Bering Sea Integrated Ecosystem Research Program

Home , Ichthyoplankton, Roe, Spinycheek starsnout

NORTH PACIFIC RESEARCH BOARD

BERING SEA INTEGRATED ECOSYSTEM RESEARCH PROGRAM

FINAL REPORT

Ichthyoplankton: horizontal, vertical, and temporal distribution of larvae and juveniles of Walleye Pollock, Pacific Cod, and Arrowtooth Flounder, and transport pathways between nursery areas

NPRB BSIERP Project B53 Final Report

Janet Duffy-Anderson1, Franz Mueter2, Nicola Hillgruber2, Ann Matarese1, Jeffrey Napp1, Lisa Eisner3, T. Smart4, 5, Elizabeth Siddon2, 1, Lisa De Forest1, Colleen Petrik2, 6

1Alaska Fisheries Science Center, National Oceanic and Atmospheric Administration, 7600 Sand Point Way NE, Seattle, WA 98115, USA

2University of Alaska Fairbanks, School of Fisheries and Ocean Sciences, 17101 Point Lena Loop Road, Juneau, AK 99801 USA

3Ted Stevens Marine Research Institute, Alaska Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration, 17109 Pt. Lena Loop Road, Juneau, AK 99801, USA

4School of Aquatic and Fishery Sciences, University of Washington, Seattle, WA 98195-5020, USA

5Present affiliation: Marine Resources Research Institute, South Carolina Department of Natural Resources, Charleston, South Carolina 29422, USA

6Present affiliation: UC Santa Cruz, Institute of Marine Sciences, 110 Shaffer Rd., Santa Cruz, CA 95060, USA

December 2014

Table of Contents

Page

Abstract ...... 3

Study Chronology ...... 6

General Introduction ...... 8

BSIERP Hypotheses ...... 11

Project Objectives ...... 12

Chapter 1 ...... 14

Chapter 2 ...... 52

Chapter 3 ...... 88

Chapter 4 ...... 115

Chapter 5 ...... 148

Chapter 6 ...... 189

Chapter 7 ...... 213

Overall Conclusions ...... 267

BSIERP and Bering Sea Project Connections ...... 270

Management Implications ...... 272

Publications ...... 273

Poster and Oral Presentations ...... 275

Outreach ...... 278

Acknowlegements ...... 280

Abstract

This project component (B53) of the Bering Sea Integrated Ecosystem Research Program (BSIERP), a six-year multidisciplinary research effort sponsored by the North Pacific Research Board to study the Bering Sea ecosystem, was designed to examine linkages between physical oceanographic variables, biotic modulators, and distribution and abundance of three target fish species in the eastern Bering Sea (EBS): Walleye Pollock (Gadus chalcogrammus, previously described as Theragra chalcogramma), Pacific Cod (Gadus macrocephalus), and Arrowtooth Flounder (Atheresthes stomias). Effort focused primarily on Walleye Pollock due to limited data availability on Pacific Cod and Arrowtooth Flounder. Studies were based on historical zooplankton, ichthyoplankton, and physical data (1985-2010) collected by the NOAA/Alaska Fisheries Science Center over the Bering Sea shelf as well as on a series of seasonal, collaborative cruises that occurred 2008-2010 which were part of the BSIERP. Data derived from the above investigations were applied to a biophysical model developed to examine interannual patterns in Walleye Pollock distribution and abundance. The model work was funded, in part, with monies from project B53 and with funds provided through the Bering Ecosystem STudy (BEST) Synthesis Program, a Bering Sea research synthesis program sponsored by the National Science Foundation. Cumulative results of this work provide a better understanding of the potential effects of hydrographic variations in rearing conditions, transport, dispersal, and distribution of early life stages of Walleye Pollock in the eastern Bering Sea.

In the first part of the project we examined factors affecting distribution, abundance and community composition of larval fish assemblages, which included all three target species, over the Bering Sea shelf and determined: 1) A strong cross-shelf gradient delineates slope and shelf assemblages which is influenced by water masses emanating from the Gulf of Alaska, 2) Larval species assemblages in the Bering Sea differ between warm and cold periods, with larval abundances (including Walleye Pollock) being generally greater in warm years, and 3) Community-level patterns in larval fish composition reflect species-specific responses to climate change.

The next part of the project involved a comprehensive suite of studies designed to examine factors influencing distribution and abundance of young Walleye Pollock over the Bering Sea shelf.

First, abiotic and biotic variables were examined for effects on distribution and abundance of Walleye Pollock larval stages. From this project it was found that: 1) The influence of temperature on abundances of larval and juvenile Walleye Pollock increases with fish ontogeny, 2) Winds enhance the transport of early life stages from initial spawning locations to shallower depths over the continental shelf, 3) Localized measurements of zooplankton prey are a better indicator of young pollock abundance than

broad-scale measurements that are integrated over the shelf, and 4) Temperature is a major driving force structuring variability in abundance of Walleye Pollock in their first year of life.

Second, the effects of temperature on young Walleye Pollock were examined in greater detail in a study designed to determine whether early life stages undergo spatial shifts in response to changing temperature conditions, and test whether temperature affects the phenology of developmental events. It was found that: 1) Walleye Pollock early life stages are distributed further east over the continental shelf in warm years compared to cold years, and 2) Differences in the timing of density peaks support the hypothesis that the timing of spawning, hatching, larval development, and juvenile transition are temperature-dependent.

Third, a study of the factors affecting vertical distributions was undertaken. Results from this investigation found: 1) Walleye Pollock demonstrate a decrease in the depth of occurrence following hatching, indicating an ontogenetic change in vertical distribution, 2) Walleye Pollock vertical distributions are related to the date of collection, water column depth, and thermocline depth, 3) Non- feeding stages (eggs and yolksac larvae) do not exhibit diel vertical migration, 4) Flexion and postflexion stage larvae, 10.0–24.5 mm SL undergo regular diel migrations (0–20 m, night; 10–40 m, day), 5) Vertical distributions and diel migration a trade-off between prey access and predation risk for postflexion larvae, and 6) Vertical distributions of Walleye Pollock eggs, yolksac larvae, and preflexion larvae in the Bering Sea are different from previously-documented distributions in other ecosystems.

Fourth, we sought to determine mechanisms driving the observed spatial shifts in distribution of young Walleye Pollock in warm and cold years. To accomplish this we developed a biophysical model of spawning and dispersal, utilizing the data derived from the above investigations as input data, to examine spawning, drift and connectivity of Walleye Pollock populations over the Bering Sea shelf in warm and cold years. Model results indicate that: 1) Neither interannual variations in advection nor advances or delays in spawning time adequately represent the field-observed differences in distribution between warm and cold years, 2) Changes to spawning areas, particularly spatial contractions of spawning areas in cold years, resulted in modeled distributions that were most similar to observations, and 3) The location of spawning Walleye Pollock in reference to cross-shelf circulation patterns is important in determining the distribution of eggs and larvae.

Another part of the project sought to resolve taxonomy of Arrowtooth (Atheresthes stomias) and Kamchatka (Atheresthes evermanni) Flounder larvae in the eastern Bering Sea. Results show that: 1) Arrowtooth Flounder and Kamchatka Flounder can be successfully identified to the species level through molecular approaches, and 2) Arrowtooth Flounder 6.0–12.0 mm SL and ≥ 18.0 mm SL can be identified to the species level using morphological techniques alone, but species identification of individuals 12.1– 4

17.9 mm SL remains confounded by morphological similarities between the two species. Results indicate that molecular approaches must be used to conclusively identify Arrowtooth Flounder from Kamchatka Flounder in the 12.1 – 17.9 mm SL size range. Work on the ecology of larvae of the two flounder species suggests that: 1) the distribution of larvae (< 25.0 mm SL) of both Arrowtooth Flounder and Kamchatka Flounder is similar in the eastern Bering Sea; however, juvenile (≥ 25.0 mm SL) Kamchatka Flounder occur closer to the shelf break and in deeper water than juvenile Arrowtooth Flounder, and 2) Kamchatka Flounder larvae and Arrowtooth Flounder larvae have similar lipid content in the larval and juvenile stages when corrected for length. Results show that Kamchatka Flounder and Arrowtooth Flounder spatially co-occur over the Bering Sea slope and shelf, but Arrowtooth Flounder appears to have a wider distribution over the shelf than Kamchatka Flounder.

Taken collectively, our results indicate that abundance and distribution of fish larvae in the Bering Sea are influenced by abiotic factors such as temperature, oceanographic currents, light levels, water column stratification, seascape topography, water depth, and winds, as well as by biotic variables including prey availability, initial spawning location, and timing of spawning. Moreover, we found there is ecological plasticity in the responses of young fish to these various forcing factors depending on overall ecosystem state (warm vs. cold conditions).

Keywords: Walleye Pollock, Bering Sea, larvae, ichthyoplankton, distribution, abundance, Arrowtooth Flounder, Pacific Cod, survey, model

Citation: Duffy-Anderson, J.T., Mueter, F.J., Hillgruber, N., Matarese, A., Napp, J., Eisner, L., Smart, T., Siddon, E., De Forest, L., Petrik, C. 2014. Horizontal, vertical, and temporal distribution of larvae and juveniles of Walleye Pollock, and transport pathways between nursery areas. NPRB BSIERP Project B53 Final Report, 274 p.

Study Chronology

Date What Comments Projects PIs: N. Hillgruber (UAF), Janet Duffy- 2007 B53 project funded Anderson (AFSC), Jeffrey Napp (AFSC), Ann Matarese (AFSC), Lisa Eisner (AFSC) N. Hillgruber, UAF: graduate student advisor UAF doctoral student (E. J. Horne, UW: postdoctoral co-advisor 2008 Siddon), UW postdoctoral J. Duffy-Anderson (NOAA): postdoctoral co- researcher (T. Smart) recruited advisor 2008 February, May, BSIERP field season cruises Planned, executed, synthesized July, September 2008 Progress Report Submitted on time October 2008 Laboratory work, collaboration & Plankton sorting, larval identifications, genetics autumn integration with other projects Diet analyses of larval P. Cod and Walleye Graduate student Wess Pollock begun (work funded by PCCRC). 2009 autumn Strasburger recruited. Relevant data made available to BSIERP B53 UAF advisor: N. Hillgruber project Laboratory work, collaboration & 2009 Plankton sorting, larval identifications, genetics integration with other projects Work on larval fish assemblages 2009 Data from historical collections & BSIERP cruises over the EBS shelf Biological science technician Expedite plankton sample processing from 2009 January recruited BSIERP cruises All relevant data supplied to BSIERP project B53. Separately-funded NPRB/FOCI Cruise sampled an area of presumed Atheresthes 2009 April research cruise to Bering and spp. spawning. Data abundance and vertical and Pribilof Canyons horizontal distributional data on Atheresthes spp. made available to B53 project 2009 Progress Report Submitted on time April Work on vertical distribution of 2009 spring Data from historical collections & BSIERP cruises pollock, ATF, P. Cod larvae F. Mueter assumed Hillgruber responsibility of Departure of PI Hillgruber, PI 2009 spring UAF portion of project, including supervision of Mueter recruited graduate student E. Siddon 2009 May, BSIERP field season cruises Planned, executed, synthesized July, September Laboratory work, collaboration & 2009 autumn integration with other projects, Plankton sorting, larval identifications, genetics data analyses 2009 Progress Report Submitted on time October All relevant data supplied to BSIERP project B53. Cruise sampled an area of presumed Atheresthes FOCI-funded research cruise to spp. spawning. Data abundance and vertical and 2010 February Bering and Pribilof Canyons horizontal distributional data on Atheresthes spp. made available to B53 project

Laboratory work, collaboration & 2010 spring integration with other projects, Plankton sorting, larval identifications, genetics data analyses Work on horizontal distribution 2010 spring From historical collections & BSIERP cruises of pollock, ATF, P Cod larvae 2010 Progress Report Submitted on time April 2010 May, BSIERP field season cruises Planned, executed, synthesized July, September Laboratory work, collaboration & 2010 autumn integration with other projects, Plankton sorting, larval identifications, genetics data analyses 2010 autumn Warm cold year comparison work 2010 Progress Report Submitted on time October Laboratory work, collaboration & Plankton sorting, larval identifications, 2011 spring integration with other projects, collaborations with E. Farley (B90), R. Heintz data analyses, synthesis efforts (B54) Periodic data and metadata All data submissions were occurred as data 2008-2010 submissions to NPRB became available and were completed by 2010 No Cost Extension requested and 2012 October Revised end date: Dec. 31, 2013 granted Laboratory work, collaboration & Plankton sorting, larval identifications, 2011 autumn integration with other projects, collaborations with P. Stabeno (B52) data analyses, synthesis efforts 2012 spring Data synthesis, manuscripts 2012 April Progress Report Submitted on time Recruited postdoctoral researcher Developed and implemented biophysical model to 2012 autumn C. Petrik (BSIERP-BEST examine farcing factors underlying warm-cold Synthesis funded) spatial shifts in Walleye Pollock distribution 2012 October Progress Report Submitted on time 2013 spring Data synthesis, manuscripts Funds reallocated from NOAA to UAF to support Budget reallocation requested and 2013 January postdoctoral researcher C. Petrik’s pollock granted biophysical modeling efforts 2013 spring Data synthesis, manuscripts 2013 April Progress Report Submitted on time 2013 autumn Data synthesis, manuscripts 2013 autumn Project Headline completed (B53) “Young Fish in a Warm Bering Sea” “Simulating the Dispersal of Walleye Pollock Project Headline completed 2013 autumn Eggs and Larvae Over the Eastern Bering Sea (B53/BEST Synthesis) Shelf in Warm and Cold Years” 2013 October Progress Report Submitted on time 2013 December Final Report Submitted on time

General Introduction

Walleye Pollock (Gadus chalcogrammus, also described as Theragra chalcogramma) are sub- Arctic gadids found from Japan to the Chukchi Sea to central California. Walleye Pollock (also refered to as pollock, herein) are a semi-pelagic species that supports a major fishery in the rich waters of the Pacific subarctic, where catches have ranged from approximately 0.7 – 1.7 million metric tons in the Eastern Bering Sea (EBS) and Gulf of Alaska (GOA) since 1984. Ex-vessel value for these catches is approximately 400 million US dollars per year in the last decade. Fishery targets include roe in the winter and fillets throughout the spring and summer. Not only are pollock of significant commercial interest, they are also a central component of the food web in the eastern Bering Sea, serving as prey for fish, marine mammals, and seabirds.

Walleye Pollock population fluctuation has been linked to broad-scale climate variability including oscillating phases of the Pacific Decadal Oscillation and the Aleutian Low. Hunt et al. (2002) drew on these observed relationships to develop the Oscillating Control Hypothesis (OCH), a conceptual theory predicting ecosystem responses to climate phase shifts. Specifically, the hypothesis predicted that warm climate conditions would promote an early ice retreat, stratifying waters and maintaining production in the pelagic realm. It was hypothesized that these conditions enhanced survival of pelagic species, including Walleye Pollock. However, after a series of poor Walleye Pollock recruitment events followed an extended warm period, it was recognized that a revision of the OCH hypothesis was necessary. A revised hypothesis (Hunt et al. 2011) currently under investigation now speculates that a lack of large, lipid-rich copepod species during successive warm years is a factor in poor body condition of age-0 pollock in autumn, contributing to high over winter mortality and poor survival to age-1.

The OCH hypothesis recognizes that events occurring during the first year of life influence recruitment success of Walleye Pollock. Indeed, there is a large body of work demonstrating that events that occur during the early life phases influence recruitment because high abundances of small-sized offspring are more vulnerable to mortality than older, more established life stages. Therefore, in an effort to better understand recruitment in Walleye Pollock, it is appropriate to take a careful look at events occurring during the first year of life. BSIERP Project (B53) examines spatial and temporal dynamics in the ecology of early life stages of groundfishes, in particular Walleye Pollock, in the EBS to better evaluate potential consequences of climate change on Walleye Pollock and the broader ecosystem.

The goal of the first chapter of Project B53 was to quantify how spring larval fish assemblages respond to environmental variability, in particular temperature variability, and to examine what delineates community composition in the EBS. Characterizing patterns in larval fish community composition for the waters north of the Alaska Peninsula is of particular interest because this region includes known 8

spawning and nursery areas for a variety of ecologically and economically important groundfish species, including Walleye Pollock. In addition, the influx of larvae advected through Unimak Pass from the Gulf of Alaska may have important ecological consequences due to their potential impacts on local populations.

The second chapter examines the influence of broad- and fine-scale environmental conditions on abundance of Walleye Pollock early life history stages using time series data collected from the southeastern Bering Sea. For Bering Sea Walleye Pollock, there has been no comprehensive examination of all early life stages and the relative importance of environmental conditions at each stage. The influence of environmental conditions on early life stages (ELS) was then compared to the influence of spatial position and time of year, which have been identified as important predictors of Walleye Pollock ELS abundance in the Bering Sea.

The third chapter of Project B53 examines the most influential factor identified in Chapter Two in greater detail, temperature. Temperature can affect the annual distribution and development of Walleye Pollock ELS through the influences on the location of spawning, post-spawning transport, the timing of spawning, and/or rates of development. In this chapter we coupled data from BSIERP collections with NOAA/Alaska Fisheries Science Center historical data to compare distributions of Walleye Pollock ELS in years when the sea surface temperature anomaly was positive (warm years) and negative (cold years) to determine whether variations in temperature influenced distribution of early life stages over the Bering Sea shelf.

The fourth chapter examines environmental effects on vertical distributions of Walleye Pollock eggs and larvae. Ontogenetic vertical migration (OVM) is a pattern in which vertical distribution changes with stage of development. Diel vertical migration (DVM) is a behavioral trend in which depth of occurrence changes with time of day and light intensity. Accurate, stage-specific information on both OVM and DVM is necessary to understanding drift, transport, and connectivity of Walleye Pollock ELS from spawning to juvenile nursery grounds. This work examined abiotic and biotic factors influencing vertical position of Walleye Pollock ELS over all domains of the eastern Bering Sea.

The fifth chapter describes work to develop, test, and implement a biophysical model to describe larval Walleye Pollock transport trajectories over the eastern Bering Sea shelf. A physical model, the Regional Ocean Model System parameterized for the North East Pacific (ROMS-NEP6) was coupled with an individual-based model (TRACMASS), to test mechanisms that might underlie observed warm year-cold year spatial variability of Walleye Pollock ELS over the eastern Bering Sea shelf (Chapter 3). We utilized this model to decompose the transport process to test whether differences in distribution could be the result of oceanographic variations in physical transport, or the result of biological responses 9

to physical variation; specifically, climate-mediated variations in spawning distribution and/or temporal variations in spawning time. The sixth chapter describes work on another, less well-studied groundfish species, Arrowtooth Flounder (Atheresthes stomias). Biomass of Arrowtooth Flounder has been increasing in the eastern Bering Sea, and the species is a known predator of Walleye Pollock. Accordingly, there is considerable interest in understanding the ecology and recruitment dynamics of Arrowtooth Flounder. However, efforts to develop work focusing on ELS, work that is similar to that described above for Walleye Pollock, have been frustrated by an inability to conclusively identify larvae to species. Chapter six describes our efforts to conclusively identify Arrowtooth Flounder larvae collected from the Bering Sea to the species level using complimentary approaches including genetics, meristics, and morphology. We then apply that information to BSIERP-collected and NOAA/AFSC archival data to begin describing the ecology and life history dynamics of Arrowtooth and Kamchatka (Atheresthes evermanni) Flounder in the first year of life in the Bering Sea. The final chapter, chapter seven, re-focuses on the relationship between Walleye Pollock early life history and recruitment dynamics in the Bering Sea. Chapter seven reviews and summarizes the literature on Walleye Pollock ecology during the first year of life, adding and integrating data and results from the BSIERP project. Chapter seven revisits recruitment hypotheses that address population fluctuation as a function of mortality of ELS, identifies gaps in knowledge, and makes recommendations for future effort.

BSIERP Hypotheses

Research component (B53) was executed within the framework of the following BSIERP hypotheses:

BSIERP Hypothesis 4a. Climate-ocean conditions impact advection of larvae and juveniles to favorable nursery habitat (increased prey availability).

i. Shoreward wind-driven advection to favorable nursery habitat increases larval and juvenile walleye pollock survival. BSIERP Hypothesis 4b. Climate-ocean conditions impact predator-prey spatial and temporal overlaps.

Onshore currents separate juvenile fish from outer domain piscivores by transporting larvae inshore, away from adults.

i. Strength of frontal boundaries will weaken due to absence of the summer cold pool and low summer winds. A weakened inner front will open gateways to the inner domain for predators from the middle and outer domains.

BSIERP Hypothesis 4c. Climate-ocean conditions impact the strength of fronts between domains and the sizes of the domains.

ii. Strength of frontal boundaries will weaken due to absence of the summer cold pool and low summer winds. Out-migrations of anadromous species will shift away from shore due to the weakening of the inner front. iii. Strength of the inner front will weaken, allowing expansion of the inner domain, which will increase the carrying capacity of the inner domain for juveniles.

BSIERP Hypothesis 6. Climate and ocean conditions influencing circulation patterns and domain boundaries of the eastern Bering Sea shelf will affect the distribution, frequency, and persistence of fronts and other prey-concentrating features and thus the foraging success of marine birds and mammals.

Objectives

Objectives of the project were:

1. Describe the horizontal and vertical distribution of early life stages of Walleye Pollock, Pacific Cod, and Arrowtooth Flounder in spring, summer, and fall over the EBS shelf;

2. Determine the connectivity between larval and juvenile rearing areas and potential mechanisms by which offspring are transported from spawning areas to favorable juvenile nursery grounds for Walleye Pollock, Pacific Cod, and Arrowtooth Flounder;

3. Assess effects of changing climate on larval dispersal, juvenile settlement, and overall recruitment success for Walleye Pollock, Pacific Cod, and Arrowtooth Flounder.

All objectives were successfully met for Walleye Pollock, but paucity of historical observations to draw from, coupled with scant collections during BSIERP field years, precluded complete syntheses for Arrowtooth Flounder and Pacific Cod (Gadus macrocephalus). Specifically,

Objective 1: Information on Walleye Pollock spatial and temporal distribution can be found in Chapters 1, 2, 3, 4, and 5. Selected spatial and temporal information for Arrowtooth Flounder is available in Chapter 6. Selected spatial and temporal data are available for Pacific Cod in Chapter 1.

Objective 2: Information on connectivity of Walleye Pollock and relationships with environmental forcing variables can be found in Chapters 1, 2, 3, and 5. Selected information for Arrowtooth Flounder is presented in chapter 6. Selected spatial and temporal data are available for Pacific Cod in Chapter 1.

Objective 3: Assessment of effects of climate shifts on early life ecology of Walleye Pollock can be found in chapters 1, 2, 3, 5, and 7. Selected information for Arrowtooth Flounder is presented in chapter 6. Selected spatial and temporal data are available for Pacific Cod in Chapter 1.

Circumstances that precluded syntheses of Arrowtooth Flounder and Pacific Cod early life ecology as part of project B53:

Unfortunately, despite a 20-year history of sampling in the eastern Bering Sea led by the National Oceanic and Atmospheric Administration, collections of Pacific Cod and Arrowtooth Flounder larvae are either very few (cod) or not to species level (flounder). Instances of catch of Pacific Cod larvae in bongo tows collected from the Bering Sea are fewer than 300 (http://access.afsc.noaa.gov/ichthyo/index.php), and catches during the BSIERP program years were less than 20. Records of Arrowtooth Flounder are greater than for Pacific Cod, and catches were higher during BSIERP years as well, though historical difficulties with taxonomic resolution between this species and a congeneric, Kamchatka Flounder, Atheresthes evermanni, required that samples be identified only to the species level. We made great 12

progress in reclassifying historical samples to the species level using a combination of genetic and morphometric approaches, but ultimately were not successful at species-specific resolution across all larval size classes. This limited what data could be put into ecological context and the conclusions that could be drawn. Accordingly, data for Pacific Cod and Arrowtooth Flounder were deemed insufficiently robust to be utilized to the extent we originally envisioned. Accordingly, we focused on understanding the ecology of Walleye Pollock early life, and provided information on Arrowtooth Flounder and Pacific Cod where possible.

Chapter 1: Community-level response of fish larvae to environmental variability in the southeastern Bering Sea

Elizabeth C. Siddon1, Janet T. Duffy-Anderson2, Franz J. Mueter1

1University of Alaska Fairbanks, School of Fisheries and Ocean Sciences, Juneau, Alaska 99801, USA

2RACE Division, Recruitment processes Program, Alaska Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration, Seattle, WA 98115 USA

Citation: Siddon, E. C., Duffy-Anderson, J.T., and Mueter, F. 2011. Community-level response of ichthyoplankton to environmental variability in the eastern Bering Sea. Mar. Ecol. Prog. Ser. 426: 225- 239.

Abstract

Oceanographic conditions in the southeastern Bering Sea are affected by large-scale climatic drivers (e.g. Pacific Decadal Oscillation, Aleutian Low Pressure System). Ecosystem changes in response to climate variability should be monitored, as the Bering Sea supports the largest commercial fishery in the USA (Walleye Pollock, Gadus chalcogrammus). This analysis examined shifts in larval fish community composition in the southeastern Bering Sea in response to environmental variability across both warm and cold periods. Larvae were sampled in spring (May) during 5 cruises between 2002 and 2008 using oblique 60 cm bongo tows. Non-metric multidimensional scaling (NMDS) was used to quantify variability and reduce multi-species abundance data to major modes of species composition. Generalized additive models (GAMs) characterized spatial and temporal differences in assemblage structure as a function of environmental covariates. We identified a strong cross-shelf gradient delineating slope and shelf assemblages, an influence of water masses from the Gulf of Alaska on species composition, and the importance of nearshore areas for larval fish. Species assemblages differed between warm and cold periods, and larval abundances were generally greater in warm years. High abundances of Walleye Pollock in warm years contributed most to differences in Unimak Pass, outer domain, and shelf areas (geographic areas in the study region defined based on bathymetry). Sebastes spp. contributed to differences over the slope with increased abundances in cold years. We propose that community-level patterns in larval fish composition may reflect speciesspecific responses to climate change and that early life stages may be primary indicators of environmental change.

Introduction

Climate variability affects marine ecosystems through direct effects on ocean temperatures; an underlying warming trend (IPCC 2007) is therefore likely to affect commercial, recreational, and subsistence fisheries. Community-level consequences of environmental variability arise because species have different temperature tolerances (physiological optima and limits) and mobility to stay within their preferred thermal range (Pörtner et al. 2001). Populations or species with higher temperature optima will have a competitive advantage in warm conditions, resulting in species turnover and changes in community composition (e.g., Chavez & Messie 2009). In addition to direct responses of fish and other organisms, temperature changes are modulated by simultaneous changes in food availability and predation pressure, which are more difficult to predict because they interact in non-linear ways (Ciannelli et al. 2004).

Most previous studies have focused on temperature effects to adult demersal fish and shellfish communities (Brander et al. 2003, Perry et al. 2005, Mueter et al. 2007, Mueter & Litzow 2008, Spencer 2008). Less work has been done to investigate changes in the pelagic community structure or early life stages of fishes (Duffy-Anderson et al. 2006, Brodeur et al. 2008, Doyle et al. 2009). The pelagic distribution of ichthyoplankton is related to the spawning locations of adult fish (Doyle et al. 2002). After spawning, larval drift is subject to advection of water masses (Lanksbury et al. 2005), which is strongly influenced by wind stress and varies interannually as a result of basin-scale climate variability. Transport pathways can lead to differential survival of larvae based on life history characteristics (Doyle et al. 2009), predator abundances (Hunt et al. 2002), or availability of suitable juvenile habitat (Wilderbuer et al. 2002). Understanding variability in ichthyoplankton assemblage structure may indicate ecosystem- level and/or species-specific responses to climate change.

The southeastern Bering Sea has experienced both warm and cold conditions (as defined in Hunt et al. 2002, 2011) in recent years, offering an opportunity to examine changes in larval fish community compositions. Underlying this variability is a long-term warming trend of approximately 0.1°C per decade, with the most pronounced increases occurring during summer months (F. Mueter unpubl. data). Historically, sea surface temperatures (SSTs) in the Bering Sea were cool in the early 20th century followed by a relatively warm period from 1925 to the mid- to late 1940s. Temperatures in the 1950s to early 1970s were also cool, but increased after the 1976–77 regime shift (Hare & Mantua 2000). The Bering Sea has been generally warmer following this regime shift, and the highest summer temperatures since the beginning of the last century were observed between 2002 and 2005. However, the most extensive ice cover and coldest water column temperatures since the early 1970s were observed from 2006 to at least the end of 2010. While water-column temperatures have been much lower recently, average SSTs over the shelf during late summer have stayed relatively high (Mueter et al. 2009). 16

The goal of this work is to quantify how spring larval fish assemblages respond to environmental variability, in particular temperature variability, and to examine what delineates community composition in the southeastern Bering Sea. Characterizing patterns in larval fish community composition for the waters north of the Alaska Peninsula is of particular interest because this region includes known spawning and nursery areas for a variety of ecologically and economically important groundfish species (Lanksbury et al. 2007, Bacheler et al. 2010). In addition, the influx of larvae advected through Unimak Pass from the Gulf of Alaska (e.g., Northern Rock Sole Lepidopsetta polyxystra) (Lanksbury et al. 2007) may have important ecological consequences as these species interact with local populations.

Study Region

The southeastern Bering Sea is characterized by a broad continental shelf (>500 km wide) with an average depth of only about 70 m and supports a highly productive ecosystem owing to on-shelf flow of nutrient-rich waters. From spring to early fall, persistent oceanographic fronts (Hunt & Stabeno 2002) separate the shelf into 3 domains: the inner shelf domain (inside of the 50 m isobath), the middle domain (between 50 and 100 m isobaths), and the outer domain (between 100 and 200 m isobaths) (Iverson et al. 1979, Coachman 1986).

Predominant currents onto the southeastern Bering Sea shelf include the Alaska Coastal Current (ACC) that transports lower salinity waters from the Gulf of Alaska through Unimak Pass, and the Aleutian North Slope Current that brings higher salinity oceanic waters to the slope (Schumacher & Stabeno 1998, Stabeno et al. 2006). Current trajectories over the shelf are generally northwestward with the Bering Slope Current flowing along the shelf break and ACC waters following either the 50 or 100 m isobath (Stabeno et al. 2001).

The ACC flows counterclockwise around the Gulf of Alaska and southwestward along the Alaskan Peninsula; it branches through Unimak Pass, which represents the major conduit of flow between the Gulf of Alaska and the Bering Sea shelf (Ladd et al. 2005). The volume of ACC water advected through Unimak Pass varies seasonally and interannually (Stabeno et al. 2002). Freshwater discharge into the Gulf of Alaska can be used as a proxy for the strength of the ACC and, presumably, flow through 3 –1 Unimak Pass (Weingartner et al. 2005). Average discharge in March for 2002 to 2005 was 9764 m s 3 –1 versus 1872 m s for 2006 to 2008 (T. Royer unpubl. data based on formulae in Royer 1982) suggesting greater flow through Unimak Pass in warm years. The direction of ACC waters entering the Bering Sea varies based on differences in forcing mechanisms (e.g. wind speed and direction) that affect water column structure and front formation. The onset and location of fronts affect water current trajectories 17

(Kachel et al. 2002) and, therefore, transport pathways of larvae (Duffy-Anderson et al. 2006).

Materials and methods

Biological sampling

Data on spring larval fish assemblage structure were collected during 5 research cruises in the southeastern Bering Sea (Figure 1.1) between 2002 and 2008 using 60 cm bongo nets fitted with either 335 µm (2008) or 505 µm (2002, 2003, 2005, 2006) mesh; previous research determined that abundances of collected larvae are comparable between the 2 mesh sizes (Shima & Bailey 1994, Boeing & Duffy- Anderson 2008, Duffy-Anderson et al. 2010). Cruises occurred in May of each year (Table 1.1). During all cruises, quantitative oblique tows were made to a maximum depth of 300 m (or to within 10 m of the substratum), allowing for vertically integrated estimates of larval fish abundance. The ship speed was monitored and adjusted (1.5 to 2.5 knots) throughout each tow to maintain a wire angle of 45° from the ship to the bongo net. The nets were equipped with a calibrated 40 m flow meter; therefore, catch rates –2 were standardized to catch per unit effort (CPUE; number · 10 m ). Sampling occurred 24-hours a day and it was assumed that vertically integrated abundance estimates were not affected by diel vertical migrations. The geographic coverage of the sampling grid varied each year; to investigate changes in larval fish assemblage structure over time, only those stations sampled in at least two years were included in the analyses (‘common stations’; Table 1.1; Figure 1.1).

After retrieval of the bongo nets, all fish larvae were removed from the codends and a volume displacement measurement of remaining zooplankton (including small gelatinous zooplankton; large jellyfish were removed so as not to bias the displacement volume) was taken as a coarse measure of zooplankton wet weight biomass and an index of overall production at each station (Napp et al. 2002, Coyle et al. 2008, 2011). All samples were preserved at sea in 5% buffered formalin seawater solution. Fish larvae were sorted, identified to the lowest possible taxonomic level, measured (mm standard length [SL]), and enumerated at the Plankton Sorting and Identification Center in Szczecin, Poland. Identifications were verified at the Alaska Fisheries Science Center, NOAA (National Oceanic and Atmospheric Administration) in Seattle, Washington, USA.

Physical environment sampling

A Sea-Bird SBE 19 CTD was attached in-line between the bongo nets and the wire terminus to provide real-time estimates of temperature, conductivity, and pressure over the towed path. An estimate 18

of the water temperature within the study area each year was calculated by averaging the sampled water column temperature across all stations in a given year. Temperature and salinity measurements were averaged throughout the sampled water column at each station for comparison with the vertically integrated larval fish abundances to determine the characteristics of the water column when larvae were present. Larvae likely resulted from different water masses (e.g., above and below the pycnocline), but any effect of averaging was consistent across the study region (Duffy-Anderson et al. 2006). Temperature and salinity were also averaged within the top 20 m (surface layer) to visualize and identify water mass characteristics by geographic areas (see below). Water column profiles varied from well-mixed nearshore stations to more stratified offshore stations. Surface water characteristics best captured broad differences by area and provided a reasonable metric to track the less-saline (i.e., less dense) ACC water through Unimak Pass and subsequent mixing on the shelf.

Four distinct geographic areas were examined for the analyses, and stations were grouped as follows: Unimak Pass, slope (outside of 200 m isobath), outer domain (between 100 and 200 m isobaths), and shelf (within 100 m isobath). Very few stations were sampled within the inner domain (inside of 50 m) therefore these were combined with the middle domain (between 50 and 100 m) stations and comprised the shelf area. Surface (top 20 m) measurements of temperature and salinity distinguished unique water masses within each geographic area. Unimak Pass and the outer domain water masses had intermediate salinities, with Unimak Pass stations having relatively colder temperatures. Slope waters had the highest salinities and warmest temperatures while shelf waters had lower salinities and colder water temperatures (Figure 1.2). Community analyses

To quantify variability in species composition over time and space, we used non-metric multidimensional scaling (NMDS) to reduce multi-species abundance data to their major modes of variability (PRIMER 6, v6.1.11) (Clarke & Gorley 2006). NMDS allowed us to detect patterns in the biological data first and then interpret those patterns in relation to the environmental data (Field et al. 1982) using generalized additive models (GAMs). NMDS is also more robust to violations of assumptions than other methods (e.g. detrended correspondence analysis or principle components analysis) (Minchin 1987). Stations at which no larval fish were caught (n = 8) and rare species, defined as those present at less than 5% of the stations across all years, were removed from the analyses. Rare species likely do not contribute to broad-scale temporal and spatial patterns (Duffy-Anderson et al. 2006), therefore our approach allowed for detection of substantial shifts in species composition between years.

Larval fish abundance data were highly right-skewed, therefore a 4th root transformation 0.25 (CPUE ) was used to reduce the influence of samples with very high abundances. Transformed data 19

0.25 were standardized to species maxima (i.e. each value was divided by the maximum CPUE value for the corresponding species) to give equal weight to all species, regardless of their average numerical abundance (Field et al. 1982). Bray-Curtis similarity matrices were then computed to examine differences in assemblage structure among (1) individual stations and (2) by geographic area based on larval fish composition, followed by ordinations using NMDS to visualize similarities in species composition among stations or areas. The NMDS algorithm attempts to arrange samples (either stations or areas) such that pairwise distances in the ordination plot match Bray-Curtis similarities as closely as possible; thus, samples closer together in the ordination plot have a more similar species composition than samples farther apart. The final configuration of stations (areas) was determined by minimizing Kruskal’s stress statistic (Kruskal 1964), and the number of dimensions for the final ordinations was chosen as the smallest number of dimensions that achieved a stress of no more than 0.2. A stress of 0.1 or lower is considered a good fit (Kruskal 1964) and we defined a stress of less than 0.2 as acceptable.

NMDS by station

The final station-by-species matrix included 318 stations (Table 1.1) and 31 prevalent species (or species complexes) (see Table 1.2). The ordination axes in the NMDS plot, consisting of dimensionless values or scores for each station, were used as the response variable for modeling differences in assemblage structure in space and as a function of environmental covariates using GAMs. Spearman rank correlations were used to identify those species whose abundances were most strongly correlated (positively or negatively) with the axis scores and which therefore contributed most to the observed patterns of species composition. Only species for which the absolute correlation with a given ordination axis was equal to or larger than 0.4 were further examined.

A GAM approach was used for modeling species composition to avoid pre-specifying a functional relationship between the response and predictor variables. GAMs quantify the relationship between a set of predictors and the response through non-parametric smooth functions of the predictor variables (e.g. a smooth spatial surface can be fit as a function of latitude and longitude). The optimum amount of smoothing was chosen through a cross-validation approach as implemented in the R package ‘mgcv’ (Wood 2006). Appropriate (biologically meaningful) covariates (year, temperature, salinity, zooplankton displacement volume, latitude, and longitude) were selected to explain variability in larval fish assemblage structure. Station depth (bathymetry) is strongly confounded with the spatial term (latitude and longitude), and the estimated spatial surface captures any effects of location whether related to bathymetry, distance from shore, or other variables. Therefore, we did not include station depth as a covariate in the model.

The full model included a categorical year term to allow for differences in the average response between years (subscript t denotes different years), a smooth function (ƒ) of temperature and salinity to allow for possible interactions, a smooth function of zooplankton displacement volume, and a smooth spatial surface (interaction term for latitude and longitude):

Axis 1 = Yeart + f1(temperature, salinity) + f2(zooplankton displacement volume) + f3(latitude, longitude)

+  (Eq. 1)

Alternative models were considered that included separate smooth terms for temperature and salinity or eliminated one or more variables from the model (e.g., no zooplankton displacement volume term). Based on Akaike’s Information Criterion (AIC) (Akaike 1973, Burnham & Anderson 2002) and the amount of variability explained by each model (adjusted R2 values), a best fit model was selected for characterizing the estimated effects of environmental variability on species composition for each axis.

NMDS by geographic area

We compared species composition by geographic area by averaging the CPUE for each species across all stations within a given area, which resulted in an area-by-species matrix that included 20 year- area combinations (n = 5 years; n = 4 areas) and 31 species. The PRIMER routine MVDISP, which measured the relative dispersion of yearly values within each area, was used to compare the variability in species composition by area across the study period. To examine differences in species composition between the warm period (years 2002, 2003, and 2005) and cold period (years 2006 and 2008), a 1-way analysis of similarity (ANOSIM) tested for pairwise differences between each area-period combination. Separate ANOSIM tests were performed for each area to further test whether species compositions were significantly different between warm and cold periods. A SIMPER (similarity percentages) analysis was then performed using the full station-by-species matrix to determine the contribution of individual species responsible for the dissimilarity between areas and periods.

