrate to the surface to feed at night (Blaxter 1985).

  • 4

    Like many fish species, the population structure of wNA herring has been

    variable over time. Due to heaving fishing pressure, the Georges Bank herring fishery

    declined drastically in the 1970s and the fishery completely failed in 1977 (DFO 2008).

    However, dramatic drops in quotas have allowed for a substantial recovery so that the

    current population is at 1965 levels (TRAC 2006). A drop in weight-at-age was detected

    in the early 1980s, but current weight-at-age for herring in the wNA remains constant

    (TRAC 2006). The dramatic drop and subsequent increase in the population size of

    herring may have caused a decline in genetic variability, but this does not seem to have

    affected the ability of wNA herring to recover (McPherson et al. 2001).

    Herring Lipid Structure and Metabolism

    The lipid content of a fish can be used as an indicator of its nutritional value and

    overall body condition (Sargent et al. 1988). The main lipid class found in herring is

    triacylglycerol which is a glycerol backbone (HOCH2CH(OH)CH2OH) with three fatty

    acids connected to it by ester linkages. The individual fatty acid components of herring

    lipids also have the potential to provide information about the nutritional value of a given

    fish. The fatty acid tails are usually straight carbon chains 14 to 24 carbons in length

    with varying degrees of saturation. Because of biochemical limitations, many fatty acids

    are incorporated into the tissues of predators with little or no modification. This

    characteristic makes them ideal for use as indicators or markers of dietary source (Budge

    et al. 2006).

  • 5

    The relative proportions of the ~80 different fatty acids that are present in a

    marine organism determine the fatty acid “signature” of that organism (Iverson 1993).

    Fatty acid signatures have already been used to distinguish species of fish (based on

    differences in fatty acid signatures) in several regions, including the Scotian Shelf,

    Canada and Prince William Sound, Alaska (Budge et al. 2002, Iverson et al. 1997) and to

    distinguish the diet of Hawaiian monk seals (Monachus schauinslandi, Goodman-Lowe

    et al. 1999). Fatty acid signature analysis is also being used extensively in dietary studies

    of upper trophic level predators through models such as QFASA (quantitative fatty acid

    signature analysis, see Iverson et al. 2004). QFASA provides quantitative estimates of

    the proportion of various prey species in a predator’s diet using fatty acid signatures.

    This model requires the creation of a library of fatty acid profiles of all potential prey

    species. Often, the information available for individual species is limited to a small

    number of samples collected in a narrow range of space and time. The significant spatial

    and temporal variability present in the marine environment can limit the broad

    application of these snapshot characterizations of the fatty acid signatures of prey species

    in models such as QFASA, which attempt to obtain generalized results for prey. Data

    from this study will provide detailed information on the fatty acid signature of BoF

    herring over broad spatial and temporal scales which can characterize the species more

    accurately than studies that are more limited in space and time. In addition to

    determining the types of fatty acids present in the lipids of Atlantic herring, this study

    will provide an important foundation for future studies aimed at estimating dietary

    composition of predators.

  • 6

    In this study, the nutritional value of individual fish was determined using two

    complementary methods, total lipid content and fatty acid composition. It is not only the

    lipid content, but also the fatty acid composition of each fish that determines its worth to

    its predators. For example, if fish from summer and winter have similar caloric values

    but different fatty acid compositions they may have very different nutritional values to

    their predators. There is now mounting evidence supporting the positive health effects of

    consuming fatty acids of the omega-3 and monounsaturated families (e.g. Schacky and

    Harris 2006, Willett 2006), which are major constituents of herring. Such fatty acids

    include alpha-linolenic acid (18:3n-3), eicosapentaeic acid (20:5n-3), and

    docosahexaenoic acid (22:6n-3). Fish are known to have high levels of these omega-3

    and omega-6 fatty acids (Sargent 1988, Iverson et al. 1997, Budge et al. 2002, Iverson et

    al. 2002) and data from this study will provide information on the percent composition of

    these important fatty acids in this population.

    Commercial Importance of Herring

    Atlantic Canada and the Northeast coast of the United States are large producers

    of canned herring, with $US 19.4 million sold in the US in 2007 (NMFS 2008) and

    $CAN 36.5 million sold in Canada in 2007 (Statistics Canada 2008a). Ground herring are

    a major component of the feed that is used at hundreds of salmon aquaculture sites each

    day in the BoF and Gulf of Maine (GoM). Farm-raised salmon represents a huge

    industry, with 2007 annual sales reaching $37.7 million in the US and $32.7 million in

    Canada. (NMFS 2008, Statistics Canada 2008b). Herring is also the main bait used in

    both the U.S. and Canadian lobster fisheries. The lobster fishery is one of the most

  • 7

    lucrative fisheries in the wNA, representing nearly 1 billion dollars in 2007 (NMFS 2008,

    DFO 2008). Recent evidence suggests that consumption of bait herring by lobsters

    represents 35-55% of the diet of these crustaceans (Eastman 2003). Further, the addition

    of herring biomass onto the sea floor as bait serves to provision lobsters and this bait can

    potentially support 25-33% of the current lobster population that is landed annually (Salia

    et al. 2002).

    Ecological Importance of Herring

    Herring are an ecologically important species throughout the Northern

    hemisphere. They are considered a critical prey item to upper trophic level predators in

    many ecosystems, including the wNA, the Pacific, and the Baltic Sea (e.g.: Wiawood and

    Majkowski 1984, Linko et al. 1985, Recchia and Read 1989, Haug et al. 1995, Gannon et

    al. 1998, Budge et al. 2002, Iverson et al. 2002,). Herring, in addition to other forage fish

    species such as capelin (Mallotus villosus) and Atlantic mackerel (Scomber scombrus),

    are an important dietary item for many predators in the wNA. For example, the stomach

    contents of harbor porpoises (Phocoena phocoena) from the wNA have been extensively

    examined (e.g.: Recchia and Read 1989, Fontaine et al. 1994, Gannon et al. 1998) and

    revealed that herring comprise at least 80% of the diet. Herring is also a major

    component of the diets of other apex predators in the North Atlantic, such as minke

    whales (Balenoptera acutorostrata), bluefin tuna (Thunnus thynnus) (Haug et al. 1995,

    Chase 2002), Atlantic cod (Gadus morhua) greater than 41cm in fork length (Powels

    1958, Waiwood and Majkowski 1984, Schwalme and Chouinard 1999), greater and sooty

    shearwaters (Puffinus gravis, P. griseus, Brown et al. 1981), Arctic and common terns

  • 8

    (Sterna paradisaea, S. hirundo), razorbills (Alca torda) and Atlantic puffins (Fratercula

    arctica, Diamond and Devlin 2003). Herring are clearly an important prey item in the

    wNA, however recent work on the nutritional value of herring in this ecosystem is

    limited to fatty acid signature analysis of herring from the Scotian Shelf, as part of a

    larger study looking at variation in lipids across and within species during a single year

    (Budge et al. 2002). In the Budge et al. (2002) study, herring (n=74) were one of 28

    species analyzed to determine inter- and intra-specific variability in fatty acid signatures.

    The study found that fish size (measured by length) had the strongest effect on fatty acid

    signatures and that herring had seasonally variable lipid content as they built up fat

    reserves in their tissues in the fall.