Results

Biological sampling

A total of 31 species or species complexes (e.g. Sebastes spp.) representing 14 different families were collected during 5 cruises over the 7 yr sampling period and were included in the community analyses. Walleye Pollock Gadus chalcogrammus was numerically the most abundant species in the assemblage (66% of total catch), followed by Pacific Sand Lance Ammodytes hexapterus, rockfishes Sebastes spp., Northern Rock Sole, and Pacific Cod Gadus macrocephalus (Table 1.2). Individual species abundances varied interannually; for example, Walleye Pollock comprised a maximum of 85% of the

total catch in 2002 to a minimum of 29% in 2006 (Figure 1.3). Fish larvae were generally more abundant in the warm years than in the cold years, especially Flathead Sole Hippoglossoides elassodon, northern rock sole, and Pacific Sand Lance. Rare species, though not included in the analyses, were sampled in either warm years (e.g. High Cockscomb Anoplarchus purpurescens and Greenland Halibut Reinhardtius hippoglossoides) or cold years (e.g. Arctic Cod Boreogadus saida). Physical environment sampling

The average water column temperature varied considerably between the warmer years of 2002, 2003, and 2005 (3.86, 4.75, and 4.16°C, respectively) and the colder years of 2006 and 2008 (3.39 and 2.0°C, respectively). This provided an environmental continuum against which to investigate changes in larval fish species composition. Water mass characteristics were unique at slope, outer domain, and shelf stations in cold years (Figure 1.2D), but the outer domain and shelf waters were not as clearly separated in warm years (Figure 1.2C). Unimak Pass stations generally displayed similar characteristics to outer domain waters; however, in 2002 and 2005, stations on the east side of Unimak Pass displayed characteristics of shelf waters, whereas stations on the west side of Unimak Pass were more similar to slope waters based on differences in salinity, indicating flow in both directions through Unimak Pass.

Community analyses

NMDS by station

The ordination of individual stations (Figure 1.4) condensed information on the abundance of each species and afforded both community-level and species-specific gradients to be described across the study area. GAMs illustrated 3 patterns in species composition that captured important habitat attributes for larval fish distributions as described below. Spearman rank correlations with the NMDS axis scores showed which individual species contributed most to the observed gradients. The first axis captured the greatest amount of variability in species composition, which was corroborated by strong species’ correlations, both positive and negative. Several species were strongly positively correlated with the second and third axes; however, no species were strongly negatively correlated with these axes (Table 1.3).

Generalized Additive Models Axis 1 The first axis described a gradient between a slope assemblage (species positively correlated with Axis 1) and a shelf assemblage (negatively correlated with Axis 1) that was resilient to interannual differences in species abundances (Figure 1.5A). The slope assemblage was characterized by Sebastes spp. and Atheresthes spp., as well as deeper-water species such as Pacific Blacksmelt (Bathylagus pacificus). In 22

contrast, the shelf assemblage was characterized by Alaska Plaice (Pleuronectes quadrituberculatus), Pacific Sand Lance, Walleye Pollock, and Northern Rock Sole (Table 1.3).

The best model for Axis 1 scores was described as:

Axis 1 ~ Yeart + f(Temperature, Salinity) + f(Latitude, Longitude) (Eq. 2) and included a significant categorical year term denoting a difference in the average value of the response among years, a significant smooth term of temperature and salinity, and a smooth spatial term (latitude and longitude) (Table 1.4). Although temperature and salinity were confounded with the spatial term, the latter largely captured residual variability not explained by either temperature or salinity. Zooplankton displacement volume was not significant in the full model described by Eq. (1) and was dropped from the best model. The model explained a significant proportion of the variability in species composition along 2 the first axis (adjusted R = 0.865; n = 318).

Both temperature and salinity had a strong influence on species composition (Figure 1.5B). The slope assemblage (positive correlations) was more common at higher temperatures and at higher salinities, while the shelf assemblage (negative correlations) was found at lower temperatures and salinities, corroborating the cross-shelf spatial pattern described above. In addition, we found significant variability in species composition among years (Figure 1.5C) that was not explained by local water mass characteristics or spatial patterns. Species abundances were generally higher in warm years, driven by shelf species such as Pacific sand lance, flathead sole, and northern rock sole

Axis 2 The second axis identified a plume of similar species composition originating in Unimak Pass and extending onto the shelf. Figure 1.6A shows the average spatial pattern across all years, though the spatial extent of the plume varied between years. Species strongly correlated with this plume of water included flathead sole, Pacific Cod, and Northern Rock Sole (Table 1.3).

The best model for Axis 2 was described as: Axis 2 ~ f(Temperature, Salinity) + f(Latitude, Longitude) (Eq. 3) and included a significant smooth term of temperature and salinity and a smooth spatial term (Table 1.4), however the year and zooplankton displacement volume terms were not significant. The model explained 2 additional variability in species composition along the second axis of the NMDS ordination (adjusted R = 0.423; n = 318).

The water mass associated with the plume of species originating from the Unimak Pass region 23

had warmer temperatures and lower salinities than surrounding waters (Figure 1.6B). The ACC carries lower salinity waters from the Gulf of Alaska through Unimak Pass (Stabeno et al. 2002) and may have influenced the spatial distribution (i.e. plume) of species assemblages. After accounting for the effects of temperature and salinity, as well as the spatial pattern, there was no significant effect of year in Axis 2 scores, suggesting that interannual variability in species compositions was fully accounted for by interannual differences in water mass characteristics.

Axis 3 The third axis delineated species that were associated with nearshore habitats in waters north of the Alaska Peninsula (Figure 1.7A). Species strongly positively correlated with Axis 3 included Pacific Cod, Pacific Sand Lance, and Sturgeon Poacher Podothecus acipenserinus (Table 1.3).

The best model for Axis 3 was described as: Axis 3 ~ f(Temperature, Salinity) + f(Latitude, Longitude) (Eq. 4) and included a significant smooth term of temperature and salinity and a smooth spatial term (latitude and longitude) (Table 1.4), whereas the year and zooplankton displacement volume terms were not significant. The model explained additional variability in species composition along the third axis 2 (adjusted R = 0.458; n = 318).

Temperature and salinity helped to distinguish the spatial pattern. Nearshore waters had warmer temperatures and lower salinities, whereas offshore waters had cooler temperatures and higher salinities (Figure 1.7B). The nearshore species assemblage may reflect spawning habitat preferences of the adult stages.

NMDS by geographic area The ordination by area (Figure 1.8) allowed detection of changes in species composition across broader geographic areas by year. The ordination showed a clear gradient in species assemblages from the slope to the shelf. Unimak Pass assemblages were more similar to outer domain assemblages in warm years, but less so in cold years. Species compositions were more variable between warm and cold periods for Unimak Pass, the outer domain, and shelf areas. In contrast, the slope assemblage was less variable across the study period based on average rank dissimilarity (MVDISP; Unimak Pass = 1.44; slope = 0.63; outer domain = 0.8; shelf = 1.13). Pairwise comparisons showed the highest dissimilarity among years for Unimak Pass and the lowest dissimilarity slope stations (Index of Multivariate Dispersion [IMD] = -0.76, where -1 would indicate maximum difference).

All unique area-by-period combinations showed significant differences in species assemblages (ANOSIM; p < 0.05), except slope assemblages between warm and cold periods (p = 0.08) and between Unimak Pass and outer domain assemblages in warm (p = 0.19) and cold (p = 0.1) periods. Within each area, species compositions were significantly different between warm and cold periods (p < 0.05 for all), though the difference on the slope was only weakly significant (p = 0.041). SIMPER then identified which individual species contributed most to assemblage differences between the warm and cold periods for each area. The average abundance of Sebastes spp. contributed most to differences in the slope assemblage, with more Sebastes spp. in cold years. The abundance of Walleye Pollock contributed most to assemblage differences in Unimak Pass, outer domain, and shelf areas with greater abundances of Walleye Pollock during the warm period (Table 1.5).

Discussion

Larval fish community composition in the southeastern Bering Sea was delineated by strong spatial patterns related to differences in water column temperature and/or salinity. Interannual differences in assemblage composition were attributed to species-specific responses to warm or cold conditions. Larval abundances were generally higher in warm years with high abundances of Walleye Pollock contributing most to differences in Unimak Pass, outer domain, and shelf areas between warm and cold periods. Assemblages over the slope were less variable between years and may be somewhat insulated from interannual variability. The slope assemblage was consistently dominated by Sebastes spp. with increased abundances in cold years. Therefore, community-level patterns in larval fish composition may reflect species-specific responses to environmental variability.

Cross-shelf assemblage structure was primarily associated with a geographic and/or salinity gradient that distinguished slope and shelf communities. Salinities are higher over the slope due to the oceanic influence of the Aleutian North Slope Current and lower on the shelf due to increased freshwater from the mainland and from the ACC flowing through Unimak Pass. The advection of slope waters onto the shelf can be seen in the spatial pattern of predicted species composition according to Axis 1 (Figure 1.5A). A finger of slope-derived species extended onto the shelf indicating larval transport through Bering Canyon. The cross-shelf gradient, largely driven by differences in spawning habitat for slope and shelf species, appears resilient to environmental variability between warm and cold years.

The observation of unique slope and shelf assemblages corroborates previous patterns (Doyle et al. 2002) and provides information on spawning habitats of adult fish. Although larval Sebastes spp. (slope assemblage) cannot easily be identified to species, many adult distributions follow the shelf break and slope habitats in the southeastern Bering Sea (e.g. Pacific Ocean Perch Sebastes alutus; Brodeur

2001). Juvenile Atheresthes spp., comprising Arrowtooth Flounder A. stomias and Kamchatka Flounder A. evermanni, are widely distributed on the continental shelf and begin recruiting to the slope habitat after about age-4 (Wilderbuer et al. 2009). In recent years, their abundance has increased, leading to a greater trophic impact; adult Arrowtooth Flounder are known to be voracious predators on juvenile Walleye Pollock (Livingston & Jurado-Molina 2000, Knoth & Foy 2008, Ianelli et al. 2009). Larval pollock, however, were predominant in the shelf assemblage in our study, indicating spatial separation from adult Arrowtooth Flounder and from larval aggregations of Atheresthes spp. over the slope. Alaska Plaice spawn along the north side of the Alaska Peninsula in April and May, and eggs and larvae drift north and northeast over the shelf (Duffy-Anderson et al. 2010). While the drift trajectories vary interannually, the general current flow retains Alaska plaice within the shelf habitat.

The advection of ACC waters through Unimak Pass (Ladd et al. 2005) may affect the distribution of larval fish on the southeastern Bering Sea shelf. Water in Unimak Pass is similar to the outer domain water mass, especially in cold years, indicating directional flow of ACC water onto the outer Bering Sea shelf. Warm years with greater inflow of ACC water (T. Royer unpubl. data) may result in increased mixing and subsequent blending of water mass characteristics over the shelf (Figure 1.2C). In cold years, inflow of ACC water is reduced, resulting in a clearer distinction of water masses (Figure 1.2D).

Species entrained in, or advected by, ACC waters within Unimak Pass and the Bering Sea shelf included Pacific Cod and northern rock sole, with higher overall abundances of these species in warm years. The trawling grounds around Unimak Pass are some of the most productive fishing areas for Pacific Cod in the Bering Sea (Conners & Munro 2008), and just northeast of Unimak Pass is a major spawning area (Shimada & Kimura 1994). Pacific Cod larvae caught in and near Unimak Pass in this study may reflect these well-known spawning areas and/or reflect the contribution of Pacific Cod spawned in the Gulf of Alaska to Bering Sea populations. Previous research on northern rock sole has identified spawning areas west of Unimak Pass along the Aleutian Islands and in the Gulf of Alaska with advection through Unimak Pass. Transport pathways follow the middle and outer shelf or flow eastward along the Alaska Peninsula (Lanksbury et al. 2007). Differential survival of Northern Rock Sole depends on transport to adequate nursery grounds in the coastal domain (Wilderbuer et al. 2002, Lanksbury et al. 2007). Unfortunately, our sampling design cannot resolve whether these larvae originated in the Gulf of Alaska or were entrained in ACC waters within Unimak Pass and nearby spawning grounds. The impact of Gulf of Alaska larvae on Bering Sea populations, and the degree to which the populations are connected, are important ecological (i.e. competition, predation) and fisheries management (number of sub- populations) questions. To address the connectedness of these populations, future work tracking larvae from different spawning grounds using genetic markers, otolith microchemistry, or differential growth rates could improve the resolution of Gulf of Alaska larval contributions to Bering Sea populations. 26

The importance of nearshore habitats to Pacific Cod, Bathymaster spp., and Pacific Sand Lance could reflect preferred spawning grounds of adult fish (e.g. Pacific Cod; Shimada & Kimura 1994). The onshore-offshore gradient in species composition was more difficult to interpret because correlations with individual species’ CPUEs were weaker than for the other axes. In addition, the third NMDS axis captured residual variability not already accounted for in the first or second axes. However, the importance of nearshore habitat and an onshore-offshore gradient in species composition are biologically reasonable; therefore, we believe our interpretation of this axis is realistic.

The 3 spatial patterns of larval fish assemblages identified from the NMDS ordination axes are not exclusive; individual species can be correlated with more than one gradient, thereby capturing different influences on larval distribution. For example, Pacific sand lance was strongly correlated with the first and third axes. The first axis described Pacific sand lance as a shelf species, while the third axis further associated larval sand lance with the nearshore environment of the shelf habitat. Pacific Cod was correlated with the second and third axes. The second axis highlighted the importance of ACC waters in the distribution of larval Pacific Cod while the third axis identified the nearshore environment as important, likely due to the spawning preferences of adult fish.

The analytical approach of multivariate ordination followed by GAMs as an exploratory regression technique successfully highlighted the main delineations of species compositions and modeled the response of the assemblage to environmental covariates. However, caution should be used when interpreting such results, as spurious (i.e. non-biologically relevant) outcomes are possible due to the flexible nature of GAMs. We are confident in our interpretations of the model results based on current knowledge of the Bering Sea ecosystem and believe our approach captured underlying mechanisms that determine larval fish species compositions in the southeastern Bering Sea.

While the timing of surveys used for this study was consistent across years, differential temperature effects on early life history events (e.g. spawning) could affect our interpretations. If colder temperatures result in delayed adult spawning activities or reduced rates of ichthyoplankton development, the fixed timing of our surveys could have been mismatched to the variable timing of larval production. Further, the timing of front formation in the region can also affect the distribution of larvae. For example, the Bering Sea Inner Front is a seasonally established hydrographic front that sets up in the vicinity of the 40 m isobath in spring and persists through late autumn (Schumacher & Stabeno 1998, Kachel et al. 2002). We hypothesize that if cold conditions persist over the shelf into late spring, the timing of the set up of the Inner Front would be delayed, resulting in continued retention of larvae in northward moving currents along the 100 and 200 m isobaths and potentially out of our east-west survey area (Lanksbury et 27

al. 2007).

The Oscillating Control Hypothesis (Hunt et al. 2002; revised in Hunt et al. 2011) provides a theoretical framework within which to predict ecosystem responses to warm and cold regimes in the southeastern Bering Sea. In warm regimes with early ice retreat, stratified waters maintain production within the pelagic system (Mueter et al. 2006), resulting in enhanced survival of species such as Walleye Pollock (Hunt & Stabeno 2002, Mueter et al. 2006, Moss et al. 2009). This is supported by the observation in the current study of high larval Walleye Pollock abundances in the warm years of 2002, 2003, and 2005. However, recent data show that in warm regimes, larger zooplankton taxa (e.g. large calanoid copepods and euphausiids) are less abundant, thus reducing growth rates and lipid reserves of young-of-year Walleye Pollock and thereby increasing predation risk and decreasing overwinter survival (Hunt et al. 2011). Therefore, a discontinuity exists between early spring conditions (i.e. water column temperature and prey availability), larval abundance, and the abundance of age-1 Walleye Pollock observed following the first winter. Although higher abundances of larval Walleye Pollock may not be indicative of eventual year-class strength, community-level analyses may provide information on ecological interactions affecting specific populations.

Our study was the first to look at changes in larval fish community composition within the southeastern Bering Sea over a time period that included both warm and cold periods. Significant differences in assemblage structure were detected, supporting the hypothesis that early life stages may be primary indicators of environmental change. The biological shifts between warm and cold regimes are difficult to predict due to direct and indirect species responses; a better understanding of non-linear environmental effects will increase predictive and management capabilities. The eastern Bering Sea Walleye Pollock fishery averaged 1.31 million tons annually between 2000 and 2009 (Ianelli et al. 2009), representing the largest commercial fishery in the USA by weight. Therefore, it is important to understand the mechanisms underlying interannual variability in this stock.

Acknowledgements

We thank the officers and crew of NOAA’s RVs ‘Miller Freeman’ and ‘Oscar Dyson’. Funding was provided through NOAA’s NPCREP and EcoFOCI programs, as well as the North Pacific Research Board (NPRB) Bering Sea Integrated Ecosystem Research Program (BSIERP). We thank 3 anonymous reviewers for providing helpful comments that improved the manuscript. This research is contribution EcoFOCI-0759 to NOAA’s Fisheries-Oceanography Coordinated Investigations, NPRB 285, and BEST- BSIERP 17.

Literature Cited

Akaike H (1973) Information theory as an extension of the maximum likelihood principle. In BN Petrov and F Csaki (eds) Second International Symposium on Information Theory. Akademiai Kiado, Budapest, p. 267-281.

Bacheler NM, Ciannelli L, Bailey KM, Duffy-Anderson JT (2010) Spatial and temporal patterns of Walleye Pollock (Theragra chalcogramma) spawning in the eastern Bering Sea inferred from egg and larval distributions. Fish Oceanogr 19(2): 107-120.

Boeing WJ, Duffy-Anderson JT (2008) Ichthyoplankton dynamics and biodiversity in the Gulf of Alaska: Responses to environmental change. Ecol Indicators 8: 292-302.

Brander KM, Blom G, Borges MF, Erzini K, Henderson G, MacKenzie BR, Mendes H, Santos AMP, Toresen P (2003) Changes in fish distribution in the eastern North Atlantic: are we seeing a coherent response to changing temperature? ICES Mar Sci Symp 219: 260-273.

Brodeur RD (2001) Habitat-specific distribution of Pacific ocean perch (Sebastes alutus) in Pribilof Canyon, Bering Sea. Cont Shelf Res 21: 227-224.

Brodeur RD, Peterson WT, Auth TD, Soulen HL, Parnel MM, Emerson AA (2008) Abundance and diversity of coastal fish larvae as indicators of recent changes in ocean and climate conditions in the Oregon upwelling zone. Mar Ecol Prog Ser 366: 187-202.

Burnham KP, Anderson DR (2002) Model selection and multi-model inference: a practical information- theoretic approach, Springer-Verlag, New York, NY.

Chavez FP, Messié M (2009) A comparison of Eastern Boundary Upwelling Ecosystems. Prog Oceanogr 83(1-4): 80-96.

Ciannelli L, Chan K-S, Bailey KM, Stenseth NC (2004) Nonadditive effects of the environment on the survival of a large marine fish population. Ecology 85(12): 3418-3427.

Clarke KR, Gorley RN (2006) PRIMER v6: User manual/tutorial. PRIMER-E: Plymouth.

Coachman LK (1986) Circulation, water masses, and fluxes on the southeastern Bering Sea shelf. Cont Shelf Res 5(1-2): 23-108.

Conners ME, Munro P (2008) Effects of commercial fishing on local abundance of Pacific cod (Gadus macrocephalus) in the Bering Sea. Fish Bull 106: 281-292.

Coyle KO, Pinchuk AI, Eisner LB, Napp JM (2008) Zooplankton species composition, abundance and biomass on the eastern Bering Sea shelf during summer: The potential role of water-column stability and nutrients in structuring the zooplankton community. Deep-Sea Res II 55: 1775-1791.

Coyle KO, Eisner LB, Mueter FJ, Pinchuk AI, Janout MA, Cieciel KD, Farley EV, Andrews AG (2011) Climate change in the southeastern Bering Sea: impacts on pollock stocks and implications for the Oscillating Control Hypothesis. Fish Oceanogr. 20: 139-156.

Doyle MJ, Mier KL, Busby MS, Brodeur RD (2002) Regional variation in springtime ichthyoplankton assemblages in the northeast Pacific Ocean. Prog Oceanogr 53: 247-281.

Doyle MJ, Picquelle SJ, Mier KL, Spillane MC, Bond NA (2009) Larval fish abundance and physical forcing in the Gulf of Alaska, 1981-2003. Prog Oceanogr 80: 163-187.

Duffy-Anderson JT, Busby MS, Mier KL, Deliyanides CM, Stabeno PJ (2006) Spatial and temporal patterns in summer ichthyoplankton assemblages on the eastern Bering Sea shelf 1996-2000. Fish Oceanogr 15: 80-94.

Duffy-Anderson JT, Doyle MJ, Mier KL, Stabeno PJ, Wilderbuer TK (2010) Early life ecology of Alaska plaice (Pleuronectes quadrituberculatus) in the eastern Bering Sea: seasonality, distribution, and dispersal. J Sea Res 64: 3-14.

Field JG, Clarke KR, Warwick RM (1982) A practical strategy for analyzing multispecies distribution patterns. Mar Ecol Prog Ser 8: 37-52.

Hare SR, Mantua NJ (2000) Empirical evidence for North Pacific regime shifts in 1977 and 1989. Prog Oceanogr 47: 103-145.

Hunt GL Jr, Stabeno PJ (2002) Climate change and the control of energy flow in the southeastern Bering Sea. Prog Oceanogr 55: 5-22.

Hunt GL Jr, Stabeno PJ, Walters G, Sinclair E, Brodeur RD, Napp JM, Bond NA (2002) Climate change and control of the southeastern Bering Sea pelagic ecosystem. Deep-Sea Res Part II 49: 5821- 5853

Hunt Jr., G.L., Coyle, K.O., Eisner, L., Farley, E.V., Heintz, R., Mueter, F.J., Napp, J.M., Overland, J.E., Ressler, P.H., Salo, S., Stabeno, P.J., 2011. Climate impacts on eastern Bering Sea foodwebs: A synthesis of new data and an assessment of the Oscillating Control Hypothesis. ICES J. Mar. Sci. 68(6): 1230-1243.

Ianelli JN, Barbeaux S, Honkalehto T, Kotwicki S, Aydin K, Williamson N (2009) Assessment of the Walleye Pollock stock in the Eastern Bering Sea. In: Stock assessment and fishery evaluation report for the groundfish resources of the Bering Sea/Aleutian Islands regions. North Pacific Fishery Management Council, 605 W. 4th Ave., Suite 306, Anchorage, AK, p 49-148.

IPCC (Intergovernmental Panel on Climate Change) (2007) Climate Change 2007: the physical science basis. Solomon S, Qin D, Manning M, Chen Z and others (eds) Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge.

Iverson RL, Coachman LK, Cooney RT, English TS, Goering JJ, Hunt GL Jr, Macauley MC, McRoy CP, Reeburgh WS, Whitledge TE (1979) Ecological significance of fronts in the southeastern Bering Sea. In RJ Livingston (ed) Ecological Processes in Coastal and Marine Systems. Plenum Press, New York, NY. p. 437-466.

Kachel NB, Hunt GL Jr, Salo SA, Schumacher JD, Stabeno PJ, Whitledge TE (2002) Characteristics and variability of the inner front of the southeastern Bering Sea. Deep-Sea Res Part II 49: 5889-5909.

Knoth BA, Foy RJ (2008) Temporal variability in the food habits of Arrowtooth Flounder (Atheresthes stomias) in the western Gulf of Alaska. NOAA Tech Memo NMFS-AFSC 184.

Kruskal JB (1964) Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis. Psychometrika 29: 1-27.

Ladd C, Hunt GL Jr, Mordy CW, Salo SA, Stabeno PJ (2005) Marine environment of the eastern and central Aleutian Islands. Fish Oceanogr 14(Suppl. 1): 22-38.

Lanksbury JA, Duffy-Anderson JT, Mier KL, Wilson MT (2005) Ichthyoplankton abundance, distribution, and assemblage structure in the Gulf of Alaska during September 2000 and 2001. Est Coast Shelf Sci 64: 775-785.

Lanksbury JA, Duffy-Anderson JT, Mier KL, Busby MS, Stabeno PJ (2007) Distribution and transport patterns of northern rock sole, Lepidopsetta polyxystra, larvae in the southeastern Bering Sea. Prog Oceanogr 72: 39-62.

Livingston PA, Jurado-Molina J (2000) A multispecies virtual populations analysis of the eastern Bering Sea. J Mar Sci 57: 294-299.

Minchin PR (1987) An evaluation of the relative robustness of techniques for ecological ordination. Vegetatio 69: 89-101. 31

Moss JH, Farley EV, Feldman AM, Ianelli JN (2009) Spatial distribution, energetic status, and food habits of eastern Bering Sea age-0 Walleye Pollock. Trans Am Fish Soc 138: 497-505

Mueter FJ, Litzow MA (2008) Sea ice retreat alters the biogeography of the Bering Sea continental shelf. Ecol Appl 18(2): 309-320.

Mueter FJ, Ladd C, Palmer MC, Norcross BL (2006) Bottom-up and top-down controls of walleye pollock (Theragra chalcogramma) on the Eastern Bering Sea shelf. Prog Oceanogr 68:152–183.

Mueter FJ, Hunt GL Jr, Litzow MA (2007) The Eastern Bering Sea shelf: a highly productive seasonally ice-covered sea. ICES CM2007/D:04, available at www.ices.dk/products/ CMdocs/CM- 2007/D/D0407.pdf.

Mueter FJ, Broms C, Drinkwater KF, Friedland KD, Hare JA, Hunt GL Jr, Melle W (2009) Ecosystem responses to recent oceanographic variability in high-latitude Northern Hemisphere ecosystems. Prog Oceanogr 81: 93-110.

Napp JM, Baier CT, Brodeur RD, Coyle KO, Shiga N, Mier K (2002) Interannual and decadal variability in zooplankton communities of the southeast Bering Sea shelf. Deep-Sea Res II 49: 5991-6008.

Perry AL, Low PJ, Ellis JR, Reynolds JD (2005) Climate change and distribution shifts in marine fishes. Science 308: 1912-1915.

Pörtner HO, Berdal B, Blust R, Brix O, Colosimo A, Wachter BD, Giuliani A, Johansen T, Fischer T, Knust R, Lannig G, Naevdal G, Nedenes A, Nyhammer G, Sartoris FJ, Serendero I, Sirabella P, Thorkildsen S, Zakhartsev M (2001) Climate induced temperature effects on growth performance, fecundity and recruitment in marine fish: developing a hypothesis for cause and effect relationships in Atlantic cod (Gadus morhua) and common eelpout (Zoarces viviparus). Cont Shelf Res 21: 1975-1997.

Royer TC (1982) Coastal freshwater discharge in the Northeast Pacific. J Geophys Res 87(C3): 2017– 2021.

Schumacher JD, Stabeno PJ (1998) The continental shelf of the Bering Sea. In AR Robinson and KH Brink (eds) The Sea: The global coastal ocean: regional studies and synthesis. John Wiley and Sons, New York, NY. p. 789-822.

Shima M, Bailey KM (1994) Comparative analysis of ichthyoplankton sampling gear for early life stages of walleye pollock (Theragra chalcogramma). Fish Oceanogr 3(1): 50-59.

Shimada AM, Kimura DK (1994) Seasonal movements of Pacific cod, Gadus macrocephalus, in the eastern Bering Sea and adjacent waters based on tag-recapture data. Fish Bull 92: 800-816

Spencer PD (2008) Density-independent and density-dependent factors affecting temporal changes in spatial distributions of eastern Bering Sea flatfish. Fish Oceanogr 17: 396-410.

Stabeno PJ, Bond NA, Kachel NB, Salo SA, Schumacher JD (2001) On the temporal variability of the physical environment over the south-eastern Bering Sea. Fish Oceanogr 10(1): 81-98.

Stabeno PJ, Reed RK, Napp JM (2002) Transport through Unimak Pass, Alaska. Deep-Sea Res Part II 49: 5919-5930.

Stabeno PJ, Hunt GL Jr, Napp JM, Schumacher JD (2005) Physical forcing of ecosystem dynamics on the Bering Sea shelf. In AR Robinson and KH Brink (eds) The Sea: The global coastal ocean: Interdisciplinary regional studies and syntheses, Part B. Harvard University Press, Cambridge, MA. p. 1177–1212.

Weingartner TJ, Danielson SL, Royer TC (2005) Freshwater variability and predictability in the Alaska Coastal Current. Deep-Sea Res Part II 52: 169-191.

Wilderbuer TK, Hollowed AB, Ingraham WJ Jr, Spencer PD, Conners ME, Bond NA, Walters GE (2002) Flatfish recruitment response to decadal climatic variability and ocean conditions in the eastern Bering Sea. Prog Oceanogr 55: 235-247.

Wilderbuer TK, Nichol DG, Aydin K (2009) Arrowtooth Flounder. In: Stock assessment and fishery evaluation report for the groundfish resources of the Bering Sea/Aleutian Islands regions. North Pacific Fishery Management Council, Anchorage, AK, p 677-740.

Wood SN (2006) Generalized Additive Models: An introduction with R. Chapman & Hall/CRC, Boca Raton, FL.

Table 1.1. Cruise name, year, temperature regime, and dates of cruises. The total number of stations sampled (‘bongo tows’) and the number of stations used in the analysis (‘common stations’) by year are shown

Cruise Year Temperature Dates Bongo tows Common

regime stations

3MF02 2002 Warm May 13- May 21 81 65

4MF03 2003 Warm May 18- May 24 60 58

5MF05 2005 Warm May 10- May 20 91 68

3MF06 2006 Cold May 9- May 18 90 75

3DY08 2008 Cold May 13- May 21 65 52

TOTALS 387 318

Table 1.2. Percent of total catch (based on number per 10m2) for species (or species complex) observed in greater than

5% of stations across the study period 2002 to 2008

Family Taxon Common name % Total catch

Gadidae Gadus chalcogrammus Walleye Pollock 66.44

Ammodytidae Ammodytes hexapterus Pacific Sand Lance 10.72

Scorpaenidae Sebastes spp. Rockfishes 8.86

Pleuronectidae Lepidopsetta polyxystra Northern rock sole 4.55

Gadidae Gadus macrocephalus Pacific Cod 1.86

Pleuronectidae Hippoglossoides elassodon Flathead Sole 1.53 Pleuronectidae Platichthys stellatus Starry Flounder 1.32

Bathymasteridae Bathymaster spp. 0.95

Gadidae Unidentified Gadidae 0.86

Pleuronectidae Pleuronectes quadrituberculatus Alaska Plaice 0.72

Pleuronectidae Atheresthes spp. 0.29

Stichaeidae Poroclinus rothrocki Whitebarred Prickleback 0.26

Bathylagidae Bathylagus pacificus Pacific Blacksmelt 0.17

Cottidae Icelus spp. 0.17

Liparidae Liparis spp. 0.13

Cottidae Myoxocephalus spp. 0.13

Stichaeidae Anoplarchus insignis Slender Cockscomb 0.11

Cryptacanthodidae Cryptacanthodes aleutensis Dwarf Wrymouth 0.10

Pleuronectidae Hippoglossus stenolepis Pacific Halibut 0.10

Bathylagidae Leuroglossus schmidti Northern Smoothtongue 0.10

Stichaeidae Anoplarchus spp. 0.09

Pleuronectidae Lepidopsetta bilineata Southern Rock Sole 0.09

Myctophidae Stenobrachius leucopsarus Northern Lampfish 0.08

Cottidae Icelinus spp. 0.07

Cottidae Hemilepidotus hemilepidotus Red Irish Lord 0.05

Hexagrammidae Hexagrammos decagrammus Kelp Greenling 0.05

Agonidae Bathyagonus alascanus Gray Starsnout 0.04

Agonidae Bathyagonus infraspinatus Spinycheek Starsnout 0.04

Agonidae Podothecus acipenserinus Sturgeon Poacher 0.04

Cottidae Artedius harringtoni Scalyhead Sculpin 0.03

Psychrolutidae Psychrolutes paradoxus Tadpole Sculpin 0.03

Table 1.3. Spearman rank correlations for the 3 axes (dimensions) interpreted from the non-metric multidimensional scaling (NMDS) ordination by station. Only those species with correlations ≥0.4 are shown.

Axis 1 Axis 2 Axis 3

Species Correlation Species Correlation Species Correlation

Hippoglossoides Gadus Sebastes spp. 0.82 elassodon 0.70 macrocephalus 0.48

Gadus Bathymaster Atheresthes spp. 0.66 macrocephalus 0.51 spp. 0.46

Lepidopsetta Ammodytes Bathymaster spp. 0.57 polyxystra 0.41 hexapterus 0.41

Podothecus Bathylagus pacificus 0.55 acipenserinus 0.40

Leuroglossus schmidti 0.53

Pleuronectes quadrituberculatus -0.58

Ammodytes hexapterus -0.57

Gadus chalcogrammus -0.48

Lepidopsetta polyxystra -0.40

Table 1.4. Model terms with corresponding significance values for each axis in the generalized additive model (GAM) analyses. Temp: temperature; Sal: salinity; Lat: latitude; Long: longitude; Yeart: year

Axis Term df F-value p-value Adjusted R2

1 0.865

Yeart 4 6.952 <0.001

Temp x Sal 16.16 8.851 <0.001

Lat x Long 6.59 5.778 <0.001

2 0.423

Temp x Sal 11.79 8.28 <0.001

Lat x Long 13.92 4.531 <0.001

3 0.458

Temp x Sal 10.94 2.815 0.0017

Lat x Long 17.27 8.281 <0.001

Table 1.5. Results from the PRIMER routine SIMPER used to identify differences in relative species composition based on geographic area and period using the full station-by-species matrix. The average abundance (catch per unit effort, CPUE, number per 10 m2) is shown for species that account for a significant amount of observed dissimilarity between periods; species’ abundances in bold account for approximately 60% of dissimilarity for the given area. For each geographic area separately, a 1-way analysis of similarity (ANOSIM) was used to test for significant differences in species composition between warm and cold periods using the Bray-Curtis resemblance matrix; the ANOSIM test statistic (R) and significance (p-value) are shown.

Geographic Period Gadus Sebastes Ammodytes Bathymaster Gadus ANOSIM ANOSIM Area chalcogrammus spp. hexapterus spp. macrocephalus (R) (p-value)

Unimak Pass Warm 270.8 46.9 66.5 49.1 218.8 0.45 0.001

Cold 8.1 73.1 19.6 2.1 24.5

Slope Warm 67.4 495.3 33.9 37.7 16.2 0.04 0.041

Cold 48.5 717.5 1.3 41.8 4.2

Outer domain Warm 259.6 32.6 28.8 26.6 51.8 0.13 0.003

Cold 87.1 62.2 9.1 0.8 14.9

Shelf Warm 3079.6 0.3 308.9 1.4 19.7 0.22 0.001

Cold 463.2 0 343.7 0 5.6

Figure 1.1. (A) Study region showing the location of sampling stations (). To investigate changes in larval fish assemblage structure over time, only those stations sampled in at least two years were included in the analyses (‘common stations’; Table 1.1). Depth contours are shown for the 40, 100, 200, and 1000 m isobaths. (B) Predominant currents in the study region include the Aleutian North Slope Current (ANSC), the Bering Slope Current (BSC), and the Alaska Coastal Current (ACC).

Figure 1.2. Stations sampled in (A) 2002 and (B) 2008, and corresponding temperature and salinity plots (averaged across the top 20 m) for (C) 2002 and (D) 2008. Samples were collected in 4 geographic areas based on bathymetry: Unimak Pass (), slope (; outside of 200 m isobath), outer domain (+; between 100 m - 200 m isobaths), and shelf (; out to 100 m). 2002 was a warm year showing increased mixing of water masses from Unimak Pass to the shelf; 2008 was a cold year with greater distinction of water masses. Note the difference in the x-axis scale in C and D.

A B

C D

Figure 1.3. Percent contribution to total catch (based on catch per unit effort) of the 5 overall most abundant species by year. The most abundant species were Walleye Pollock (Gadus chalcogrammus), Pacific Sand Lance (Ammodytes hexapterus), Sebastes spp., Northern Rock Sole (Lepidopsetta polyxystra), and Pacific Cod (Gadus macrocephalus).

90 80 Walleye pollock 70 Pacific sand lance

60 Sebastes spp. 50 Northern rock 40 sole

30 Pacific cod % contribution % 20 10 0 2002 2003 2005 2006 2008

Figure 1.4. Non-metric multidimensional scaling (NMDS) ordination, based on a Bray-Curtis similarity matrix, depicting the relative similarity in species composition among individual stations sampled across 5 years. Data were 4th root transformed and standardized to species maximum.

Figure 1.5. (A) Predicted spatial gradient of species composition as indicated by Axis 1 scores from non- metric multidimensional scaling (NMDS) ordination of species-by-station matrix, based on the generalized additive model (GAM) described by Eq. (2). The spatial surface was estimated as a smooth term of latitude and longitude; other covariates were fixed at their mean values. Species composition is predicted to be similar along contours; changes in species composition occur when moving across contours (color gradient). Spearman rank correlations of species positively or negatively correlated with these values were used to determine the main species of the slope versus shelf assemblage, respectively. Depth contours are shown for the 100 and 200 m isobaths.

Figure 1.5. (B) Predicted measure of species composition (Axis 1 scores from NMDS ordination) as a smooth function of temperature and salinity based on the GAM described by Eq. (2). Stations with salinities less than 29 (n = 3) were removed for better visualization of the relative effects of temperature and salinity. Cool colors correspond to the shelf habitat and negative species correlations; warm colors correspond to the slope habitat and positive species correlations (see A). Years are distinguished as follows: 2002 = red, 2003 = brown, 2005 = orange, 2006 = light blue, 2008 = purple.

Figure 1.5. (C) Estimated differences in species composition among years (Axis 1 scores from NMDS ordination) based on the GAM described by Eq. (2). Solid lines reflect the partial response of Axis 1 scores, on a normalized scale, when all other covariates are fixed at their mean values. Dashed lines denote 95% confidence intervals.

Figure 1.6. (A) Predicted spatial gradient of species composition as indicated by Axis 2 scores from non- metric multidimensional scaling (NMDS) ordination of species-by-station matrix, based on the generalized additive model (GAM) described by Eq. (3). Spearman rank correlations of species positively correlated with these values were used to determine the main species of the Alaska Coastal Current assemblage.

Figure 1.6. (B) Predicted measure of species composition (Axis 2 scores from NMDS ordination) as a smooth function of temperature and salinity based on the GAM described by Eq. (3). Warm colors correspond to the Alaska Coastal Current waters and positive species correlations (see A). See Figure 1.5 for additional description.

Figure 1.7. (A) Predicted spatial gradient of species composition as indicated by Axis 3 scores from non- metric multidimensional scaling (NMDS) ordination of species-by-station matrix, based on the generalized additive model (GAM) described by Eq. (4). Spearman rank correlations of species positively correlated with these values were used to determine the main species of the nearshore assemblage.

Figure 1.7. (B) Predicted measure of species composition (Axis 3 scores from NMDS ordination) as a smooth function of temperature and salinity based on the GAM described by Eq. (4). Warm colors correspond to the nearshore habitat and positive species correlations (see A). See Figure 1.5 for additional description.

Figure 1.8. Non-metric multidimensional scaling (NMDS) ordination, based on a Bray-Curtis similarity matrix, depicting the relative similarity in species composition among geographic areas by year. Outer domain: between 100 and 200 m isobaths; Shelf: within 100 m isobath; Slope: outside of 200 m isobath; Unimak Pass (see Figure 1.1). Data were 4th root transformed and standardized to species maximum.

Chapter 2: Influence of environment on Walleye Pollock eggs, larvae, and juveniles in the southeastern Bering Sea

Tracey I. Smart1, Janet T. Duffy-Anderson2, John K. Horne1, Edward V. Farley3, Christopher D. Wilson2, and Jeffrey M. Napp2

1School of Aquatic and Fishery Sciences, University of Washington, Box 355020, Seattle, WA 98195, USA

2RACE Division, Recruitment Processes Program, Alaska Fisheries Science Center, National Marine Fisheries Service, NOAA, 7600 Sand Point Way, Seattle, WA 98115, USA

3Auke Bay Laboratories, Alaska Fisheries Science Center, National Marine Fisheries Service, NOAA, 17109 Point Lena Loop Rd., Juneau, AK 99801, USA

Citation: Smart, T., Duffy-Anderson, J.T., Horne, J., Farley, E., Wilson, C., and Napp, J. 2012. Influence of environment on Walleye Pollock eggs, larvae, and juveniles in the Southeastern Bering Sea. Deep-Sea Research II: Topical Studies in Oceanography. 65-70: 196-207. 52

Abstract

We examined the influence of environmental conditions on Walleye Pollock (Gadus chalcogrammus) early life history in discrete stages at two ecological scales using a 17-year time series from the southeastern Bering Sea. Generalized additive models (GAMs) were used to quantify relationships between Walleye Pollock stages (eggs, yolksac larvae, preflexion larvae, late larvae, and juveniles), the fine-resolution environment (temperature, wind speed, salinity, and copepod concentration), and the broad-resolution environment (annual spawning stock biomass, temperature, zooplankton biomass, and wind mixing). Early stages (eggs, yolksac larvae, and preflexion larvae) were associated with high spawning stock biomass, while late stages (late larvae and juveniles) were not associated with spawning stock biomass. The influence of temperature increased with ontogeny: high egg abundance was associated with temperatures from -2 to 7 oC and negative annual temperature anomalies and high juvenile abundance was associated with temperatures from 4 to 12 oC and positive temperature anomalies. Winds enhanced the transport of early stages from spawning locations to shallower sampling depths, but did not affect feeding stages (preflexion larvae, late larvae, and juveniles) in a manner consistent with the encounter-turbulence hypothesis. Feeding stages were positively associated with localized copepod concentrations but not zooplankton biomass anomaly, suggesting that the localized measurements of potential prey is a better indicator compared to broad-scale conditions measured in areas where these stages do not necessarily occur. Broad-resolution covariates, however, explained a greater portion of the overall variation than did fine-resolution models. Of the environmental conditions examined, temperature explained more variation in abundance of Walleye Pollock early life stages than any other covariate. Temperature is likely a major driving force structuring variability in populations of Walleye Pollock in their first year of life, acting directly upon them and indirectly upon their physical habitat and prey community.

Introduction

The southeastern Bering Sea (SEBS) is part of a productive, high-latitude ecosystem that provides critical spawning and larval habitat for demersal fish species. In particular, the SEBS serves as important spawning, nursery, and forage habitat for Walleye Pollock (Gadus chalcogrammus, hereafter referred to as pollock), a target species for one of the world’s largest commercial fisheries (FAO, 2007). Pollock account for about 60 % of the total landings of groundfish in Alaskan waters (Ianelli et al., 2009) and also serve key ecological roles acting as top predators, prey for other top predators, and consumers of zooplankton. For Bering Sea pollock, female spawning biomass has fluctuated widely over the last four decades (from 0.85 to 4.1 million metric tons) and recruitment variability in the species is high, similar to other gadids (Ianelli et al., 2009). In the Gulf of Alaska (GOA) spawning biomass changes the initial production of eggs and availability of hatched larvae (Bacheler et al., 2009). Climate and hydrography influence the extent of transport from spawning grounds (Ciannelli et al., 2005; Wilson, 2009) and the survival and growth of larvae and juveniles (Porter et al., 2005; Logerwell et al., 2010).

Recently, Ciannelli et al. (2004) used a stage-specific approach to determine the density- dependent structure of pollock survival in the GOA, and Bacheler et al. (2010) used a stage-specific approach to define the temporal distributions of pollock eggs and early stage larvae over the SEBS shelf. Bacheler et al. (2010) identified three unique spawning areas with distinct egg and early larval phenologies, but did not examine interannual or seasonal changes in the physical environment experienced by eggs, larvae, or juveniles. Both studies demonstrated the importance of stage-specific responses in our ability to account for population trends and variation.

It is probable that a combination of biological and oceanographic conditions influences all fish life history stages and that the importance of any factor varies among stages. Objectives of the current study were to identify key factors that affect pollock early life stage (ELS) abundances in the Bering Sea and quantify the influence these factors have during the first year of development. Of course, environmental conditions may act on growth and survival of ELS either singly or in combination. For example, pollock egg development, time to hatch, and use of yolk reserves are controlled by temperature (Blood, 2002). Early larvae are more successful under both moderate wind-driven mixing conditions and good prey conditions (Walline, 1985; Bailey and Macklin, 1994; Mueter et al., 2006). Olla et al. (1996) suggested that mixing and temperature interact to determine larval pollock feeding success and subsequent growth and survival. Prey availability and temperature also interact to determine prey capture rates and survival of larvae, particularly when prey is scarce (Fukuchi, 1976). Adult pollock biomass, temperature, and wind conditions influence early juvenile and age-1 abundances in the Bering Sea (Wespestad et al., 2000; Duffy-Anderson et al., 2005; Mueter et al., 2006).

Environmental data are available over a range of spatial and temporal resolutions in the northeast Pacific. This range enables the examination of the relative importance of biological and physical factors at fine- and broad-scales. We define fine-resolution environmental data as those factors measured in association with individual ichthyoplankton tows at daily or weekly frequencies. Broad-resolution environmental data are those factors integrated over space and time across the eastern Bering Sea and within months or years. Broad-resolution indices have the advantage of larger sample sizes, while fine- resolution measurements address the immediate environmental conditions, which are more relevant to growth and survival of an early stage pollock. For Bering Sea pollock, there has been no comprehensive examination of all ELS and the relative importance of environmental conditions at each stage. The potential influence of the adult population on pollock ELS abundances was accounted for by testing for temperature-induced changes in spawn timing, with may influence which ELS we capture during surveys, and by including a spawning population index, which can constrain the number of ELS produced annually. The influence of environmental conditions on pollock ELS was compared to the influence of spatial position and time of year, which have been identified as important predictors of pollock ELS abundance in the Bering Sea (Bacheler et al., 2010).

Methods

Ichthyoplankton sampling

Depth-integrated abundance (number collected per 10 m2 sampled) of pollock ELS were measured during ichthyoplankton surveys conducted by NOAA’s Alaska Fisheries Science Center (AFSC, see Matarese et al., 2003). Surveys targeted pollock spawning grounds, areas adjacent to spawning grounds, and areas of downstream transport. Catch records were selected to include only samples south of 60 N latitude and east of the Aleutian Basin as samples were collected rarely outside of this area (Figure 2.1). Catch records also were limited to those between February and October due to limited sampling outside of this window and alignment with the spawning season (Bacheler et al., 2010). Surveys within the target area and dates occurred in 17 years within the time series (1988, 1991, 1992, 1994-2000, and 2002-2008, Table 2.1). It should be noted that Bacheler et al. (2010) identified three unique pollock spawning areas within our geographic range, but we opted to integrate the data to: 1) provide a synoptic look at factors influencing pollock ELS in the SEBS, and 2) provide sufficient geographic coverage to ensure that all ELS were represented.

Catch records were divided by developmental stage to determine if stage-specific responses to environmental conditions existed. Separate analyses were conducted for eggs, yolksac larvae (≤ 4.4 mm SL), preflexion larvae (4.5 and 9.9 mm SL), late larvae (10.0 and 24.9 mm SL), and juveniles (25.0 and 65.0 mm SL). Larval categories were based on standard length (SL) and key developmental attributes 55

(outlined in the Ichthyoplankton Information System, http://access.afsc.noaa.gov/ichthyo/index.cfm). Juveniles > 65 mm were not considered due to confounding effects of capture efficiency at larger sizes. Sampling for eggs through preflexion larvae was conducted using obliquely towed bongo nets (333 or 505 µm mesh). The number of tows containing late larvae and juveniles was rare relative to the other three stages. To increase the available sample size for these stages, we also examined catch records from the 1-m2 MOCNESS (333 or 505-µm mesh) and the 1-m2 Tucker trawl (333 or 505-µm mesh). Differences in selectivity between bongo, MOCNESS, and Tucker gears have been shown to be minor for all stages examined (Wiebe et al., 1976; Shima and Bailey, 1994). Volume filtered by each gear type was determined using calibrated flow meters. Table 2.1 lists the number of tows conducted with each sampling gear.

Adult maturity

To address the potential for variation in onset of spawning to impact abundance estimates from field surveys, gonad maturity data was collected by AFSC’s Midwater Assessment and Conservation Engineering (MACE) program. Gonad maturity stage was assessed macroscopically by readers on the NOAA ships Miller Freeman and Oscar Dyson each winter using samples from trawls fished over pollock spawning grounds. Maturity data collected during pollock surveys from 1992 to 1996 were based on a 5-stage categorical scale developed specifically for Walleye Pollock (Williams, 2007). From 1996 to the present this scale was expanded to an 8-stage scale based on Maier’s (1908) general teleost maturity classification (in Williams, 2007). Data from 1996 to 2007 were converted back to the 5-stage scale, where immature gonad was represented by stage 1 up to spent gonad represented by stage 5 (Stahl and Kruse, 2008). Since we were most interested in the possibility of delayed spawning related to sea ice or temperature conditions, we limited our examination to females larger than the presumptive spawning size (35 cm fork length, FL; Stahl and Kruse, 2008) captured between March 1 and 10. These dates were sampled most consistently across years. Samples were restricted to hauls occurring in the eastern Aleutian Islands outlined in Figure 2.1 to account for differences in spawning time across areas (Bacheler et al., 2010). We compared the stage of gonad maturity across temperature anomalies (ºC) based on net- mounted temperature sensor data collected during each tow. Maturity stages were compared by multiple linear regressions with fork length and gear temperature anomaly as independent variables.

2.3 Environmental conditions: fine- and broad-resolution

Environmental conditions were examined at fine- and corresponding broad-resolutions when possible (Table 2.2). Fine-resolution analyses utilized point-source data from net tows, either measured at the same location and time or within one day of the tow. Fine-resolution variables examined included water temperature, salinity, wind speed, copepod concentration, day of year (DOY), location (an 56

interactive term consisting of latitude and longitude), and year (as a factor). Temperature (ºC) and salinity data were derived from vertical CTD (Sea Bird 25, Sea-Bird Electronics, Bellevue, Washington, USA) profiles collected concurrently with ichthyoplankton samples from 2002 to 2008. Prior to 2002, vertical profiles matching ichthyoplankton tows in space and time were obtained from the EPIC data archive maintained by the Pacific Marine Environmental Laboratory (www.epic.noaa.gov/epic/). Wind speeds (m s-1) on each day of sampling were obtained from the National Data Buoy Center (www.ndbc.noaa.gov) for buoys (46021, 46782, and 46020) on the SEBS shelf (Figure 2.1). Measurements of wind speeds were not always available for each day; therefore, weekly averages were used in these cases (n = 336). Concentration (number m-3) of small-sized zooplankton (2 mm – 15 mm) were estimated from copepods collected in zooplankton tows made concurrently with ichthyoplankton sampling using a 20 cm bongo net with 153 µm mesh equipped with a flowmeter. Copepod concentrations were included only in the models for feeding stages (preflexion larvae, late larvae, and juveniles).

Broad-resolution analyses utilized published long-term anomalies and indices of environmental conditions derived from monthly or annual averages over the large spatial areas. Indices included temperature anomaly, wind mixing, zooplankton biomass anomaly, and a female spawning stock biomass anomaly. Spring sea surface temperature anomaly (SSTa, ºC), summer bottom temperature anomaly (BTa, ºC), and spring and summer wind mixing indices (m s-1) were obtained from the NCEP/NCAR Reanalysis project (Kalnay et al. 1996, http://www.esrl.noaa.gov/psd/data/reanalysis/). Indices for the month of May (spring) were used to examine effects on early stages (eggs, yolksac and preflexion larvae), while indices for June-July (summer) were used to examine effects on late larvae and juveniles. Zooplankton biomass anomalies (mg m-3) were derived from a time series of summer zooplankton over the SEBS middle shelf collected by the Hokkaido University (Hunt et al., 2002, 2008; A. Yamaguchi (Hokkaido University) and J. Napp, unpublished data). Size of the spawning population was represented by annual female spawning stock biomass anomalies (SSB, tons). SSB values were derived from a statistical age-structured population abundance model based on acoustic and bottom trawl surveys conducted by AFSC around Bogoslof Island and in the eastern Aleutian Islands (Ianelli et al., 2009). The Bogoslof / eastern Aleutians spawning area occurs within the bounds of the geographic region (slope, shelf) encompassed in this study, and can be viewed as a proxy for SSB over the shelf (Ianelli et al., 2009).

We used generalized additive models (GAMs) to quantify the influence of environmental factors on the abundance of pollock ELS (Hastie and Tibshirani, 1993; Wood, 2006). GAMs were chosen over other types of models, such as generalized linear models, because preliminary inspection of the data suggested that non-linear relationships were common between stages and environmental factors. Also,

surveys varied in spatial coverage, temporal coverage, and in the number of tows conducted in each year; GAMs provided a mechanism to account for this variation. Two types of models were used for each life history stage: fine-resolution and broad-resolution. Zero abundance was common in all data sets, particularly in later stages, requiring the use of the over-dispersed Poisson error distribution (Wood, 2006).

Fine- and broad-resolution covariates included in initial models for each response variable and each stage are shown in Table 2.2. Final model selection involved the choice of covariates and the level of smoothness for each covariate. For both model types, the cubic spline smoother s was used (Hamming, 1973). The level of smoothing was minimized to fewer than four knots to assure interpretability of results and to minimize effects of spatial autocorrelation in fine-resolution models (Stige et al., 2006; Heinenan et al., 2008). A final set of significant covariates was selected by backward stepwise elimination when p-values > 0.05 (Burnham and Anderson, 1998; Johnson and Omland, 2004). Interactions between covariates were also tested for significance and improving model fit within each model using the “by” function. The final model was determined based on the lowest GCV score. AIC scores also were compared for each model and were in agreement with selection based on GCV. Final- selection models were checked for temporal autocorrelation with the ACF function in R (ACF < 0.4). All models were coded and analyzed using the mgcv library (version 1.4-1; Wood, 2008) in R version 2.7.2 (R Development Team, 2008).

We tested for spatial autocorrelation in the fine-resolution models using Moran’s I (spdep library 0.4-56). Spatial autocorrelation increases the likelihood of making a Type I error with covariates that explain little deviance (labeled “questionable” covariates). We analyzed spatially autocorrelated models with and without location. If a questionable covariate still explained a significant portion of the variance without location in the model, it remained in the final-selection model. In all cases examined, questionable covariates remained in the model.

Results

Ichthyoplankton

Eggs and larvae were collected in all years. Juveniles were rare in the time-series and absent in 1988, 1991, 2006, and 2008. Over 1 million eggs, 40,000 larvae, and nearly 5,000 juveniles were collected over the course of this study. Eggs occurred in 63 % of bongo tows, yolksac larvae in 27 %, and preflexion larvae in 60 %. Late larvae occurred in 20 % and juveniles in 4 % of tows of all gear types. Most tows that caught ELS were collected over the continental shelf at depths less than 200 m, where a 58

majority of the sampling occurred. Eggs, yolksac larvae, and preflexion larvae were collected in three distinct areas: Bering Canyon, Unimak Island and the Alaska Peninsula, and the Pribilof Islands (2.2A- C). Late larvae were collected in two areas: Unimak Island and the Alaska Peninsula and the Pribilof Islands (Figure 2.2D). Juveniles were collected in the Pribilof Islands, primarily, and across the middle and inner domains (Figure 2.2E).

Adult Maturity

Gonad maturity stage of female fish > 35 cm FL from cruises conducted from March 1 to 10, 1992 until 2007, was compared to fork length and gear temperature anomaly using multiple linear regression. Maturity stage was not related to gear temperature anomaly or fork length, but showed an increasing trend with increasing temperature (Figure 2.3, model r2 = 0.255, p > 0.05).

Environmental conditions

Environmental factors were included in 66 – 100 % of the final models. Fine-resolution models explained more deviance in the abundance of each stage than did broad-resolution models (Tables 2.2 and 2.3). This difference can be attributed to the inclusion of year, location, and DOY in the fine-resolution models, rather than the contribution of environmental covariates. Broad-resolution covariates explained a greater portion of the deviance for all stages combined than did the fine-resolution environmental covariates (Table 2.4).

Fine-resolution

Fine-resolution models explained between 28.8 and 59.3 % of the deviance for all stages of pollock development (Table 2.3). Interactions between fine-resolution covariates were not significant (p > 0.05). Salinity did not significantly contribute to the explanation of deviance for any stage. The influence of fine-resolution covariates on ELS was related to ontogeny, with either stage-specific changes, or different responses by early stages compared to late stages (Figure 2.4). The partial effect of location was dependent on stage and area examined. The spatial effect on egg abundance was highest in the north and lowest along the shelf break. The spatial effect on yolksac larval abundance was highest in the inner domain and lowest in the outer domain. The spatial effect of location on preflexion larval abundance was highest in the outer domain and lowest in the northeast. The spatial effect on late larval abundance was highest in the inner domain and lowest along the shelf break. The spatial effect on juvenile abundance was highest in the north and along the peninsula. Temporal distribution varied across stages, consistent with the temporal evolution of a cohort (Figure 2.5). Positive effects of time on abundance were seen in eggs from DOY 100 – 150 (April – June), in yolksac larvae from DOY 100 – 175 (April – June), in preflexion larvae from DOY 100 – 200 (May – July), in late larvae from DOY 125 – 59

225 (May – August), and in juveniles from DOY 140 – 240 (May – Sept). A common trait for early stages was a low abundance of samples and high catch variability at the end of the summer, and high variability and few positive catches in the spring for juveniles.

Fine-resolution environmental covariates generally explained a smaller portion of the model variation than location or DOY. Temperatures associated with positive effects on abundance increased with ontogenetic stage (Figure 2.6). The partial effect of temperature on egg abundance was positive at low temperatures (-2 – 7 ºC) with most positive samples collected between 1 and 6 ºC, and catch variability highest above this range. The temperature effect was highest between -2 and 5 ºC for yolksac larvae, between 1 and 6 ºC for preflexion larvae, between 3 and 8 ºC for late larvae, and between 4 and 12 ºC for juveniles. The relationship with wind speed effect strength was positive for the early stages (eggs, yolksac larvae), but became negative for late larvae (Figure 2.7). Most samples containing each stage were collected at low wind speeds, and variability in effect strength increased above wind speeds of 8 m s-1. The relationship between copepod concentration and feeding stages (preflexion larvae, late larvae, and juveniles) was positive across the entire data range (Figure 2.8). The number of positive samples decreased above 4,000 copepods / m3. Variability in effect strength also increased above this threshold.

Broad-resolution

Broad-resolution environmental covariates and spawning biomass explained between 14.5 and 29.8 % of the deviance in the abundance of pollock ELS (Table 2.4). Interactions between broad- resolution covariates were not significant (p > 0.05). Spawning biomass influenced the abundance of eggs, yolksac larvae, and preflexion larvae (Figure 2.9). The partial effect of SSB was highest at the extremes for all three stages (either very low or high SSB). Variability was highest in 2008, the year of lowest SSB. The partial effect of temperature anomaly was highest at negative anomalies for eggs and yolksac larvae, at positive anomalies for preflexion larvae, and positive anomalies (except 2008) for late larvae and juveniles (Figure 2.10). Similar to fine-resolution temperature, catch variability was highest at the extreme values of temperature anomalies, even though sampling was distributed evenly across the temperature range. The partial effect of wind mixing was highest at high values for eggs, yolksac larvae, and juveniles (Figure 2.11). The effect of mixing on late larvae was highest at very low or high values. The partial effect of zooplankton biomass was highest in less than average and average years for preflexion larvae, late larvae, and juveniles (Figure 2.12).

Discussion

Spawning population density, environmental conditions, location, and time influence the abundance of Walleye Pollock (Gadus chalcogrammus) early life stages in the southeastern Bering Sea in conjunction with location and time. Pollock eggs are produced in the eastern Aleutian Islands and the 60

Pribilof Islands, while late larvae occur north of the Alaska Peninsula and in the Pribilof Islands. Juveniles occurred in the northern portion of the sampling area, suggesting dominant transport pathways north and east. The temporal effect suggests that juveniles reached peak abundance about 75 days after eggs and 25 days after late larvae. Location, time of year, and year accounted for most of the variation in abundance of pollock ELS relative to localized environmental covariates (Table 2.5). Spatiotemporal factors incorporate the aggregative and seasonal nature of spawning. As expected, SSB influenced egg, yolksac larval, and preflexion larval abundances. Within each suite of environmental conditions examined, temperature was the leading predictor of pollock ELS abundance, explaining more variation for more stages than either wind or zooplankton (Table 2.5).

Similar to pollock eggs in Shelikof Strait (Bacheler et al., 2009), SEBS eggs were positively associated with SSB, with the exception of 2008, a low SSB and cold year. Although we would predict few eggs in 2008 based on SSB, the potential for spawning and development to be delayed under cold conditions would retain eggs in the water column for longer periods of time (Blood, 2002). We did not find a significant relationship between maturity and temperature to confirm changes in onset of spawning. The positive trend between maturity stage and temperature suggests that onset of spawning plays a role in egg abundance, but more directed work is needed to test this hypothesis. SSB becomes less important in later stages, with no relationship between SSB and the abundance of the last two stages. Given cannibalism in pollock (Bailey, 1989; Wespestad et al., 2000; Duffy-Anderson et al., 2003), one might expect a negative relationship between SSB and late larval or juvenile abundance, but we found no support for this hypothesis within the current time series. One problem with the argument that cannibalism drives abundance of late stages is the assumption that the abundance of pollock larvae and juveniles can surpass the intra-specific predation capacity of the system. Late larvae and juveniles are several orders of magnitude less abundant than the eggs produced each year and are unlikely to survive in high enough numbers to surpass predation pressure without depleting their own food source (Duffy- Anderson et al., 2002).

We found that temperature and temperature anomaly with the greatest impact on stage-specific abundance increased with ontogeny. We predicted that the abundance of Bering Sea eggs would be positively related to temperature based on a previous study that found enhanced development and hatching with high temperature (Blood, 2002). Given the high numbers of eggs collected at low SSTa, it could be construed that production of pollock is higher at cooler than average temperatures (Hunt et al., 2002). We suggest that high production is not the case, but that high abundance of eggs in cold conditions results from one of two other possible scenarios. First, colder water temperatures slow rates of growth and development of ELS (Canino, 1994; Blood, 2002), so in cold conditions slow-developing eggs accumulate in the water column along with recently-spawned eggs. Second, pollock spawning

phenology is delayed at colder temperature, increasing temporal overlap between eggs and surveys. Spawning and recruitment among several gadid species has been shown to be temperature-dependent (O’Brien et al., 2000; Sundby, 2000; Poltev, 2008). Ichthyoplankton surveys start at about the same date and occur during a narrow temporal window relative to the timing of spawning each year. Delayed hatching under cool conditions coupled with slow growth and slow rates of starvation also would extend the yolksac stage, increase the temporal overlap of yolksac larvae with surveys, and reduce the overlap between surveys and feeding stages, and potentially increase mortality through predation or cannibalism.

Previous studies in the Gulf of Alaska found contrasting relationships between water temperature and abundance of pollock larvae. In the GOA, Doyle et al. (2009) found a negative relationship between pollock larval abundance and winter temperature, and no relationship between larval abundance and spring temperature. Larvae from that study were not sub-divided into length classes, which may have masked stage-specific patterns of abundance. Bailey et al. (1995) found reduced larval abundance in Shelikof Strait in an anomalously cold year relative to an average year. Although Chan et al. (2010) found a negative relationship between larval abundance and pre-spawning temperature due to late spawning and/or slow development, they found that higher than normal temperatures during the larval period in the GOA increased larval survival rates relative to those measured at lower temperatures. Bering Sea feeding-stage larvae and juveniles were associated with higher local and summer temperature conditions, consistent with either advanced growth rates or high survival.

Wind speed measurements provide a proxy for small-scale turbulent mixing (Oakey and Elliot, 1982), which can impact feeding and growth of pollock larvae (Megrey and Hinckley, 2001). The turbulence-encounter rate theory predicts optimal turbulence intensity for successful attacks by fish larvae (MacKenzie and Leggett, 1991), resulting in a dome-shaped relationship between turbulence or wind speed and feeding (Sundby and Fossum, 1990). We expected to find a similar curvilinear relationship between wind speeds and larval pollock abundance, assuming that growth and survival are dependent on prey capture ability alone. We did not find supporting evidence for this theory other than in the negative effect on abundance of late larvae at wind speeds greater than 12 ms-1. However, the number of samples within this range was low and overall wind speed did not have a large affect of abundances of feeding stages. We did find evidence of the potential for turbulence to bring early stages up from spawning depth into the range of our collection gear (shallower than 300 m) in the positive relationships between wind speed and mixing and egg and yolksac larval abundances, as would be predicted based on the prevalence of upwelling in the outer domain.

Chan et al. (2010) linked early larval survival in the GOA to enhanced transport from the spawning area when sea surface winds were high. Typically, eggs spawned offshore or near Unimak Island would be transported either north in the Bering Slope Current or east by the Aleutian North Slope 62

Current (Figure 2.1). Winds along the peninsula tend to move from east to west and strong winds from the east are associated with high northward flow along the slope from Unimak Pass (Kalnay et al., 1996). Wind speed data from buoys examined in this study were located relatively close to the peninsula, and although we were unable to examine wind direction, high wind speeds likely indicate enhanced northward flow along the slope due to Eckman transport. Wind-enhanced transport would result in a reduction in the larval population toward the northern end of our survey area over time. One potential result would be a reduction in later stages that would be either missed by our surveys or occur at the northern end only. Indeed, both late larvae and juveniles were most common toward the outer shelf and northern half of our survey area. In a parallel observation, Wespestad et al. (2000) suggested that wind-driven transport promoted the movement of age-1 pollock to the north and outer shelf.

As expected, feeding stages were positively related to copepod concentration as this indicates favorable feeding conditions. Copepod eggs and nauplii constitute a high proportion of the pollock larval diet and are expected to be available prey in the spring and summer, particularly to the critical first- feeding stage included in the preflexion size range (Theilacker et al., 1996; Napp et al. 2000). Surprisingly, feeding stages were not positively related to zooplankton biomass. Zooplankton biomass anomaly was based on zooplankton abundance over the middle shelf in the summer. The feeding stages were collected primarily in other areas, such as the peninsula and the Pribilof Islands. The spatial discrepancy between where feeding stages occurred and where zooplankton biomass was estimated may explain the lack of the expected relationship. Rather than dome-shaped relationships, zooplankton biomass suggested that poor zooplankton conditions support high abundances of pollock ELS. For preflexion larvae, there also is a temporal mis-match with zooplankton biomass anomaly. Preflexion larvae were most abundant in the spring and the anomaly was based on summer zooplankton estimates. In the case of prey availability, the fine-resolution copepod concentration may be a more appropriate predictor of abundance because it is estimated at the location and time at which the feeding stages are present and responding.

Conclusions

The influence of the immediate environment and overall conditions for the Bering Sea was stage- specific. Previously, environmental influence was examined at one stage or at one scale (e.g. Blood, 2002; Ciannelli et al., 2005; Doyle et al., 2009). With the exception of Mueter et al. (2006), very little work has specifically addressed the Bering Sea. In the SEBS, spawning stock biomass and maturation cycles can impact initial abundances and timing of hatching, but have no effect on developed larvae or juveniles. Following initial production, pollock ELS are most affected by temperature, followed by wind conditions, and prey production over the shelf. Spatial distributions of individual stages indicated that north- and eastward flowing currents are responsible for drift pathways. Of particular interest, stage- 63

specific changes in abundance occurred at different temperatures, with an oscillation between early stages at low temperatures and late stages at high temperatures. When available, fine-resolution environmental data are useful for modeling and predicting responses of individual stages to changing conditions. When fine-resolution data is unavailable or limited, patterns from broad-resolution indices paralleled those of their fine-scale counterparts, with the caveats that broad-resolution factors may not have been measured in similar locations or at similar times in which pollock ELS occur. Our results emphasize the importance of including ontogeny when predicting responses to environmental change, as ecological constraints imposed on a given stage do not necessarily act upon other stages in a similar manner.

Acknowledgements

Many thanks to the members of NOAA’s Ecosystems and Fisheries Oceanography Coordinated Investigations (EcoFOCI) who were involved in the collection and processing of the ichthyoplankton samples. L. Ciannelli assisted with statistical analyses and interpretation. A. Yamaguchi of the Hokkaido Univ. Graduate School of Fisheries provided zooplankton data for biomass anomalies. Earlier versions of this manuscript benefited from comments by Kevin Bailey, Jerry Berger, and three anonymous reviewers. This research was supported by the BSIERP program of the North Pacific Research Board (NPRB). This paper is EcoFOCI Contribution No. N753-RAOA-0 to NOAA's North Pacific Climate Regimes and Ecosystem Productivity research program, BEST-BSIERP Publication No. 35, and NPRB Publication No. 337, and BSIERP publication number 51.

Literature Cited

Bacheler, N.M., Bailey, K.M., Ciannelli, L., Bartolino, V., Chan, K.S., 2009. Density-dependent, landscape, and climate effects on spawning distribution of walleye pollock Theragra chalcogramma. Mar. Ecol. Prog. Ser. 391, 1-12.

Bacheler, N.M., Ciannelli, L., Bailey, K.M., Duffy-Anderson, J.T., 2010. Spatial and temporal patterns of walleye pollock spawning in the eastern Bering Sea inferred from egg and larval distributions. Fish. Oceanogr.19, 107-120. Bailey, K.M., 1989. Interaction between the vertical distribution of juvenile walleye pollock Theragra chalcogramma in the eastern Bering Sea and cannibalism. Mar. Ecol. Prog. Ser. 53, 205-213. Bailey, K.M., Macklin, S.A., 1994. Analysis of patterns in larval Walleye Pollock Theragra chalcogramma survival and wind mixing events in Shelikof Strait, Gulf of Alaska. Mar. Ecol. Prog. Ser. 113, 1-12. Bailey, K.M., Canino, M.F., Napp, J.M., Spring, S.M., Brown, A.L., 1995. Contrasting years of prey levels, feeding conditions and mortality of larval walleye pollock Theragra chalcogramma in the western Gulf of Alaska. Mar. Ecol. Prog. Ser. 119, 11-23. Bakkala, R.G., 1993. Structure and historical changes in the groundfish complex of the eastern Bering Sea. NOAA Tech. Rep. 114, 91 pp.

Bivand, R.S., Pebesma, E.J., Gómez-Rubio, V., 2008. Applied Spatial Data Analysis with R. New York, Springer. Blood, D.M., 2002. Low-temperature incubation of walleye pollock (Theragra chalcogramma) eggs from the southeastern Bering Sea shelf and Shelikof Strait, Gulf of Alaska. Deep-Sea Res. II 49, 6095– 6108.

Burnham, K.P., Anderson, D.R., 1998. Model Selection and Inference: A Practical Information- Theoretical Approach. Springer, New York.

Canino, M. 1994. Effects of temperature and food availability on growth and RNA/DNA ratios of walleye pollock Theragra chalcogramma (Pallas) eggs and larvae. J. Exp. Mar. Biol. Ecol. 17, 1-16.

Chan, K-S., Zhang, T., Bailey, K.M., 2010. Otolith biochronology reveals factors underlying dynamics of marine fish larvae. Mar. Ecol. Prog. Ser. 412, 1-10.

Ciannelli, L., Chan, K.S., Bailey, K.M., Stenseth, N.C., 2004. Nonadditive effects of the environment on the survival of a large marine fish population. Ecology 85, 3418-3427.

Ciannelli, L., Bailey, K.M., Chan, K.S., Belgrano, A., Stenseth, N.C., 2005. Climate change causing phase transitions of walleye pollock (Theragra chalcogramma) recruitment dynamics. Proc. R. Soc. Lond. B Biol. Sci. 272, 1735-1743.

Doyle, M.J., Picquelle, S.J., Mier, K.L., Spillane, M.C., Bond, N.A., 2009. Larval fish abundance and physical forcing in the Gulf of Alaska, 1981-2003. Prog. Oceanogr. 80, 163-187.

Duffy-Anderson, J.T., Bailey, K., Ciannelli L., 2002. Consequences of a superabundance of larval walleye pollock Theragra chalcogramma in the Gulf of Alaska in 1981. Mar. Ecol. Prog. Ser. 243, 179-190.

Duffy-Anderson, J.T., Ciannelli, L., Honkalehto, T. Bailey, K.M., Sogard, S.M., Springer, A., Buckley, T., 2003. Distribution of age-1 and age-2 walleye pollock in the Gulf of Alaska and Eastern Bering Sea: sources of variation and implications for higher trophic levels. In: Browman, H., A. Skiftesvik (eds.), The big fish bang: Proceedings from the 26th annual larval fish conference, 381- 394.

Duffy-Anderson, J.T., Bailey, K., Ciannelli L., Cury, P., Belgrano, A., Stenseth N.C., 2005. Phase transitions in marine fish recruitment processes. Ecological Complexity 2, 205-218.

FAO, 2007. 15 December 2009. Yearbooks of Fisheries Statistics. [http://www.fao.org/fishery/statistics/en].

Fukuchi, M., 1976. Some aspects of bioenergetics of walleye pollock Theragra chalcogramma (Pallas) at early life stages. Ph. D. thesis, Univ. of Hokkaido, Hokkaido, Japan.

Hamming, R.W., 1973. Numerical Methods for Scientists and Engineers. McGraw-Hill, New York.

Hastie, T., Tibshirani, R., 1993. Varying-coefficient models. J. R. Stat. Soc. B 55, 757-796.

Heinenän, S., Rönkä, M., von Numers, M., 2008. Modeling the occurrence and abundance of a colonial species, the arctic tern Sterna paradisaea in the archipelago of SW Finland. Ecography 31, 601- 611.

Hunt, G.L., Stabeno, P.J., Walters, G., Sinclair, E., Brodeur, R.D., Napp, J.M., Bond, N.A., 2002. Climate change and control of the southeastern Bering Sea pelagic ecosystem. Deep-Sea Res. II 49, 5821- 5853.

Hunt, G.L., Stabeno, P.J., Strom, S., Napp, J.M., 2008. Patterns of spatial and temporal variation in the marine ecosystem of the southeastern Bering Sea, with special reference to the Pribilof Domain. Deep-Sea Res. II 55, 1919-1944.

Ianelli, J.N., Barbeaux, S., Honkalehto, T., Kotwicki, S., Aydin, K., Williamson, N., 2009. Assessment of the Walleye Pollock stock in the Eastern Bering Sea. In: Stock assessment and fishery evaluation report for the groundfish resources of the Bering Sea/Aleutian Islands regions. North Pac. Fish. Mgmt. Council, Anchorage, AK, section 1, 49-148.

Ichthyoplankton Information System. 23 July 2009. National Oceanic and Atmospheric Administration. [http://access.afsc.noaa.gov/ichthyo/index.cfm].

Johnson, J.B., Omland, K.S., 2004. Model selection in ecology and evolution. Trends Ecol. Evol. 19, 101- 108. Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L., Iredell, M., Saha, S., White, G., Woollen, J., Zhu, Y., Leetmaa, A., Reynolds, B., Chelliah, M., Ebisuzaki, W., Higgins, W., Janowiak, J., Mo, K., Ropelewski, C., Wang, J., Jenne, R., Joseph, D., 1996. The NCEP/NCAR 40-year reanalysis project. Bull. Am. Meteorological Soc. 77, 437–471. Logerwell, E., Duffy-Anderson, J.T., Wilson, M.T., McKelvey, D. 2010. The physical and biological processes resulting in habitat partitioning by two potential competitors, juvenile pollock and capelin, in the Gulf of Alaska. Fish. Oceanogr. 19, 262-278.

MacKenzie, B.R., Leggett, W.C., 1991. Quantifying the contribution of small-scale turbulence to the encounter rates between larval fish and their zooplankton prey: Effects of wind and tide. Mar. Ecol. Prog. Ser. 73, 149-160.

Matarese, A.C., Blood, D.M., Picquelle, S.J., Benson, J.L., 2003. Atlas of abundance and distribution patterns of ichthyoplankton from the northeast Pacific Ocean and Bering Sea ecosystems based on research conducted by the Alaska Fisheries Science Center (1972-1996). NOAA Prof. Pap. NMFS No. 1, NOAA, Seattle, WA.

Megrey, B.A., Hinckley, S., 2001. Effect of turbulence on feeding of larval fishes: a sensitivity analysis using an individual-based model. ICES J. Mar. Sci. 58, 1015-1029. Mueter, F.J., Ladd, C., Palmer, M.C., Norcross, B.L., 2006. Bottom-up and top-down controls of walleye pollock (Theragra chalcogramma) on the eastern Bering Sea shelf. Prog. Oceanogr. 68, 152-183.

Oakey, N.S., Elliott, J.A., 1982. Dissipation within the surface mixed layer. J. Phys. Oceanogr. 12, 171- 185.

O’Brien, C.M., Fox, C.J., Planque, B., Casey, J., 2000. Climate variability and North Sea cod. Nature 404, 142.

Olla, B.L. Davis, M.W., Ryer, C.H., Sogard, S.M., 1996. Behavioural determinants of distribution and survival in early stages of walleye pollock, Theragra chalcogramma: a synthesis of experimental studies. Fish. Oceanogr. 5, 167-178.

Poltev, Y.N., 2008. Some issues related to reproduction of Pacific cod Gadus macrocephalus in waters of the eastern coast of the North Kuril Islands and the southern extremity of Kamchatka. J. Ichthyol. 48: 345-355.

Porter, S.M., Ciannelli, L., Hillgruber, N., Bailey, K.M., Chan, K-S., Canino, M.F., Haldorson, L.J., 2005. Environmental factors influencing larval walleye pollock Theragra chalcogramma feeding in Alaskan waters. Mar. Ecol. Prog. Ser. 302, 207-217.

Potts, J.M., Elith, J., 2006. Comparing species abundance models. Ecol. Modelling 199, 153-163.

R Development Core Team, 2008. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna. Shima, M., Bailey, K.M., 1994. Comparative analysis of ichthyoplankton sampling gear for early life stages of walleye pollock (Theragra chalcogramma). Fish. Oceanogr. 3, 50-59.

Stige, L.C., Ottersen, G., Brander, K., Chan, K.S., Stenseth, N.C., 2006. Cod and climate: effect of the North Atlantic Oscillation on recruitment in the North Atlantic. Mar. Ecol. Prog. Ser. 325, 227– 241.

Stahl, J.P., Kruse, G.H., 2008. Spatial and temporal variability in size at maturity of walleye pollock in the eastern Bering Sea. Trans. Am. Fish. Soc. 137, 1543-1557.

Sundby, S., 2000. Recruitment of Atlantic cod stocks in relation to temperature and advection of copepod populations. Sarsia 85, 277–298.

Sundby, S., Fossum, P., 1990. Feeding conditions of Arcto-Norwegian cod larvae compared with the Rothschild–Osborn theory on small-scale turbulence and plankton contact rates. J. Plankton Res. 12, 1153–1162.

Theilacker, G.H., Bailey, K.M., Canino, M.F., Porter, S.M., 1996. Variations in larval walleye pollock feeding and condition: a synthesis. Fish. Oceanogr. 5, 112-123.

Walline, P.D., 1985. Growth of larval Walleye Pollock related to domains within the SE Bering Sea. Mar. Ecol. Prog. Ser. 21, 197-203.

Wespestad, V. G., Fritz, L. W., Ingraham, W. J., Megrey, B. A., 2000. On relationships between cannibalism, climate variability, physical transport, and recruitment success of Bering Sea walleye pollock (Theragra chalcogramma). ICES J. Mar. Sci. 57, 268–274.

Wiebe, P.H., Burt, K.H., Boyd, S.H., Morton, A.W., 1976. A multiple opening/closing net and environmental sensing system for sampling zooplankton. J. Mar. Res. 34, 313-326.

Williams, K., 2007. Evaluation of the macroscopic staging method for determining maturity of female Walleye Pollock Theragra chalcogramma in Shelikof Strait, Alaska. Alaska Fish. Res. Bull. 12, 252-263

Wilson, M.T., 2009. Ecology of small neritic fishes in the western Gulf of Alaska. I. Geographic distribution in relation to prey abundance and the physical environment. Mar. Ecol. Prog. Ser. 392, 223-237.

Wood, S.N., 2006. Generalized Additive Models: An Introduction with R. Chapman & Hall/CRC, Boca Raton, FL.