    Pacific herring (Clupea pallasi) play a similar role in the North Pacific marine

    ecosystem to Atlantic herring in the wNA. They are one of a suite of forage fishes, such

    as capelin (Mallotus villosus), sandlance (Ammodytes hexapterus), and juvenile pollock

    (Theragra chalcogramma) that support the apex predators in that system (Springer and

    Speckman 1997, Bowen and Siniff 1999). In contrast to the limited data available from

    the North Atlantic herring population, the lipid content and composition of North Pacific

    herring have been examined in detail (e.g. Iverson et al. 1997, 2002). Iverson et al.

    (2002) found that age class had the strongest effect on fatty acid composition (n=360),

    and that fat content was over three times higher in the fall (11.4% ± 0.80) than in the

    spring (3.5% ± 0.48) for fish between 9 and 25cm. This variation shows that the

    nutritional quality of Pacific herring changes depending on the size of the fish or the

  • 9

    season in which it was caught. This study aims to examine variation in the nutritional

    quality of wNA herring, using data from the Pacific as background information.

    Baltic herring (Clupea harengus membras) play a different role in the marine food

    chain than do Atlantic or Pacific herring. Due to low salinity, reduced tides, human

    predation and habitat degradation, the Baltic Sea does not support many upper trophic

    level predators (Elmgren 1989). Baltic herring are a predominant predatory species in

    this environment (Cardinale and Arrhenius 2000) due to the important zooplanktivorous

    role these fish play. The nutritional value of Baltic herring has been examined, but not in

    as much detail as the data available for the North Pacific. The Baltic studies that exist

    have either pooled numbers of fish together (Linko et al. 1985) or only looked at fish that

    were already cooked (Aro et al. 2000). Thus the literature shows the widespread

    importance of herring in many marine ecosystems but also underscores large gaps in our

    current knowledge.

    Although Atlantic herring (Clupea harengus) are an important and abundant prey

    species in the wNA, surprisingly little is known about their nutritional condition and how

    it varies ontogenetically, seasonally, or annually. The purpose of this study was to

    determine the lipid content and composition of herring from the BoF over a longer time

    period than previously investigated, from 2005-2008. Specifically, this study examined

    the total lipid and fatty acid composition of herring for significant differences on

    ontogenetic, annual, seasonal and spatial scales. Finally, the data collected in this study

  • 10

    were used to consider the implications of these differences on the diets and energetics of

    upper level predators in the BoF.

    MATERIALS AND METHODS

    Herring Sample Collection and Pre-processing

    The herring samples for this project were collected from around Grand Manan,

    New Brunswick, Canada between 2005-2008. Field work was based out of the Grand

    Manan Whale & Seabird Research Station, located on the island. During each summer

    (June-September), herring sampling took place at least three times a week from at least

    five different locations around Grand Manan with the help of local fishermen. During the

    winter (October-May), samples were provided at least once monthly by Connors Bros,

    Inc. from their Blacks Harbour, N.B. processing plant. In all cases, herring were obtained

    from commercial fishing operations, either purse seines or herring weirs.

    During the summer months herring are caught in nets called weirs around Grand

    Manan. A weir is a stationary, kidney-shaped net that is located close to shore (see

    Figure 1). The opening (mouth) of the weir is positioned to face the shore. Coming off

    of the mouth is another net (fence) that runs perpendicular to the shore. Schools of fish

    are directed into the weir during the night when they encounter the fence as they move

    along the shore to feed. Once the fish are captured in the weir, the fisherman will close

    off the mouth so that the fish do not escape. Herring are harvested by deploying a large

    seine net inside the weir.

  • 11

    During summer, local weir fishermen collected a representative sample (20 fish)

    of their catch and placed the fresh fish in a labeled Ziploc bag in a cooler that was

    provided. These fish were collected soon after they were removed from the water and

    processed (see below) within 24 hours. During winter, fish caught around Grand Manan

    (usually by a purse seine) were collected by a plant manager at Connors Bros, Inc. and

    frozen directly after size sorting. These samples were then collected from the processing

    plant by the Canadian Department of Fisheries and Oceans (DFO) as a component of a

    larger DFO herring monitoring study. Fish were characterized as “inshore” if they were

    retrieved from stationary weirs located near the shore of Grand Manan and as “offshore”

    if they were caught by purse-seine boats at locations throughout the Bay of Fundy (See

    Figure 2 for a map of catch locations).

    During pre-processing, fish were weighed (to the nearest milligram), fork length

    was recorded (to the nearest millimeter) and the otoliths were collected for future age

    studies. When possible, sex was determined by visual examination of the gonads. If

    gonads were not mature, they were still removed and weighed, but sex was not

    determined. After the initial measurements, 10 fish per collection were homogenized

    whole in a food processor (Kitchen Aid). A sub-sample of this homogenate was frozen at

    -20ºC in a 20 mL vial sealed with Parafilm. Samples were transported back to UNCW in

    a sealed cooler with dry ice to ensure that they remained frozen.

  • 12

    Figure 1. A typical herring weir off Grand Manan. Fish swim along the shore at night and are directed into the mouth by the fence. Once fish are captured, they are held in the main weir until removal by the fisherman.

  • 13

    Figure 2. Grand Manan and surrounding waters. Red circles represent “offshore” catch locations while green stars represent “inshore” weirs. Connors Bros., Inc. herring plant is also included on the map. Offshore fish from all years were caught in locations similar to those indicated on the map, red circles indicate the only locations were exact coordinates were reported.

  • 14

    Lipid Extraction and Content and Composition Determination

    Total lipid was extracted from herring samples using a modified Folch et al.

    (1957) chloroform: methanol extraction method as described in Budge et al. (2002).

    Each weighed sample (approximately one gram) was placed in 20 ml of 2:1 chloroform

    methanol 0.01% BHT (butylated hydroxytoluene) and left to soak overnight to ensure

    complete extraction of lipids. The sample was filtered to remove solids and anhydrous

    Na2SO4 was added to ensure complete removal of water from the solution. The solvent

    was evaporated under N2 gas to yield percent lipid content of the tissue sampled. Lipid

    was re-suspended at 50 mg of lipid/mL hexane and stored under nitrogen gas at -20 ˚C.

    For gas chromatography (GC) analysis, fatty acid butyl esters (FABE) were

    prepared from total lipid extracts. Fatty acids were separated and analyzed by GC using a

    Varian capillary GC (3800) with a flame ionization detector (ID) in a fused silica column

    (30 x 0.25mm ID) (Zebron FFAP; Phenomenex). Helium was used as the carrier gas and

    the gas line was equipped with an oxygen/water scrubber. The following temperature

    program was used: 65°C for 2 min, hold at 165°C for 0.40 min after ramping at

    20°C/min, hold at 215°C for 6.6 min after ramping at 2°C/min, and hold at 250°C for 5

    min after ramping at 5°C/min. Up to 50 different fatty acids were identified following

    Iverson et al. (1997, 2002). Each fatty acid was described using the nomenclature of

    A:Bn-X, where A is the number of carbon atoms, B is the number of double bonds, and X

    is the position of the double bond closest to the terminal methyl group. Identification was

    confirmed and results were integrated using Galaxie GC software (version 1.8.501.1,

    Varian, Inc., Palo Atlo, CA).

  • 15

    Statistical Analyses

    Statistical analysis was conducted using SPSS 16.0 (SPSS, Inc., Chicago, IL) and

    Plymouth Routines In Multivariate Ecological Research (PRIMER 6, Primer-E, Ltd.,

    Ivybridge, UK) statistical software with 0.05 as the significance level for each program.

    Analysis of covariance (ANCOVA) and analysis of variance (ANOVA) tests were

    evaluated for equality of variance using Levene’s test at the 0.05 significance level.

    Total Percent Lipid Analyses

    Kolmogorov-Smirnov and Shapiro-Wilk normality tests were conducted on raw

    total percent lipid data. Square root, arcsin and reflect transformations were explored

    when necessary. Corrected Akaike information criteria (AICc) scores were calculated for

    linear, quadratic, and cubic distributions to determine the best curve fit for the

    relationship between lipid content and fish size (Burnham and Anderson 2002). Below is

    the equation that was used for calculating AICc scores:

    AICc n logSSE

    n 2K

    2K K 1n K 1

    The change in AICc (∆) score represents the difference between the calculated score and

    the smallest score. The likelihood of any model fitting the data was calculated by the

    following equation:

    likelihood

    The probability of any curve fitting the data compared to any other curve is represented

    by the statistic (wi). This statistic was calculated for each possible curve from the

    following equation:

  • 16

    wi 1

    The wi statistic was used to evaluate the likelihood of each curve fitting the data.

    The relationship between total wet-weight percent lipid and fish fork length over

    the entire size range of fish collected was analysed using linear regression. Since size

    had a significant effect on total percent lipid (see results), the remaining analyses were

    conducted using ANCOVA to account for this co-variation. The annual variation in total

    percent lipid of herring was examined using ANCOVA (year = fixed independent

    variable, square root transformed total percent lipid = dependent variable, fork length =

    covariate). The seasonal variation in total percent lipid of herring was also examined

    using ANCOVA (season = fixed independent variable, square root transformed total

    percent lipid = dependent variable, fork length = covariate). Only samples from 2006-

    2007 were included in seasonal analyses because the winter of 2006-2007 was the only

    winter available for analysis. “Summer fish” were categorized as fish caught between

    June and September and “winter fish” were considered fish caught between October and

    May. The regression of square root transformed total percent lipid over the entire size

    range of fish collected was conducted separately for each year and season. The

    ANCOVA of transformed total percent lipid by year was conducted separately for each

    season.

    The variation in total percent lipid of herring by month was examined using

    ANCOVA (month = fixed independent variable, square root transformed total percent

    lipid = dependent variable, fork length = covariate). This test was repeated separately for

  • 17

    each year and season with the same variables listed above. Since annual and seasonal

    variation was detected in the analysis of all samples (see results), this test was repeated

    for each year and season separately with the same variables listed above. The variation in

    total percent lipid of herring was examined spatially using ANCOVA (inshore/offshore =

    fixed independent variable, square root transformed total percent lipid = dependent

    variable, fork length = covariate). This test was repeated for each year and season

    separately with the same variables listed above. Variation in total percent lipid of herring

    by sex was examined using ANCOVA (sex = fixed independent variable, square root

    transformed total percent lipid = dependent variable, fork length = covariate). Due to the

    methods by which fish were sexed (visual examination of gonads); only mature fish were

    included in this analysis (19.9 - 28.2cm in fork length). Also, only summer fish were

    included in this analysis due to the small number of winter fish for which sex could be

    determined.

    To examine the variation in total percent lipid in a way that was comparable to

    other studies, all fish analyzed were divided into three equal groups (n =~ 297),

    representing small (9.0 - 16.5cm), medium (16.5 – 20.4cm) and large (20.4 – 28.2cm)

    size classes of fish, according to Iverson et al. (2002). An ANOVA was conducted with

    size class as the random independent variable and square root transformed total percent

    lipid as the dependent variable. When significant differences between size classes were

    found, the Tukey multiple comparison test was conducted to determine which size classes

    differed from each other. The ANOVA of transformed percent lipid by size class was

    conducted for each year and season separately with the same variables listed above.

  • 18

    Fatty Acid Composition Analyses

    Individual fatty acids were analyzed for relationships with fork length, year,

    season, and catch location (inshore/offshore). Fatty acids that were present in

    concentrations of at least 5% of the total sample plus omega-3 and -6 fatty acids and

    essential fatty acids (those not modified from prey to predator; 18:3n-3, 18:2n-6, 20:4n-6

    and 20:5n-3), were examined in more detail. The relationships of the fatty acids listed

    above with fork length were examined using linear regression for all samples combined

    as well as separately for each year. Fish were divided into two size categories based on

    the patterns of relative abundance of the fatty acids that exhibited strong patterns with

    fork length (15.5cm was used as the cut off between small and large fish; see results).

    These two size categories were used in later analyses to determine the relationship

    between fatty acid signature and fish size.

    Seasonal and spatial effects on percent composition of individual fatty acids were

    examined using ANCOVA with percent composition of an individual fatty acid as the

    dependent variable, season or catch location as the fixed independent variable and fork

    length as the covariate. Only samples from 2006-2008 were included in yearly and

    spatial analyses due to the unbalanced sample distribution by fork length in 2005 (only

    samples >15.5cm were available; see Table 8). Seasonal effects were only examined in

    samples from 2006-2007 because the winter between these two years was the only time

    in which winter fish were collected.

  • 19

    Fatty acid signature analysis was first conducted using discriminant function

    analysis in SPSS 16.0. Fatty acids that were analyzed using discriminant models to

    determine if significant differences in the fatty acid signatures of fish exist on

    ontogenetic, annual, seasonal and spatial scales. Since the purpose was to characterize

    fish based on their entire fatty acid signature, all of the fatty acids that were present in at

    least 95% of all individuals were included in this analysis. If one of the above mentioned

    fatty acids was not detected in an individual fish, the concentration of that fatty acid was

    changed to 0.005% (Iverson et al. 2002). This value was chosen because it was below

    the minimum detectable level (0.01%), but was not so small that it would result in

    extreme outliers (Iverson et al. 2002).

    Although the fatty acid data violate the assumptions of discriminant analysis

    (equality of covariance matrices, normality, independence), discriminant analysis was

    conducted for ease of comparison to the general body of fatty acid literature (e.g.: Budge

    et al. 2002, Iverson et al. 2002). Box’s M was used to evaluate equality of covariance

    matrices. Only functions with Eigenvalues > 1 were accepted as accurate determinants of

    group differences. Wilk’s λ was used to indicate the ability of the analysis to

    discriminate between groups (low values indicated high success in discrimination). The

    percent of cases correctly classified were used to determine the power of the

    classification functions and these values were cross validated using a jackknife procedure

    (leave-one-out cross-validation). Discriminant functions were analyzed using ANOVA

    and post-hoc Tukey tests to evaluate ability of the functions to classify samples.

  • 20

    Since the number of individual samples in any group must be greater than the

    number of variables used in a discriminant analysis, principal components analysis (PCA)

    was conducted prior to discriminant analyses when groups contained less than 23

    individuals. The 23 fatty acids used in the discriminant analysis were analyzed with PCA

    to create components that adequately represented the data. These components were then

    analyzed with discriminant analysis as described above.

    Fatty acid signature analysis was also conducted using Primer 6 software (Clarke

    1993, Clarke and Warwick 2001, Clarke and Gorley 2006) by examining the same fatty

    acids that were used in the discriminant analysis. As with the parametric analysis, the

    purpose was to characterize fish based on their entire fatty acid signatures. Therefore, all

    of the fatty acids that were present in at least 95% of all individuals were included in the

    model. If a certain fatty acid was not detected in a single individual, the concentration of

    that fatty acid was changed from zero to 0.005% (Iverson et al. 2002) as discussed above.