Wood, S.N., 2008. Fast stable direct fitting and smoothness selection for generalized additive models. J. R. Stat. Soc. Ser. B Stat. Method. 70, 495-518.

Zarauz, L., Irigoien, X., Urtizberea, A., Gonzalez, M., 2007. Mapping plankton distribution in the Bay of Biscay during three consecutive spring surveys. Mar. Ecol. Prog. Ser. 345, 27-39.

Table 2.1. Summary of Gadus chalcogrammus early life stage surveys in the southeastern Bering Sea.

Number of cruises, gear types used, number of tows (n), and sampling date range. Tows examined were limited to bongo nets (BON), MOCNESS (MOC), and Tucker trawls (TUCK).

Year Cruises Gear n Dates

1988 2 BON 100 03/17 – 04/25

TUCK 4

1991 2 BON 71 03/11 – 05/08

1992 2 BON 35 04/16 – 07/14

MOC 7

1994 4 BON 101 04/16 – 09/21

TUCK 8

MOC 9

1995 5 BON 256 02/22 – 09/24

TUCK 46

MOC 5

1996 5 BON 16 03/06 – 09/15

TUCK 9

MOC 19

1997 4 BON 125 04/16 – 09/17

TUCK 22

MOC 2

1998 2 BON 17 04/07 – 09/14

1999 4 BON 108 04/14 – 09/14

MOC 6

2000 5 BON 40 02/17 – 09/22

2002 5 BON 35 05/13 – 10/06

TUCK 3

2003 4 BON 125 03/04 – 09/27

MOC 9

2004 5 BON 54 07/29 – 10/02

2005 7 BON 121 03/04 – 10/06

MOC 20

2006 4 BON 189 05/09 – 09/22

MOC 12

2007 7 BON 161 04/11 – 10/08

2008 6 BON 116 02/18 – 09/26

MOC 13

Table 2.2. Variables considered for inclusion in generalized additive models. Fine-resolution covariates were measured in association with ichthyoplankton tows and broad-resolution covariates were monthly or annual counterparts to the fine-resolution measurements integrated over the eastern Bering Sea.

Fine- Broad- resolution resolution

Year ---

Location ---

Day of Year ---

Surface Water Water Temperature Temperature Anomaly

Wind Speed Wind Mixing

Zooplankton Zooplankton Biomass Concentration Anomaly

Salinity ---

Spawning Stock Biomass --- Anomaly

Table 2.3. Generalized additive model results for fine-resolution (ichthyoplankton-associated) covariates. Percent of the total model deviance explained attributable to each covariate included for each stage and final model output. Deviance attributed to individual covariates was determined from comparison of full- and reduced-models. Covariates that were not considered in a given model are noted by n/a and those that were not significant in the final model are noted by n. sig.

Egg Yolksac Preflexion Late Juvenile

Year 3.4 5.0 7.4 10.0 9.2

Location 2.5 1.3 1.1 1.4 3.2

Day of Year 4.9 5.1 4.8 7.9 0.7

Temperature 0.6 0.9 1.9 1.7 1.1

Wind Speed 1.0 1.0 n. sig. 0.9 n. sig.

Zooplankton n/a n/a 0.8 0.2 0.2

Salinity 1.3 n. sig. n. sig. n. sig. n. sig.

r2 0.543 0.229 0.403 0.551 0.154

% Deviance 51.2 28.8 44.8 59.3 41.3

# Tows 1393 1393 1393 1479 1479

Table 2.4. Generalized additive model results for broad-resolution (annual and monthly) covariates, spawning stock biomass (SSB), spring or summer temperature, wind mixing index, and zooplankton biomass. Percent of the total model deviance explained attributable to each covariate included for each stage and final model output. Deviance attributed to individual covariates was determined from comparison of full- and reduced-models. Covariates that were not considered in a given model are noted by n/a and those that were not significant in the final model are noted by n. sig.

Egg Yolksac Preflexion Late Juvenile

SSB 7.1 5.0 0.7 n. sig. n. sig.

Temperature 0.9 2.8 8.2 16.0 0.8

Mixing 2.6 0.4 n. sig. 5.4 8.8

Zooplankton n/a n/a 4.9 3.3 3.8

r2 0.226 0.149 0.213 0.272 0.081

% Deviance 16.3 14.5 17.3 29.8 19.0

# Tows 1671 1671 1671 1902 1902

# Years 17 17 17 17 17

Table 2.5. Covariate ranks derived from GAMs. Ranks are based on the total possible models that covariate was considered for, the number of models each covariate was included in, and the weighted deviance explained by that covariate. Weighted deviance is the deviance explained by a covariate in each model weighted by the deviance explained by that model summed across model types and stages.

Rank Covariate Possible Included Weighted

Models Models Deviance

1 Temperature Anomaly 5 5 0.26

2 Zooplankton Biomass 3 3 0.20

3 Wind Mixing 5 4 0.17

4 Year 5 5 0.16

5 SSB 5 3 0.16

6 DOY 5 5 0.11

7 Location 5 5 0.04

8 Temperature 5 5 0.03

9 Wind Speed 5 4 0.02

10 Copepods 3 2 0.01

Figure 2.1. The southeastern Bering Sea showing depth contours, major currents, locations of wind buoys, and ichthyoplankton sampling intensity (tows per 10x10 km grid cell). The box outlined by the broken lines indicates samples included in the current analyses and the box outlined by the solid lines represents locations of maturity surveys in the eastern Aleutian Islands.

Figure 2.2. Walleye Pollock early life stages in the southeastern Bering Sea. Relative abundances of A) eggs, B) yolksac larvae, C) preflexion larvae, D) late larvae, and E) juveniles. The size of the bubbles is scaled to the largest catch within each stage.

Figure 2.3. Walleye Pollock gonad maturity. Mean maturity stage and gear temperature anomaly in each year (1992 – 2007) from March 1 – 10 in the Aleutian Islands area outlined in Figure 2.1.

Figure 2.4. Partial effect of spatial location. Warm colors indicate predicted increases and cool colors indicate predicted decreases in abundance of Walleye Pollock A) eggs, B) yolksac larvae, C) preflexion larvae, and D) late larvae and in presence of E) juveniles. Contours and numbers represent amplitude and direction of location effect strength, positive or negative.

Figure 2.5. Partial effect of day-of-year. Positive values indicate predicted increases and negative values indicate predicted decreases in abundance of Walleye Pollock A) eggs, B) yolksac larvae, C) preflexion larvae, D) late larvae, and E) juveniles. Shaded areas are 95% confidence intervals, tick marks on the x- axis indicate sampling intensity, and k is the number of knots.

Figure 2.6. Partial effect of surface water temperature. Positive values indicate predicted increases and negative values indicate predicted decreases in abundance of Walleye Pollock A) eggs, B) yolksac larvae, C) preflexion larvae, and D) late larvae and presence of E) juveniles. Shaded areas are 95% confidence intervals, tick marks on the x-axis indicate sampling intensity, and k is the number of knots.

Figure 2.7. Partial effect of wind speed. Positive values indicate predicted increases and negative values indicate predicted decreases in abundance of Walleye Pollock A) eggs, B) yolksac larvae, and C) late larvae. Shaded areas are 95% confidence intervals, tick marks on the x-axis indicate sampling intensity, and k is the number of knots.

Figure 2.8. Partial effect of copepod concentration. Positive values indicate predicted increases and negative values indicate predicted decreases in abundance of Walleye Pollock A) preflexion larvae, B) late larvae, and C) juveniles. Shaded areas are 95% confidence intervals, tick marks on the x-axis indicate sampling intensity, and k is the number of knots.

Figure 2.9. Partial effect of spawning stock biomass anomaly. Positive values indicate predicted increases and negative values indicate predicted decreases in abundance of Walleye Pollock A) eggs, B) yolksac larvae, and C) preflexion. Shaded areas are 95% confidence intervals, tick marks on the x-axis indicate sampling intensity, and k is the number of knots.

Figure 2.10. Partial effect of temperature anomaly. Positive values indicate predicted increases and negative values indicate predicted decreases in abundance of Walleye Pollock A) eggs, B) yolksac larvae, C) preflexion larvae, D) late larvae, and E) juveniles. Shaded areas are 95% confidence intervals, tick marks on the x-axis indicate sampling intensity, and k is the number of knots.

Figure 2.11. Partial effect of wind mixing. Positive values indicate predicted increases and negative values indicate predicted decreases in abundance of Walleye Pollock A) eggs, B) yolksac larvae, C) late larvae, and D) juveniles. Shaded areas are 95% confidence intervals, tick marks on the x-axis indicate sampling intensity, and k is the number of knots.

Figure 2.12. Partial effect of zooplankton biomass anomaly. Positive values indicate predicted increases and negative values indicate predicted decreases in abundance of Walleye Pollock A) preflexion larvae, B) late larvae, and C) juveniles. Shaded areas are 95% confidence intervals, tick marks on the x-axis indicate sampling intensity, and k is the number of knots.

Chapter 3: Alternating Temperature States Influence Walleye Pollock (Gadus chalcogrammus) Early Life Stages in the Southeastern Bering Sea

T. I. Smart1, 3,*, J. T. Duffy-Anderson2, J. K. Horne1

1School of Aquatic and Fishery Sciences, University of Washington, Seattle, Washington 98195, USA

2Recruitment Assessment and Conservation Engineering Division, Recruitment Processes Program, Alaska Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration, Seattle, Washington 98115, USA

3Present address: Marine Resources Research Institute, South Carolina Department of Natural Resources, Charleston, South Carolina 29422, USA

Citation: Smart, T.I., Duffy-Anderson, J.T., Horne, J.K. 2012. Alternating climate states influence Walleye Pollock life stages in the southeastern Bering Sea. Mar. Ecol. Prog. Ser. 455: 257-267.

Abstract

Water temperatures in the southeastern Bering Sea influence the density of Walleye Pollock Gadus chalcogrammus early life stages, potentially influencing spatial distributions and the phenology of reproduction and development. We quantified stage-specific changes in spatial and temporal distributions under cold- and warm-water conditions using generalized additive models. Analyses showed that Walleye Pollock egg and yolksac larval spatial distributions are unaffected by temperature, suggesting that spawning locations are stable. Preflexion larvae, late larvae, and juveniles shift onto the shelf under warm conditions, similar to spatial shifts observed in distributions of sub-adults and adults. Temporal distributions were used to address the hypothesis that timing of the density peak at each stage is delayed under cold conditions. Differences in the timing of density peaks supported the hypothesis that the timing of spawning, hatching, larval development, and juvenile transition are temperature-dependent. The current analysis represents the best support available for the importance of temperature to Walleye Pollock in determining early life stage development and population trends in the eastern Bering Sea. Our data indicate that future changes in water temperatures could influence the early life stages of an ecologically dominant member of the Bering Sea community by changing phenology and habitat use in the first several months of life.

Introduction

In the eastern Bering Sea, interannual variation in the extent of sea ice over the continental shelf determines water temperatures during the spring and summer when many fish species reproduce (Hunt & Stabeno 2002). Gadus chalcogrammus (Walleye Pollock, hereafter pollock) is one of the most common spring and summer shelf-spawning species in the Bering Sea ecosystem. Juvenile and adult populations change in abundance and distribution in response to changes in water temperatures. Juvenile pollock (< 130 mm fork length) densities in surface trawls were higher and spatial distributions were broader during two recent warm summers relative to two recent cold summers (Moss et al. 2009). The authors hypothesized that warm conditions promote growth and survival in this population. Sub-adult (< 30 cm fork length) and adult biomasses and spatial distributions have been shown to correlate with summer temperature during the pollock post-spawning feeding season (Kotwicki et al. 2005). Biomass increased with summer temperatures and the feeding migration progressed farther north and inshore (east) from spawning grounds as temperature increased. Kotwicki et al. (2005) hypothesized that warmer water temperatures advance the post-spawning feeding migration by advancing the timing of spawning. They did not address the alternate hypothesis that spawning distribution changes in response to temperature conditions.

Colder-than-average years are characterized by late sea-ice retreat from the shelf in the spring, water column temperatures < 2 ºC over the shelf during spring and summer, and abundant large-bodied zooplankton (Hunt et al. 2002, Coyle et al. 2011). Warmer-than-average years are characterized by early ice retreat, continental shelf water temperatures > 2 ºC during spring and summer, and small-bodied zooplankton. Sea-ice conditions also determine the extent and location of a pool of cold bottom water in the summer over the middle shelf (50 – 100 m depth, Stabeno et al., 2012a). Pollock distributions generally are restricted to areas > 2 ºC, such as those found outside the cold pool (Wyllie-Echeverria & Wooster 1998). Therefore, in warm years, a larger area of thermally-suitable habitat is available to pollock over the continental shelf, particularly over the middle shelf.

Smart et al. (2012) identified temperature in spring and summer as a leading environmental factor influencing stage-specific densities of pollock early life stages (ELS) in the southeastern Bering Sea (SEBS). Lower than average temperature years have high densities of eggs and newly hatched yolksac larvae, regardless of the size of the adult spawning population. Average and higher-than-average temperature years have high densities of feeding larvae (pre- and postflexion) and early juveniles (< 65 mm standard length). One hypothesis to explain the oscillation between early and developed stages is high egg production under cold conditions. Spawning stock biomass in recent cold years has been low, while egg densities have been high (Ianelli et al. 2009, Smart et al., 2012). An alternate hypothesis is that the oscillation between early and late stages results from shifts in spawning and development phenology. 90

Maturity stage of SEBS pollock females in late winter tends to be advanced in warm relative to cold years (Smart et al., 2012), although the impact of temperature on spawning has not been rigorously tested. Temperature influences time to hatching in laboratory-reared Bering Sea pollock eggs with eggs at 2 ºC, typical of cold conditions, requiring almost a week longer to hatch than eggs raised at 3.8 ºC, typical of warm conditions (Blood 2002). In the Gulf of Alaska, lower than average temperatures contribute to high mortality rates of field-observed larvae (Bailey et al. 1995).

The current study has two objectives: 1) to determine whether pollock ELS undergo spatial shifts in response to changing temperature conditions and 2) to test whether temperature affects the phenology of developmental events. Temperature can affect the annual distribution and development of pollock ELS through the location of spawning, post-spawning transport, the timing of spawning, and rates of development.

Materials and Methods

Field Surveys

The SEBS is bordered to the east by the Alaska mainland, to the south by the Alaska Peninsula and eastern Aleutian Islands, to the west by the Aleutian Basin, and to the north by a change in vertical structure of the water column and hydrography near 60º N latitude (Stabeno et al. 2012, b, Figure 3.1). The shelf in the SEBS is very broad (~500 nautical miles). Northwest flow is driven by the Bering Slope Current and flow through Aleutian passes (Napp et al. 2000). East-west flow is driven by the Aleutian North Slope Current. The SEBS shelf can be divided into 3 bathymetric domains, each with its own characteristic hydrography (Coachman 1986). The inner shelf (or coastal domain, < 50 m) is weakly stratified and influenced by freshwater run-off; the middle shelf (50 – 100 m) is strongly stratified and home to the cold pool in summer; and the outer shelf (100 – 200 m) is an area of intermittent upwelling in the spring and summer, high productivity, and stratification (Hunt et al. 2002).

Depth-integrated densities (number 10 m-2) of pollock ELS were determined from a time series of ichthyoplankton surveys conducted by the NOAA Alaska Fisheries Science Center EcoFOCI program (Seattle, WA, cf. Matarese et al. 2003). Surveys varied in spatial and temporal coverage and in the number of tows conducted and gears used. To account for these variations, we limited the tows considered for analysis to include pollock spawning areas, areas of potential transport, and the spawning season. Data were limited to tows conducted south of 60 N latitude and east of the Aleutian basin, where a majority of spawning occurs (Figure 3.1). Pollock spawning in the SEBS occurs from February through November (Bacheler et al. 2010), but since sampling at the start or end of the spawning season is rare, we

limited catch records to between March and September (Figure 3.2). Fifteen years in the time series met these spatial and temporal criteria (see Table 3.1 for details).

Gear types examined included obliquely towed bongo nets, a 1-m-2 multiple opening and closing net and environmental sensing system (MOCNESS), and 1-m Tucker trawls, all equipped with 333 or 505 µm mesh. Differences in selectivity between bongo, MOCNESS, and Tucker gears are generally minor and are appropriate for the length ranges of pollock examined in this study (Wiebe et al. 1976, Shima & Bailey 1994). Volume filtered by each gear type was determined by calibrated flow meters mounted in the mouth of each net. Catches were divided into 5 developmental stages: eggs, yolksac larvae (< 4.4 mm standard length, SL), preflexion larvae (4.5 – 9.9 mm SL), late larvae (10.0 – 24.5 mm SL), and early juveniles (25 – 64.5 mm SL). Divisions were based on standard length and developmental attributes outlined in the Ichthyoplankton Information System (http://access.afsc.noaa.gov/ichthyo/index.cfm). The upper size limit for early juveniles was selected based on the reduced catchability of larger fish with the current gear types.

Monthly mean sea surface temperatures in May are a good indicator of overall conditions throughout the pollock spawning period, correlating with both sea-ice state and summer temperatures (Moss et al. 2009, Stabeno et al. 2012a). Sea surface temperature anomalies were derived from means of monthly sea surface temperatures in May averaged over the area 54.3 – 60.0 ºN and 161.2 – 172.5 ºW (NCEP/NCAR Reanalysis, Kalnay et al. 1996). The anomaly used in the current analysis is the deviation from the mean value (2.11 ºC) for the survey years from 1988 – 2009, normalized by the standard deviation (0.82 ºC).

Spatial Distributions

Stage-specific spatial distributions were determined for each ELS using variable-coefficient generalize additive models (VCGAMs). VCGAMs allow the formulation of a nonparametric, nonlinear regression model in which the effect of predictor variable can be determined for a specific location (Hastie & Tibshirani 1993, Bacheler et al. 2009). In our case, we examined if density at each ELS was expected to increase or decrease with an increase in annual sea surface temperature anomaly for a given location. VCGAMs quantified changes in density for each ELS using the following model structure:

Ci = offset(volume) + a + s(SSB) + s(DOY) + s(location) + SSTa + s(location) · SSTa + ε (1)

where Ci is the count at stage i in each tow standardized (offset) by the volume of water filtered in each tow (volume, m3), a is the model intercept, SSB is the annual female spawning stock biomass (tons) included to account for annual changes in spawning effort, DOY is the day of year of each tow, location is a combined term of the latitude and longitude for each tow, SSTa is the annual sea surface temperature 92

anomaly, and ε is the model error. SSB was based on summer surveys for adult pollock over the eastern Bering Sea shelf (Ianelli et al. 2009). In preliminary examination of the data, there was no relationship between temperature and SSB (T. Smart, unpubl.). The negative binomial error distribution was used because of overdispersion caused by a high number of zeroes in the data set (Dupont 2002, Maunder and Punt 2004). The negative binomial error distribution also produced the best model fit using the Akaike Information Criterion (AIC) compared to Poisson, gamma, or Gaussian error distributions. The identity link function was used in all VCGAMs. The level of smoothing was minimized to between 1 and 4 knots to minimize the effects of spatial autocorrelation (Stige et al. 2006, Heinenän et al. 2008). All models were coded and analyzed using the mgcv library (version 1.4-1, Wood 2006) in R (version 2.7.1, R Development Core Team 2008). VCGAM predictions were plotted for each 0.1° by 0.1° degree cell for which data were available in our study area.

Temporal Distributions

Generalized additive models (GAMs) were used to describe the non-linear temporal distributions for each stage to estimate temporal shifts between temperature categories. Years of positive SSTa were categorized as warm years (n = 6) and years of negative SSTa were categorized as cold years (n = 9). This method was chosen over the Kolmogorov-Smirnoff test, as the Kolmogorov-Smirnoff test cannot separate changes in peaks in the distribution from changes in the shapes of the distribution. GAMs allow a visual comparison of the shapes and peaks in density among collection dates. The use of GAMs also allowed us to account for the spatial component of our surveys, as tow locations often varied among months and years. GAMs quantified anomalies in density for each ELS using:

Ci = offset(volume) + a + s(SSB) + factor(temperature) + s(location) + s(DOY) + ε (2)

where Ci is the count of stage i in each tow standardized (offset) by the volume of water filtered in each tow (volume, m3), a is the model intercept, SSB is the annual female spawning stock biomass (tons), temperature is the temperature category (cold or warm), location is a combined term of the latitude and longitude for each tow, DOY is the day of year of each tow, and ε is the model error. The negative binomial error distribution produced the best model fit based on AIC compared to the Poisson, gamma, or Gaussian error distributions. The identity link function was used in all models. The level of smoothing was minimized to between 1 and 4 knots to minimize the effects of spatial autocorrelation (Stige et al. 2006, Heinenän et al. 2008). When the effect of temperature category on density anomaly was significant (α = 0.05), we derived the partial effect of DOY on the density anomaly within each temperature category to define the temporal distribution. The switch from negative to positive density anomalies indicated the DOY when each stage first appeared in surveys. The switch back to negative anomalies indicated the DOY when each stage of development was complete (i.e. the season for each stage). The maximum 93

positive density anomaly predicted from DOY in GAMs indicated when each stage reached its maximum density within each temperature category.

Results

Spatial Distributions

Egg densities were high across much of the study area in cold years, with highest catches near Unimak Pass and the Pribilof Islands (Figure 3.3A). Tows with high egg density were common off the shelf in 1988 and 1990 and sporadic over the shelf in the rest of the warm years. High-density catches of yolksac larvae were common over the shelf and around the Pribilof Islands in cold years and either over the basin or over the shelf in warm years (Figure 3.3B). Preflexion larval densities were highest over the shelf and near the Pribilof Islands in cold years and over the shelf and basin in warm years (Figure 3.3C). Highest densities of late larvae were found near the Pribilof Islands in cold years and over the shelf in warm years (Figure 3.3D). Very few tows collected juveniles in high numbers in cold years, but these tows occurred mostly near the Pribilof Islands (Figure 3.3E). In warm years, high-density juvenile catches occurred over the shelf and near the Pribilof Islands.

For all stages, annual temperature anomaly had a significant impact on location-specific density (Table 3.2). Eggs and yolksac larvae were predicted to decrease in density as temperature increased in all study area locations (Figure 3.4A, B). Preflexion larvae were predicted to increase in density as temperature increased in the central portion of our study area (shelf break, outer shelf, middle shelf) and decrease in density over the basin and in the inner shelf (Figure 3.4C). Late larval density decreased with warming over the basin and to the north and increased over the shelf and to the south with warming (Figure 3.4D). Warming did not impact juvenile density over the shelf, but was predicted to decrease densities along the shelf break, over the basin, and to the north (Figure 3.4E).

Temporal Distributions

Temperature category affected the density of all pollock ELS in GAM analyses (Table 3.3). Egg density anomalies peaked prior to DOY 150 (< June 1) in cold years and before DOY 140 (< May 20) in warm years (Figure 3.5A). The yolksac larval season occurred between DOY 100 and 190 (Apr 10 – July 10) in cold years and prior to DOY 175 (< June 25) in warm years (Figure 3.5B). Among preflexion larvae, density anomalies peaked between DOY 100 and 195 (Apr 10 – July 15) in cold years and prior to DOY 185 (< July 5) in warm years (Figure 3.5C). The late larval season occurred between DOY 125 and 240 (May 5 – Aug 25) in cold years and between DOY 100 and 215 (Apr 10 – Aug 1) in warm years (Figure 3.5D). For juveniles, density anomalies peaked after DOY 130 (> May 10) in cold years and after DOY 120 (> May 1) in warm years (Figure 3.5E). For the stages that we can predict start or end dates of their 94

seasons, all seasons started or finished earlier in warm years than in cold years. However, as sampling does not typically occur prior to March, we are unable to predict start dates for eggs, yolksac larvae in warm years, and preflexion larvae in warm years. As sampling rarely occurs after October, we are unable to predict end dates for juveniles. With these limitations, we cannot at present determine whether or not the span of each stage is shortened in warm years as hypothesized.

Temporal shifts were estimated for each stage using the density maximum predicted by DOY. Differences in maxima indicated delays in peak temporal distributions of all ELS in cold compared to warm years (Table 3.4). These differences suggested a 40-d delay in the egg stage maxima (spawning and hatching), a 45-d delay in the yolksac larval stage maxima (hatching and absorption of yolk reserves), a 20-d delay in the preflexion stage maxima (feeding commences), a 30-d delay in the late larval stage maxima (flexion of the notochord), and a 25-d delay in the juvenile stage maxima (completion of the juvenile transition and growth to sizes above which our collection gear are avoided).

Discussion

Water temperatures in the southeastern Bering Sea influenced Walleye Pollock (Gadus chalcogrammus) ELS in several ways. Spatial distributions of eggs and yolksac larvae were not affected by temperature, while feeding stages (preflexion larvae, late larvae, and juveniles) were affected. Variable-coefficient GAMs predicted decreases in the density of eggs and yolksac larvae with warming, suggesting that there is no discernible shift in spawning locations. Variable-coefficient GAMs predicted that densities of preflexion larvae and late larvae would increase over the shelf, while densities of preflexion larvae, late larvae, and juveniles would decrease offshore, suggesting a movement of the population onto the shelf. A likely mechanism to explain these patterns is that spawning location remains relatively constant, but larval transport varies with temperature. Temporal distributions predicted by GAMs indicated a delay in the timing of reproductive and ontogenetic events in cold temperatures. Earlier spawning, hatching, and onset of feeding could explain the low densities of eggs and yolksac larvae observed in warm years. We were unable to determine rates of development because of uncertainty at the beginning and end of the sampling season. Our analysis suggests that in warm years this population reaches the juvenile stage earlier and would have more time for feeding and growth prior to the oncoming winter.

Spatial Distribution Shifts

Bacheler et al. (2012) found indications of shifts in pollock spawning location across the shelf using an earlier time series and variable-coefficient GAMs. In support of this pattern, Anchovy (Engraulis encrasicolus), Sardine (Sardina pilchardus), and Common Sole (Solea solea) have all been reported to shift spawning locations in response to environmental factors (Bellier et al. 2007, Eastwood et 95

al. 2001). We were unable to discern the same pattern for pollock eggs. An alternate explanation is that SEBS pollock spawning location does not, but subsequent transport of ELS does. The current study supplements the findings of Bacheler et al. (2012) with spatial shifts in subsequent ELS, which can isolate differential transport from shifts in spawning location. In cold years, surface currents are generally westward and faster relative to warm years (Stabeno et al., 2012 a), which would transport larvae to the outer shelf or offshore if they were spawned over the shelf. In warm years, larvae are more likely to be retained over the shelf. Although we found only a small increase in juvenile density over the inner shelf with warming, we did observe a large decrease in density offshore. Together, these results suggest that juveniles move onshore in warm years although densities may be low and distributions are widespread. We attribute distributions of larvae and early juveniles to shifts in subsequent transport rather than shifts in spawning location (see also Duffy-Anderson et al. 2006).

Spatial shifts observed in pollock larvae and juveniles determine their use of two SEBS shelf domains which vary in productivity and zooplankton communities. The shelf tends to have a greater concentration of smaller copepods (more appropriate prey items for small larvae), while larger copepods (appropriate prey items for large larvae) are generally found in spring over the deeper outer shelf (Cooney & Coyle 1982, Vidal & Smith 1986, Baier & Napp 2003). Zooplankton present in the SEBS during cold years are larger, energetically richer, and are concentrated over the outer shelf and shelf break (Coyle et al. 2011). Cold temperatures favor a larger-bodied prey for late larvae and juveniles and are predicted to increase juvenile survival if they can exploit this resource. Warm water temperatures with a smaller-sized prey field are predicted to benefit the growth of preflexion larvae, which are gape-limited (Nakatani 1988). In warm years, preflexion larvae are more likely to be in the vicinity of small prey over the shelf. In cold years, late larvae and juveniles are more likely to be in the vicinity of large prey over the shelf break. This hypothesis is supported by recent evidence that the few juvenile pollock collected in recent cold years in the SEBS were energetically richer than those sampled in recent warm years (Hunt et al., 2011). The spatial distribution shifts observed in the current study combined with the observation of spatially-distinct zooplankton communities supports the hypothesis that prey fields in cold years should support higher densities and energetically richer pollock larvae and juveniles.

Phenological Shifts

Pollock ELS oscillate between high densities of early stages (eggs and yolksac larvae) in cold years and high densities of developed stages (preflexion and late larvae and juveniles) in warm years (Smart et al. 2012). One explanation for high numbers of early-staged individuals in cold years is a combination of delayed spawning and slower rates of development. Egg density temporal distribution is dependent on both the timing of spawning and the time required to develop to the hatching stage. Based on observed egg temporal distributions, either the timing of spawning events or development to hatching 96

is temperature-dependent in the SEBS. There are currently no data available on the temperature- dependence of gametogenesis in pollock. As a comparison, maturation rates and spawning in Atlantic Cod (Gadus morhua) are temperature-dependent, and completion of maturation is advanced by several weeks in warm temperatures (Beaugrand et al. 2003, Yoneda & Wright 2005). Extrapolating from laboratory data, Blood (2002) estimated that egg incubation period could be delayed by 13 d during a cold (1997) compared to a warm (1998) year. In the field, the estimated delay in peak egg density is 40 d, suggesting that both timing of spawning and development could be delayed, accounting for the interval between temperature categories in the current study.

Comparing temperature categories, we found evidence of temperature-dependence in the timing of hatching, onset of feeding, flexion of the notochord, juvenile transition, or some combination of temperature-dependence among any of these developmental events. Delayed hatching increases the likelihood of collecting eggs during our primary sampling period (May through September). Delayed onset of feeding retains larvae in the non-feeding yolksac stage. Jung et al. (2006) estimated that hatching was delayed by 12 – 17 d in 1976 (cold) compared to 1977 (warm) based on the staging of field-caught eggs. Under warm conditions, early-spawned eggs should rapidly develop and hatch. The newly hatched larvae could consume their yolk reserves in perhaps as little as a week (Walline 1985, Porter & Theilacker 1999). Temperature-dependent development rates in larval stages have consequences for matching with prey production cycles and for growth-dependent mortality (Gallego & Heath 1997). Currently, we cannot estimate start and end points and by extension the time span of development, for all stages. Therefore, the timing of events must stand as a proxy supporting the hypothesis that development is delayed in cold temperatures relative to warm water temperatures.

In summary, our data demonstrated that Walleye Pollock ELS in the southeastern Bering Sea responded to changes in oceanographic conditions (i.e. water temperature and flows) by shifting density distributions in time and, in some cases, space. The temperature-associated oscillation between high egg and yolksac larval density and high density of more developed larvae and juveniles can be explained by phenological shifts in spawning and development. Distributions of pollock feeding larvae and juveniles also spatially shifted between two habitats, the continental shelf and the shelf break or offshore, which has the potential to affect access to appropriate sized prey. We hypothesize that the lack of spatial shifts observed in pollock eggs and early larvae indicate that adults do not adjust the location of spawning in response to temperature. We further hypothesize that spatial shifts observed in larval and juvenile densities reflect differences in transport between cold and warm years. Comparable trends among life stages suggest shared habitats. As temperatures continue to change in the North Pacific, identification of spatiotemporal shifts in habitat use will be crucial to understanding short- and long-term population trends in this ecologically and economically important species.

Acknowledgements

Thanks to the members of NOAA’s Ecosystems and Fisheries Oceanography Coordinated Investigations (EcoFOCI) who were involved in the collection and processing of the ichthyoplankton samples. L. Ciannelli (Oregon State University) assisted with statistical analyses and interpretation. This research was supported by the Bering Sea Integrated Ecosystem Research program (BSIERP) of the North Pacific Research Board and the North Pacific Climate Regimes and Ecosystem Productivity (NPCREP) program of the National Oceanographic and Atmospheric Administration. This paper is EcoFOCI Contribution No. N754 – RAOA – 0, BEST-BSIERP Publication No. 35. We appreciate comments by Jeffrey Napp, Ann Matarese, and three anonymous reviewers. The findings and conclusions in the paper are those of the authors and do not necessarily represent the views of the National Marine Fisheries Service.

Literature Cited

Bacheler NM, Ciannelli L, Bailey KM, Duffy-Anderson JT (2010) Spatial and temporal patterns of Walleye Pollock spawning in the eastern Bering Sea inferred from egg and larval distributions. Fish Oceanogr 19:107-120 Bacheler NM, Ciannelli L, Bailey K, Bartolino V (2012) Do Walleye Pollock exhibit flexibility in where or when they spawn based on variability in water temperature? Deep-Sea Res II 65-70: 208-216 Baier CT, Napp JM (2003) Climate-induced variability in Calanus marshallae populations. J Plank Res 25:771-782 Bailey KM, Canino MF, Napp JM, Spring SM, Brown AL (1995) Contrasting years of prey levels, feeding conditions and mortality of larval Walleye Pollock Theragra chalcogramma in the western Gulf of Alaska. Mar Ecol Prog Ser 119:11-23 Beaugrand G, Brander KM, Souissi JALS, Reid PC (2003) Plankton effect on cod recruitment in the North Sea. Nature 426:661-664

Bellier E, Planque B, Petitgas P (2007) Historical fluctuations in spawning location of Anchovy (Engraulis encrasicolus) and Sardine (Sardina pilchardus) in the Bay of Biscay during 1967-73 and 2000-2004. Fish Oceanogr 16:1-15

Blood DM (2002) Low-temperature incubation of Walleye Pollock (Theragra chalcogramma) eggs from the southeastern Bering Sea shelf and Shelikof Strait, Gulf of Alaska. Deep-Sea Res II 49:6095- 6108

Coachman LK (1986) Circulation, water masses, and fluxes on the southeastern Bering Sea shelf. Continental Shelf Res 5(1-2):23-108

Cooney RT, Coyle KO (1982) Trophic implications of cross-shelf copepod distributions in the southeastern Bering Sea. Mar. Biol. 70:187-196 Coyle KO, Eisner LB, Mueter FJ, Pinchuk AI, Janout MA, Cieciel KD, Farley EV, Andrews AG (2011) Climate change in the southeastern Bering Sea: impacts on pollock stocks and implications for the oscillating control hypothesis. Fish Oceanogr 20:139-156 Duffy-Anderson JT, Busby MS, Mier KL, Deliyanides CM, Stabeno PJ (2006) Spatial and temporal patterns in summer ichthyoplankton assemblages on the eastern Bering Sea shelf 1996-2000. Fish Oceanogr 15:80-94

Dupont WD (2002) Statistical Modelling for Biomedical Researchers: A Simple Introduction to the Analysis of Complex Data. Cambridge Press, New York, 522 pp.

Eastwood PD, Meaden GJ, Grioche A (2001) Modeling spatial variations in spawning habitat suitability for the sole Solea solea using regression quantiles and GIS procedures. Mar Ecol Prog Ser 224:251-266

Gallego A, Heath M (1997) The effect of growth-dependent mortality, external environment and internal dynamics of larval fish growth: An individual-based modeling approach. J Fish Biol 51:121-134

Hastie T, Tibshirani R (1993) Varying-coefficient models. J R Stat Soc B 55:757-796

Heinenän S, Rönkä M, von Numers M (2008) Modelling the occurrence and abundance of a colonial species, the arctic tern Sterna paradisaea in the archipelago of SW Finland. Ecography 31:601- 611

Hunt GL, Stabeno PJ (2002) Climate change and the control of energy flow in the southeastern Bering Sea. Prog Oceanogr 55:5-22

Hunt GL, Stabeno PJ, Walters G, Sinclair E, Brodeur RD, Napp JM, Bond NA (2002) Climate change and control of the southeastern Bering Sea pelagic ecosystem. Deep-Sea Res II 49:5821-5853

Hunt GL, Coyle KO, Eisner LB, Farley E, Heintz R, Mueter F, Napp JM, Overland JE, Ressler PH, Salo S, Stabeno PJ (2001). Climate impacts on eastern Bering Sea food webs: A synthesis of new data and an assessment of the Oscillating Control Hypothesis. ICES J Mar Sci 68: 1230-1243

Ianelli JN, Barbeaux S, Honkalehto T, Kotwicki S, Aydin K, Williamson N (2009). Assessment of the Walleye Pollock stock in the Eastern Bering Sea. In: Stock assessment and fishery evaluation report for the groundfish resources of the Bering Sea/Aleutian Islands regions. North Pac. Fish. Mgmt. Council, Anchorage, AK, section 1, 49-148

Ichthyoplankton Information System. National Oceanic and Atmospheric Administration. Accessed 23 July. http://access.afsc.noaa.gov/ichthyo/index.cfm

Jung KM, Kang S, Kim S, Kendall AW (2006) Ecological characteristics of Walleye Pollock eggs and larvae in the southeastern Bering Sea during the late 1970s. J Oceanogr 62:859-871

Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L, Iredell M, Saha S, White G, Woollen J, Zhu Y, Leetmaa A, Reynolds B, Chelliah M, Ebisuzaki W, Higgins W, Janowiak J, Mo K, Ropelewski C, Wang J, Jenne R, Joseph D (1996) The NCEP/NCAR 40-year reanalysis project. Bull Am Meteorolog Soc 77:437–471

100

Kotwicki S, Buckley TW, Honkalehto T, Walters G (2005) Variation in the distribution of Walleye Pollock (Theragra chalcogramma) with temperature and implications for seasonal migration. Fish Bull 103:574-587 Matarese AC, Blood DM, Picquelle SJ, Benson JL (2003) Atlas of abundance and distribution patterns of ichthyoplankton from the northeast pacific Ocean and Bering Sea ecosystems based on research conducted by the Alaska Fisheries Science Center (1972-1996). NOAA Prof Pap NMFS No. 1, NOAA, Seattle, WA

Moss JH, Farley EV, Feldmann AM, Ianelli JN (2009) Spatial distribution, energetics status, and food habits of eastern Bering Sea age-0 Walleye Pollock. Trans Am Fish Soc 138:497-505

Maunder MN, Punt AE (2004) Standardizing catch and effort data: a review of recent approaches. Fish Res 70:141-159 Nakatani T (1988) Reproductive strategy of Walleye Pollock, Theragra chalcogramma. Jap Soc Fish Oceanogr 25:48 Napp JM, Kendall AW, Shumacher JD (2000) A synthesis of biological and physical processes affecting the feeding environment of larval Walleye Pollock (Theragra chalcogramma) in the eastern Bering Sea. Fish Oceanogr 9:147-162

Porter SM, Theilacker GH (1999) The development of the digestive tract and eye in larval Walleye Pollock, Theragra chalcogramma. Fish Bull 97:722-729

R Development Core Team (2008) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna

Shima M, Bailey KM (1994) Comparative analysis of ichthyoplankton sampling gear for early life stages of Walleye Pollock (Theragra chalcogramma). Fish Oceanogr 3:50-59

Smart TI, Duffy-Anderson JT, Horne JK, Farley EV, Wilson CD, Napp JM (2012) Influence of environmental conditions on walleye pollock eggs, larvae, and juveniles in the southeastern Bering Sea. Deep-Sea Res II

Stabeno PJ, Moore S, Napp J, Sigler M, Zerbini A (2012a). Comparison of warm and cold years on the southeastern Bering Sea shelf. Deep-Sea Res 65-70: 31.45

Stabeno PJ, Farley E, Kachel N, Moore S, Mordy C, Napp JM, Overland JE, Pinchuk AI, Sigler M (2012b). A comparison of the physics, chemistry, and biology of the northern and southern shelf of the eastern Bering Sea. Deep-Sea Res 65-70: 14-30

101

Stige LC, Ottersen G, Brander K, Chan KS, Stenseth NC (2006) Cod and climate: effect of the North Atlantic Oscillation on recruitment in the North Atlantic. Mar Ecol Prog Ser 325:227–241

Vidal J, Smith SL (1986) Biomass, growth, and development of populations of herbivorous zooplankton in the southeastern Bering Sea during spring. Deep-Sea Res 33: 523-556

Walline PD (1985) Growth of larval Walleye Pollock related to domains within the SE Bering Sea. Mar Ecol Prog Ser 21:197-203

Wiebe PH, Burt KH, Boyd SH, Morton AW (1976) A multiple opening/closing net and environmental sensing system for sampling zooplankton. J Mar Res 34:313-326

Wood SN (2006) Generalized additive models: an introduction with R. Chapman & Hall/CRC, Boca Raton, FL, 391 pp

Wyllie-Echeverria T, Wooster WS (1998) Year-to-year variations in Bering Sea ice cover and some consequences for fish distributions. Fish Oceanogr 7:159-170

Yoneda M, Wright PJ (2005) Effects of varying temperature and food availability on growth and reproduction in first-time spawning female Atlantic Cod. J Fish Biol 67:1225-1241

102

Table 3.1. Gadus chalcogrammus. Walleye Pollock survey years included in analysis as either cold (negative sea surface temperature anomaly, SSTa) or warm (positive SSTa). Cruise start dates, end dates, mean day of year (DOY), and the number of tows collected are included. Growth and mortality rates were derived from sequential cruises in similar geographic areas within each year

Temperature Year SSTa Cruise Start Cruise End Mean DOY # Tows

Category

Cold 1994 -0.16 4/15 4/30 113 101

9/5 9/13 253 19

1995 -0.52 4/17 5/1 114 152

5/4 5/18 131 139

9/11 9/18 258 14

1997 -0.18 4/16 4/25 111 32

5/4 5/13 128 31

7/1 7/13 189 86

1999 -1.09 4/14 4/18 138 43

5/15 5/20 138 15

7/13 7/23 199 56

2000 -0.05 5/7 5/11 131 14

6/22 7/2 178 6

7/28 7/31 212 8

9/18 9/22 265 12

2006 -1.01 5/9 5/18 134 96

6/22 6/25 174 21

9/11 9/22 259 84

2007 -1.09 4/11 5/11 114 35

4/24 4/25 114 6

5/8 5/18 133 89

6/8 6/22 167 12

103

7/1 7/11 184 11

7/25 7/29 208 8

2008 -1.86 5/13 5/21 138 66

6/3 6/16 162 14

6/19 6/28 175 5

7/4 7/15 192 43

2009 -1.95 4/27 5/3 120 12

5/8 5/18 133 89

6/5 6/17 163 9

6/20 7/6 177 10

7/12 7/30 201 10

9/25 9/30 274 60

Warm 1988 0.16 3/17 4/4 84 64

4/11 4/26 110 40

1991 0.01 3/11 3/15 72 19

4/14 5/8 115 52

1996 0.88 4/23 4/25 115 4

5/15 5/16 136 5

5/19 5/20 140 5

7/21 8/4 210 12

9/7 9/15 256 19

2002 0.71 5/13 5/21 136 77

8/2 8/8 218 6

8/14 8/28 234 8

8/20 9/21 241 55

9/8 9/30 264 8

2003 1.73 3/4 3/6 64 13

5/18 5/24 141 69

7/21 7/24 204 5

104

9/9 9/27 256 28

2005 1.21 3/3 3/5 63 4

5/10 5/22 134 110

5/16 5/27 140 5

7/15 7/18 197 9

9/22 9/25 266 8

105

Table 3.2. Gadus chalcogrammus. Summary of variable coefficient generalized additive models used to develop spatial distributions of each Walleye Pollock early life stage within each temperature category. Number of tows containing each stage (positive tows), number of tows lacking each stage (zero tows), % Deviance explained by the model, and estimated degrees of freedom (edf) for spawning stock biomass (SSB), day of year (DOY), location, sea surface temperature anomaly (SSTa), and the spatially-explicit SSTa term. Asterisks denote significance at the following alpha levels: *0.05; **0.01; ***0.001.