    Primer operates under few, if any, assumptions about the form of the data because

    non-parametric ordination and permutation tests are fundamental to the analyses in the

    package (some analyses require random, independent samples). In contrast to the

    parametric models formerly used to examine the fatty acid data (e.g.: discriminant

    function analysis), the number of variables entered into the Primer models is not

    restricted by the number of samples in the dataset. The philosophy of Primer is based on

    the fact that the relationship between observations is based on the ranks they have in a

    resemblance matrix, created on the basis of similarity between sample points. The above

  • 21

    mentioned characteristics make Primer ideal for examining patterns in fatty acid

    signatures due to the non-normal and non-independent nature of these data, and the use of

    Primer in the fatty acid literature is becoming more common (e.g.: Stubing et al. 2003,

    Kelly et al. 2008, Cooper et al. 2009). Individual fatty acids were standardized prior to

    analyses by dividing the value of each fatty acid in each sample by the standard deviation

    of that fatty acid in all samples, making the standard deviation of each fatty acid equal to

    one. Resemblance matrices were created based on Bray-Curtis similarity. A Bray-

    Curtis similarity resemblance matrix represents the similarity between samples as

    determined by the following equation:

    Similarity 100 1∑ |y y |

    ∑ y ∑ y

    Where yi1 and yi2 are the values of individual observations (after transformation and/or

    standardization).

    Non-metric multi-dimensional scaling (MDS) analyses were conducted on the

    fatty acid signatures of all samples and separately for large and small fish. MDS

    constructs a map or configuration of the samples in two or three dimensions, and attempts

    to satisfy all of the conditions imposed by the dis/similarity matrix. The result of this

    analysis is a plot that places samples with high similarity close to one another and

    samples with high dissimilarity far from each other on the plot. The MDS algorithm is

    iterative and therefore the “best” solution may not be obtained each time; this property

    requires a large number of restarts of the model to obtain the best representation of the

    matrix. A stress value is produced each time the model is run, which indicates how

    faithfully the relationships from the dis/similarity matrix are represented in the plot.

  • 22

    Stress values range from 0-1, and low stress values indicate that the model placed the

    samples in the same place relative to one another for many of the restarts and is confident

    in its output. The MDS ordination is assumed to adequately represent the relationships

    between samples if the stress value is less than 0.2 (Clark and Warwick 2001).

    Analyses of similarities (ANOSIM) were conducted on all samples and separately

    for large and small fish. ANOSIM is the approximate analogue of the analysis of

    variance (ANOVA) test. No covariate can be applied to the model, requiring that fish of

    different sizes be tested separately. The main purpose of ANOSIM is to determine if

    assemblage differences exist between groups of samples specified by a specific factor

    (e.g.: size, year, season or catch location). Just as with MDS, the model for ANOSIM is

    based on a specified number of permutations/randomisations that are based on the ranks

    in the resemblance matrix. This is consistent with the philosophy of Primer, in which the

    relationship between observations is based on the ranks they have in the resemblance

    matrix. The test statistic, global R, becomes more significant the farther from zero the

    value is (range = 0 – 1). The significance level of the test statistic is determined by the

    number of permutations the model conducts and still achieves a test statistic greater than

    predicted if no differences exist between groups (the null hypothesis). For example, 999

    permutations tests the null hypothesis at the 0.1 significance level, while 9999

    permutations tests the null hypothesis at the 0.01 significance level.

    One-way similarity percentages analysis (SIMPER) was conducted on all samples

    and separately for large and small fish. SIMPER analyses are conducted after ANOSIM

  • 23

    determines that significant differences between groups exist. The purpose of SIMPER is

    to determine the role of individual “species” (fatty acids) in contributing to the separation

    between two groups of samples (size classes, years, seasons, locations). The model

    operates by decomposing the average dissimilarity between all pairs of samples, one from

    each group, into percentage contributions from each “species” (fatty acid). The average

    dissimilarity of samples between pairs of groups was collected and the percent

    contribution of each fatty acid in the model to the dissimilarity of samples between

    groups was also collected. Although SIMPER is a useful tool, some multivariate patterns

    cannot be explained by a few main “species” (fatty acids) and therefore results should be

    interpreted with care.

    RESULTS

    Total Lipid Content Results

    Descriptive Statistics

    889 individual fish collected between 2005-2008 were analyzed for total lipid

    content. Table 1 shows the distribution of samples over the four year period. Raw data

    failed (P

  • 24

    Curve Fit

    Corrected Akaike information criteria (AICc) scores were calculated for linear,

    quadratic, and cubic distributions to determine the best curve fit for the relationship

    between percent lipid and fish length (Bernum and Anderson 2002, see Table 2). Since

    the wi statistics for each model were not significantly different, it was determined that no

    curve fit the data better than the linear curve, and for ease of analysis, a linear model was

    used to analyze the data for regression purposes.

    Total Percent Lipid by Fork Length

    The linear regression of square root transformed total percent lipid versus fork

    length was significant (P

  • 25

    Table 1. Distribution of fish analyzed for total lipid content in this study. Size represents the fork length of individual fish. Year Total Fish Size (cm) Summer Winter Inshore Offshore 2005 113 9.0-25.0 113 0 113 0 2006 330 8.8-27.3 292 38 239 91 2007 334 10.7-28.2 214 120 178 156 2008 112 10.8-28.0 112 0 69 43

  • 26

    Figure 3. Histogram of square root transformed total percent lipid data for all samples. Although normality tests failed with square root transformed data, the square root transformed data appeared to be the most normal. Therefore, square root transformed data were used in the remaining analyses.

  • 27

    Table 2. Results of AICc calculations to determine the model that best fit the data. K = number of parameters in the model, SSE = error sums of squares for each model, AICc = corrected AIC score, wi = 1 minus the proportion of likelihood of the model compared to all other models (%).

    Model K R2 SSE AICc wiLinear 2 0.130 1.222 2540.15 0.671

    Quadratic 3 0.223 1.091 2581.92 0.665Cubic 4 0.242 1.065 2589.22 0.664

  • 28

    Figure 4. Bivariate scatterplot of total percent lipid by fork length for all samples. Although the linear R2 value is only 0.147 is lower than the R2 values for both cubic (R2 = 0.263) and quadratic (R2 = 0.248) curves, the AIC analysis determined that no curve was any better fit to the data than the linear curve. The significance of this regression is very high (P

  • 29

    Total Percent Lipid by Year

    The ANCOVA of transformed total percent lipid between years revealed

    significant differences (P

  • 30

    Table 3. Size-corrected mean total percent lipid and standard error (in raw data format for ease of interpretation) as well as results of individual regressions of square root transformed total percent lipid by fork length for 2005-2008. All samples from all years were included.

    Year Mean % Lipid (wet wt.) ± SE R2 P-value Slope b n 2005-2008 8.63±0.13 0.147

  • 31

    5a. 5b.

    5c. 5d. Figure 5a-d. Bivariate scatterplots of fork length (cm) vs. total lipid content for each year separately. 5a highlights 2005 data, 5b highlights 2006 data, 5c highlights 2007 data and 5d highlights 2008 data.

    n = 112

    n = 113 n = 330

    n = 334

  • 32

    Table 4. Size-corrected mean percent lipid and standard error for each season and year from 2006-2007. Results of individual regressions of transformed total percent lipid by fork length for season and for each year combined for samples from 2006-2007.

    Season Mean% Lipid (wet wt.) ± SE

    R Square P-value Slope b n

    Summer 2006 8.66 ± 0.19 0.316

  • 33

    Total Percent Lipid by Month

    As season had a significant effect, the data were further examined by considering

    month. The ANCOVA of percent lipid by month for all samples returned significant

    differences (P

  • 34

    able to be sexed in this study (19.9 – 28.2cm in fork length), so these results can only be

    applied to mature fish.