Stage Egg Yolksac Preflexion Late Juvenile

Positive tows 1032 404 862 345 74

Zero tows 1003 1631 1173 1690 1971

% Deviance 47.9 50.9 53.0 65.8 75.1

SSB 2.995*** 2.995*** 2.986*** 2.999*** 2.978***

DOY 2.998*** 2.998*** 2.997*** 2.995*** 2.938***

Location 2.965*** 2.993*** 2.999*** 2.001*** 2.971***

SSTa 1* 1*** 1* 1** 1*

Location · SSTa 29.937*** 28.930*** 29.980*** 29.963*** 29.996***

106

Table 3.3. Gadus chalcogrammus. Summary of generalized additive models used to develop temporal distributions of each Walleye Pollock early life stage within each temperature category. Number of tows is the same as in Table 3.3. % Deviance explained by the model and estimated degrees of freedom (edf) for model covariates. Asterisks denote significance at the following alpha levels: *0.05; **0.01; ***0.001.

Stage Egg Yolksac Preflexion Late Juvenile

% Deviance 41.5 51.4 51.1 60.1 54.3

SSB 8.980** 8.981*** 8.980** 8.964*** 8.953**

Location 2.997*** 2.991*** 2.997*** 2.939*** 2.987***

Temperature 1*** 1*** 1*** 1*** 1**

DOY 2.997*** 2.997*** 2.997*** 2.995*** 2.995***

107

Table 3.4. Gadus chalcogrammus. Temporal distributions of Walleye Pollock early life stages derived from generalized additive models. Day of year (DOY) when each stage appears in surveys prior to the peak (start), DOY at the peak in temporal distributions (peak), and DOY when each stage appears for the final time in surveys after the peak (end). NA denotes when sampling did not occur early or late enough to determine these values.

Cold Warm

Stage Start Peak End Start Peak End

Eggs NA 105 150 NA 65 130

Yolksac Larvae 100 145 180 NA 100 170

Preflexion Larvae 100 145 195 NA 125 185

Late Larvae 125 180 240 100 150 210

Juveniles 135 225 NA 120 200 NA

108

Figure 3.1. Gadus chalcogrammus. Dominant eastern Bering Sea currents include the Aleutian North Slope Current (ANSC), the Bering Slope Current (BSC), and inflow through passes driven by the Alaska Coastal Current (ACC). The eastern Bering Sea is divided into 3 shelf domains based on hydrography: outer, middle, and inner. Right: cumulative sampling in 10 cold (upper) and 7 warm (lower) years from 1988 – 2009.

109

Figure 3.2. Gadus chalcogrammus. Monthly sampling frequency for Walleye Pollock in the southeastern Bering Sea. Percent of tows conducted within each month in a given year for early life stages during the Walleye Pollock spawning season in (A) cold and (B) warm years from 1988 – 2009.

110

Figure 3.3. Gadus chalcogrammus. Walleye Pollock early life stages in the southeastern Bering Sea. Relative abundances of A) eggs, B) yolksac larvae, C) preflexion larvae, D) late larvae, and E) juveniles within cold (left) and warm (right) temperature categories. The size of the bubbles is scaled to the largest catch within each stage.

111

Figure 3.4. Gadus chalcogrammus. Spatially-explicit impacts of temperature on density of Walleye Pollock A) eggs, B) yolksac larvae, C) preflexion larvae, D) late larvae, and E) juveniles. Red bubbles indicate an increase and blue bubbles indicate a decrease in density as temperature increases in the southeastern Bering Sea. Bubble size is relative to the size of the change in density.

112

Figure 3.5. Gadus chalcogrammus. Temporal distributions of Walleye Pollock A) eggs, B) yolksac larvae, C) preflexion larvae, D) late larvae, and E) juveniles within cold (left) and warm (right) temperature categories derived from the relationship between count and day of year in generalized additive models. Shaded areas are 95% confidence intervals and tick marks on the x-axis indicate

113

sampling intensity on each day of year (DOY). The vertical line denotes the estimated DOY of the density maxima.

114

Chapter 4. Vertical distributions of the early life stages of Walleye Pollock (Gadus chalcogrammus) in the southeastern Bering Sea

Smart, T.I.1,4, Siddon, E.C.2, Duffy-Anderson, J.T.3

1School of Aquatic and Fishery Sciences, University of Washington, Seattle, WA, 98195 USA 2Fisheries Division, School of Fisheries and Ocean Sciences, University of Alaska Fairbanks, Juneau, AK, 99801 USA 3RACE Division, Recruitment Processes Program, Alaska Fisheries Science Center, NOAA Fisheries, Seattle, WA, 98115 USA 4Current address: Marine Resources Research Institute, Charleston, South Carolina 29422 USA

Citation: Smart, T., Duffy-Anderson, J.T., and Siddon, E. 2013. Vertical distribution of early life stages of walleye pollcok and implications for transport and connectivity. Deep-Sea Research II: Topical Studies in Oceanography. 94: 201-201.

115

Abstract

The present study examines vertical distributions of the early life stages of Walleye Pollock (Gadus chalcogrammus) in the Southeastern Bering Sea to assess ontogenetic and dielvertical migration in relation to development and habitat. Walleye Pollock demonstrated a decrease in the depth of occurrence following hatching, indicating an ontogenetic change in vertical distribution. Eggs occurred deepest in the water column and early juveniles occurred shallowest. Vertical distributions were related to the date of collection, water column depth, and thermocline depth. Non-feedingstages (eggs and yolksac larvae, <4.5 mm standard length [SL]) did not exhibit diel vertical migration. Feeding larvae exhibited diel vertical migration, although patterns varied between two feeding stages. Preflexion stage larvae (4.5–9.9 mm SL) were concentrated between 10 and 20 m during the day and deeper at night. Postflexion stage larvae (flexion and postflexion, 10.0–24.5 mm SL) underwent regular diel migrations (0–20 m, night; 10– 40 m, day). Low sample sizes limited our ability to assess diel vertical migration in early juveniles, but this stage tends to occur in the upper 20 m of the water column, regardless of time of day. These results suggest that vertical distributions and diel migration potentially are driven by prey availability at sufficient light levels for preflexion larvae to feed and a trade-off between prey access and predation risk for postflexion larvae. Vertical distributions of eggs and preflexion larvae varied with habitat examined (on the continental shelf versus over the continental slope). Vertical distributions of Walleye Pollock eggs, yolksac larvae, and preflexion larvae in the Bering Sea are different from distributions in other ecosystems, which can impact transport and modeling efforts.

116

Introduction

Information on vertical distributions of early life history stages of fishes is critical for accurate modeling of larval transport to nursery habitats as transport can differ depending on vertical position in the water column (Tanaka 1991, Stenevik et al. 2003, Fiksen et al. 2007, Miller 2007, Kristiansen et al. 2009). Ontogenetic vertical migration (OVM) is a pattern in which vertical distribution changes with stage of development. Typically, OVM involves a shoaling in the depth of occurrence as eggs and larvae develop, which allows larvae to exploit high food concentrations and fast currents in surface waters (Fortier and Leggett 1983, Norcross and Shaw 1984, Hare and Govoni 2005). Often OVM is accompanied by increasing complexity of vertical behaviors, such as responsiveness to changes in light intensity (Heath et al. 1988, Hare and Govoni 2005). Diel vertical migration (DVM) is a behavioral trend in which depth of occurrence changes with time of day and light intensity. DVM typically is exhibited by feeding stages. Larval fish are visual predators, consuming a variety of zooplankton, and as such will orient themselves in the water column to overlap with prey vertical distribution and with light levels sufficient to facilitate feeding (Heath et al. 1988, Porter et al. 2005). Larval fish also are preyed upon by a variety of visual predators and may move down to darker depths during the daytime to avoid predation (Hunter and Sanchez 1976, Yamashita et al. 1985). Regular DVM is the pattern in which larvae move deeper in the water column during the day to avoid visual predators and shallower at night to feed upon zooplankton in surface waters (Kerfoot 1985, Ohman 1990). Reverse DVM is the pattern in which larvae migrate to the surface during the day and migrate to depth at night, often as a response to tidal currents or the presence of non-visual predators.

The continental shelf and shelf break areas of the southeastern Bering Sea (SEBS) are important spawning and nursery grounds for commercially valuable pelagic and demersal fishes, such as Walleye Pollock (Gadus chalcogrammus, Matarese et al. 2003). The shelf in the SEBS is very broad (~300 nmi), providing a wide area of shallow habitat for developing larvae and juveniles. Walleye Pollock spawn in several areas near islands in the SEBS, along the Alaska Peninsula, and in deep-water canyons along the continental slope (Hinckley 1987, Bacheler et al. 2010). Coupled with the locations of spawning, the dominant currents in the SEBS can deliver Walleye Pollock early life stages to several different habitats over the continental shelf and slope, each with unique hydrographic and biological characteristics (Coachman 1986, Stabeno et al. 66 2001). A coastal domain (< 50 m water depth) surrounds all islands and the Alaska mainland and peninsula. The coastal domain is well-mixed. The middle shelf domain (50 – 100 m water depth) is strongly stratified in summer and characterized by a pool of cold (< 2 ºC) bottom water in summer that is detrimental to the development of larvae (Napp et al. 2000). The coastal and middle domains are habitat for several species of copepod whose naupliar stages are preferred prey items

117

for larval Walleye Pollock (Hillgruber et al. 1995, Coyle et al. 2011). The outer shelf domain (100 – 200 m water depth) is an area of intermittent upwelling in spring and summer, high productivity, strong stratification, and abundant potential predators (Springer et al. 1996, Hunt et al. 2002, Coyle et al. 2011). The slope domain (> 200 m water depth) adjoins the Aleutian Basin and is predicted to provide lower growth potential to larvae due to lower prey availability and temperature (Napp et al. 2000). Among the hydrographic domains, Walleye Pollock larvae are exposed to depth strata with distinct flow regimes, thermal regimes, predation pressures, and prey availability.

Although Walleye Pollock spawn in a variety of habitats and water depths in the North Pacific, ontogenetic and diel vertical migrations are known for relatively few areas and we have limited knowledge of how habitat interacts with vertical distributions. Vertical distributions have been studied most extensively in the Gulf of Alaska (GOA). In the GOA, Walleye Pollock early life stages undergo OVM. Eggs occur between 150 and 200 m depth (Kendall et al. 1994), yolksac larvae rise gradually to the surface where feeding larvae are found above the thermocline (Davis and Olla 1994), and juveniles are primarily pelagic (Brodeur and Wilson 1996, Laurel et al. 2007). By comparison, in the shallow-water Funka Bay, Japan, eggs and larvae are found at depths less than 50 m with no indication of OVM (Kamba 1977, Kendall et al. 1987). In the laboratory, regular DVM is initiated once GOA larvae reach 6 mm standard length (SL) and are feeding (Olla and Davis 1990 A, Davis and Olla 1994). Kendall et al. (1994) found limited DVM in feeding larvae in Shelikof Strait, while larvae in Auke Bay, Alaska, responded to patches of copepod nauplii with regular DVM (Haldorsen et al. 1993). Juvenile Walleye Pollock undergo regular DVM in the western GOA (Brodeur and Rugen 1994, Olla and Davis 1990 B).

Our knowledge of vertical distributions of SEBS Walleye Pollock early life stages is limited relative to the GOA (but see Walline 1981, Hillgruber et al. 1995, and Brase 1996). Egg distribution over the basin is much deeper (400-500 m) than over the shelf (≤ 100 m) (Serobaba 1974, Nishiyama et al. 1986). Juvenile Walleye Pollock near the Pribilof Islands undergo regular DVM (similar to the western GOA), presumably in response to prey movement (Schabetsberger et al. 2000). There is currently no assessment of OVM or DVM in SEBS Walleye Pollock early life stages other than this example. The preferred prey of SEBS larvae, copepod eggs and nauplii, can be found in surface waters and are unlikely to undergo DVM themselves. Fish larvae in the SEBS are exposed to a variety of visual predators (e.g. Walleye Pollock, Pacific Cod) that could drive DVM (Ohman 1990). Fundamental differences in Walleye Pollock early life ecology exist between the GOA and SEBS (Bailey 1989, Kendall et al. 1994, Duffy- Anderson et al. 2003), and it is likely that there are differences in vertical position as well. These differences may be critical, especially in efforts to model transport and habitat use in the SEBS.

118

The purposes of the present study were: (1) to describe the stage-specific vertical distribution patterns of the early life stages of Walleye Pollock in the SEBS and (2) to examine potential drivers of differences in distribution such as physical forcing, ontogeny, and trade-offs for survival. To this end, we examined the vertical distribution among life stages for evidence of ontogenetic vertical migration, we compared vertical distributions of each stage among time of day to test for diel vertical migration, and we assessed the interactions between the physical water column and habitat and vertical distribution patterns.

Materials and Methods

Study Area

The SEBS is bordered to the east by Alaska, to the south by the Alaska Peninsula and eastern Aleutian Islands, to the west by the Aleutian Basin, and to the north by Nunivak Island. Walleye Pollock spawning areas included in this study were north of Unimak Island (Bering Canyon), the Alaska Peninsula and near the Pribilof Islands and Pribilof Canyon (Figure 4.1). Shelf domains include the coastal domain (< 50 m), the middle domain (50 – 100 m), the outer shelf domain (100 – 200 m), and the slope domain (depths > 200 m, Figure 4.1, Coachman 1986, Stabeno et al. 2001).

Sampling for Ichthyoplankton Vertical Distributions

Vertical distributions were determined from depth-specific densities (number per 1000 m3 of water sampled) derived from sampling with a 1-m-2 Multiple Opening and Closing Net and Environmental Sensing System (MOCNESS, 333 or 505 μm mesh equipped with a flow meter to estimate volume filtered by each net). The larger mesh size was used when large phytoplankton blooms were present and clogged the smaller mesh. In a comparative study, Wiebe et al. (1976) found no difference in catchability for Walleye Pollock early life stages between the two mesh sizes so samples were pooled across mesh size. Walleye Pollock early life stages were collected in MOCNESS tows in 13 years between 1992 and 2009 (Table 4.1) by NOAA’s Fisheries-Oceanography Coordinated Investigations (FOCI) program. A tow is defined as the unit of sampling with multiple nets within a given tow. Concentration records were divided into stages based on standard length and developmental attributes outlined in the Ichthyoplankton Information System (http://access.afsc.noaa.gov/ichthyo/index.cfm). Catches were divided into 5 life stages: eggs, yolksac larvae (< 4.5 mm SL), preflexion larvae (4.5 - 9.9 mm SL), postflexion larvae (10.0 - 24.9 mm SL), and early juveniles (25.0 - 64.9 mm SL). MOCNESS tows are not ideal for sampling the very surface of the water column, which can contain buoyant eggs. To supplement the information derived from MOCNESS tows, egg densities from Sameoto neuston tows (30 x 50 cm mouth opening, 333 or 505 μm mesh), which fished the upper 25 cm of the water column, also were examined in 2003 and 2005-2009.

119

Temperature and density data were derived from vertical CTD (Sea Bird 19 or 25, Sea- Bird Electronics, Bellevue, Washington, USA) profiles collected concurrently with ichthyoplankton samples from 2002 to 2009. Prior to 2002, CTD vertical profiles matching ichthyoplankton tows in space and time were obtained from the EPIC data archive maintained by the Pacific Marine Environmental Laboratory (www.epic.noaa.gov/epic/). Thermo- and pycnocline depths were extracted from CTD casts for each MOCNESS tow. The thermo- and pycnocline depths were defined as the depth at which the greatest rate of change in temperature or density occurred (Coyle and Pinchuk 2005). Only thermocline depth was examined because pycnocline depth was correlated with thermocline depth. Years were assigned to either cold or warm temperature categories based on May sea surface temperature anomalies (see Smart et al. 2012, for details) to test for differences in vertical distribution with prevailing annual conditions in the study area.

Ontogenetic Vertical Migration

Net depth intervals of MOCNESS tows were inconsistent across years and cruises, ranging from 10-m to 100-m intervals. Tows with net depth intervals greater than 20-m in the upper 50 m of the water column were removed from analysis. For the remaining tows, we converted depth-specific concentrations of each stage to catch-weighted-mean depths (CWMD) to provide a comprehensive view of vertical distribution. CWMD was calculated by the following equation:

n n CWMD =  xidi /  xi i1 i1 where xi is the concentration of each stage at each depth interval i, di is the midpoint of each depth interval, and n is the total number of depth intervals in each tow. CWMDs were compared by a two-way analysis of covariance (ANCOVA) with stage and temperature category as fixed factors and day of year (DOY), thermocline depth (m), and bottom depth (m) as covariates (Sokal and Rohlf 1995). The concentration of eggs in neuston tows was compared by one-way ANCOVA with temperature category as a fixed factor and thermocline depth and bottom depth as covariates. Neuston samples were collected primarily during two weeks in May, so we did not examine DOY as a covariate.

Diel Vertical Migration

Tows that sampled in 10-m net depth intervals in the upper 50 m of the water column (high- resolution tows) were used to examine diel vertical migration behavior of early life stages. Depth strata of high-resolution tows were 0-10 m, 10-20 m, 20-30 m, 30-40 m, 40-50 m, 50-100 m, 100-200 m, and 200- 300 m. Concentrations collected at each depth strata varied widely among samples and tows. Depth- 120

specific concentrations from high-resolution tows were standardized among tows by conversion to the proportion of the total concentration in each tow collected at each depth stratum to remove any effect of differences in concentration among domains, years, or times of day. Each tow was assigned to either day time or night time (time of day, TOD) based on time and date of collection (Brodeur and Rugen 1994). Tows collected in February between 0900 and 1900 were considered day time tows and tows collected between 1900 and 0900 were considered night time tows, in April and May, day time was 0630 – 2130 and night time was 2130 – 0630, and in September day time was 0700 – 2100 and night time was 2100- 0700. Dusk and dawn categories were not used because there was no replication available for these time periods. Each tow also was assigned to a shelf domain to compare distributions among habitats.

The effects of depth stratum, TOD, and domain on proportion of each stage were examined using generalized additive mixed models (GAMMs) with TOD as a fixed factor, depth stratum and domain as continuous covariates, and tow as a random variable (Zuur et al. 2009). Temperature and bottom depth were initially examined as continuous covariates but were not chosen during the model fitting process. Models were fitted by comparing Akaike’s Information Criterion values and removing non-significant variables until the best fit model was selected (Akaike 1974). The negative binomial error distribution was a better fit to the data compared to other alternates, such as the Poisson, Gaussian, and lognormal distributions. Interactions between depth stratum and TOD were included to test for diel vertical migration. Interactions between depth stratum and domain were included to test for differences in vertical distributions among areas or habitats.

Results Sampling for Ichthyoplankton Vertical Distributions

Catches ranged from no early life stages collected to depth-specific concentrations of up to 540,000 individuals 1000 m-3. Over 200,000 Walleye Pollock eggs, 34,000 larvae, and 125 early juveniles were collected by MOCNESS sampling since 1992.

Ontogenetic Vertical Migration

Walleye Pollock CWMD and variability in CWMD decreased with ontogeny (Table 4.2, Table 4.3). CWMD was influenced by DOY, bottom depth, and thermocline depth. CWMD of all stages decreased with DOY (Table 4.3). CWMD of all stages except postflexion larvae increased with bottom depth (Table 4.3). CWMD of all stages except early juveniles increased with thermocline depth (Table 4.3). There was no difference in stage-specific CWMD or overall CWMD with temperature category. Eggs were collected at higher densities in surface neuston tows in cold years than in warm years (Table 4.4).

121

Diel Vertical Migration There was no evidence of DVM in Walleye Pollock eggs (Table 4.5, Figure 4.2), but the domain in which they were collected interacted with the depth stratum in which proportion of total egg concentration was highest (Table 4.5, Figure 4.3). Eggs were found ≤ 30 m in all three shelf domains but ≥ 100 m over the slope. Walleye Pollock eggs occurred throughout the water column, but proportion of concentration was higher in the upper 20 m or below 100 m relative to the middle of the water column. There was no difference in yolksac larval depth distribution between TODs (Table 4.5, Figure 4.4). There were not enough replicate samples in multiple domains to assess for yolksac larvae. Yolksac larvae occurred at depths less than 100 m, and proportion of concentration was highest from 10 – 40 m relative to other depth strata.

For the two feeding larval stages, differences evidence for DVM were found. Preflexion larval depth distribution differed between TODs although the level of significance was marginal (Table 4.5, Figure 4.5). Overall, preflexion larvae were shallower and more concentrated in the day time than at night time. Preflexion larvae were shallower over the shelf (10-20 m) than over the slope (20-30 m, Figure 4.6). The interactions between depth strata and TOD were significant for postflexion larvae (Table 4.5). Postflexion larvae exhibited regular DVM; deeper during the day (10 – 40 m) than at night (0 – 20 m, Figure 4.7). Domain did not interact with depth stratum for postflexion larvae (Figure 4.8). Proportion of postflexion larval concentration was highest above 30 m and lowest below 30 m relative to other strata.

Sample sizes for early juveniles were low relative to the other stages. Early juveniles tended to occur deeper at night than during the day, but the interaction between TOD and depth strata was not significant (Table 4.5, Figure 4.9). Early juvenile vertical distribution was not affected by domain (Figure 4.10). Early juveniles occurred at depths less than 100 m with greatest concentration from 0 to 20 m.

Discussion

Ontogenetic and diel vertical migrations have adaptive significance for planktonic organisms, including optimal transport, energy conservation, access to prey, and avoidance of predators (Hunter and Sanchez 1976, Fortier and Leggett 1983, Hare and Govoni 2005). For Walleye Pollock in the SEBS, spawning occurs in either deep-water canyons or over the shelf, eggs are buoyant, and juveniles are abundant over the shelf. Here, prey concentrations are highest in the upper water column and visual predators occur in the water column and near the benthos. Based on these characteristics, we would expect vertical migration strategies that maximize on-shelf transport from spawning grounds over the basin or retention over the shelf, such as a decrease in depth of occurrence to provide access to prey and 122

minimize predation risk, such as OVM and regular DVM (Kerfoot 1985). For Walleye Pollock early life stages in the SEBS, depth distribution became shallower and variability in depth distributions decreased with ontogeny supporting OVM, similar to the GOA but unlike Funka Bay. Weighted mean depths decreased with the progression of summer as stratification tended to increase. Weighted mean depths also decreased as bottom depth decreased and as the thermocline depth decreased (except juveniles), suggesting that distributions mirror the breadth of the available water column. Two of the five early life stages examined exhibited evidence of DVM: reverse in preflexion larvae and regular in postflexion larvae. These two stages are active feeders, while eggs and yolksac larvae are nonmotile or weak swimmers and would not be expected to exhibit distribution patterns typically associated with active behaviors. Sample sizes for early juveniles were too small to assess DVM adequately.

OVM in Walleye Pollock provides several advantages for the early life stages. First, the majority of eggs over the slope were found deeper than 40 m, where they are exposed to the deep water Aleutian North Slope Current through Bering Canyon or the Bering Slope Current through Pribilof Canyon. Eggs spawned offshore, therefore, can be transported onto the shelf by these currents rather than advected further over the Aleutian Basin by cyclonic flow above these currents (Reed and Stabeno 1999). Second, high egg densities were observed in the neuston layer over the shelf in cold years when surface temperatures were comparable to temperatures below in deep water in warm years (Stabeno et al. 2012), allowing for extended development times and, by extension, the potential for increased time for on-shelf transport in cold years, regardless of the vertical position. High egg densities in the surface layer in cold years could be the result of a passive rise to the surface in the absence of a distinct density structure in the water column or the result of delayed hatching at low temperatures. Third, yolksac larvae in the cooler waters below 20 or 30 m depth will conserve energy and extend the period of time before yolk reserves are exhausted and exogenous feeding must begin, which can be advantageous if larvae have not been transported far enough over the shelf to where prey are available. Fourth, following the rise to the upper 20 m of the water column, feeding larvae have access to prey in the upper water column (Coyle and Pinchuk 2005) and warmer water, both of which reduce development time.

DVM typically is exhibited by feeding stages reacting to changes in time of day and light intensity, and Walleye Pollock conform to this pattern. As expected, neither eggs nor yolksac larvae exhibited a pattern consistent with DVM. Vertical distributions of nonmotile or nonfeeding stages are driven generally by physics such as buoyancy or passive mixing). These two early stages displayed passive patterns which also suggests that diel differences in vertical distribution exhibited by feeding stages was driven by behavior rather than by diel changes in the physical water column. In preliminary analyses, we found no differences in the thermocline depth between day and night. Surprisingly, we were

123

unable to conclude that early juvenile Walleye Pollock undergo DVM, although they tended to be deeper at night in the few tows in which they were collected. There is evidence from both laboratory (Sogard and Olla 1996) and field (Bailey 1989) studies that vertical migration in shorter juveniles (< 60 mm) in the GOA is weak and the migration to deeper layers intensifies as the fish gain locomotory and sensory capabilities. Schabetsberger et al. (2000) documented DVM in juvenile Walleye Pollock between 30 and 92 mm SL near the Pribilof Islands. The juveniles in the present study were all shorter than 65 mm SL, supporting that DVM in these shorter fish is weak in the SEBS and consistent with the previous findings in the GOA. However, we were unable to address any impacts of predator presence or net avoidance in the current study.

Walleye Pollock feeding larvae appear to take advantage of surface food concentrations, as they primarily occurred shallower than 30 or 40 m at the onset of feeding. We did not have samples to determine directly the vertical distributions of the preferred prey items, copepod eggs and nauplii. Walleye Pollock undergo a transition from predominantly daytime feeding as larvae (Canino and Bailey 1995) to nocturnal feeding as juveniles (Brodeur et al. 2000). This is allowed by increased visual acuity and sensitivity to light with ontogeny (Miller et al. 1993, Carvalho et al. 2004). Preflexion larval distributions were consistent with reverse DVM (up during the day, down at night) and postflexion larvae underwent regular DVM (down and variable during the day, up at night), suggesting that different trade- offs between prey access, predator avoidance, and perhaps physical forces could be acting on the two stages. Other studies have found that copepod nauplii do not undergo diel vertical migrations (Haldorson et al. 1993, Irigoien et al. 2004). By moving into surface waters during the day when light levels are high and where their preferred prey likely occur (Hillgruber et al. 1995, Brase 1996), preflexion larvae could have higher success capturing prey since visual acuity is low relative to later stages. Postflexion larvae would have higher capture success in surface waters at night at reduced light levels than preflexion larvae because of their improved visual acuity. Postflexion larvae are large enough to be of interest to visual predators such older age class Walleye Pollock (Juanes 2003), which could lead to the pattern of avoiding surface waters during the day. Alternately, postflexion larvae in surface waters during the day may have been better able to avoid our collection gears due to their visual acuity and swimming abilities.

Walleye Pollock vertical distributions vary among habitats both in the SEBS and the GOA. Forward et al. (1996) found three different patterns of vertical distribution for Atlantic menhaden (Brevoortia tyrannus) in three separate studies and suggested that vertical behaviors are flexible in order to incorporate necessary trade-offs that vary between ecosystems or habitats. Eggs in the SEBS were concentrated either below 100 m (slope) or in the upper 30 m (shelf domains), indicative of where they were spawned and probably their buoyancy (Kendall and Nakatani 1992). SEBS yolksac larvae were

124

concentrated around 30 m. In the GOA, eggs are spawned at or below 150 m, followed by an increase in depth prior to hatching (Kendall et al. 1994). GOA yolksac larvae remain at depths > 150 m for several more days. One obvious difference between these systems is the depth of the water column and depth of spawning activity. Olla et al. (1996) proposed that remaining at the spawning depth in the GOA provided a predator refuge for eggs and yolksac larvae, an option that is not available in most of the SEBS habitats due to their relatively shallow water column. SEBS preflexion larvae underwent reverse DVM in most habitats and no DVM in the coastal domain. DVM is either regular or absent for preflexion larvae in the GOA (Olla and Davis 1990 A, Davis and Olla 1994). In the shallow-water coastal domain, the stimulus to move up during the day (e.g. sufficient light for hunting) may not be in place for SEBS preflexion larvae because sufficient light levels are available throughout the shallow (≤ 50 m) water column (Kendall and Nakatani 1992). Similar to larvae in the GOA, coastal domain SEBS preflexion larvae also could respond to very high light levels at the surface during the day with negative phototaxis (Olla and Davis 1990 A). Postflexion larvae in the SEBS undergo regular DVM, in accordance with behavior observed near Auke Bay, Alaska, GOA (Haldorson et al. 1993). In Auke Bay, larvae migrated in response to the trade-off between the vertical distribution of nauplii and avoidance of predators. For postflexion larvae, visual predators occur in all areas in which this stage was collected and the same potential predators occur in the SEBS, suggesting that the trade-off between feeding and predation risk could be a common factor across habitats and ecosystems for this stage. We did not find support for DVM in SEBS early juveniles most likely due to small sample sizes, but juveniles undergo regular DVM in the GOA (Olla and Davis 1990 B, Brodeur and Rugen 1994) and larger juveniles near the Pribilof Islands in the SEBS undergo regular DVM in response to prey movement (Schabetsberger et al. 2000).

Smart et al. (2012) found shifts in the spatial distributions of Walleye Pollock larvae and juveniles between cold years and warm years. The authors hypothesized that spatial shifts were driven by changes in area-specific mortality or transport. Satellite-tracked drifters and hydrographic models show high variation in on- and off-shore transport among years (Danielson et al. 2011, Stabeno et al. 2012), which could be linked to temperature conditions (Sohn et al. 2010) and might explain the differences in spatial distribution. Annual differences in vertical distributions could impact the amount of transport offshore if larvae respond to variations in conditions with adjustments in their vertical distributions (Napp et al. 2000). We found no support for differences in vertical distributions related to categorizing years as either cold or warm.

Conclusions

125

Current speeds in the SEBS middle shelf tend to be slow relative to other spawning areas for Walleye Pollock (< 5 cm s-1, Napp et al. 2000). Because of these slow speeds, the probability of retention over the spawning grounds is high. Some off-shelf spawning grounds such as Bering Canyon may not have the highest growth potential, thus selection for an ontogenetic migration toward the surface where transport onto the shelf would be enhanced is likely in the SEBS. Walleye Pollock early life stages underwent ontogenetic vertical migration and feeding stages were found in the upper portion of the water column where prey availability is high typically. Feeding larvae also exhibited diel vertical migrations that suggest trade-offs occurred between access to prey and exposure to predators. Characteristics of the habitat and ecology of each stage suggested that some determinants of vertical distribution are common between the SEBS and GOA (i.e. prey, light levels, predators) while others are not (i.e. depth refuges, spawning depth). Several hypotheses developed for the SEBS have linked variation in recruitment to the level of overlap between juveniles and their predators or juveniles and their prey, which in turn may be related to the extent and direction of transport. For example, Wespestad et al. (2000) found that strong year classes were linked to high spatial segregation of juveniles and cannibalistic older age classes. The authors proposed that juvenile distribution was closely tied to the transport and distribution of eggs and larvae. One way to address the connection between these various life stages is to model transport. Our results clearly demonstrated that pollock larvae are not passive particles, early life stages are not distributed randomly throughout the water column, and vertical distributions from the GOA are not comparable to all stages in the SEBS. Accurate modeling needs to account for variation in vertical distribution and behavior among stages and habitats, and these data are now available for the SEBS.

Acknowledgments

Thanks to the members of NOAA’s Ecosystems and Fisheries Oceanography Coordinated Investigations (EcoFOCI) who were involved in the collection and processing of the ichthyoplankton samples. This research was supported by the Bering Sea Integrated Ecosystem Research program (BSIERP) of the North Pacific Research Board and the North Pacific Climate Regimes and Ecosystem Productivity (NPCREP) program of the National Oceanographic and Atmospheric Administration. This paper is EcoFOCI Contribution No. N754 – RAOA – N789, BEST-BSIERP Publication No. 89, and NPRB Publication No. 407. We appreciate comments by Jeffrey Napp, Morgan Busby, Thomas Hurst, and three anonymous reviewers.

126

Literature Cited

Akaike, H. 1974. A new look at the statistical model identification. IEEE Transactions on Automatic Control 19, 716 – 723.

Bacheler, N.M., Ciannelli, L., Bailey, K.M., Duffy-Anderson, J.T., 2010. Spatial and temporal patterns of Walleye Pollock spawning in the eastern Bering Sea inferred from egg and larval distributions. Fish. Oceanogr. 19, 107 - 120.

Bailey, K.M., 1989. Interaction between the vertical distribution of juvenile Walleye Pollock Theragra chalcogramma in the eastern Bering Sea and cannibalism. Mar. Ecol. Prog. Ser. 53, 205 – 213.

Brase, A.L.J., 1996. Temporal variation in feeding of larval Walleye Pollock (Theragra chalcogramma) in the Southeastern Bering Sea shelf region. M.Sc. Thesis, Univ. Alaska, Fairbanks, AK.

Brodeur, R.D., Rugen, W.C., 1994. Diel vertical distribution of ichthyoplankton in the northern Gulf of Alaska. Fish. Bull. 92, 223 – 235.

Brodeur, R.D., Wilson, M.T., 1996. A review of the distribution, ecology, and population dynamics of age-0 Walleye Pollock in the Gulf of Alaska. Fish. Oceanogr. 5, 148 – 166.

Brodeur, R.D., Wilson, M.T., Ciannelli, L., 2000. Spatial and temporal variability in feeding and condition of age-0 Walleye Pollock (Theragra chalcogramma) in frontal regions of the Bering Sea. ICES J. Mar. Sci. 57, 256 – 264.

Canino, M.F., Bailey, K.M., 1995. Gut evacuation of Walleye Pollock larvae in response to feeding conditions. J. Fish. Biol. 46, 389 – 403.

Carvalho, P.S.M., Noltie, D.B., Tillitt, D.E., 2004. Biochemical, histological and behavioural aspects of visual function during early development of rainbow trout. J. Fish Biol. 64, 833 – 850.

Coachman, L.K., 1986. Circulation, water masses, and fluxes on the southeastern Bering Sea shelf. Cont. Shelf Res. 5, 23-108.

Coyle, K.O., Pinchuk, A.I., 2005. Seasonal cross-shelf distribution of major zooplankton taxa on the northern Gulf of Alaska shelf relative to water mass properties, species depth 414 preferences and vertical migration behavior. Deep-Sea Res. II 52, 217 – 245.

Coyle, K.O., Eisner, L.B., Mueter, F.J., Pinchuk, A.I., Janout, M.A., Cieciel, K.D., Farley, E.V., Andrews, A.G., 2011. Climate change in the southeastern Bering Sea: impacts on pollock stocks and implications for the oscillating control hypothesis. Fish. Oceanogr. 20, 139-418 156. 127

Danielson, S., Eisner, L., Weingartner, T., Aagaard, K., 2011. Thermal and haline variability over the central Bering Sea shelf: Seasonal and interannual perspectives. Cont. Shelf 421 Res. 31, 539 – 554.

Davis, M.W., Olla, B.L., 1994. Ontogenetic shift in geotaxis for Walleye Pollock, Theragra chalcogramma, free embryos and larvae: potential role in controlling vertical distribution. Environ. Biol. Fishes 39, 313 – 318.

Duffy-Anderson, J.T., Ciannelli, L., Honkalehto, T., Bailey, K.M., Sogard, S.M., Springer, A.M., 426 Buckley, T., 2003. Distribution of age-1 and age-2 Walleye Pollock in the Gulf of Alaska and eastern Bering Sea: sources of variation and implications for higher trophic levels. Institute of Marine Research, Postboks 1870 Nordnes N-5817, Bergen Norway.

Fiksen, O., Jorgensen, C., Kristiansen, T., Vikebo, F., Huse, G., 2007. Linking behavioral ecology and oceanography: larval behaviour determines growth, mortality and dispersal. Mar. Ecol. Prog. Ser. 347, 195 – 205.

Fortier, L., Leggett, W.C., 1983. Vertical migrations and transport of larval fish in a partially mixed estuary. Can. J. Fish. Aquat. Sci. 40, 1543 – 1555.

Forward, R.B., Jr., Tankersley, R.A., Burke, J.S., 1996. Endogenous swimming rhythms of larval Atlantic menhaden Brevoortia tyrannus Latrobe: Implications for vertical migration. J. Exp. Mar. Biol. Ecol. 204, 195 – 207.

Haldorson, L., Prichett, M., Paul, A.J., Ziemann, D., 1993. Vertical distribution and migration of fish larvae in a Northeast Pacific bay. Mar. Ecol. Prog. Ser. 101, 67 – 81.

Hare, J.A., Govoni, J.J., 2005. Comparison of average larval fish vertical distributions among species exhibiting different transport pathways on the southeast United States continental shelf. Fish. Bull. 103, 729 – 736.

Heath, M.R., Henderson, E.W., Baird, D.L., 1988. Vertical distribution of herring larvae in relation to physical mixing and illumination. Mar. Ecol. Prog. Ser. 47, 211 – 228.

Hewitt, R.P., Methot, R.D., 1982. Distribution and mortality of northern anchovy larvae in 1978 and 1979. CalCOFI Report 23, 226 – 245.

Hillgruber, N., Haldorson, L.J., Paul, A.J., 1995. Feeding selectivity of larval Walleye Pollock 447 Theragra chalcogramma in the oceanic domain of the Bering Sea. Mar. Ecol. Prog. Ser. 448 120, 1 – 10. 128

Hinckley, S., 1987. The reproductive biology of Walleye Pollock, Theragra chalcogramma, in the Bering Sea, with reference to spawning stock structure. Fish. Bull. 85, 481 - 498.

Hunter, J.R., Sanchez, C., 1976. Diel changes in swim bladder inflation of the larvae of the northern anchovy Engraulis mordax. Fish. Bull. 74, 847 – 855.

Irigoien, X., Conway, D.V.P., Harris, R.P., 2004. Flexible diel vertical migration behaviour of zooplankton in the Irish Sea. Mar. Ecol. Prog. Ser. 267, 85 – 97.

Joyeux, J.C., 1998. Spatial and temporal entry patterns of fish larvae into North Carolina estuaries: comparisons among one pelagic and two demersal species. Est. Coast. Shelf Sci. 47, 731 – 752.

Juanes, F., 2003. The allometry of cannibalism in piscivorous fishes. Can. J. Fish. Aquat. Sci. 60, 594 – 602.

Kamba, M., 1977. Feeding habits and vertical distribution of Walleye Pollock, Theragra chalcogramma (Pallas), in early life stage in Uchiura Bay, Hokkaido. Res. Inst. N. Pac. Fish., Hokkaido Univ., Spec. Vol., 175 – 197.

Kendall, A.W., Jr., Clarke, M.E., Yoklavich, M.M., Boehlert, G.W., 1987. Distribution, feeding, and growth of larval Walleye Pollock, Theragra chalcogramma, from Shelikof Strait, Gulf of Alaska. Fish. Bull. 85, 499 – 521.

Kendall, A.W., Jr., Incze, L.S., Ortner, P.B., Cummings, S.R., Brown, P.K., 1994. The vertical distribution of eggs and larvae of Walleye Pollock (Theragra chalcogramma) in Shelikof Strait, Gulf of Alaska. Fish. Bull. 92, 540 – 554.

Kerfoot, W.C., 1985. Adaptive value of vertical migration: Comments on the predation hypothesis and some alternatives. Contrib. Mar. Sci. 27, 91–113.

Kristiansen, T., Joergensen, C., Lough, R.G., Vikeboe, F., Fiksen, O., 2009. Modeling rule- based behavior: habitat selection and the growth-survival trade-off in larval cod. Behav. Ecol. 20, 490- 500.

129

Laurel, B.J., Stoner, A.W., Ryer, C.H., Hurst, T.P., Abookire, A.A., 2007. Comparative habitat associations in juvenile Pacific Cod and other gadids using seines, baited cameras and laboratory techniques. J. Exp. Mar. Biol. Ecol. 351, 42 – 55.

Matarese A.C., Blood, D.M., Picquelle, S.J., Benson, J.L., 2003. Atlas of abundance and distribution patterns of ichthyoplankton from the northeast Pacific Ocean and Bering Sea ecosystems based on research conducted by the Alaska Fisheries Science Center (1972-483 1996). NOAA Prof Pap NMFS No. 1, NOAA, Seattle, WA.

Miller, T.J., Crowder, L.B., Rice, J.A., 1993. Ontogenetic changes in behavioural and histological measures of visual acuity in three species of fish. Environ. Biol. Fish. 37, 1 –8.

Miller, T.J., 2007. Contribution of individual-based coupled physical-biological models to understanding recruitment in marine fish populations. Mar. Ecol. Prog. Ser. 347, 127 –138.

Napp, J.M., Kendall, A.W., Shumacher, J.D., 2000. A synthesis of biological and physical processes affecting the feeding environment of larval Walleye Pollock (Theragra chalcogramma) in the eastern Bering Sea. Fish. Oceanogr. 9, 147 – 162.

Nishiyama, T., Hirano, K., Haryu, T., 1986. The early life history and feeding habits of larval Walleye Pollock Theragra chalcogramma (Pallas) in the southeast Bering Sea. INPFC Bull. 45, 177-227.

Norcross, B.L., Shaw, R.F., 1984. Oceanic and estuarine transport of fish eggs and larvae: a review. Trans. Am. Fish. Soc. 113, 153 – 165.

Ohman, M.D., 1990. The demographic benefits of diel vertical migration by zooplankton. Ecol. Monogr. 60, 257 – 281.

Olla, B.L., Davis, M.W., 1990 A. Effects of physical factors on the vertical distribution of larval Walleye Pollock Theragra chalcogramma under controlled laboratory conditions. Mar. Ecol. Prog. Ser. 63, 105-122.

Olla, B.L., Davis, M.W., 1990 B. Behavioral responses of juvenile Walleye Pollock Theragra chalcogramma Pallas to light, thermoclines and food: Possible role in vertical distribution. J. Exp. Mar. Biol. Ecol. 135, 59 – 68.

Olla, B.L., Davis, M.W., Ryer, C.H., Sogard, S.M., 1996. Behavioural determinants of distribution and survival in early stages of Walleye Pollock, Theragra chalcogramma: a synthesis of experimental studies. Fish. Oceanogr. 5, 167-178.

130

Porter, S.M., Ciannelli, L., Hillgruber, N., Bailey, K.M., Chan, K.-S., Canino, M.F., Haldorson, L.J., 2005. Environmental factors influencing larval Walleye Pollock Theragra chalcogramma feeding in Alaskan waters. Mar. Ecol. Prog. Ser. 302, 207 - 217.

Reed R.K., Stabeno, P.J., 1999. The Aleutian North Slope Current. In: T.R. Loughlin and K. Ohtani [eds.], Dynamics of the Bering Sea. University of Alaska Sea Grant, AK-SF-99-03, Fairbanks.

Schabetsberger, R., Brodeur, R.D., Ciannelli, L., Napp, J.M., Swartzman, G.L., 2000. Diel vertical migration and interaction of zooplankton and juvenile Walleye Pollock (Theragra chalcogramma) at a frontal region near the Pribilof Islands, Bering Sea. ICES J. Mar. Sci. 57, 1283 – 1295.

Serobaba, I.I., 1974. Spawning ecology of the Walleye Pollock, Theragra chalcogramma, in the Bering Sea. J. Ichthyol. 14, 544 -552.

Shannon, L.V., Nelson, G., Crawford, R.J.M., Boyd, A.J., 1996. Possible impacts of environmental change on pelagic fish recruitment: modeling anchovy transport by advective processes in the southern Benguela. Global Change Biology 2, 407 – 420.

Shima M., Bailey, K.M., 1994. Comparative analysis of ichthyoplankton sampling gear for early life stages of Walleye Pollock (Theragra chalcogramma). Fish. Oceanogr. 3, 50 - 59.

Smart T.I., Duffy-Anderson, J.T., Horne, J.K., Farley, E.V., Wilson, C.D., Napp, J.M., 2012. Influence of environmental conditions on Walleye Pollock eggs, larvae, and juveniles in the southeastern Bering Sea. Deep-Sea Res. II. 65 - 70, 196 – 207.

Smart T.I., Duffy-Anderson, J.T., Horne, J.K., 2012. Alternating temperature states influence Walleye Pollock early life stages in the southeastern Bering Sea. Mar. Ecol. Prog. Ser. 455, 257 – 267.

Sogard S.M., Olla, B.L., 1996. Food deprivation affects vertical distribution and activity of a marine fish in a thermal gradient: potential energy-conserving mechanisms. Mar. Ecol. Prog. Ser. 133, 43 – 55.

Sohn, D., Ciannelli, L., Duffy-Anderson, J.T., 2010. Distribution and drift pathways of Greenland halibut (Reinhardtius hippoglossoides) during early life stages in the eastern Bering Sea and Aleutian Islands. Fish. Oceanogr. 19, 339-353.

Sokal, R.R., Rohlf, F.J., 1995. Biometry: The Principles and Practice of Statistics in Biological Research. W.H. Freemand and Co., New York.

131

Springer, A., McRoy, C.P., Flint, M., 1996. The Bering Sea Green Belt: shelf-edge processes and ecosystem production. Fish. Oceanogr. 5, 205 – 223.

Stabeno, P.J., Bond, N.A., Kachel, N.B., Salo, S.A., Schumacher,.J.D., 2001. On the temporal variability of the physical environment over the south-eastern Bering Sea. Fish. Oceanogr. 10, 81-98.

Stabeno P.J., Moore, S., Napp, J., Sigler, M., Zerbini, A., 2012. Comparison of warm and cold years on the southeastern Bering Sea shelf and some implications for the ecosystem. Deep-Sea Res. II 65 – 70, 31 - 45.

Stenevik, E.K., Skogen, M., Sundby, S., Boyer, D., 2003. The effect of vertical and horizontal distribution on retention of sardine (Sardinops sagax) larvae in the Northern Benguela –observations and modelling. Fish. Oceanogr. 12, 185-200.

Tanaka, Y., 1991. Change of specific gravity of fish eggs and larvae and their distribution and movement in relation to current structure, p. 60 – 78. In: E. Kawai [ed.], Current and life - Lecture of fisheries oceanography. Kyoto University Press, Kyoto.

Walline, P.D., 1981. Hatching dates of Walleye Pollock (Theragra chalcogramma) and vertical distribution of ichthyoplankton from the Eastern Bering Sea, June – July 1979. NWAFSC Processed Report 81-05. NOAA Fisheries, Department of Commerce.

Wespestad, V.G., Fritz, L.W., Ingraham, W.J., Megrey, B.A., 2000. On relationships between cannibalism, climate variability, physical transport, and recruitment success of Bering Sea Walleye Pollock (Theragra chalcogramma). ICES J. Mar. Sci. 57, 268–274.

Wiebe, P.H., Burt, K.H., Boyd, S.H., Morton, A.W., 1976. A multiple opening/closing net and environmental sensing system for sampling zooplankton. J. Mar. Res. 34, 313-326.

Yamashita, Y., Kitagawa, D., Aoyama, T., 1985. Diel vertical migration and feeding rhythm of the larvae of the Japanese sand-eel Ammodytes personatus. Bull. Japan. Soc. Sci. Fish. 51, 1 – 5. Zuur, A.F., Ieno, E.N., Walker, N., Saveliev, A.A., Smith, G.M. 2009. Mixed Effects Models and Extensions in Ecology with R. Springer, New York, NY. 574 p.

132

Table 4.1. Sampling for vertical distribution of Walleye Pollock (Gadus chalcogrammus) in the southeastern Bering Sea 1992 – 2009. One to three cruises each year conducted MOCNESS tows with a wide range of depth intervals (MOCNESS tows), tows with depth intervals less than 20 m used for catch- weighted mean depths (CWMD tows), tows with depth intervals of 10 m (high-resolution tows), and/or neuston tows.

Year Cruise Date Range MOCNESS CWMD High- Neuston Tows Tows Resolution Tows (month/day) Tows

1992 2MF92 4/16 – 4/22 20 6 0 0

1993 3MF93 4/17 – 4/28 17 4 0 0

1994 4MF94 4/16 – 4/27 9 5 0 0

1995 6MF95 4/23 – 4/30 5 2 2 0

7MF95 5/5 – 5/16 9 3 3 0

1996 10MF96 9/7 – 9/15 21 19 18 0

1997 9MF97 9/11 – 9/17 13 2 2 0

1999 1MF99 4/17 – 4/18 6 6 6 0

2003 4MF03 5/18 – 5/24 9 9 9 53

2005 5MF05 5/10 – 5/18 20 20 20 71

2006 3MF06 5/9 – 5/18 12 12 10 90

6MF06 9/11 – 9/22 9 9 6 86

2007 4MF07 5/8 – 5/17 2 2 2 83

2008 1MF08 2/19 – 2/26 8 8 8 0

3DY08 5/12 – 5/21 7 7 7 64

2HE08 7/3 – 7/17 44 44 0 0

2009 3DY09 5/8 – 5/17 2 2 2 77

1KN09 6/13 – 7/10 91 91 0 0

133

Table 4.2. Ontogenetic vertical distributions. Summary of catch-weighted-mean depths (m; mean and standard deviations [Std. Dev.]) of Walleye Pollock (Gadus chalcogrammus) early life stages in the southeastern Bering Sea.

Eggs Yolksac Preflexion Postflexion Early Larvae Larvae Larvae Juveniles Mean 33.1 29.3 21.5 20.0 18.5 Std. Dev. 46.7 15.4 11.4 10.1 8.9

134

Table 4.3. Ontogenetic vertical distribution analyses. Comparisons of catch-weighted-mean depths of Walleye Pollock (Gadus chalcogrammus) early life stages and the physical characteristics that influence them. Degrees of freedom (df), sum of squares (SS), mean squares (MS), and f-ratios (F) are shown for all stages. P-values (p) in bold were significant at α = 0.05.

Factor df SS MS F p

Stage 4 1.05 x 104 2.60 x 103 2.743 0.029

Temperature 1 3.00 x 102 3.00 x 102 0.314 0.575

Stage x 4 1.14 x 103 2.84 x 102 0.299 0.878 Temperature

DOY 1 1.76 x 104 1.76 x 104 18.592 < 0.001

Bottom Depth 1 1.14 x 104 1.14 x 104 11.991 0.001

Thermocline 1 9.09 x 103 9.09 x 103 9.577 0.002 Depth

Error 340 3.23 x 105 9.49 x 102

135

Table 4.4. Ontogenetic vertical migration. Comparisons of Walleye Pollock (Gadus chalcogrammus) egg densities in the surface neuston layer and the physical characteristics that influence them. Degrees of freedom (df), sum of squares (SS), mean squares (MS), and f-ratios (F) are shown. P-values (p) in bold were significant at α = 0.05.

Factor df SS MS F p

Temperature 1 4.03 x 109 4.03 x 109 5.745 0.017

Bottom Depth 1 1.04 x 109 1.04 x 109 1.481 0.224

TC Depth 1 1.12 x 109 1.12 x 109 1.598 0.207

Error 513 3.60 x 1011 7.02 x 108

136

Table 4.5. Diel vertical distributions. Generalized additive mixed models for differences in proportion of total density of Walleye Pollock (Gadus chalcogrammus) early life stages among depth strata (Depth), times of day (TOD), and domains. Degrees of freedom (df) or estimated degrees of freedom, f-ratios (F), and p-values (p). P-values in bold were significant at α = 0.05.

Stage Factor df F p

Eggs Depth 2.85 19.30 <0.001

TOD 1 0.60 0.439

Domain 1 9.65 0.002

Depth x TOD 3.91 3.88 0.100

Depth x Domain 8 8.25 0.001

Yolksac Depth 2.72 8.08 0.001

Larvae TOD 1 0.11 0.738

Depth x TOD 4.10 2.48 0.349

Preflexion Depth 2.84 23.40 <0.001

Larvae TOD 1 0.34 0.561

Domain 1 0.17 0.678

Depth x TOD 3.26 7.12 0.034

Depth x Domain 4.21 3.89 0.046

Postflexion Depth 2.42 6.67 0.001

Larvae TOD 1 0.04 0.847

Domain 1 0.56 0.455

Depth x TOD 2.78 4.48 0.014

Depth x Domain 2.65 2.01 0.131

Early Depth 2.49 3.39 0.030

Juvenile TOD 1 0.01 0.951

Domain 1 0.01 0.963

Depth x TOD 2.83 2.44 0.144

Depth x Domain 2.63 0.65 0.616

137

Figure 4.1. Vertical distribution tows collected by MOCNESS (gray circles) from 1992 to 2009. Major currents (gray arrows) in this area include the Bering Slope Current (BSC) and the Aleutian North Slope Current (ANSC). Grey lines indicate the 50, 100, 200, and 1000 m isobaths. Walleye Pollock spawning areas include the Alaska Peninsula, Bering Canyon (west of Unimak Island), and Pribilof Canyon (south of the Pribilof Islands). Hydrographic domains include three shelf domains (outer, middle, and coastal) and one off-shelf domain (slope). Inset: neuston samples (open circles) collected in 2003 and 2005-2009.

138

Figure 4.2. Vertical distribution of southeastern Bering Sea Walleye Pollock (Gadus chalcogrammus) eggs during day and night from high-resolution MOCNESS tows. Line represents predicted proportions for each depth interval with 95% confidence intervals (shaded area).

Day Night

2 2

1 1

0 0

-1 -1

Predicted Egg Proportion EggPredicted Proportion EggPredicted

-2 -2

0-10 20-30 40-50 100-200 0-10 20-30 40-50 100-200

Depth Depth

139

Figure 4.3. Vertical distribution of southeastern Bering Sea Walleye Pollock (Gadus chalcogrammus) eggs among domains from high-resolution MOCNESS tows. Line represents predicted proportions for each depth interval with 95% confidence intervals (shaded area).

Inner Shelf Middle Shelf

2 2

1 1

0 0

-1 -1

-2 -2

-3 -3

Predicted Predicted Egg Proportion Predicted Egg Proportion

-4 -4

0-10 20-30 40-50 100-200 0-10 20-30 40-50 100-200

Depth Depth

Outer Shelf Continental Slope

2 2

1 1

0 0

-1 -1

-2 -2

-3 -3

Predicted Predicted Egg Proportion Predicted Egg Proportion

-4 -4

0-10 20-30 40-50 100-200 0-10 20-30 40-50 100-200

Depth Depth

140

Figure 4.4. Vertical distribution of southeastern Bering Sea Walleye Pollock (Gadus chalcogrammus) yolksac larvae during day and night from high-resolution MOCNESS tows. Line represents predicted proportions for each depth interval with 95% confidence intervals (shaded area).

Day Night

2 2

1 1

0 0

-1 -1

-2 -2

-3 -3

-4 -4

Predicted YolksacPredictedLarvalProportion Predicted YolksacProportion PredictedLarval 0-10 20-30 40-50 0-10 20-30 40-50

Depth Depth

141

Figure 4.5. Vertical distribution of southeastern Bering Sea Walleye Pollock (Gadus chalcogrammus) preflexion larvae during day and night from high-resolution MOCNESS tows. Line represents predicted proportions for each depth interval with 95% confidence intervals (shaded area).

Day Night

2 2

1 1

0 0

-1 -1

-2 -2

-3 -3

-4 -4 Predicted Preflexion Larval Proportion PreflexionLarvalPredicted Predicted Preflexion Larval Proportion PreflexionLarvalPredicted 0-10 20-30 40-50 0-10 20-30 40-50

Depth Depth

142

Figure 4.6. Vertical distribution of southeastern Bering Sea Walleye Pollock (Gadus chalcogrammus) preflexion larvae among domains from high-resolution MOCNESS tows. Line represents predicted proportions for each depth interval with 95% confidence intervals (shaded area).

Inner Shelf Middle Shelf

-5

-10

-2 -15

0-10 20-30 40-50 0-10 20-30 40-50

Predicted Predicted Preflexion Proportion Larval Predicted Preflexion Proportion Larval

Depth Depth

Outer Shelf Continental Slope

-2

-2 -6

0-10 20-30 40-50 0-10 20-30 40-50

Predicted Predicted Preflexion Proportion Larval Predicted Preflexion Proportion Larval

Depth Depth

143

Figure 4.7. Vertical distribution of southeastern Bering Sea Walleye Pollock (Gadus chalcogrammus) postflexion larvae during day and night from high-resolution MOCNESS tows. Line represents predicted proportions for each depth interval with 95% confidence intervals (shaded area).

Day Night

2 2

1 1

0 0

-1 -1

-2 -2 Predicted Postflexion Larval Proportion LarvalPostflexion Predicted Predicted Postflexion Larval Proportion LarvalPostflexion Predicted 0-10 20-30 40-50 0-10 20-30 40-50

Depth Depth

144

Figure 4.8. Vertical distribution of southeastern Bering Sea Walleye Pollock (Gadus chalcogrammus) postflexion larvae among domains from high-resolution MOCNESS tows. Line represents predicted proportions for each depth interval with 95% confidence intervals (shaded area).

Inner Shelf Middle Shelf

-1 -2

0-10 20-30 40-50 0-10 20-30 40-50

Predicted Predicted Postflexion Proportion Larval Predicted Postflexion Proportion Larval

Depth Depth

Outer Shelf

-5

-15 -25

0-10 20-30 40-50 Predicted Predicted Postflexion Proportion Larval

Depth

145

Figure 4.9. Vertical distribution of southeastern Bering Sea Walleye Pollock (Gadus chalcogrammus) early juveniles during day and night from high-resolution MOCNESS tows. Line represents predicted proportions for each depth interval with 95% confidence intervals (shaded area).

Day Night

4 4

2 2

0 0

-2 -2

-4 -4

-6 -6

Predicted Early JuvenileEarlyPredicted Proportion JuvenileEarlyPredicted Proportion 0-10 10-20 20-30 30-40 40-50 0-10 10-20 20-30 30-40 40-50

Depth Depth

146

Figure 4.10. Vertical distribution of southeastern Bering Sea Walleye Pollock (Gadus chalcogrammus) early juveniles among domains from high-resolution MOCNESS tows. Line represents predicted proportions for each depth interval with 95% confidence intervals (shaded area).

Inner Shelf Middle Shelf

-5

-10

-2

-15

Predicted Predicted Juvenile Proportion Predicted Predicted Juvenile Proportion 0-10 10-20 20-30 30-40 40-50 0-10 10-20 20-30 30-40 40-50

Depth Depth

Outer Shelf

-2

-4 Predicted Predicted Juvenile Proportion 0-10 10-20 20-30 30-40 40-50

Depth

147

Chapter 5: Biophysical transport model suggests climate variability determines distribution of Walleye Pollock early life stages in the Eastern Bering Sea through effects on spawning

Colleen M. Petrik1,2, Janet T. Duffy-Anderson3, Franz Mueter1, Katherine Hedstrom4, and Enrique N. Curchitser5

1University of Alaska Fairbanks, School of Fisheries and Ocean Sciences, 17101 Point Lena Loop Rd., Juneau, AK 99801, USA 2Present address: UC Santa Cruz, Institute of Marine Sciences, 110 Shaffer Rd., Santa Cruz, CA 95060, USA 3Alaska Fisheries Science Center, NOAA NMFS, 7600 Sand Point Way NE, Seattle, WA 98115, USA 4Univeristy of Alaska Fairbanks, Arctic Region Supercomputing Center, 105 West Ridge Research Bldg., P.O. Box 756020, Fairbanks, AK 99775, USA 5Department of Environmental Sciences and Institute of Marine and Coastal Sciences, Rutgers University, 14 College Farm Rd., New Brunswick, NJ 08901, USA

Citation: Petrik, C., Duffy-Anderson, J.T., Mueter, F.J., Hedstrom, K., and Curchitser, E. In press. Modeling the effect of climate variations on the transport and distribution of Walleye Pollock early life stages in the eastern Bering Sea. Progress in Oceanography. http://www.sciencedirect.com/science/article/pii/S0079661114001128 148

Abstract

The eastern Bering Sea recently experienced an anomalously warm period followed by an anomalously cold period. These periods varied with respect to sea ice extent, water temperature, wind patterns, and ocean circulation. The distributions of Walleye Pollock early life stages also differed between periods, with stages found further eastward on the shelf in warm years. Statistical analyses indicated that these spatial distributions were more closely related to temperature than to other covariates, though a mechanism has not been identified. The objective of this study was to determine if variable transport could be driving the observed differences in pollock distributions. An individual-based model of pollock early life stages was developed by coupling a hydrodynamic model to a particle-tracking model with biology and behavior. Simulation experiments were performed with the model to investigate the effect of wind on transport, ice presence on time of spawning, and water temperature on location of spawning. This modeling approach benefited from the ability to individually test mechanisms to quantitatively assess the impact of each on the distribution of pollock. Neither interannual variations in advection nor advances or delays in spawning time could adequately represent the observed differences in distribution between warm and cold years. Changes to spawning areas, particularly spatial contractions of spawning areas in cold years, resulted in modeled distributions that were most similar to observations. The location of spawning pollock in reference to cross-shelf circulation patterns is important in determining the distribution of eggs and larvae, warranting further study on the relationship between spawning adults and the physical environment. The different distributions of pollock early life stages between warm and cold years may ultimately affect recruitment by influencing the spatial overlap of pollock juveniles with prey and predators.

149

Introduction

The eastern Bering Sea shelf is one of the most biologically productive marine ecosystems in the world with marine resources that are integral to the culture and diet of native Alaskans and that comprise roughly 50% of the US commercial fish harvest. The continental shelf extends approximately 500 km westward from the Alaskan mainland coast to the Aleutian Basin shelfbreak and 1000 km northward from the Alaska Peninsula to the Bering Strait. The shelf can be divided into three regions based on bathymetry: inner shelf (<50 m), middle shelf (50-100 m), and outer shelf (>100 m; Coachman 1986). The inner shelf is weakly stratified and influenced by freshwater runoff, while the middle and outer shelves are strongly stratified (Coachman 1986). A large and highly variable portion of the shelf is ice- covered during winter, cooling the entire water column and resulting in a bottom layer of very cold water over much of the middle shelf that persists through the summer (the “cold pool”). Offshore, the Bering Slope Current (Figure 5.1) transports nutrient-rich waters along the slope to the northwest, replenishing nutrients on the shelf through cross-shelf exchanges associated with eddies (Mizobata et al. 2008), intrusions of water in canyons (Stabeno et al. 2008), and wind-forced cross-shelf flows (Stabeno et al. 2001, Danielson et al. 2011). Mean circulation over the shelf is dominated by the Alaska Coastal Current, which is a seasonal current that flows roughly parallel to the 50-m isobath northeast along the Alaska Peninsula, around Bristol Bay, and continues to the northwest off of mainland Alaska towards the Bering Strait (Figure 5.1). Cross-shelf and along-shelf flows provide important pathways for the planktonic stages of many species that spawn on the outer shelf or slope to reach suitable nursery areas (Lanksbury et al. 2007, Wespestad et al. 2000, Wilderbuer et al. 2013).

Walleye Pollock (Gadus chalcogrammus; hereafter pollock) is an ecologically and commercially important gadid in the eastern Bering Sea, supporting one of the largest single-species fisheries in the world. Adults are semi-demersal and occur primarily in regions 50-300 m deep (Duffy-Anderson et al. in review). Females are iterative spawners with up to 10 batches per female per year (Duffy-Anderson et al. in review). Pollock show fidelity to at least two spawning sites over the southeastern Bering Sea shelf (Figure 5.1). Spawning begins nearshore north of Unimak Island in March and April and later near the Pribilof Islands from April through August (Jung et al. 2006, Bacheler et al. 2010). Egg production from these two locations peaks in April or May (Bacheler et al. 2010). Eggs can be found as deep as 300 m, but the center of the vertical distribution is < 30 m (Smart et al. 2012). Egg incubation lasts 18-34 days and yolksac larvae hatch at 4.6-5.7 mm standard length (SL), both depending on temperature (Blood 2002). Larvae develop in the upper 100 m of the water column and the depth of maximum abundance shifts deeper with age. Depth differences are related to flexion, which occurs between 10 and 17 mm SL, after which larval swimming ability increases (Dunn & Matarese 1987). Pollock transition from larva to

150

pelagic juvenile when they are 30-40 mm SL (Matarese et al. 1989) at age-0 and recruit into the fishery at age-3 to age-4 (Ianelli et al. 2012a).

The eastern Bering Sea recently experienced a prolonged warm period (2001-2005), followed by a prolonged cold period (2007-2012; Stabeno et al. 2012). During colder than average years, winter ice extends farther south and offshore, creating a cold pool of bottom water <2○C that influences the distribution (Mueter & Litzow 2008, Barbeaux 2012, Ianelli et al. 2012a) and potentially the spawning ecology of pollock and other demersal fishes. The timing and location of pollock spawning affect the initial distribution of eggs, while their subsequent advection is a function of prevailing atmospheric and hydrographic conditions. Therefore both the initial distribution and advective forcing vary from year to year due to climatic variability. For example, strong northward flow and/or weaker cross-shelf flow have been observed at a hydrographic mooring over the middle shelf during the recent warm period compared to recent cold years that had strong westward flow (Stabeno et al. 2012), which could affect the dispersal of pelagic early life stages (ELS). Modeling results suggest generally enhanced on-shelf transport when winds blow predominantly from the southeast during winter (Danielson et al. 2011), which coincided with warm temperatures from 2001 to 2005, while the following cold years were characterized by winds from the northwest. It has been speculated that these differences in advection influence the distributions of pollock ELS, whose centers are further inshore in warm years than cold years (Smart et al. 2012).

The observed variability in distribution could be the result of differences in physical transport as hypothesized, or they could be the result of biological responses to physical variation. Preliminary analysis of roe fishery harvest data and pollock fishery observer maturity data suggest that spawning extends further onshore in warm years (S. Barbeaux unpub. data) and is delayed in time by as much as 40 d in cold years (Jung et al. 2006, Smart et al. 2012). Differences in water temperature could not only impact where adults spawn, but also the development rates of ELS.

Mechanisms behind the distribution differences have not been identified, and effects of climate variation on the dispersal of pollock ELS are poorly understood. In order to gain a clearer understanding of the forcing mechanisms that underlie observed spatial differences in pollock egg and larval distribution, we developed an individual-based model of pollock biology and behavior (TRACMASS) coupled to a hydrodynamic model (ROMS). Our objective was to test the effects of atmospheric (wind), oceanographic (ice, water column temperature), and biological (time and location of spawning) conditions on the distribution, development, and transport of Walleye Pollock eggs and larvae. Model results are used to elucidate the dominant physical mechanisms responsible for observed changes. We present a

151

description of the coupled biophysical model, its validation with observations, and its use to evaluate mechanisms that determine the spatial distribution of propagules during warm and cold years.

Methods Observational data Ichthyoplankton surveys

Observational data from the Fisheries Oceanographic Coordinated Investigations (FOCI) surveys and Bering Ecosystem Study and Bering Sea Integrated Ecosystem Research Project (BEST-BSIERP) cruises were used for model-data comparisons. Observational data included egg and larval samples collected with Bongo nets, Multiple Opening/Closing Net and Environmental Sensing System (MOCNESS), Modified Bottom Trawls (MBTs), and Tucker trawls. Larval stages were segregated by length: yolksac < 6 mm; 6 mm ≤ preflexion < 10 mm; 10 mm ≤ late < 40 mm (Matarese et al. 1989). Data from all four gear types were used to characterize horizontal distributions, while only depth-stratified MOCNESS samples were used for the vertical distributions since the other gear types did not provide enough vertical resolution. Horizontal distributions were determined for each stage and for each available month and year from 1995 to 2012, by aggregating all observations within a given month and year. To quantify horizontal distributions, vertically integrated concentrations (number per 10 m2) were computed over the entire depth range sampled by depth-stratified tows (MOCNESS, MBT, Tucker) so that they were comparable to non-stratified samples (Bongo). All available data sets from 1995 and 2007 were used for model selection. Only data from the cruises listed in Smart et al. (2012) were used for comparison between observed and modeled distributions in warm and cold years.

Spawning data Two different data sets were used to provide information on potential locations for initializing particles as eggs in the model. The first data set was derived from collections of female pollock for the roe fishery (S. Barbeaux unpub. data). These data included the day of fishing, the centroid latitude and longitude of where the boat was fishing that day, and the amount of roe of each quality caught that day for the years 2001-2006. The market quality of roe deteriorates as the adults near spawning and is ranked as immature, mature, overmature, or other. Hydrated roe is within a few days of spawning and of the worst quality (S. Barbeaux pers. comm.). We used only presence and absence data of the hydrated, overmature roe as a proxy for spawning times and locations.

The second data set consisted of maturity index classifications from the NOAA Fisheries North Pacific groundfish observer program. These data were obtained on pollock fillet fishery vessels from

152

2008-2012 and contained date, latitude, longitude, bottom depth, fishing depth, and a maturity index. Observers classify the maturity level of a random subsample of the catch using the maturity indices: immature, developing, pre-spawn, spawning, and spent. We used only the dates and locations where female pollock in spawning condition were recorded as a proxy for spawning times and locations.

These data were used together with available literature on dominant spawning areas (Bacheler et al. 2010, Hinckley 1987) to infer generalized spawning distributions for warm and cold years, respectively, as described below. We were unable to use annual observations on spawning times and locations for model initialization because neither data set spanned the whole time period of interest from 2000 to 2012. Furthermore, the two data sets could not be combined because they used data from different fisheries and used different classification methods.

Physical model

We used an implementation of the Regional Ocean Modeling System (ROMS; Shchepetkin & McWilliams 2009) as the hydrodynamic model to force the Lagrangian particle-tracking model. ROMS is a free-surface, hydrostatic primitive equation ocean circulation model. It is a terrain-following, finite volume (Arakawa C-grid) model with the following advanced features: high-order, weakly dissipative algorithms for tracer advection; a unified treatment of surface and bottom boundary layers (e.g., K-Profile Parameterization; Large et al. 1994); atmosphere-ocean flux computations based on the ocean model prognostic variables using bulk-formulae. ROMS has been coupled to a sea-ice model (Budgell 2005) consisting of the elastic-viscous-plastic (EVP) rheology (Hunke & Dukowicz 1997) and the Mellor and Kantha (1989) thermodynamics. The sea ice code is fully explicit and implemented on the ROMS Arakawa C-grid and is therefore fully parallel, just as ROMS is. The model also includes frazil ice growth in the ocean being passed to the ice (Steele et al. 2004). It currently follows a single ice category, which provides accurate results in a marginal ice zone such as the Bering Sea. Large-scale climate signals are propagated to the regional domain through the model boundaries. This global-to-regional downscaling via open boundary conditions has several desirable features for the implementation of regional models: for multi-decadal integrations, climate signals project onto the high-resolution inner domains through boundary forcing; tidal forcing is naturally implemented on the domain open boundaries; for extensive integrations a tidal potential correction is applied to ensure proper tidal phasing (Curchitser et al. 2005, Danielson et al. 2011a).

We used the ROMS model for the northeast Pacific version 6 (NEP6), an update of version 5 (NEP5) (Hermann et al. 2009, Danielson et al. 2011a). The NEP model domain extends from approximately 20°N to 71°N, reaching about 2250 km offshore from the North America west coast at a 153

nominal horizontal resolution of 10 km and with 50 terrain-following vertical levels stretched towards the surface boundary. The grid (a rectangle in a Lambert Conical projection) is rotated relative to lines of constant longitude so as to minimize computations over land. The coupled ocean-sea ice model was integrated in hindcast mode for the period from 1994-2012. These hindcasts derived the surface forcing from the Modern Era Retrospective-Analysis for Research and Applications (MERRA; Rienecker et al. 2011), which consists of 3-hourly winds, air temperatures, sea level pressure, specific humidity, short- wave and downwelling long-wave radiation, precipitation, and daily albedo. The air-sea fluxes were computed using bulk formulae (Large & Yeager 2009). Riverine inputs were implemented using the Dai and Trenberth (2002) method as a surface fresh water flux. Boundary and initial conditions for this domain were derived from the Simple Ocean Data Assimilation (SODA) ocean reanalysis (Carton & Giese 2008) for the early years. Later years used boundary conditions from global HYCOM assimilative product (HYCOM Ocean Prediction website). Key model outputs have recently been validated against observations at relevant spatial and temporal scales (Curchitser et al. 2010, Danielson et al. 2011a).

The output of the hindcast was saved as daily averages to force the offline particle-tracking model, as described below. Specifically, the particle-tracking model used ROMS-generated velocities, temperature, and mixed layer depth.

Particle-tracking model TRACMASS We used the particle-tracking model TRACMASS for the individual-based model. TRACMASS calculates Lagrangian trajectories from Eulerian velocity fields. The coupling is offline, using stored output from general circulation model (GCM) simulations. Offline coupling is less computationally expensive, thus it allows for more calculations of trajectories in comparison with online coupling (simultaneously with the GCM). TRACMASS accepts output from many GCMs, including ROMS. TRACMASS interpolates the GCM three-dimensional grid to its own grid and solves the trajectory path through each grid cell with an analytical solution of a differential equation, which depends on the velocities on the grid cell walls. This novel scheme was originally developed by Döös (1995) and Blanke and Raynaud (1997) for stationary velocity fields and further developed by de Vries and Döös (2001) for time-dependent fields. With time-dependent fields it solves a linear interpolation of the velocity field both in time and in space over each grid box, in contrast to the Runge-Kutta method where the trajectories are iterated forward in time with short time steps. The TRACMASS code has been further developed over the years and used in many atmospheric and oceanographic studies of large global scales (e.g. Drijfhout et al. 2003) and regional scales (e.g. Engqvist et al. 2006).

154

The particle-tracking time step was one hour. We chose to use the turbulence subroutine in TRACMASS to incorporate a sub-grid scale parameterization. This scheme adds a random horizontal turbulent velocity to the horizontal velocity from ROMS to each trajectory and each horizontal grid wall every time step (Döös & Engqvist 2007). The amplitude of the random turbulent velocity is set to the same size as the ROMS velocity. We tested simulations of a small cluster of particles with and without sub-grid turbulence. Including turbulence resulted in less patchy distributions of ELSs, which were assumed to be more realistic, although fine-scale data to evaluate this assumption are lacking. In addition to particle trajectories, TRACMASS calculated surface light as a function of latitude, longitude, date, and time of day.

Number of particles Often results from stochastic models are presented as the average of several repeated simulations with the same initial and forcing conditions. For particle-tracking models, an alternate method exists by releasing multiple particles at each time and location, each representing one possible outcome of the simulation. The results of a rigorous particle-tracking model should not change significantly between repeated simulations (Brickman et al. 2009). The number of particles released at each time and location (number of simulation repetitions) was determined by calculating the fraction of particles at 4 random locations downstream of the initial start locations, and finding the minimum number of particles for which those fractions did not change appreciably. Ten particles per 10 m depth increment per spawning location were deemed appropriate for producing stable results.

Biological model Model parameterization The literature on Walleye Pollock ELS was reviewed to determine appropriate formulations for the growth and vertical behavior routines in the biological model. When available, data from the Bering Sea population was used over data from the Gulf of Alaska.

Growth Egg development was implemented as a temperature-dependent function parameterized from laboratory data for Bering Sea pollock eggs (Blood 2002). Time to hatch, hatchhrs (hr) (Figure 5.2a), was

-0.194×T hatchhrs = 895.97×e , with temperature, T, in C. At each time step, an egg accumulated a fraction of the hatch time given the temperature. When the accumulated hatch time reached 1, the egg transitioned to the yolksac stage with a random hatch length drawn from a normal distribution with a mean of 5.125 mm SL and a standard deviation of 0.46 mm (Blood 2002). 155

Both a temperature-dependent and a temperature-independent growth model were developed for larvae to allow for tests of the effect of temperature on growth and the resulting distributions of different

ELS. The temperature-independent routine was an empirical function for length, Llarva (mm), fit to otolith- estimated age of larvae collected in the Gulf of Alaska by Yoklavich and Bailey (1990),

Llarva = Lhatch ×exp(7.854×(1-exp(-0.004×age)))

where Lhatch is length at hatch (mm), and age is days post hatch (Figure 5.2c solid line). The temperature- based growth model fitted growth rates from laboratory studies as a function of temperature (Canino 1994 for yolksac, Porter & Bailey 2007 for feeding larvae).

-1 Yolksac larvae growth, gy (mm d ; Figure 5.2b), was:

gy = 0.0686×log(T)+0.0594 .

-1 Growth of feeding larvae (preflexion and late larvae), gf (mm d ), was additively adjusted to account for differences in laboratory growth rates of larvae fed a natural assemblage of zooplankton vs. those fed Artemia. Growth (Figure 5.2b, 5.2c dotted lines) was calculated as:

gf = 0.0902×log(T)-0.0147.