    Total Percent Lipid by Size Class

    The ANOVA analysis of the data by size class (small: 9.0 – 16.5cm (n=297), medium:

    16.5 – 20.4cm (n=296), large: 20.4 – 28.2cm (n=296), Iverson et al. (2002)) showed that

    the lipid content of the fish the size classes were significant different (P

  • 35

    Table 5. Size-corrected mean percent lipid and standard error by month for each year for summer months. The ANCOVA of transformed percent lipid by month for each year was significant (P < 0.01 for each year).

    Mean % Lipid ± SE (n)

    Year June July August September 2005 n/a 6.53± 0.31(65) 6.92±0.62(16) 5.00±0.45(32) 2006 n/a 9.53±0 .46(50) 8.48±0.29(125) 8.49±0.30(117) 2007 10.34±0.88(23) 11.84±0.58(38) 10.89±0.38(87) 9.18±0.45(66) 2008 n/a 10.60±0.42(64) 7.67 ±0.56 (38) 12.85±1.05(10)

  • 36

    Table 6. Size-corrected mean percent lipid and standard error by month for each year for winter months. Winter fish were only collected during the winter of 2006-2007. The ANCOVA of transformed percent lipid by month for each year was significant (P < 0.01 for each year).

    Mean % Lipid ± SE (n) Year January March April November December October 2006 n/a n/a n/a n/a 9.34 ± 0.77

    (18) 7.11 ± 0.72

    (20) 2007 8.06 ± .82

    (20) 3.56 ± 1.13

    (10) 3.05 ± 1.14

    (10) 7.22 ± 1.11

    (10) n/a 7.13 ± 0.42

    (70)

  • 37

    Table 7. Size-corrected mean percent lipid and standard error by location for each year in which both inshore and offshore fish were collected (2006-2008). Results of the ANCOVA of transformed total percent lipid by catch location (inshore/offshore). All groups had equal variance except for winter fish.

    Mean % Lipid ±SE Year Inshore Offshore P-value # Inshore # Offshore

    2006-2008 8.97 ± 0.18 9.03 ± 0.24 0.839 486 290 Summer 2006 8.70 ± 0.23 8.88 ± 0.38 0.571 203 80

    All 2006 8.68 ± 0.22 8.40 ± 0.36 0.574 231 90 Winter ‘06-‘07 6.64 ± 0.36 6.70 ± 0.61 0.978 109 49 Summer 2007 11.17 ± 0.43 10.14 ± 0.39 0.174 98 117

    All 2007 8.99 ± 0.35 9.21 ± 0.38 0.454 179 156 Summer 2008 9.25 ± 0.49 10.71 ± 0.66 0.179 69 43

  • 38

    Fatty Acid Composition Results

    Descriptive Statistics

    551 individual fish from 2005-2008 were analyzed for lipid composition in this

    study. Fatty acids that were found in concentrations of at least 5 wt% in all samples were

    14:0, 16:0, 16:1n-7, 18:1n-9, 20:1n-9, 20:5n-3, 22:1n-11 and 22:6n-3. These fatty acids,

    as well as omega-3 and -6 fatty acids and essential fatty acids (18:3n-3, 18:2n-6, 20:4n-6)

    were then examined in more detail. The distributions, mean percent composition and

    standard errors of each of these groups of fatty acids in all samples as well as separately

    for each year are shown in Tables 8 and 9.

    Fatty Acid Composition by Fish Size

    The bivariate scatterplots of various fatty acids versus fork length (16:0, 20:1n-9,

    20:5n-3, 22:1n-11, 22:6n-3 and 20:4n-6) are shown in Figure 6. Although all linear

    regressions of individual fatty acids by fork length produced significant results

    (P

  • 39

    signatures with size. The average percent composition of the essential fatty acids by fish

    size is shown in Tables 9 and 10. This table is consistent with the patterns observed in

    the regressions of the individual fatty acids by fork length, with small fish having a

    higher concentration of most omega-3 and omega-6 fatty acids examined.

    It was apparent upon visual inspection of the data that many of the fatty acids that

    exhibited a pattern with fork length did not do so in a linear manner. Certain fatty acids

    showed a sharp initial decline in percent composition in small fish and then reached an

    asymptote at relatively low percent composition in large fish (e.g.: 16:0, 22:6n-3, see

    Figure 6). Other fatty acids showed the opposite trend, with low percent composition in

    small fish increasing to an asymptote at relatively high percent composition in large fish

    (e.g.: 20:1n-9, 22:1n-11, see Figure 6). These two relationships underscore the variability

    in fatty acid composition present in small fish and the relative consistency in fatty acid

    composition in large fish.

    Fatty Acid Composition Annually and Seasonally

    Individual fatty acids were examined individually by year, but no meaningful

    patterns were identified in the data. The analysis of entire fatty acid signatures by year

    was more meaningful (see below). The results of the ANCOVA of individual fatty acids

    by season showed patterns with 16:1n-7, 20:5n-3 and 22:1n-11 when percent composition

    of the fatty acid was corrected for differences in fish sizes between seasons. The percent

    composition of 16:1n-7 was significantly higher in summer samples (4.86% ± 0.06) than

    in winter samples (3.85% ± 0.12). The percent composition of 20:5n-3 was significantly

  • 40

    higher in summer samples (7.00% ± 0.07) than in winter samples (3.93% ± 0.15). The

    percent composition of 22:1n-11 was significantly higher in winter samples (26.99% ±

    0.39) than in summer samples (20.46% ± 0.18). The ANCOVA analyses of percent

    composition of individual fatty acids by catch location did not reveal any significant

    results.

  • 41

  • 42

    Table 9. Mean percent composition and standard error of omega-3 and omega-6 fatty acids in herring from 2005-2008 and separately for each year in the study.

    .

    Year n Size n - 3 n - 6 05-08 551 all 19.40±0.31 1.64±0.05

    415 ≥15.5cm 18.41±0.13 1.60±0.02 136

  • 43

    6a. 6b.

    6c. 6d.

    6e. Figure 6. Bivariate scatterplots of fatty acids that exhibited a pattern with fork length upon visual inspection. These fatty acids were: a) 16:0, b) 20:1n-9, c) 20:5n-3, d) 22:1n-11 and e) 22:6n-3. Although all linear regressions of individual fatty acids by fork length produced very significant results (P = 0.000), only the regression of 22:6n-3 by fork length was interpreted as meaningful (R2 = 0.24, P

  • 44

  • 45

    Fatty Acid Signature Analysis

    Twenty-three fatty acids were analyzed by discriminant and principal components

    analysis (PCA) in 551 individual fish to determine whether patterns existed in fatty acid

    signatures in herring of different sizes, years, seasons and locations. These fatty acids

    were: 14:0, 16:0, 16:1n-11, 16:1n-9, 16:1n-7, 16:1n-5, 18:1n-11, 18:1n-9, 18:2n-6, 18:3n-

    3, 18:4n-3, 20:1n-11, 20:1n-9, 20:1n-7, 20:4n-6, 20:4n-3, 20:5n-3, 22:1n-11, 22:1n-9,

    22:1n-7, 22:5n-3, 22:6n-3 and 24:1n-9.

    Discriminant Analysis

    The discriminant analysis of fatty acid signatures of all fish by size class did not

    produce functions that classified the samples better than the raw data alone (Eigenvalue =

    0.923. The discriminant analysis of all fish detected significant differences (Wilk’s λ<

    0.001) in fatty acid signatures annually. Three significant discriminant functions were

    created that correctly classified fish from 2005 78.6% of the time, fish from 2006 93.5%

    of the time, fish from 2007 78.6% of the time and fish from 2008 71.6% of the time. The

    first two functions accounted for 79.5% of the variance in the analysis. Function 1

    strongly separated the fatty acid signatures of fish from 2005 from the fish from the other

    three years (Eigenvalue = 2.700, see Figure 7) and this result was confirmed by the

    Tukey test from the ANOVA of discriminant function 1 by year. Function 2 separated

    the fatty acid signatures of fish from all years (Eigenvalue = 1.392), according to the

    results of the Tukey test from the ANOVA of discriminant function 2 by year. However,

    upon inspection of the discriminant plot (see Figure 7), function 2 separates 2006, 2007

  • 46

    and 2008 fish. Function 3 separated the fatty acid signatures of fish from 2005 and 2006

    from fish from 2007 and 2008 (Eigenvalue = 1.053).

    The discriminant analysis revealed also significant differences in the fatty acid

    signatures of fish from different seasons (Wilk’s λ

  • 47

    from 2005 and 2006 from fish from 2007 and 2008 (Eigenvalue = 1.328, see Figure 9,

    confirmed by the Tukey test). Function 3 also separated the fatty acid signatures of fish

    from 2005 and 2006 from 2007 and 2008 (Eigenvalue = 1.214). The discriminant

    analysis did not reveal any significant differences in the fatty acid signatures of fish from

    different seasons or catch locations (inshore/offshore).

    PCA and Discriminant Analysis of Small Fish

    The sample size of small fish from 2008 was less than the number of variables in

    the discriminant analysis (n = 12). Therefore, a principal components analysis (PCA)

    was conducted on small fish with the 23 fatty acids that were analyzed through the

    discriminant analyses mentioned above. The PCA revealed 5 components with

    Eigenvalues greater than 1 that explained 80.3% of the variance in the fatty acid

    signatures of small fish combined. These 5 components were then analyzed through

    discriminant analysis to determine if patterns exist in the fatty acid signatures of small

    fish annually (see Figure 10). Fish from 2005 were excluded from this analysis because

    no fish less than 15.5cm were collected during that year. Two discriminant functions

    were created based on the principal components analyzed in the model and the first

    function accounted for 71.0% of the variance in the analysis, while the second function

    accounted for the remaining 29.0%. The functions correctly classified fish from 2006

    89.7% of the time, fish from 2007 89.1% of the time and fish from 2008 58.3% of the

    time. The low classification results from 2008 may be due to the low sample size from

    that year (n =12). Function 1 (Eigenvalue = 1.381, Wilk’s λ

  • 48

    Wilk’s λ

  • 49

    Figure 7. Plot of discriminant functions 1 and 2 created from the fatty acid signatures of all fish (n=551) by year. The analysis revealed significant differences in the fatty acid signatures of fish between years (Wilk’s λ

  • 50

    Figure 8. Plot of discriminant function 1 vs. fork length for all fish by season (summer = 329, winter = 71). Only fish from 2006-2007 were used in this analysis. The analysis revealed significant differences in the fatty acid signatures of fish between seasons (Wilk’s λ

  • 51

    Figure 9. Plot of discriminant functions 1 and 2 created from the fatty acid signatures of large fish (≥15.5cm) by year (n = 415). The analysis revealed significant differences in the fatty acid signatures of fish between years (Wilk’s λ

  • 52

    Figure 10. Plot of discriminant functions 1 and 2 created from the fatty acid signatures of small fish (

  • 53

    Figure 11. Plot of discriminant function 1 vs. fork length for small fish (

  • 54

    Multivariate Non-parametric Primer 6 Models

    Twenty-three fatty acids were analyzed in 549 individual fish using non-

    parametric Primer models to determine whether patterns existed in fatty acid signatures in

    herring of different sizes, years, seasons and locations. These fatty acids were: 14:0,

    16:0, 16:1n-11, 16:1n-9, 16:1n-7, 16:1n-5, 18:1n-11, 18:1n-9, 18:2n-6, 18:3n-3, 18:4n-3,

    20:1n-11, 20:1n-9, 20:1n-7, 20:4n-6, 20:4n-3, 20:5n-3, 22:1n-11, 22:1n-9, 22:1n-7,

    22:5n-3, 22:6n-3 and 24:1n-9.

    The results of the ANOSIM analysis of all fish by size class showed a significant

    difference in fatty acid signatures between small (

  • 55

    or equal to 14.0cm, the ANOSIM analysis revealed significant differences in the fatty

    acid signatures of these two groups of fish (global R = 0.315, P

  • 56

    year. The average dissimilarity of comparisons including fish from 2005 were higher

    than comparisons only including fish from the other three years (2005/2008: dissimilarity

    = 17.96%, 2005/2007: 16.10%, 2005/2006: 15.58%, 2006/2007: 11.26%, 2006/2008:

    14.40%, 2007/2008: 13.99%). The SIMPER analysis also revealed that 16:1n-11 and

    16:1n-5 were the most important fatty acids contributing to the differences in fatty acid

    signatures of fish from 2005 compared to the other three years. The ANOSIM analysis

    did not reveal significant differences in the fatty acid signatures of large fish by season

    (global R = 0.053, P>0.05) or by catch location (global R = 0.041, P>0.05).

  • 57

    Figure 12. MDS plot of all fish by size class. The stress of 0.18 shows the model is confident of its placement of each sample in relative space. The ANOSIM analysis of these fish showed a significant difference in fatty acid signatures between small (

  • 58

    Figure 13. MDS plot of large fish by year. The stress value of 0.18 indicates the model is confident of its placement of the samples in relative space. The ANOSIM analysis revealed significant differences in the fatty acid signatures of fish by year (global R = 0.319, P

  • 59

    Fatty Acid Signature Analysis of Small Fish

    The ANOSIM analysis indicated significant differences in the fatty acid

    signatures of small fish by year (global R = 0.301, P

  • 60

    Figure 14. MDS plot of small fish by year. The stress value of 0.1 indicates the model is confident of its placement of the samples in relative space. The ANOSIM analysis revealed significant differences in the fatty acid signatures of small fish by year (global R = 0.301, P

  • 61

    Figure 15. MDS plot of small fish by season. The stress value of 0.1 indicates the model is confident of its placement of the samples in relative space. The ANOSIM analysis revealed significant differences in the fatty acid signatures of small fish by season (global R = 0.439, P

  • 62

    DISCUSSION

    The potential implications of the variation found in total percent lipid and fatty

    acid composition in Atlantic herring are vast. Herring are the main source of food for

    many upper trophic level predators in the BoF and variation in their lipid, and therefore

    energy, content can directly affect the quality of food available to higher levels of the

    food chain. Herring are also a main consumer of the dominant zooplankton species

    (Calanus finmarchicus) in the BoF, directly linking the fish to lower trophic level

    consumers. Since herring are intimately connected to primary producers via zooplankton

    and apex predators, variation in energy content can have implications throughout the food

    chain.

    Ontogenetic Variation in the Nutritional Value of Herring

    The significant ontogenetic differences found in both the total lipid content and

    fatty acid signatures of fish indicate differences in the nutritional value of fish by size.