Vertical behavior

Vertical behaviors were determined from depth-stratified observations of pollock ELS (Smart et al. 2013). Five different vertical behavior routines were developed and tested. They included 1) passive (neutrally buoyant) individuals of all stages, 2) passive eggs and yolksac larvae, and preflexion and late larvae that move to the middle of the mixed layer, 3) all stages move to the middle of the mixed layer, 4) eggs, yolksac, and preflexion larvae move to the middle of the mixed layer, and late larvae make diel vertical migrations (DVMs) between 20 m during the day and 5 m during the night, 5) eggs and yolksac larvae move to the middle of the mixed layer, and preflexion and late larvae make DVMs. For DVM, day was defined as times when surface light was greater than zero. Swimming speed (w) was parameterized as:

w = wmax ×(-tanh(0.2×(z - zpref ))) 156

where z is depth (m), zpref (m) is the preferred depth (middle of the mixed layer or day-time/night-time -1 preferred depths), and the maximum vertical swimming speed, wmax (m s ), is

-3 wmax = 0.5× Llarva ×10 .

Swimming speeds of larval cod (Gadus morhua), a related gadoid fish, range from 0.3-0.9 body lengths s- 1 (Peck et al. 2006), thus a maximum speed of 0.5 body lengths s-1 was a conservative estimate for sustained swimming.

Model initialization

Spawning polygons were created by generalizing areas with fishery-based observations of adult pollock in spawning condition. Only the dominant spawning regions on the eastern Bering Sea shelf, as identified from the literature (Hinckley 1987, Bacheler et al. 2010), were considered. Spawning observations in regions deeper than 250 m were not used, as these are likely to consist of spawners from the Bogoslof Island population, which is considered to be a separate population from that over the eastern Bering Sea shelf (Ianelli et al. 2012b). Polygons were created for 2-week periods from the middle of January to the end of April for a total of 7 periods (Figure 5.3; release dates Jan 15, Feb 1, Feb 15, Mar 1, Mar 15, Apr 1, Apr 15). Spawning was initialized at all ROMS grid points within each spawning polygon. Spawning depths occurred every 10 m from surface to bottom at each grid point. Individuals were followed for 90 d after each spawning time or until they reached 40 mm. At approximately 30-40 mm SL pollock larvae transition to pelagic juveniles with different growth rates and enhanced swimming abilities (Matarese et al. 1989), which was not represented within the model.

Model selection Two years with the most spatially and temporally resolved observations (1995, 2007; both cold years) were used to choose the growth and vertical behavior routines to be implemented in all other simulations of the model. Ten different growth-behavior combinations (2 growth functions, 5 vertical behaviors) were run for each year, for a total of 20 different simulations.

The observational data were used to calculate the average monthly concentration for each ELS in 0.25 x 0.25 grid cells over the eastern Bering Sea (-175W to -160W, 53N to 61N) for comparisons with model output aggregated at the same spatial scale. When multiple observations occurred within the same grid cell, the mean concentration was used. A grid cell without any observations was set to “missing” (NaN). 157

To account for uneven spatial coverage of samples among years and to reduce the impact of small-scale sampling variability, resulting from patchiness and/or sampling errors, observed concentrations were spatially smoothed using a General Additive Modeling (GAM) approach. We modeled log-transformed concentrations of each ELS in each month as a smooth function of latitude and longitude using a flexible thin-plate regression spline (Wood 2006). Residuals were assumed to be independent and normally distributed, an assumption that was visually assessed for each model.

To minimize extrapolation beyond the observations, contours of the resulting model standard error were mapped and a threshold was chosen visually by selecting the smallest standard error contour that encompassed all of the observations. For all grid cells within the spatial domain defined by this standard error threshold, egg or larval concentrations within each grid cell were predicted from the GAM fit. The predicted concentrations were then re-scaled to correspond to the fraction of each ELS in each grid cell in each month for comparison with model output. Fractions within each grid cell were used for comparison because the model produces relative concentrations that cannot directly be compared to observed densities.

Model output (longitude, latitude, depth, egg/larval length, time) was treated in the same way as observational data. The output for all 7 spawning times was aggregated, since the spawning times of eggs and larvae collected in the field were unknown. The model output was separated by month and pollock ELS. Model output for each stage, month, and year was restricted to those particles that fell within the same spatial domain over which GAM-predicted concentrations were obtained. The fraction of particles within each of the corresponding grid cells was calculated for all stages, months, and years.

The growth and vertical behavior routines were selected using multiple skill metrics (Stow et al. 2009), including correlation coefficient (R), root mean square error (RMSE), and modeling efficiency (MEF), calculated from the GAM-predicted fractions and the modeled fractions for all simulated months (Jan-Jul) and years (1995 and 2007). All three skill metrics (Table 5.1) indicated the same two behaviors as the best model: behaviors 1 (passive individuals of all stages) and 2 (passive eggs and yolksac larvae, preflexion and late larvae that move to the middle of the mixed layer). Combined with the two different growth formulas, 3 of the 4 combinations were the only simulations that produced results better than using the average of the observations, as denoted by positive modeling efficiencies (Table 5.1).

Model simulations

158

Based on the skill results, the warm and cold year simulations were run with temperature- dependent growth and vertical behavior 2 (passive eggs and yolksac larvae, preflexion and late larvae that move to the middle of the mixed layer). After selecting the growth and vertical behavior routines based on results of the 1995 and 2007 simulations, the biophysical model was run for each of the warm (1996, 2002, 2003, 2005) and cold (1997, 1999, 2000, 2006, 2008-2012) years using model output from ROMS NEP6 hindcasts. Years were identified as either warm or cold based on the sign of the sea surface temperature anomaly following Smart et al. (2012). The modeled egg and larval distributions during warm and cold years were contrasted and related to physical forcing to determine potential mechanisms for the observed differences. A series of simulations were conducted to elucidate if differences in horizontal distributions of pollock ELS between warm and cold years are attributable to spawning timing, spawning location, physical transport, and/or development during transport. All simulations used year- specific physical forcing, but different scenarios for spawning time and location:

(1) Physical transport – The same spawning time and location for both warm and cold years, thus differences in distribution arise from differences in advection alone (2) Delay spawning time – The same spawning locations, delay initialization of cold year spawning by 40 d (3) Advance spawning time – The same spawning locations, advance initialization of warm year spawning by 40 d (4) Contract spawning location – Contract spawning polygons in cold years by shifting the eastern edge of spawning polygons offshore to the southwest (by 0.5 to the South and 0.25 to the West around the Pribilof Islands; by 1.0 to the South and 0.50 to the West around Unimak Island). (5) Expand spawning location – Expand spawning polygons in warm years by shifting the eastern edge of spawning polygons onshore to the northeast (0.50 to the North, 0.25 to the East).

The advance and delay in spawn timing for scenarios 2 and 3 (40 d) were chosen based on the difference in peak egg abundance between warm and cold years (Smart et al. 2012). The simple contraction and expansion of spawning polygons were used to simulate the exclusion of spawning from areas with sea surface temperatures < 2.4C and >3.8C based on the statistical model of Barbeaux (2012) that found positive adult pollock winter abundances at these temperatures. The spawning polygons developed for the physical transport run described above were compared to the distribution of ROMS- generated SST for a few warm and cold years. General shifts were created to remove spawning from areas that tended to be <2.4C in cold years and to expand spawning in regions usually <3.8C. With 14 years and 4 additional simulation experiments, a total of 56 simulations were completed.

159

Comparisons and statistics The same 0.25 x 0.25 grid over the eastern Bering Sea was used to constrain the model output of the scenario tests to the spatial regions containing observations in both cold and warm years. Model results were restricted to particles that fell within a grid cell with at least one observation in both a cold and warm year. The fraction of particles within each grid cell containing observations in both cold and warm years was calculated for each stage and each year, from all individuals in a given stage during the whole simulated period. The mean fraction of particles in each grid cell was then computed for cold and warm years separately for visualization of the horizontal distributions.

In addition to horizontal distributions, the center of gravity and major and minor axes were calculated for each life stage by employing the approach described in Woillez et al. (2009) and implemented in R (R Development Core Team 2011). This calculation used the longitudes, latitudes, and fractions of each stage in a grid cell from all years that were either cold or warm (not the mean fraction). For comparison, the center of gravity and major and minor axes were similarly calculated for the observations from cold and warm years, also constrained to the same grid cells with observations in both thermal regimes. In contrast, longitude and latitude was weighted by the mean concentration of each stage in each grid cell instead of fraction of individuals.

Results Sensitivity of physical transport simulation In both cold and warm years, there was a noticeable difference only in the late larval centers of gravity in simulations with passive individuals compared to simulations with vertical behavior (Table 5.2). Including behavior in the model resulted in the center to be more to the east (on-shelf), with a slightly larger difference in warm years. Simulated centers of gravity were more sensitive to the growth formulation than to implementation of vertical behavior. In both cold and warm years, temperature- dependent growth resulted in different centers of gravity for yolksac, preflexion, and late larvae compared to temperature-independent growth (Table 5.2) due to differences in stage duration. The centers of gravity of yolksac and preflexion larvae with temperature-dependent growth were more on-shelf, whereas they were more off-shelf for late larvae. The differences between temperature-dependent and independent centers of gravity were greater in cold years than warm years (Table 5.2). This stems from the larger difference in growth rates between the temperature-independent model and temperature-dependent model at lower temperatures (Figure 5.2c), resulting in longer stage durations in cold years when growth is temperature-dependent.

160

The effect of temperature-dependent growth alone on the differences in distribution between warm and cold can be seen by comparing simulations with temperature-dependent growth and either passive or vertical behavior. When passive, eggs and yolksac larvae were more off-shelf while preflexion and late larvae were more on-shelf in warm years (Table 5.2). The distance between centers of gravity was greater for the yolksac stages than the preflexion and late stages (Table 5.2). Adding behavior moved the centers of gravity of the egg, yolksac, and late stages eastwards in warm years, resulting in smaller differences between the eggs and yolksac stages, but late larvae shifted even further on-shelf in comparison to cold years (Table 5.2). Conversely, the center of gravity of preflexion larvae with behavior moved westward in warm years, producing a negligible difference between warm and cold years (Table 5.2).

Cold vs. warm years Physical transport When initial conditions (egg release locations and timing) were held constant across years, such that all differences were due to variation in ocean circulation driven by climate variability, there do not appear to be large differences between the mean horizontal distributions of simulated eggs, yolksac larvae, and preflexion larvae between cold (Figure 5.4 center) and warm years (Figure 5.5 center). The distributions of the late larval stage were noticeably different, with late larvae more widespread on the outer and middle shelves in warm years (Figure 5.4 center, Figure 5.5 center). The differences in the centers of gravity quantify that, on average, the egg, yolksac, and preflexion stages were more westward (off-shelf) in warm years, whereas the late stages were more eastward (Table 5.3). The distance between the centers of gravity in cold and warm years is small for the egg stage, increases for the yolksac stage, reaches its minimum with the preflexion stage, then increases markedly for the late larval stage (Table 5.3).

Delay spawning time Like the physical transport scenario, there are no obvious differences between the mean modeled stage distributions between warm and cold years when spawning is delayed 40 d in cold years (Figure 5.4 left, Figure 5.5 center). In this case, the largest differences occur for the preflexion larval stage (Figure 5.4 left) that show a similar pattern of being more widespread over the shelf in warm years (Figure 5.5 center). Preflexion and late larvae were more on-shelf in warm years, whereas the eggs and yolksac larvae were slightly more off-shelf (Table 5.3). The differences between the centers of gravity for warm and cold years were marginal for the egg and yolksac stages, and were largest for the preflexion stage (Table 5.3). Compared to the physical transport case in cold years, a 40 d delay in spawning changed the distributions of the preflexion and late larvae (Figure 5.4 left, center). With a delay, the distribution of the late larvae is

161

similar to the distribution of preflexion larvae in the physical transport simulation, while the preflexion distribution after delayed spawning is unlike any of the stage distributions from the transport scenario (Figure 5.4 left, center).

Advance spawning time Spawning 40 d early in warm years did not perceptibly change the modeled early life stage mean distributions (Figure 5.5 left, center). Though the variations in the horizontal distribution from the physical transport case may not be apparent visually, advancing the spawning times resulted in greater differences in the centers of gravity of all stages except the yolksac larvae (Table 5.3). The yolksac, preflexion, and late stages were more on-shelf in warm years with early spawning, and distance between the centers of gravity increased with stage (Table 5.3).

Contract spawning location A large difference in the mean horizontal distributions occurred when spawning locations were contracted in cold years. All stages were concentrated on the outer shelf and over the slope in cold year simulations (Figure 5.4 right), with none over the middle shelf as seen in warm years (Figure 5.5 center) or physical transport simulations of cold years (Figure 5.4 center). The centers of gravity of all stages were more eastward in warm years by approximately 1.0-1.5 longitude (Table 5.3). The centers of gravity grew further apart with stage, with a slight decrease in distance for the late larval stage (Table 5.3). Spawning contraction in cold years caused the largest changes in center of gravity longitude and total distance between cold and warm years (Table 5.3).

Expand spawning location Expansion of the spawning areas onto the shelf in warm years was reflected in the modeled distributions of pollock ELS (Figure 5.5 right). The eggs and yolksac larvae were present over a larger area to the northeast of Unimak Island, whereas the preflexion and late larvae increased in relative concentration over the middle shelf and decreased over the slope and outer shelf (Figure 5.5 center, right). The yolksac stage exhibited the smallest difference in center of gravity between cold and warm years (Table 5.3). The centers of gravity of all stages were more northeastward in warm years, though the differences between the warm and cold centers of gravity were not as large as when spawning areas were contracted in cold years (Table 5.3).

Comparison with observations Distributions of pollock eggs and larvae from the FOCI survey observations used in Smart et al. (2012) exhibit large differences in the centers of gravity between cold and warm years that ranged from

162

77 to 186 km (Table 5.3). The centers of all the larval stages were more on-shelf to the northeast in warm years, whereas the egg stage showed the reverse pattern (Table 5.3, Figure 5.6). The simulation with spawning locations contracted in cold years produced results with longitudinal differences and absolute differences most like the observations for the larval stages (Table 5.3). However, the contracted spawning simulation (Figure 5.6) and all other simulations generated centers of gravity to the northwest of the observed centers of gravity. Additionally, the simulated centers were over the outer shelf or slope, whereas observed centers of gravity of all larval stages in warm years were on the middle shelf (Figure 5.6).

Discussion Coupled biological-physical modeling simulations revealed the importance of variations in oceanographic conditions between anomalously cold and warm years in the eastern Bering Sea on the transport and distribution of Walleye Pollock early life stages. Model simulations suggest that the locations of spawning pollock can drive differences in the horizontal distribution of pollock eggs and larvae. It appears that the influence of sea ice on the distribution and spawning of pollock via water temperatures impacts the early life stage distributions more than the differences in currents between cold and warm years.

Vertical behavior Variations in egg buoyancy and the vertical behavior of larvae can alter dispersal and is an important consideration when developing a biophysical model of fish (Fiksen et al. 2007), which is why five different vertical behavior routines were tested. Ontogenetic changes in egg buoyancy of the Gulf of Alaska pollock population are well understood (Kendall 1994), but egg density differs between the Bering and Gulf populations (Kendall 2001), resulting in dissimilar vertical distributions. Over the Bering Sea shelf, eggs tend to be found at depths <100 m with the greatest concentrations in the upper 20 m (Nishiyama et al. 1986, Smart et al. 2013). There appear to be changes in buoyancy such that middle stage eggs occur higher in the water column than early and late stage eggs (Nishiyama et al. 1986), which is suggested in laboratory measurements of specific gravity (Kendall 2001), though the spread of laboratory estimates of buoyancy results in a weak pattern that cannot be parameterized with confidence. Egg specific gravity would be expected to affect egg depth in stratified regions, but not in well-mixed areas (Kendall 2001). Eggs were neutrally buoyant in the model formulation that produced distributions that best matched both horizontal and vertical observations of all pollock ELS. Our use of neutrally buoyant eggs in the model is reasonable since pollock eggs are found at all depths in the Bering Sea (Smart et al. 2013), though an empirical relationship between buoyancy and egg age would improve simulations.

163

The model results that corresponded best with observed horizontal and vertical distributions simulated yolksac larvae as neutrally buoyant. The swimming ability of fish larvae increases over time (Olla et al. 1996) and is rather weak until the inflation of a gas bladder by the time of first feeding (Davis & Olla 1992). As yolksac larvae have limited swimming capability and contain a buoyant yolksac, it seems reasonable to model their vertical behavior like eggs, in this case as neutrally buoyant. In addition to best model performance, this assumption is validated by observations of yolksac larvae distributed throughout the upper 100 m on the Bering Sea shelf (Smart et al. 2013). Constraining preflexion and late larvae to the mixed layer with directed swimming towards the middle of the mixed layer performed better in model-observation comparisons than either model formulation with diel vertical migration (DVM). The evidence for DVM of pollock larvae on the Bering Sea shelf is mixed and varies by larval size and location on the inner, middle, or outer shelf (Smart et al. 2013). Though an oversimplification, our vertical behavior for feeding larvae represents active swimming to stay within the depths with high prey abundance.

The routine that resulted in the highest model skill simulated eggs and yolksac larvae as neutrally buoyant, and preflexion and late larvae as active swimmers directed towards the center of the mixed layer. This routine only differed from purely passive physical transport for two stages, and as a result the centers of gravity of the stage distributions varied little between models with and without this behavior pattern. Despite multiple vertical behavior routines that tried to represent observed vertical distributions, the second best model did not include any vertical behavior. This may suggest that physical processes in the eastern Bering Sea are more important than active migration in determining the vertical distribution of pollock ELS.

Growth The model results displayed a greater sensitivity to the growth formulation than to vertical behavior. Many of the simulations with temperature-independent growth performed better than those with temperature-dependent growth; however, the model formulation with the lowest RMSE and highest modeling efficiency included temperature-based growth. Comparisons of distributions between warm and cold years were made between different ELS, often defined by length. It is possible that temperature effects on growth rates alone could be responsible for differences in distributions. For example, if transport did not differ between warm and cold years, a given stage could be found at a different location in cold years because it took longer to reach that stage, thus it would be further along on its transport trajectory. This was tested by comparing distributions of simulations with neutrally buoyant ELS with temperature-dependent growth. Indeed, there was a difference in the centers of gravity for warm and cold years, with preflexion and late larvae showing the observed pattern of being more on-shelf in warm years.

164

However, the distance between centers of gravity were not as large as those observed or those produced in other simulations (warm expansion, cold contraction), and the addition of behavior removed the pattern in the preflexion stage. However, the difference between warm and cold years increased for late larvae, thus temperature-dependent growth may play an important role in determining where the largest larvae are found.

The temperature-independent growth rate used in the model simulations was derived from a length-age relationship from larvae collected in the Gulf of Alaska from May through July. This is probably an overestimate for Bering Sea pollock because the Gulf of Alaska experiences warmer temperatures than the Bering Sea (Blood 2002) and these measurements were made at the warmest part of the larval period. The temperature-based growth equations of yolksac and feeding larvae were each parameterized from 3 data points. These growth rates were measured in laboratory studies of larvae feeding on Artemia, which produce lower growth rates compared to a natural assemblage of zooplankton (Porter & Bailey 2007). To account for the nutritional difference between Artemia and natural zooplankton, the growth rate equation of feeding larvae was increased. As yolksac larvae still gain some nutrition from their yolksac, this adjustment was not necessary. More data are needed for a better representation of larval pollock growth. Length should probably increase exponentially over time such that growth rate is a function of both temperature and larval size. Despite the coarse approximation of temperature-dependent growth, the resulting lengths at 90 d under different temperatures compare favorably with age-0 lengths sampled at different times of the year (Siddon et al. 2013). For example, mean model lengths were 14 and 21 mm on May 16, 21 and 28 mm on Jun 13, and 32 and 37 mm on Jul 14 for cold and warm years respectively (Figure 5.7). The observations show lengths 5-10 mm on May 15 and Jun 15, and 10-25 mm on Jul 15. Thus, the temperature-dependent growth rate may overestimate growth, but not as much as the age-based growth rate does.

Interannual climate variability The modeled centers of distribution of pollock eggs and larvae did not differ much between warm and cold years when spawning time and location were held constant, and when spawning times were advanced or delayed by 40 days. These results suggest that climate-related differences in ocean circulation and delays in spawning time were not sufficient to cause the observed changes in pollock early life stage distributions in the eastern Bering Sea. Simulations with the combined impact of interannual variability in ocean currents and in spatial shifts in spawning areas were the best at capturing the observed pattern of early life stages located more on-shelf in warm years compared to cold years. In particular, simulations with contraction of the spawning areas off-shelf (to the west and south) created the largest differences between warm and cold centers of gravity.

165

The modeled differences between warm and cold years in the 'physical transport' model are broadly consistent with results from other studies. Mean currents on the middle shelf are relatively weak with current speeds ranging from negligible to about 2 cm s-1 near the surface at a mooring site on the middle shelf (Stabeno et al. 2012). However, Stabeno et al. (2012) found marked differences in current speed and direction at M2 between warm and cold years with a difference in east-west velocity on the order of 1-2.5 cm s-1. This implies a difference in expected transport between warm and cold years of approximately 78-194 km over a 90-day period, compared to the estimated difference in the modeled COG for late larvae of 52.6 km (Table 5.3). A smaller difference in the center of gravity, which reflects the average endpoint, is expected because eggs and larvae generally do not follow straight-line trajectories. Modeled transport is also highly variable over time as evident in the large differences in relative transport between different life stages in our model (Table 5.3), consistent with seasonal variability in currents (Stabeno et al. 2012).

Our analyses focused on differences in transport between warm and cold years, but these differences should not be interpreted as a consequence of temperature variability. Rather, differences in both temperature and transport result from variability in wind forcing, which is associated with variability in along-isobath and cross-isobath advection (Danielson et al 2011b). Cross-isobath fluxes in particular are strongly dependent on wind direction (Danielson et al 2012) as determined by the seasonal mean zonal position of the Aleutian Low (Danielson et al. 2011b). Strong winds from the northwest are associated with larger westward transports and cold conditions, while stronger winds from the southeast are associated with enhanced eastward and northward flows and warmer conditions (Danielson et al 2011b, 2012). However, winds during warm (cold) years are not always favorable to on-shelf (off-shelf) Ekman transport, therefore the differences in modeled transport between warm and cold years in this study may underestimate the importance of physical transport.

Nevertheless, model results suggest that spatial shifts in spawning distribution underlie the shifts in egg and larval distributions. There is a significant lack of empirical information on the factors regulating where and when adults spawn. We propose that sea ice and water temperature affect the pollock spawning distribution in the eastern Bering Sea. Some support for our hypothesis comes from 30 years of summer (May-Aug) bottom trawl survey data that demonstrates an effect of the areal extent of the cold pool defined at 0C on the eastern Bering Sea pollock distribution (Kotwicki & Lauth 2013). The similarity between distributions in pairwise comparisons of all years decreased with increases in the change of the cold pool extent (Kotwicki & Lauth 2013). Resembling the larval distributions, the center

166

of gravity of adult summer distributions differed between warm and cold years, with centers located 42 km off-shelf in cold years (Kotwicki & Lauth 2013).

Thermally influenced shifts in adult distributions, spawning migrations, and ELS distributions have been noted for a number of species, such as Atlantic cod (Gadus morhua; deYoung & Rose 1993) and capelin (Mallotus villosus; Rose 2005), in addition to other pollock stocks (Bacheler et al. 2009). Bacheler et al. (2009) found spatial effects of spawning stock biomass, transport, and temperature on pollock egg density in the Shelikof Strait using a spatially explicit variable coefficient generalized additive model. With increased temperature, egg density increased around the northern edge of Kodiak Island and decreased in the southwest of Shelikof Strait (Bacheler et al. 2009), suggesting shifts in spawning areas with temperature. A review of data on Newfoundland cod by deYoung and Rose (1993) provided evidence for shifts in adult and larval distributions with temperature with a link to recruitment. They found more southerly distributions of adults in cold years that resulted in more southerly distributions of eggs and larvae, placing them in regions with lower retention (deYoung & Rose 1993). They posited that the reduced recruitment in cold years was a consequence of larvae spending less time on the shelf (deYoung & Rose 1993).

Much like the temperature-induced changes in spawning and larval distributions of cod off Newfoundland, the eastern Bering Sea pollock ELS distributions varied with respect to local circulation. The differences in centers of gravity between the simulations of warm years with regular spawning and cold years with off-shelf contracted spawning arise from the lack of eggs and larvae on the middle shelf off Unimak Island and the Alaskan Peninsula in cold years. This difference highlights the importance of the spatial variability of the currents on the eastern Bering Sea shelf. Contraction of spawning off-shelf reduces the number of eggs and larvae exposed to the cross-shelf currents near Unimak Island, including the Alaska Coastal Current, while increasing the proportion of eggs and larvae exposed to more energetic along-shelf currents over the outer shelf and slope. There is also retentive circulation around the Pribilof Islands (Kowalik & Stabeno 1999) that may have facilitated the presence of pollock on the middle shelf when spawning was shoalward of the 100-m isobath in the simulations without contracted spawning.

The model results emphasize the interdependencies of cross-shelf transport and thermal regime to pollock ELS and of spawning near cross-shelf features. Tests of spawning distributions were implemented as generalized off-shelf or on-shelf contractions or expansions because a consistent observational data set of adults in spawning condition that spanned cold and warm years was not available. The hypothetical variations in spawning areas were constructed from a relationship between SST and adult pollock winter abundance (Barbeaux 2012). The generalized shifts removed spawning from areas that tended to be

167

<2.4C in cold years and to expand spawning in regions usually <3.8C, though these shifts were imprecise. More realistic spawning locations could be generated for each year using the annual distributions of SST and initializing eggs in places with SST between 2.4 and 3.8C. A better approach would be to formulate a relationship between water temperature and the presence of adult pollock in spawning conditions. We intend to develop this relationship and use it to repeat the hindcasts of this study as well as complete forecasts with Intergovernmental Panel on Climate Change (IPCC) scenarios.

Expectations with climate change Projections with IPCC climate models predict decreased sea ice extent and increased sea surface temperatures in the eastern Bering Sea (Wang et al. 2012), conditions much like the warm years of 2002- 2005. During these warm years the abundance of large, energy-rich zooplankton decreased and abundance of small, energy-poor zooplankton increased (Coyle et al. 2011). The asymmetric effect of temperature on development and growth suggest that the typically large, energy-rich copepods would have been smaller with fewer lipid reserves in warm years as well (Coyle et al. 2011). The shift in the zooplankton community has a two-fold effect on young pollock. Firstly, the decrease in lipid-rich prey would decrease the energy density of age-0 pollock, thereby reducing the probability of overwinter survival (Moss et al. 2009, Heintz et al. in press). Secondly, age-0 pollock experience increased predation pressure from predatory fishes that would otherwise feed on the large, energy-rich zooplankton (Moss et al. 2009, Coyle et al. 2011). Combined, the decreased energy reserves for overwintering and increased predation pressure suggest lower pollock survival and recruitment in warm years.

Changes in ice extent and water temperatures are expected to alter the distribution of spawning pollock adults. Model simulations demonstrated that simplified expansions and contractions of the spawning areas affect the distributions of pollock ELS. Going beyond distributions, it is important to understand how transport pathways and connectivity differ between warm and cold years. Future studies will examine how potential prey and predators overlap spatially and temporally with ELS of pollock during transport and with the end (summer) distribution of juveniles (age-0’s) in warm and cold years. To this end, we will investigate if variations in the transport pathways between warm and cold years contribute to recruitment variability, or if recruitment is more strongly related to annual shifts in the zooplankton community and processes that occur during the transition from age-0 to age-1.

Considerations The model simulations with contracted spawning areas in cold years qualitatively captured the observed difference in horizontal distributions of pollock larvae between cold and warm years, though they were not perfect representations. Centers of gravity were calculated from the observations collected 168

on the cruises given in Smart et al. (2012), yet yielded egg centers of gravity that were more off-shelf in warm years. This was different from the lack of a temperature effect on egg distribution found by Smart et al. (2012). The observed centers of gravity of all larval stages were more on-shelf in warm years, but both cold and warm centers were located further southeast than the simulation results. Moreover, the observed centers of gravity in warm years were over the middle shelf, whereas the modeled centers were over the outer shelf in all warm year simulations.

There are a number of potential reasons that the model did not completely replicate the observed distributions of pollock eggs and larvae. One explanation is related to the differences between the observations and model results. The observations were collected in the months of March through September, while the model represented the months January through July. The monthly observations and model results were aggregated by year, and then by cold or warm period, rather than compared at particular times and locations. Observations occurred at discrete locations, whereas modeled particles essentially covered the eastern Bering Sea. Though the model results were restricted to only particles in areas that were sampled in both warm and cold years, this could contribute to divergences. Secondly, actual spawning times and locations may result in simulated distributions more similar to the observations than the generalized spawning used in this research, as mentioned above. Finally, spatial variability in both egg production and the mortality of all stages, and their seasonal and interannual changes, could significantly shape the distribution of pollock eggs and larvae. As this study was focused on interannual differences in advection, these spatial variations were not considered because they could have masked the effect of physical transport alone.

Conclusions An advantage of this modeling approach was the ability to individually test environmental forcing mechanisms to quantitatively assess the impact of each on the distribution of pollock. Interannual variations in advection and advances and delays in spawning time were insufficient at reproducing the observed differences in pollock early life stage distributions between warm and cold years in the eastern Bering Sea. Changes to spawning areas, especially offshore contractions in cold years, resulted in simulated distributions most similar to observations. The location of spawning with respect to cross-shelf circulation patterns was an important factor influencing the distribution of eggs and larvae. Further study is needed on the relationship between the spatial distribution of spawning pollock and the physical environment, and its effect on the distribution of early life stages. We are currently investigating how warm and cold year variations in distribution correlate with recruitment and how they affect age-0 survival by way of spatial overlaps with potential prey and predators.

169

Acknowledgements This manuscript benefitted from data provided by and discussions with Steve Barbeaux and Seth Danielson, in addition to all the research completed by the BEST-BSIERP program. This research was funded by NSF award 1108440. This research is EcoFOCI-0811 to NOAA's Fisheries-Oceanography Coordinated Investigations and BEST-BSIERP contribution number ###.

Literature Cited

Bacheler, N.M, Bailey, K.M., Cianelli, L., Bartolino, V., Chan, K.-S. 2009. Density-dependent, landscape, and climate effects on spawning distribution of walleye pollock Theragra chalcogramma. Marine Ecology Progress Series 391, 1-12.

Bacheler, N.M., Ciannelli, L., Bailey, K., Duffy-Anderson, J.T. 2010. Spatial and temporal patterns of walleye pollock (Theragra chalcogramma) spawning in the eastern Bering Sea inferred from egg and larval distributions. Fisheries Oceanography 19, 107-120.

Barbeaux, S.J. 2012. Scientific acoustic data from commercial fishing vessels: eastern Bering Sea walleye pollock (Theragra chalcogramma). Ph.D. Thesis, University of Washington.

Blanke, B., Raynaud, S. 1997. Kinematics of the Pacific Equatorial Undercurrent: a Eulerian and Lagrangian approach from GCM results. Journal of Physical Oceanography 27, 1038-1053.

Blood, D.M. 2002. Low-temperature incubation of walleye pollock (Theragra chalcogramma) eggs from the southeast Bering Sea shelf and Shelikof Strait, Gulf of Alaska. Deep-Sea Research II 49, 6095-6108.

Brickman, D., Adlandsvik, B., Thygesen, U.H., Parada, C., Rose, K., Hermann, A.J., Edwards, K. 2009. Particle tracking. In: Manual of recommended practices for modeling physical-biological interactions during fish early life. Eds. North, E.W., Gallego, A., Petitgas, P. ICES Cooperative Research Report No. 295, pp 9-19.

Budgell, W. 2005. Numerical simulation of ice-ocean variability in the Barents Sea region. Ocean Dynamics 55, 370–387.

170

Canino, M.F. 1994. Effects of temperature and food availability on growth and RNA/DNA ratios of walleye pollock Theragra chalcogramma (Pallas) eggs and larvae. Journal of Experimental Marine Biology and Ecology 175, 1-16.

Carton, J.A., Giese, B.S. 2008. A reanalysis of ocean climate using Simple Ocean Data Assimilation (SODA). Monthly Weather Review 136, 2999-3017.

Coachman, L.K. 1986. Circulation, water masses, and fluxes on the southeastern Bering Sea shelf. Continental Shelf Research 5, 23-108.

Coyle, K.O., Eisner, L.B., Mueter, F.J., Pinchuk, A.I., Janout, M.A., Cieciel, K.D., Farley, E.V., Andrews, A.G. 2011. Climate change in the southeastern Bering Sea: impacts on pollock stocks and implications for the oscillating control hypothesis. Fisheries Oceanography 20, 139-156.

Curchitser, E.N., Haidvogel, D.B., Hermann, A.J., Dobbins, E.L., Powell, T.M., Kaplan, A. 2005. Multi- scale modeling of the North Pacific Ocean: Assessment and analysis of simulated basin-scale variability (1996–2003). Journal of Geophysical Research 110, C11021, doi:10.1029/2005JC002902.

Curchitser, E.N., Hedstrom, K., Danielson, S., Weingartner, T.J. 2010. Modeling the circulation in the North Aleutian Basin. Department of Interior, OCS Study BOEMRE 2010-028.

Dai, A., Trenberth, K.E. 2002. Estimates of freshwater discharge from continents: Latitudinal and seasonal variations. Journal of Hydrometeorology 3, 660-687. doi:10.1175/1525– 7541(2002)003<0660:eofdfc>2.0.co;2

Danielson, S., Curchitser, E., Hedstrom, K., Weingartner, T., Stabeno, P. 2011a. On ocean and sea ice modes of variability in the Bering Sea. Journal of Geophysical Research 116, C12034, doi:10.1029/2011JC007389

Danielson, S., Eisner, L., Weingartner, T., Aagaard, K. 2011b. Thermal and haline variability over the central Bering Sea shelf: seasonal and interannual perspectives. Continental Shelf Research 31, 539-554.

171

Danielson, S., Hedstrom, K., Aagaard, K., Weingartner, T., Curchitser, E., 2012. Wind-induced reorganization of the Bering shelf circulation. Geophysical Research Letters 39, L08601-L08606.

Davis, M.W., Olla, B.L. 1992. Comparison of growth, behavior and lipid concentrations of walleye pollock Theragra chalcogramma larvae fed lipid-enriched, lipid-deficient and field-collected prey. Marine Ecology Progress Series 90, 23-30. de Vries, P., Döös, K. 2001. Calculating Lagrangian trajectories using time-dependent velocity fields. Journal of Atmospheric and Oceanic Technology 18, 1092-1101. deYoung, B., Rose, G.A. 1993. On recruitment and distribution of Atlantic Cod (Gadus morhua) off Newfoundland. Canadian Journal of Fisheries and Aquatic Sciences 50, 2729-2741.

Döös, K. 1995. Inter-ocean exchange of water masses. Journal of Geophysical Research 100, 13499- 13514.

Döös, K., Engqvist, A. 2007. Assessment of water exchange between a discharge region and the open sea: a comparison of different methodological concepts. Estuarine, Coastal and Shelf Science 74, 585- 597.

Drijfhout, S., de Vries, P., Döös, K., Coward, A. 2003. Impact of eddy-induced transport of the Lagrangian structure of the upper branch of the thermohaline circulation. Journal of Physical Oceanography 33, 2141-2155.

Duffy-Anderson, J., Smart, T.I., Siddon, E.C., Heintz, R., Farley, E., Horne, J., Parker-Stetter, S., Ressler, P., Barbeaux, S., Petrik, C.M. In review. A synthesis of the ecology of walleye pollock (Theragra chalcogramma) in the eastern Bering Sea during the first year of life. Deep-Sea Research II: Topical Studies in Oceanography.

Dunn, J.R., Matarese, A.C. 1987. A review of the early life history of northeast Pacific gadoid fishes. Fisheries Research 5, 163-184.

Engqvist, A., Döös, K., Andrejev, O. 2006. Modeling water exchange and contaminant transport through a Baltic Coastal region. Ambio 35, 435-447.

172

Fiksen, Ø., Jørgensen, C., Kristiansen, T., Vikebø, F., Huse, G. 2007. Linking behavioural ecology and oceanography: larval behavior determines growth, mortality, and dispersal. Marine Ecology Progress Series 347, 195-205.

Heintz, R.A., Siddon. E.C., Farley, Jr., E.V., Napp, J.M. In press. Correlation between recruitment and fall condition of age-0 walleye pollock (Theragra chalcogramma) from the eastern Bering Sea under varying climate conditions. Deep-Sea Research II.

Hermann, A.J., Curchitser, E.N., Dobbins, E.L., Haidvogel, D.B. 2009. A comparison of remote versus local influence of El Nino on the coastal circulation of the Northeast Pacific. Deep-Sea Research II 56, 2427-2443. doi:10.1016/j.dsr2.2009.02.005.

Hinckley, S. 1987. The reproductive biology of walleye pollock, Theragra chalcogramma, in the Bering Sea, with reference to spawning stock structure. Fisheries Bulletin U.S. 85, 481-498.

Hunke, E., Dukowicz, J. 1997. An elastic-viscous-plastic model for sea ice dynamics. Journal of Physical Oceanography 27, 1849-1867.

HYCOM Ocean Prediction website. http://hycom.org/ocean-prediction. Last accessed 25 Nov 2013.

Ianelli, J.N., Barbeaux, S.J., McKelvey, D., Honkalehto, T. 2012a. Assessment of walleye pollock in the Bogoslof Island Region. In: Stock assessment and fishery evaluation report for the groundfish resources of the Bering Sea/Aleutian Islands regions. North Pacific Fishery Management Council, 605 W. 4th Ave., Suite 306, Anchorage, AK. pp. 235-244

Ianelli, J.N., Honkalehto, T., Barbeaux, S., Kotwicki, S., Aydin, K., Williamson, N. 2012b. Assessment of the Walleye Pollock stock in the Eastern Bering Sea. In Stock assessment and fishery evaluation report for the groundfish resources of the Bering Sea/Aleutian Islands regions. North Pacific Fishery Management Council, 605 W. 4th Ave., Suite 306, Anchorage, AK 99501. pp. 51-156.

Jung, K.-M., Kang, S., Kim, S., Kendall, Jr., A.W. 2006. Ecological characteristics of Walleye Pollock eggs and larvae in the southeastern Bering Sea during the late 1970s. Journal of Oceanography 62, 859-871.

173

Kendall, Jr., A.W. 1994. The vertical distribution of eggs and larvae of walleye pollock, Theragra chalcogramma, in Shelikof Strait, Gulf of Alaska. Fisheries Bulletin 92, 540-554.

Kendall, Jr., A.W. 2001. Specific gravity and vertical distribution of walleye pollock (Theragra chalcogramma) eggs. NOASA/AFSC Processed Report 2001-01. 92p.

Kotwicki, S., Lauth, R.R. 2013. Detecting temporal trends and environmentally-driven changes in the spatial distribution of bottom fishes and crabs on the eastern Bering Sea shelf. Deep-Sea Research II 94, 231-243.

Kowalik, Z., Stabeno, P.J. 1999. Trapped motion around the Pribilof Islands in the Bering Sea. J. Geophys. Res. 104, 25667-25684.

Lanksbury, J.A., Duffy-Anderson, J.T., Mier, K.L., Busby, M.S., Stabeno, P.J. 2007. Distribution and transport patterns of northern rock sole, Lepidopsetta polyxystra, larvae in the southeastern Bering Sea. Progress in Oceanography 72, 39-62.

Large, W.G., Yeager, S.G. 2009. The global climatology of an interannually varying air-sea flux data set. Climate Dynamics 33, 341-364. doi:10.1007/s00382-008-0441-3.

Large, W.G., McWilliams, J.C., Doney, S.C. 1994. Oceanic vertical mixing: a review and a model with a nonlocal boundary layer parameterization. Review of Geophysics 29, 363-403.

Matarese, A.C., Kendall, Jr., A.W., Blood, D.M., Vinter, B.M. 1989. Laboratory guide to early life history stages of northeast Pacific fishes. NOAA/National Marine Fisheries Service, NOAA Technical Report NMFS, 80.

Mellor, G.L., Kantha L. 1989. An ice-ocean coupled model. Journal of Geophysical Research 94, 10937- 10954.