    On average the highest total lipid content found in herring occurs in fish that are about

    20cm long (Figure 4). This is below the asymptotic length that herring reach in this

    population, indicating that the largest fish may not be the fattest. This trend in fat content

    may be due to the high ratio of muscle to gonad mass in 20cm fish as compared to larger

    fish in the population. To evaluate whether muscle and gonads were different in their

    lipid content, a small subsample of fish were analyzed for total lipid content separately

    for muscle and gonads. Methods were similar to those outlined earlier in this paper, but

    muscle and gonads of individual fish were removed prior to homogenization. Data from

    this subsample of fish show that muscle (n = 88, 19.9-28.2cm) has a higher fat content

  • 63

    than gonads (n = 22) in herring (9.95wt% ± 0.49 and 5.20wt% ± 0.54, respectively), and

    the relatively small gonads coupled with high muscle content found in medium sized fish

    may partly explain why they have the highest fat content. Also, the gonads analyzed in

    this subsample are both male and female and reflect a variety of stages of sexual

    maturity, indicating that the difference observed in the lipid content between somatic and

    sexual tissue not restricted to a specific sex or size class of fish. There may also be a

    decline in total lipid content for very large fish (Figure 4), possibly because these fish

    may have just recently spawned and are still large in size but have lost much of their lipid

    reserves via the release of eggs or sperm. Although muscle has higher lipid content than

    gonads, spawning still may greatly reduce the total amount of lipid in each fish.

    Atlantic herring have a fully developed gut and the ability to eat prey that adult

    fish eat once they reach about 4cm in length (Ehrlich et al. 1976, Doyle 1977, Blaxter and

    Hunter 1982). The diet of herring in the BoF is mainly Calanus copepods of the naupliar

    through adult life stages. Studies on herring from the BoF found that the adult fish diet is

    about 73% copepods with the remaining 27% consisting of (in decreasing order of

    abundance) sea squirts (Oikopleura sp.), water fleas (Cladocera sp.), decapod larvae,

    naupliar copepods and bivalve larvae (De Silva 1973). Herring grow rapidly as they

    leave the larval stage and this rapid growth continues until the fish are developmentally

    considered adults, which can last until the fish reach 9cm in length (Blaxter and Hunter

    1982). Although all the fish in the current study are developmentally adults, the small

    fish (

  • 64

    growth stage, when the fish are allocating energy for somatic growth. This is similar to

    the growth patterns of other species with similar life-history characteristics as Atlantic

    herring, such as Hawaiian anchovy, or nehu (Stolephorus purpureus, Struhsaker and

    Uchiyama 1976), American shad (Alosa sapidissima, Leggett and Carscadden 1978) and

    North eastern Atlantic herring (Jennings and Deverton 1991). Small fish could also be

    eating prey that has a lower fat content than adults (see results, fatty acid composition

    section) such as euphasiids and amphipods, which are also present in the BoF (Murison

    and Gaskin 1989).

    The current study revealed high concentrations of omega-3 and -6 fatty acids in

    small fish coupled with low concentrations of long chain monounsaturated fatty acids

    such as 20:1n-9 and 22:1-11, while the opposite trend was observed in large fish. These

    trends are shown in Tables 9 and 10, especially with the percent composition of all

    omega-3 fatty acids (column 3, Table 9). In all years combined as well as in each year

    separately, the percent composition of omega-3 fatty acids was higher in small fish

    compared to large fish. The SIMPER analysis of fish by size class revealed that the

    fatty acids contributing to the differences in fatty acid signatures between size classes of

    fish were 22:6n-3 (docosahexaenoic acid – DHA) and 20:4n-6 (arachidonic acid – AA).

    DHA and AA, along with 20:5n-3 (eicosapentaenoic acid – EPA) are the three fatty acids

    essential for normal growth in fish (Sargent et al. 1999). Marine zooplankton, such as

    copepods, also have high concentrations of long chain storage fatty acids such as 20:1n-9

    and 22:1n-11. These long chain storage fatty acids yield more ATP per fatty acid

    molecule because the oxidation of polyunsaturated fatty acids requires more energy to

  • 65

    break the large number of double bonds than that required to oxidize monounsaturated or

    saturated fatty acids (Stryer 1988). Therefore, the differences observed in the fatty acid

    signatures of fish by size are centered on certain fatty acids that are physiologically

    important to herring growth and metabolism.

    The ontogenetic differences in fatty acid signatures observed in this study may

    point to either selective metabolism of certain fatty acids by the fish during growth and

    development or a dietary shift with age. Since small and large fish are capable of eating

    the same prey (Calanus finmarchicus copepods, De Silva 1973), selective metabolism of

    certain classes of fatty acids by different size classes of fish may be occurring. Selective

    metabolism of fatty acids has been observed in mature female Norwegian capelin

    (Mallotus villosus), with polyunsaturated fatty acids (omega-3 and -6) from muscle tissue

    mobilized to aid gonad development prior to spawning (Henderson et al. 1984). Since

    growth is more rapid early in life for herring (Blaxter and Hunter 1982), small fish may

    be mobilizing their long chain fatty acids in order to retain omega-3 and -6 fatty acids for

    membrane and tissue growth while larger fish may instead be retaining long chain

    monounsaturated fatty acids for energy storage.

    Since fatty acid signatures can be used to predict diet, the ontogenetic differences

    found in the fatty acid signatures of herring may provide evidence for differences in the

    diet of herring as they grow and mature. Data from this study show that small fish have

    high concentrations of omega-3 and omega-6 fatty acids and low concentrations of long-

    chain monounsaturated fatty acid such as 20:1n-9 and 22:1n-11, while large fish show the

  • 66

    opposite trend. 22:1n-11, 22:6n-3 and 20:5n-3 are common fatty acids in copepods from

    the BoF (Michaud and Taggart 2007, Swaim et al. 2009) and it is these fatty acids that

    contribute to the large differences in fatty acid composition of fish of different sizes.

    Possible target prey items for small fish in the BoF could be euphasiids (Lee et al. 2006)

    or Oikopleura spp. (Troedsson et al. 2005), both of which are common in the North

    Atlantic (Barnard et al. 2004) or herring eggs, all of which have higher concentrations of

    22:6n-3 and 20:5n-3 and lower concentrations of 20:1n-9 and 22:1n-11. The variation

    found in both total lipid content and fatty acid composition of fish ontogenetically points

    to differences in the nutritional value of herring as they grow and mature.

    Annual Variation in the Nutritional Value of Herring

    We observed annual variation observed in the lipid content of Atlantic herring in

    the BoF with the most dramatic decrease occurring in 2005. Low lipid content in the

    2005 sample could be due to a variety of factors, including changes in the nature or

    composition of the zooplankton community, poor quality of prey or different

    environmental conditions. The differences found in both the lipid content and fatty acid

    signatures of fish from 2005 compared to the other years may point to either a shift in the

    availability or quality of zooplankton or a shift in the diet of herring during that year.

    Since total lipid and fatty acid signatures of fish from 2005 both changed, it is more

    likely that fish during that year shifted their diet. Possible sources of the decline in

    zooplankton availability or condition (resulting in a shift in herring diet to another food

    source) are vast including environmental and anthropogenic factors. Unfortunately, this

  • 67

    study did not measure zooplankton abundance or herring diet, so conclusions are difficult

    to make.

    Seasonal Variability in the Nutritional Value of Herring

    Seasonal variability in fat content in this population was expected due to the

    seasonal shifts in zooplankton availability that have been documented in the BoF

    (Murison and Gaskin 1989) and the timing of spawning by herring in the BoF. Herring

    from the BoF spawn in the fall and therefore build up fat reserves through the summer in

    preparation for spawning. Since not all fish in this study were of spawning age, the

    seasonal variation in lipid content observed may also be a result of seasonal variation in

    zooplankton availability. The availability of the copepods (the primary food of herring in

    this system (De Silva 1973)) has been shown to decline significantly in winter months

    (Murison and Gaskin 1989), possibly contributing to the relatively low lipid content of

    fish in the winter. This is largely due to the life cycle of copepods, which enter diapause

    in the winter and therefore build up energy reserves in the summer to survive this

    overwintering life stage (Comita et al. 1966). Lipid depletion during periods of low

    resource availability has previously been shown in herring (Wilkins 1967) as well as

    other marine and freshwater species such as tilapia (Satoh et al. 1984) and rainbow trout

    (Jezierska et al. 1982).

    The ANOSIM analysis revealed significant differences in the fatty acid signatures

    of small fish by season while large fish had seasonally consistent fatty acid signatures.

    The differences found in the fatty acid signatures of fish by season were not surprising,

  • 68

    given the likely diet shift described above. The absence of seasonal differences in fatty

    acid signatures of large fish could be due to the consistent metabolism of all fatty acids

    throughout the year by large fish, which are at a stable point in their life history. The

    seasonal differences found in the fatty acid signatures of small fish could be due to the

    variable life stages that are encompassed in the “small” size range (8.8-15.4cm). The

    seasonal differences found in the fatty acid signatures of small fish are likely a reflection

    of a diet shift that occurs as small fish grow. Since growth is so rapid when fish are

    small, large changes in fish size and the ability to eat certain prey items may change

    throughout the year.

    The seasonal variability in the nutritional quality of herring could have large

    impacts on upper trophic level predators in the BoF. January through April is the heart of

    winter in the BoF, when sea surface temperatures dip to 3-5°C (GoMOOS 2009) and

    zooplankton abundance is also at the yearly low (Fish and Johnson 1937). These

    conditions are reflected in the low lipid content of fish during the winter. For example,

    the mean percent lipid for the months of March and April was 3.56 ± 1.13 and 3.05 ±

    1.14 respectively, which is quite low compared to the percent lipid recorded in the other

    winter months (see Table 6). The dramatic drop in the average percent lipid of these fish

    from January to March (8.06 ± 0.82 and 3.56 ± 1.13, respectively) followed by the

    dramatic increase from April to June (3.05 ±1.14 and 10.28 ± 0.73, respectively) may

    impact the quality of food available to upper level predators seasonally (see Tables 5 and

    6). The very low lipid content of fish during these months is likely influencing the

  • 69

    distribution of upper trophic level predators, whose abundance drastically drops in the

    BoF during the winter months.

    Spatial Consistency in Lipid Content

    Previous work (McPherson et al. 2001) has found evidence for different stocks of

    herring throughout the BoF based on genetic markers, but this study found no evidence

    that they can be distinguished based on their lipid content or fatty acid composition.

    Samples from this study represent fish from at least three spawning grounds (Scots Bay,

    George’s Bank and Grand Manan) that have been distinguished based on genetic

    evidence (McPherson et al. 2001). The lack of differences in lipid content and

    composition implies that although there must be some form of reproductive isolation

    between these stocks, they all have similar feeding ecology and are likely exploiting

    similar prey items. This is not surprising since the abundance and distribution of

    copepods (which compose at least 73% of the herring diet (De Silva 1973) throughout the

    BoF has been found to be spatially uniform (Fish and Johnson 1937, Murison and Gaskin

    1989, Swaim et al. 2009).

    Variability in Lipid Content of Mature Fish by Gender

    Data from this study indicate that mature female fish (n = 83) have higher total

    lipid content than mature male fish (n = 78, 9.77% ± 0.38 and 8.64% ± 0.39,

    respectively). As stated previously, a small subsample of fish was pre-processed in order

    to examine the total lipid content of muscle and gonads separately. Data from this

    preliminary analysis indicate that eggs (n = 10) have a higher lipid content than sperm (n

  • 70

    = 6, 6.31% ± 0.76 and 4.27 ± 1.20, respectively) which may be responsible for the

    significantly higher percent lipid found in whole female fish as compared to male fish.

    This difference in total percent lipid by sex has also been found in other fish species.

    Anthony et al. (2000) found that female Pacific sand lance (Ammodytes hexapterus) had

    significantly higher lipid content than male fish, leading to higher contaminant loads

    passed on to predators from female fish than from male fish (whole fish lipid content,

    based on dried mass). The pattern of sexual differences in the lipid content in eastern

    North Atlantic herring (Clupea harengus L) is similar to their western counterparts that

    were examined in the current study; females also had significantly higher lipid content

    than males (Henderson and Almatar 1989).

    Ecological Consequences of Variability in Prey Quality

    The variation in prey quality found in this study, as reflected by variation in lipid

    content and fatty acid composition of fish has broad ecological implications. Herring are

    a primary, and sometimes the singular, prey item for many upper trophic level predators

    in the BoF. The annual and seasonal differences found in the fat content of herring may

    affect the condition of the predators that consume them. The Atlantic herring size range

    preferences of some of the main predators in the BoF are shown in Table 11 along with

    the mean percent lipid in 2005 and 2008 and summer versus winter. Although an

    extensive literature search was undertaken to determine the preferred size ranges of prey

    for the predators of herring, this analysis is only intended to illustrate the potential

    consequences of the differences in herring quality on predators. Values provided in

    tables are also only intended for illustrative purposes. It is apparent that the annual and

  • 71

    seasonal variation in lipids found in herring will affect the overall quality of the food

    predators are eating. For example, harbour porpoises in 2008 had fish available to them

    that were 66% more energy-rich than the fish that they ate in 2005 (see Table 11). This

    means that on average porpoises may have had to either spend more time foraging or

    cope with less energy from the same amount of prey, which could result in a decline in

    porpoise body condition. Likewise, in 2008 porpoises also obtained about 33% more

    energy from a single fish in the summer than it did in the winter (see Table 11).

    The effects of low food quality on predators include poorer body condition or a

    disproportionately large time spent foraging to obtain the energy required for survival. If

    predators spend more time foraging, consequently they devote less time to other

    behaviours such as reproduction and socializing (Anthony et al. 1981, Lokkeborg 1998,

    Dans et al. 2008). The seasonal consequences of low food quality are magnified in

    animals that are endothermic because of the fact that they have to maintain a stable body

    temperature (and therefore have increased energy requirements) as temperatures cool

    during the winter. The overall effects of low food quality are also exaggerated in females

    that are reproductively active or lactating. Mature female harbour porpoises (Phocoena

    phocoena) in the BoF reproduce annually and therefore are both pregnant and lactating

    for most of their adult lives (Read and Hohn 1995). Lactation and pregnancy place an

    added energy burden on females, who rely on the lipid they ingest rather than their thick

    blubber layer to cope with the increased energy demand (Koopman 1998). Since herring

    are the primary prey item of harbour porpoises (Recchia and Read 1989), the increasing

    energy demand of a female harbour porpoise is likely mostly filled by eating more

  • 72

    herring. It has been shown that lactating females have more food in their stomachs than

    other reproductive classes of porpoises that contain a wider variety of prey items

    (Recchia and Read 1989). This may be representative of mature female harbour

    porpoises expanding their niche breadth in order to meet their high energetic demands.

    Ontogeny