Mizobata, K., Saitoh, S., Wang, J. 2008. Interannual variability of summer biochemical enhancement in relation to Mesoscale eddies at the shelf break in the vicinity of the Pribilof Islands, Bering Sea. Deep-Sea Research II 55, 1717-1728.

174

Moss, J.H., Farley, Jr., E.V., Feldman, A.M., Ianelli, J.N. 2009. Spatial distribution, energetic status, and food habits of eastern Bering Sea age-0 walleye pollock. Transactions of the American Fisheries Society 138, 497-505.

Mueter, F.J., Litzow, M.A. 2008. Sea ice retreat alters the biogeography of the Bering Sea continental shelf. Ecological Applications 18, 309-320.

Nishiyama, T., Hirano, K., Haryu, T. 1986. The early life history and feeding habits of larval walleye pollock, Theragra chalcogramma (Pallas), in the southeast Bering Sea. INPFC Groundfish Symposium Paper No. P-10.

Olla, B.I, Davis, M.W., Ryer, C.H., Sogard, S.M. 1996. Behavioural determinants of distribution and survival in early stages of walleye pollock, Theragra chalcogramma, a synthesis of experimental studies. Fisheries Oceanography 5, 167-171.

Peck, M.A., Buckley, L.J., Bengtson, D.A. 2006. Effects of temperature and body size on the swimming speed of larval and juvenile Atlantic cod (Gadus morhua): implications for individual-based modeling. Environmental Biology of Fishes 75, 419-429. doi 10.1007/s10641-006-0031-3.

Porter, S.M., Bailey, K.M. 2007. Optimization of feeding and growth conditions for walleye pollock Theragra chalcogramma (Pallas) larvae reared in the laboratory. AFSC Processed Report 2007- 06. Alaska Fisheries Science Center, NOAA, National Marine Fisheries Service, 7600 Sand Point Way NE, Seattle WA 98115. 20p.

R Development Core Team. 2011. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Australia.

Rienecker, M.M., Suarez, M.J., Gelaro, R., Todling, R., Bacmeiseter, J., Liu, E., Bosilovich, M.G., Schubert, S.D., Takacs, L., Kim, G.-K., Bloom, S., Chen, J., Collins, D., Conaty, A., da Silva, A., Gu, W., Joiner, J., Koster, R.D., Lucchesi, R., Molod, A., Owens, T., Pawson, S., Pegion, P., Redder, C., Reichle, R., Robertson, F.R., Ruddick, A.G., Sienkiewicz, M., Woolen, J. 2011. MERRA: NASA’s modern-era retrospective analysis for research and applications. Journal of Climate 24, 3624-3648. doi:10.1175/jcli-d-11–00015.1

175

Rose, G.A. 2005. Capelin (Mallotus villosus) distribution and climate: a sea “canary” for marine ecosystem change. ICES Journal of Marine Science 62, 1524-1530.

Shchepetkin, A.F., McWilliams, J.C. 2009. Ocean forecasting in terrain-following coordinates: Formulation and skill assessment of the regional ocean modeling system. Journal of Computational Physics 228, 8985-9000. doi :10.1016/j.jcp.2009.09.002.

Siddon, E.C., Heinz, R.A., Mueter, F.J. 2013. Conceptual model of energy allocation in walleye pollock (Theragra chalcogramma) from age-0 to age-1 in the southeastern Bering Sea. Deep-Sea Research II 94, 140-149.

Smart, T.I., Duffy-Anderson, J.T, Horne, J.K. 2012. Alternating temperature states influence Walleye Pollock early life stages in the southeastern Bering Sea. Marine Ecology Progress Series 455, 257-267.

Smart, T.I., Siddon, E.C., Duffy-Anderson, J.T. 2013. Vertical distributions of the early life stages of walleye pollock (Theragra chalcogramma) in the Southeastern Bering Sea. Deep-Sea Research II 94, 201-210.

Stabeno, P. J., Bond, N. A., Kachel, K. B., Salo, S. A., Schumacher, J. D., 2001. Temporal variability in the physical environment over the southeastern Bering Sea. Fisheries Oceanography 10, 81-98.

Stabeno, P.J., Kachel, N., Mordy, C., Righi, D., Salo, S. 2008. An examination of the physical variability around the Pribilof Islands in 2004. Deep-Sea Research II 55, 1701-1716.

Stabeno, P.J., Kachel, N.B., Moore, S.E., Napp, J.M., Sigler, M., Yamaguchi, A., Zerbini, A.N. 2012. Comparison of warm and cold years on the southeastern Bering Sea shelf and some implications for the ecosystem. Deep-Sea Research II 65-70, 31-45.

Steele, M., Morison, J., Ernold, W., Rigor, I., Ortmeyer, M., Shimada, K. 2004. The circulation of summer Pacific haloclines water in the Arctic Ocean. Journal of Geophysical Research 109, C02027, 10.1029/2003JC002009.

176

Stow, C.A., Joliff, J., McGillicuddy, D.J., Doney, S.C., Allen, J.I., Friedrichs, M.A.M., Rose, K., Wallhead, P. 2009. Skill assessment for coupled biological/physical models of marine systems. Journal of Marine Systems 76, 4-15.

Wang, M., Overland, J.E., Stabeno, P.J. 2012. Future climate models of the Bering and Chukchi Seas projected by global climate models. Deep-Sea Research II 65-70, 46-57.

Wespestad, V.G., Fritz, L.W., Ingraham, W.J., Megrey, B.A. 2000. On relationships between cannibalism, climate variability, physical transport, and recruitment success of Bering Sea walleye pollock (Theragra chalcogramma). ICES Journal of Marine Science 57, 272-278.

Wilderbuer, T., Stockhausen, W., Bond, N. 2013. Updated analysis of flatfish recruitment response to climate variability and ocean conditions in the Eastern Bering Sea. Deep-Sea Research II 94, 157- 164.

Woillez, M., Rivoirard, J., Petitgas, P. 2009. Notes on survey-based spatial indicators for monitoring fish populations. Aquatic Living Resources 22, 155-164.

Yoklavich, M.M., Bailey, K.M. 1990. Hatching period, growth and survival of young walleye pollock Theragra chalcogramma as determined from otolith analysis. Marine Ecology Progress Series 64, 13-23.

177

Table 5.1. Skill metrics.

Skill metric Behavior Growth R RMSE MEF age 0.12 0.0036 0.03 1 - passive temperature 0.11 0.0040 -0.19 2 - pre & age 0.10 0.0036 0.03 late MLD temperature 0.11 0.0035 0.10 age 0.07 0.0040 -0.18 3 - all MLD temperature 0.06 0.0044 -0.43 4 - late age 0.07 0.0040 -0.17 DVM temperature 0.06 0.0044 -0.43 5 - pre & age 0.08 0.0040 -0.18 late DVM temperature 0.04 0.0060 -1.66

178

Table 5.2. Center of gravity (COG) sensitivity to behavior and growth. Values for longitude and latitude are differences in the COG (°E and °N, respectively) between scenarios: Behavior minus Passive, Temperature minus Age, or Warm minus Cold (where Behavior = vertical behavior of feeding stages; Passive = all stages neutrally buoyant; Temperature = temperature-dependent growth; Age = age-based length (temperature-independent growth); Warm = warm years combined; Cold = cold years combined). Values for Distance are the corresponding absolute distances (km) between COGs.

Stage Egg Yolksac Preflexion Late Cold -0.01 -0.01 0.03 0.09 Longitude Warm 0.00 -0.02 0.04 0.15 Behavior - Cold 0.00 0.00 0.01 0.11 Latitude Passive Warm 0.00 0.00 0.02 0.15 Cold 0.54 0.47 2.31 13.32 Distance Warm 0.01 0.97 3.00 19.56

Cold 0.00 1.33 0.35 -0.77 Longitude Warm 0.00 0.31 0.44 -0.43 Temperature - Cold 0.00 -0.04 -0.02 -0.03 Latitude Age Warm 0.00 0.00 -0.01 -0.01 Cold 0.00 83.20 22.22 48.31 Distance Warm 0.00 19.15 27.55 26.67

Passive -0.05 -0.76 0.37 0.34 Longitude Behavior -0.04 -0.23 -0.02 0.84 Warm - Cold Passive 0.00 0.00 0.00 0.07 Latitude (Temperature) Behavior 0.00 -0.01 -0.01 0.04 Passive 2.85 47.59 22.82 22.66 Distance Behavior 2.32 14.54 1.59 52.60

179

Table 5.3. Center of gravity (COG) differences between warm and cold years for observations and different simulations. Values for longitude and latitude are differences in the COG (°E and °N, respectively) as warm minus cold. Values for Distance are the corresponding absolute distances (km) between COGs. Observed = observations from the cruises listed in Smart et al. 2012; Transport = physical transport; Delay = 40 d delay in cold years; Early = 40 d early in warm years; Contract = contract off-shelf in cold years; Expand = expand on-shelf in warm years.

Stage

Egg Yolksac Preflexion Late

Observed -1.71 1.26 1.14 2.92

Transport -0.04 -0.23 -0.02 0.84

Delay -0.04 -0.01 0.63 0.16 Longitude Early -0.05 0.09 0.47 1.18

Contract 0.96 1.24 1.64 1.50

Expand 0.29 0.18 0.40 1.16

Observed -0.41 0.19 0.25 0.28

Transport 0.00 -0.01 -0.01 0.04

Delay 0.02 0.01 0.01 0.11 Latitude Early 0.02 -0.01 0.05 0.07

Contract -0.01 -0.08 -0.13 -0.03

Expand 0.08 0.08 0.10 0.07

Observed 116.98 82.00 76.69 185.85

Transport 2.32 14.54 1.59 52.60

Delay 3.32 1.57 39.26 15.84 Distance Early 4.07 5.92 29.92 73.38

Contract 59.70 77.52 102.83 92.83

Expand 20.18 14.34 26.91 72.46

180

181

Figure 5.1. The dominant currents (blue lines) and Walleye Pollock spawning areas (green ovals) of the Eastern Bering Sea. The Alaska coastline is shown in black and the 50, 100, and 200 m isobaths in gray. ACC – Alaska Coastal Current; BSC – Bering Slope Current.

182

Figure 5.2. (a) Temperature-dependent egg development time. (b) Temperature-dependent growth rates of yolksac and feeding (preflexion and late) larvae. (c) Length of larvae with temperature-independent (indep.; solid line) and temperature-dependent (dashed lines) growth rates given constant temperature.

183

Figure 5.3. Spawning initial locations and times used in the physical transport simulations.

184

Figure 5.4. Contour plots of the relative concentration of early life stages in cold year simulations. Relative concentration was calculated as the mean of the fraction of all particles found in each grid cell with observations of all cold years.

185

Figure 5.5. Contour plots of the relative concentration of early life stages in warm year simulations. Relative concentration was calculated as the mean of the fraction of all particles found in each grid cell with observations of all warm years.

186

Figure 5.6. Observed (circles) and simulated Contracted (squares) centers of gravity, major axes, and minor axes for the distributions of different pollock early life stages in cold (blue) and warm (red) years.

187

Figure 5.7. Mean standard length (mm) of larvae 90 d after each bi-weekly spawning in the transport only simulations for cold (black) and warm (gray) years.

188

Chapter 6: Ecology and taxonomy of the early life stages of Arrowtooth Flounder (Atheresthes stomias) and Kamchatka Flounder (A. evermanni) in the eastern Bering Sea

Lisa De Forest1, J.T. Duffy-Anderson1, R.A. Heintz2, A.C. Matarese3, E.C. Siddon2, T.I. Smart1, and I.B. Spies1

1Alaska Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration, Seattle, WA, USA 98115 2Alaska Fisheries Science Center, Auke Bay Laboratories, National Marine Fisheries Service, National Oceanic and Atmospheric Administration, Juneau, AK, USA 99801

Citation: De Forest, L., Duffy-Anderson, J.T., Heintz, R., Matarese, A.M., Siddon, E., Smart, T. and Spies, I. In press. Ecology and taxonomy of the early life stages of Arrowtooth (Atheresthes stomias) and Kamchatka (Atheresthes evermanni) Flounder in the eastern Bering Sea. Deep Sea Research II: Topical Studies in Oceanography. http://dx.doi.org/10.1016/j.dsr2.2014.05.005

189

Abstract

Arrowtooth Flounder (Atheresthes stomias) and Kamchatka Flounder (A. evermanni) are closely related flatfish species that co-occur in the eastern Bering Sea. As adults, Arrowtooth Flounder can be distinguished from Kamchatka Flounder; however, larvae and early juveniles can only be indentified to the genus level due to morphological similarities. This has precluded studies of ecology for the early life stages of both species in the eastern Bering Sea. In this study, we developed a genetic technique to identify the larvae and early juveniles of the two species using mtDNA cytochrome oxidase subunit I (COI). Genetically identified specimens were then examined to determine a visual identification method based on pigment patterns and morphology. Specimens 6.0–12.0 mm SL and ≥ 18.0 mm SL can be identified to the species level, but species identification of individuals 12.1–17.9 mm SL by visual means alone remains elusive. The distribution of larvae (< 25.0 mm SL) of both Arrowtooth Flounder and Kamchatka Flounder is similar in the eastern Bering Sea; however, juvenile (≥ 25.0 mm SL) Kamchatka Flounder occur closer to the shelf break and in deeper water than juvenile Arrowtooth Flounder. Condition was measured in larvae and juveniles of each species by analyzing lipid content (%) and energy density (kJ/g dry mass). Kamchatka Flounder larvae on average had higher lipid content than Arrowtooth Flounder larvae, but were also larger on average than Arrowtooth Flounder larvae in the summer. When corrected for length, both species had similar lipid content in the larval and juvenile stages.

190

Introduction

Arrowtooth Flounder are large (max. size 86 cm total length), predatory flatfish that occur from off the coast of central California, north to the Bering Sea, and west to the Kamchatka Peninsula, Russia (Mecklenburg et al., 2002). Kamchatka Flounder are similar in size, but occur primarily in the western Bering Sea with lesser occurrence in the eastern Bering Sea and along the Aleutian Islands (Mecklenburg et al., 2002). In the Bering Sea, Arrowtooth Flounder have been documented to feed primarily on juvenile and adult Walleye Pollock (Gadus chalcogrammus), euphausiids, and various shrimps (Yang and Livingston, 1986). They, in turn, are consumed by Alaska Skates (Bathyraja parmifera) and Sleeper Sharks (Somniosus pacificus) as adults, and by Pacific Cod (Gadus macrocephalus) and Walleye Pollock as juveniles (Spies et al., 2011). The diet of Kamchatka Flounder is similar and often considered identical to Arrowtooth Flounder (Yang and Livingston, 1986). Both their predation impact and their role as prey for other organisms indicate that Arrowtooth Flounder and Kamchatka Flounder are important constituents of the Bering Sea ecosystem (Aydin et al., 2007). The directed fishing effort for Arrowtooth Flounder has increased in recent years, as has the retention of adults caught in other commercial fisheries, possibly in response to an approximate eight-fold increase in adult Arrowtooth Flounder biomass in the Bering Sea since the early 1980s (Spies et al., 2011). Kamchatka Flounder also experience some fishing pressure as directed commercial fishing has developed in recent years (Wilderbuer et al., 2011).

Despite their many similarities, these two species can be separated as adults by the number of gill rakers on the second upper gill arch (one in Kamchatka Flounder, two to three in Arrowtooth Flounder) and the visibility of the dorsal-most eye from the blind side (top of eye is visible in Arrowtooth Flounder, no part of eye visible in Kamchatka Flounder) (Yang, 1988). Using electrophoretic examiniation of alleles at 32 different loci, Ranck et al. (1986) determined that Arrowtooth Flounder and Kamchatka Flounder were two genetically-distinct species and, based on the distinct distributions of certain alleles, there was no hybridization between the two species. However, adults of Arrowtooth Flounder and Kamchatka Flounder collected in scientific and commercial catches were often recorded as either all Arrowtooth Flounder or the genus Atheresthes (Yang and Livingston, 1986; Spies et al., 2011). In 1991, adult Arrowtooth Flounder and Kamchatka Flounder individuals began to be separated as distinct species by the Alaska Fisheries Science Center (AFSC) Groundfish Assessment Program (Zimmerman and Goddard, 1996). As of summer 2011, fishing management regulations dictated that Arrowtooth Flounder and Kamchatka Flounder must be separated and recorded as distinct species in commercial fishing hauls due to the emergence of a directed fishery for adult Kanchatka Flounder (Wilderbuer et al., 2011).

While adults of both species have been well described, the early life stages in the eastern Bering Sea have not been. High recruitment success may be the cause of the increase in population size of Arrowtooth Flounder in recent years (Spies et al., 2011). Understanding the early life stages is important 191

to understanding recruitment and the first step to understanding recruitment in Arrowtooth Flounder is to identify and separate the early life stages from Kamchatka Flounder. Our current study had the following objectives: 1) use genetic methods to identify to the species level larval and early juvenile Atheresthes collected in the eastern Bering Sea, 2) use the genetically identified individuals to develop a visual identification method that allows identification of historical and future samples, 3) describe the distribution and abundance of the early life stages of both Arrowtooth Flounder and Kamchatka Flounder, and 4) examine condition as indicated by length adjusted dry mass, lipid content (%), and energy density of larvae and juveniles to determine differences in nutritional quality between species.

Methods

Specimen collection

All specimens in this study were either directly collected at sea or were sorted from previously collected preserved samples collected on previous AFSC ichthyoplankton surveys. The specimens collected at sea were obtained on spring and summer cruises from 2006 to 2010 by scientists on AFSC, Bering Ecosystem Study (BEST), and Bering Sea Integrated Ecosystem Research Program (BSIERP) cruises (Table 6.1). Specimens were collected opportunistically, resulting in varied collection and preservation methods during the study. Methods of collection included using bongo nets (60-cm; 500-µm mesh), Multiple Opening/Closing Net and Environmental Sensing System (MOCNESS; 1-m2 with 500- µm mesh; Wiebe et al., 1976), and surface and midwater trawls. Specimens removed from the bongo net were either placed directly into 100% non-denatured ethanol or into 5% formalin after an eyeball was removed. Sampling from the MOCNESS involved removing specimens only from the drogue net, which was open for the duration of the tow; all specimens were immediately frozen at -80º C. The surface and midwater rope trawls sampled above and below the pycnocline; all specimens collected were immediately frozen at -80º C. Specimens sorted from previously collected preserved samples were chosen from cruises occurring from 1994 to 2010 (Table 6.1). The samples were collected on previous AFSC ichthyoplankton surveys using 60-cm bongo nets, MOCNESS tows, Methot trawls (Methot, 1986), Tucker trawls (1-m2 with 500-µm mesh; Tucker, 1951), and a modified bottom trawl (which samples the midwater). Sampling protocol during the cruises, along with sample handling and sorting, follows Matarese et al. (2003). All specimens used in this study were classified as either larvae (3.0–24.9 mm SL) or juveniles (≥ 25.0 mm SL) (Matarese et al., 1989).

Genetic identification

192

Species determination was based on a restriction enzyme digest that cuts DNA at a specific nucleotide sequence that is present in Arrowtooth Flounder but not Kamchatka Flounder. The test was based on an initial sequencing of three adult specimens from each species, followed by testing on 20 adult specimens of which the species identification was known. Fin tissue was used for DNA extraction in all adults and was performed using Qiagen DNA extraction kits (Qiagen, Inc. Valencia, CA). For larval and early juveniles, a single eyeball (from either side of head) removed from either fresh, frozen, or 100% non-denatured ethanol preserved specimens was used for DNA extraction. The rest of the specimen body was placed back in its original preservative or, if frozen, photographed for future visual identification work.

Larval DNA was extracted from a single eyeball using a standard Chelex DNA extraction protocol. The eyeball was submerged in 150 µl of 10% (w/v) Chelex ® 100 resin (BIO-RAD Laboratories, Hercules, CA), heated to 60° C for 20 min followed by 103° C for 25 min. A 750 bp segment of cytochrome oxidase subunit I (COI) was amplified using the following primers: COI_RajaF (5’- CCGCTTAACTCTCAGCCATC-3’) and COI_RajaR (5’-TCAGGGTGACCAAAGAATCA-3’; Spies et ng DNA (1 µl taken from the Chelex supernatant), 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 3.0 mM

MgCl2, 1.5 mM dNTPs, 0.5 pM of each primer, and 0.5 U Bioline Taq polymerase (Bioline USA, Inc., Boston, MA). PCR amplification using Chelex extractions can fail if the supernatant contains impurities such as proteins. Therefore, failed samples were repeated with a 1:10 or 1:100 dilution of the DNA, which typically resulted in success. The thermal cycling protocol consisted of 94° C for 2 min, followed by 40 cycles of 94° C (30 seconds), 57°C (30 sec), and 72° C (30 sec). DNA sequencing of amplified DNA from adult specimens was performed using COI_Raja primers at the University of Washington High Throughput Genomics Center (www.htseq.org).

Following PCR, DNA was digested using Bmt1 (New England Biolabs, Ipswich, MA) according to manufacturer’s instructions, in a total volume of 10:l with 2:l PCR product and 4 U of Bmt1 at 37° C for 1 h. Identification was based on either the presence of a single 750 bp fragment (Kamchatka Flounder) or two fragments consisting of 450 and 300 bp (Arrowtooth Flounder).

This protocol was performed in a laboratory or at sea using an Agilent 2100 Bioanalyzer with a DNA 1000 kit (Agilent Technologies, Santa Clara, CA) and a BIO-RAD DNA Engine Thermalcycler.

Visual identification

193

Specimens identified using genetic methods were examined to develop a visual identification method based on morphological and pigmentation characters. Specimens were examined both with and without knowledge of their genetic identification to determine species-specific characters. Each specimen was critically evaluated with regards to morphology and pigment (such as the number of melanophores in a patch or the general shape and pattern of melanophore patches). Only specimens preserved in the same preservation medium were directly compared to one another. This was to minimize the identification of preservation artifacts as distinguishing species characters.

Morphometric measurements were taken from Arrowtooth Flounder and Kamchatka Flounder larvae chosen from a single cruise. Only specimens initially preserved in formalin were used for morphometric measurements because the other preservation methods (100% ethanol and immediate freezing) shrink tissues and distort morphology. The following measurements were taken: standard length (SL), preanal length (tip of snout to anus), head length, snout length, eye diameter, and body depth (vertical measurement taken at pectoral fin; BD). The measurements were taken using a calibrated digital image analysis system consisting of a video camera attached to a stereo microscope and a computer with image analysis software. All measurements were recorded to the nearest 0.1 mm. A two sample t-test was used to identify statistically significant differences (p < 0.05) between each morphological measurement for the two species. Developmental stage terminology follows Kendall et al. (1984).

Distribution

Selected larval Atheresthes collected on previous AFSC cruises conducted from 1994 to 2010 (Table 6.1) were identified to the species level using pigmentation and morphological characters identified with the visual identification method. The locations of all specimens identified to species were mapped and a mean distribution based on abundance was calculated for each species using ArcGIS software.

Measures of Condition

Biological sampling

Larval and juvenile Atheresthes were collected from five research surveys conducted in the summer and early fall from 2008 to 2010 (Table 6.1). All specimens were measured to the nearest 0.01 mm SL. Specimens < 30.0 mm SL were identified to species using the genetic methods previously described, while specimens > 30.0 mm SL were identified morphologically using the number of gill rakers on the upper arch of the second gill raker. All specimens had an eyeball removed, even if identified visually. Stomach contents were removed prior to chemical analysis.

194

Dry mass

Larvae were individually dried at 60ºC in a drying oven until their weight was stabilized. Juveniles were dried at 135ºC using a LECO Thermogravimetric Analyzer (TGA) 601 or 701, which provided percent moisture values used to convert wet mass to dry mass equivalents.

Estimation of lipid content (%)

For larvae, a sulfo-phospho-vanillin (SPV) colorimetric analysis (Van Handel, 1985) was performed to determine lipid content (%). Dried material was sonicated in 2:1 (by volume) chloroform:methanol solvent in glass centrifuge tubes for 60 min. Washes of 0.88% KCL and 1:1 (by volume) methanol:water were performed on the extracts as in the modified Folch extraction method (Vollenweider et al., 2011). Resulting chloroform extracts were evaporated in a LabConco RapidVap for 30 min at 40 ºC and 250 mbar until reduced to approximately 1 ml in volume. Extracts were evaporated to dryness in 12 mm test tubes on a heating block at 75 ºC and then allowed to cool. Concentrated sulfuric acid was added to the tubes prior to incubation at 100 ºC for 10 min with subsequent cooling. The SPV reagent (1.2 mg/ml vanillin in 80% phosphoric acid) was added to each tube and allowed to develop for 10 min. Absorption was measured on an Agilent 8453 Spectrophotometer at 490 nm and extrapolated from species-specific calibration curves determined prior to analysis. For juveniles, lipid extraction was performed on dried material using a previously described method (Vollenweider et al., 2011) derived from the Folch extraction procedure.

Energy density

Energy density (kJ/g dry mass) was estimated using bomb calorimetry. Homogenized dry tissue was pressed into a pellet form and a Parr Instrument 6725 Semimicro Calorimeter with 6772 Precision Thermometer and 1109A Oxygen Bomb was used to measure the energy released from combustion of the sample pellets. The minimum pellet weight was set at 0.025 g of dry material based on the limits of instrument detection; individual larvae were composited within stations as needed to attain sufficient dry mass.

Statistical analysis

A general linear model was used to identify differences in the nutritional state of Arrowtooth Flounder and Kamchatka Flounder. Lengths were compared by a nested analysis of variance (ANOVA) to account for differences in the numbers of fish collected. Species was the main factor with life stage nested in species and year nested in life stage. Post-hoc comparisons of species were conducted among the life stages using Bonferonni’s adjusted t value. Analysis of covariance (ANCOVA) was employed to

195

compare dry mass and lipid content between species. This approach permitted comparisons when length distributions differed between species and the responses covaried with length. The ANCOVA used species, year, and their interaction as main factors and length as covariate. Analyses were conducted separately for larvae and juveniles because the response variables were heteroscedastic when pooled across life stages. Variables were transformed prior to analysis using logarithms (base 10), and the assumption that both species had equivalent slopes was tested. Student’s t was used to compare energy densities of larvae and juveniles.

Results

Genetics

Using mtDNA COI restriction analysis proved successful in identifying Atheresthes larvae and early juveniles to species in the laboratory (n = 337) and at sea (n = 78). Of the 415 total specimens, 165 were identified as Arrowtooth Flounder, 194 as Kamchatka Flounder, and 56 failed to amplify properly resulting in an unknown identity. Size of specimens examined was 6.0–38.0 mm SL, with the majority of specimens 6.0–11.5 mm SL.

Visual identification

Visual examination of genetically-identified specimens was successful in determining key pigmentation differences between Arrowtooth Flounder and Kamchatka Flounder larvae of 6.0–12.0 mm SL and ≥ 18.0 mm SL. These key pigmentation differences are size specific but not stage specific. Specimens 12.1–17.9 mm SL remain indistinguishable at this time because of the small sample size and a high degree of variation in pigmentation patterns.

Recently hatched Kanchatka Flounder larvae (6.0–7.4 mm SL) can be distinguished from Arrowtooth Flounder of the same size by the presence of pigment dorsal to the gut at the anus and an anterior dorsal pigment patch located halfway between the anus and the caudal fin (Figure 6.1a). Arrowtooth Flounder begin to develop the same pigment patterns at about 7.5 mm SL, resulting in these characteristics no longer being able to be used.

Larvae of 7.5–12.0 mm SL can be identified by two characters. Kamchatka Flounder larvae of 7.5–10.5 mm have pigment on the crown of the head, beginning as a few melanophores and developing to numerous melanophores that cover the entire crown of the head by approximately 9.0 mm SL (Figure 6.1a). In general, Arrowtooth Flounder larvae do not begin to develop head pigment prior to 10.5 mm SL, however, a few individuals do develop one or two two small melanophores on the crown of the head beginning at approximately 8.5 mm SL (Figure 6.1b). Thus, specimens 8.5–10.5 mm SL with only 1–3 196

melanophores on the head cannot be identified to species. In addition to crown pigment, the second character by which Kamchatka Flounder larvae 7.5–12.0 mm SL can be identified is the pigment dorsal to the gut that starts at mid gut and extends to the anus (Figure 6.1a). Arrowtooth Flounder larvae also have pigment dorsal to the gut, but it starts at ¾ gut length and extends to the anus. The combination of these two characters allows larvae 7.5–12.0 mm SL to be identified to species.

Few genetically identified specimens were larger than 12.0 mm SL. Identification methods for larger specimens began with the use of adult characters to identify specimens > 25.0 mm SL. These identified specimens were then used as a guide to begin identifying smaller specimens. The adult characters used to determine the larger specimens were the number of gill rakers on the second upper gill arch: Kamchatka Flounder have one gill raker, Arrowtooth Flounder have two or three. While these structures are not fully formed in specimens of 25.0–27.0 mm SL, fleshy protrusions that are the precursors to the gill rakers are evident. After numerous specimens were separated based on gill raker counts, a pattern in the dorsal pigment bands was seen to be different between the two species. The anterior dorsal patch on Kamchatka Flounder is generally longer (> 5 melanophores long) and the individual melanophores in the patch are close together, often touching and at times coalescing. Additionally, Kamchatka Flounder develop melanophores between the anterior and posterior bands, which eventually connect the two bands at a smaller size than Arrowtooth Flounder (beginning as early as approximately 13.0 mm SL, but generally by 18.0 mm SL) (Figure 6.1a). In contrast, Arrowtooth Flounder tend to have a short (< 5 melanophores long) anterior dorsal band with melanophores that are widely spaced apart. Melanophores do not develop between the anterior and posterior dorsal bands until approximately 25.0 mm SL (Figure 6.1b). The few large larvae and early juveniles used in the genetic technique confirmed the validity of these visual differences.

Morphological measurements were taken on 20 Arrowtooth Flounder larvae and 19 Kamchatka Flounder larvae. Specimens ranged in length from 6.3 to 12.6 mm SL. Head length and eye diameter were not statistically different between the two species, but body depth and preanal length were. Larvae of Kanchatka Flounder have a stouter body (BD = 12.0 ± 1.1% SL) than Arrowtooth Flounder larvae (BD = 9.8 ± 1.4% SL; p < 0.05). In addition to a stouter body, Kamchatka Flounder larvae also have a slightly longer preanal length than Arrowtooth Flounder larvae (36.3 ± 1.8% SL vs. 35.1 ± 1.8% SL; p < 0.05).

Distribution

Examining formalin-preserved specimens from previous AFSC cruises, a total of 1,928 larvae and juveniles were identified to species using pigmentation and morphological characters identified above, with 364 individuals identified to genus because of ambiguous characters or damage. A total of

197

1,315 Arrowtooth Flounder larvae and 480 Kamchatka Flounder larvae were identified; 125 Arrowtooth Flounder juveniles and eight Kamchatka Flounder juveniles were identified.

The majority of larval Arrowtooth Flounder and Kamchatka Flounder were collected along the shelf break in the southeastern Bering Sea between Unimak and Umnak Pass (Figure 6.2). There were also individuals collected farther north in Pribilof Canyon along the shelf break near deep water. The calculated mean distribution for both species is similar (Figure 6.2).

Juveniles of both species were collected primarily north of where the majority of the larvae were collected. The calculated mean distribution of the two species showed that Arrowtooth Flounder juveniles appear to be located farther east and over shallower waters than Kamchatka Flounder juveniles, which have a mean distribution close to Pribilof Canyon and deeper water (Figure 6.2). However, fewer Kamchatka Flounder juveniles were collected and identified (n = 8) than Arrowtooth Flounder juveniles (n = 125), so caution should be used when interpreting this result.

Measures of condition

Biological sampling

Length analysis was restricted to larvae collected in 2009 and 2010 and juveniles collected in 2010. These were stages and years in which sufficient numbers of specimens were collected. No significant difference was found in the lengths of larvae or juveniles sampled in 2009 and 2010 (F2.154 >

0.77, p = 0.466), and larvae were significantly smaller than juveniles (F2.154 > 1000, p < 0.001). Pairwise comparisons of larvae and juveniles between the two species averaged across years indicated that Kamchatka Flounder larvae averaged approximately 25% longer than Arrowtooth Flounder larvae (t = 4.6, p < 0.001), but no difference was detected among juveniles (t = 0.31, p = 0.991) (Table 6.2). Differences between the two sets of larvae therefore accounted for the overall difference detected between species (F1.154 = 6.71, p = 0.010).

Dry mass

The dry weights of larvae did not depend on year nor was there an interaction between year and species (F1.96 < 0.15, p > 0.700). However, larval Arrowtooth Flounder of a fixed length were approximately 25% lighter than similarly sized Kamchtaka Flounder (F1.96 = 4.16, p = 0.044; Figure 6.3a).

No difference in dry mass was detected among juvenile specimens in 2010 (F1.53 = 1.68, p = 0.20; Figure 6.3b). It was not possible to test for year effect due to an inadequate number of samples in 2008 and 2009.

198

Lipid content

In 2009 and 2010 the average lipid content of arrowtooth larvae was 7.5 ± 0.3% (n = 39) of dry mass compared with 10.8 ± 0.7% for Kamchatka Flounder larvae (n = 11; Figure 6.4a). While the lipid content of Kamchatka Flounder larvae averaged approximately 30% more than that of Arrowtooth Flounder larvae, this difference was not significant after controlling for the differences in average length

(F1.45 = 3.15, p = 0.083). Additionally, there was not an effect of year or interaction between year and species on larval lipid content (F1.45 < 2.37, p > 0.130). In 2010, there was no difference between juveniles (F1.25 = 0.03, p = 0.857). Arrowtooth Flounder juveniles averaged 10.2 ± 0.5% lipid (n = 21) in 2010 compared with 10.7 ± 0.7% (n = 7) for Kamchatka Flounder juveniles.

Energy density

Energy density in juveniles was only measured in 2010 and in larvae only in 2009. Only a few larvae could be measured because samples required compositing. There was no statistically significant difference between the energy densities of the two species. Only six measurements of larval energy density were taken and the difference between species was not statistically significant (t = 1.06, p = 0.366) (Figure 6.4b). Arrowtooth Flounder larvae averaged 19.2 ± 0.8 kJ/g dry mass; Kamchatka Flounder larvae averaged 19.8 ± 0.7 kJ/g dry mass. Energy densities between juveniles of the two species were nearly identical (t = 1.19, p = 0.244), averaging 20.5 ± 0.2 kJ/g dry mass for Arrowtooth Flounder (n = 16) and 20.8 ± 0.2 kJ/g dry mass for Kamchatka Flounder (n = 12).

Discussion

With the increasing population size of Arrowtooth Flounder and interest for increased directed fishing effort for both Arrowtooth Flounder and Kamchtaka Flounder, there is a need to fully identify and study both species in all life stages for management and ecological purposes. Adults of Arrowtooth Flounder have been studied with regard to their distribution (Zimmerman and Goddard, 1996), predation effects on Walleye Pollock (Ianelli et al., 2009), and their role as prey for other organisms (Livingston and Jurado-Molina, 2000). The early life stages in the Gulf of Alaska have been described by Blood et al. (2007); however, efforts to conclusively identify larvae and juveniles in the Bering Sea have been confounded by the overwhelming physical similarities between Arrowtooth and Kamchatka Flounder.

Genetics

Genetic methods have proven to be useful tools in identifying unknown specimens to the species or family level using the mtDNA gene COI (Hebert et al., 2002; Dawnay et al., 2007; Rocha et al., 2007). In Alaskan waters, Spies et al. (2006) also successfully used mtDNA COI to identify 15 different skate

199

species. In this current study we have demonstrated that this genetic technique is valid for identifying specimens of Atheresthes to the species level. In addition, the technique developed in this study proved effective for use both in the laboratory and at sea aboard a research vessel. In particular, this ability to conclusively identify specimens to the species level while at sea is quite useful as it can enable a timely adjustment to sampling location or procedure in order to maximize collection of the target species.

Morphology

In the marine environment closely related species are often indistinguishable as larvae due to overlap in pigmentation and morphological characters. Prior to this study, all larval and early juvenile Atheresthes collected in the Bering Sea were identified only to the genus level. Despite numerous physical similarities, unique pigmentation and morphological differences were noted between the two Atheresthes species in the size ranges of 6.0–12.0 mm SL and ≥ 18.0 mm SL. In general, it appears that Kamchatka Flounder acquire certain pigmentation characters at smaller sizes than Arrowtooth Flounder, thus the distinguishing characters noted in this paper can only be used for the specific size ranges to which they correspond. Specimens of 12.1–17.9 mm SL are indistinguishable at this time due because of a high degree of variability in pigment amongst individuals and a low number of genetically identified specimens in this size range. In addition to pigment, larvae are stouter than Arrowtooth Flounder larvae (BD 12.0% SL vs. 9.8% SL) and have a slightly longer snout to anus length (36.3% SL vs. 35.1% SL). These are modest morphological differences and can be difficult to determine but can be helpful when used in conjunction with pigmentation characters to support species identification.

Using the pigment characteristics described in this study, two specimens illustrated by Blood et al. (2007; Figure 6.13) should be re-examined as both specimens were collected in the Bering Sea. The 13.4 mm SL specimen is most likely Kamchatka Flounder, not Arrowtooth Flounder, based on the large amount of pigment present on the head and the pigment present on the dorsal midline between the two dorsal bands. While Arrowtooth Flounder develop pigment on the crown of the head, the amount of pigment present at this size is more indicative of Kamchatka Flounder. In addition, the presence of pigment on the dorsal midline between the two dorsal bands at this size can only be Kamchatka Flounder as arrowtooth do not develop this pigment until approximately 25.0 mm SL. The 10.0 mm SL specimen (Blood et al., 2007; Figure 6.12) may also be a Kamchatka Flounder due to the large amount of pigment located on the crown of the head and within the two dorsal patches. However, complete identification is difficult from the illustrations alone; a close visual examination of the original specimens is needed.

Distribution

200

The larval distributions of Arrowtooth Flounder and Kamchatka Flounder in the eastern Bering Sea are similar. Most of the larvae collected were located in the southern part of the eastern Bering Sea in water > 200 m along the shelf break (Figure 6.2a, b). This may indicate that in the eastern Bering Sea both arrowtooth and Kamchatka Flounder adults spawn in deep water, similar to that described for Arrowtooth Flounder in the Gulf of Alaska by Blood et al. (2007). The distribution of juveniles collected does appear to be different between the two species, but this result is tenuous due to the low number of juvenile Kamchatka Flounder collected (n = 8). However, this result may indicate that juvenile Kamchatka Flounder are migrating to their adult habitat. In general, the majority of adult Kamchatka Flounder are collected in deeper waters and farther to the west than adult Arrowtooth Flounder, which often occur in shallower waters to the east (Zimmerman and Goddard 1996; Wilderbuer et al., 2011). It is possible that the low number of juvenile Kamchatka Flounder collected is due to limited sampling in deeper waters to the west of the shelf break.

Condition

Differences in the condition between Arrowtooth Flounder and Kamchatka Flounder larvae collected in the summer may relate to differences in timing of development. For years in which sufficient numbers of larvae were collected for analysis, Kamchatka Flounder larvae were significantly larger than Arrowtooth Flounder larvae in the summer. Comparison of the lengths of larvae sampled in the summer indicates that the bulk of Arrowtooth Flounder larvae were in the preflexion or flexion stages (Blood et al., 2007). Assuming stage designation is similar for Kamchatka Flounder, the majority of Kamchatka Flounder sampled in the summer were in the flexion and postflexion stages. The difference in developmental stage could account for the differences in the length adjusted dry mass of the two species and the trend towards increased lipid content in larval Kamchatka Flounder. Energy allocation strategies can differ among developmental stages during the early life stages. This study suggests that postflexion larvae of Atheresthes may invest more energy accreting tissue mass than earlier stages. Comparison of the lengths of juveniles sampled indicates that the majority of the fish sampled from both species were in the transformation stage (Blood et al., 2007). Hence there were no differences in condition between juveniles of either species.

The results presented here are the first estimates of lipid and energy density in larval and juvenile Arrowtooth Flounder and Kamchatka Flounder. Lipid content values of juveniles in this study are similar to those reported for Atlantic Halibut (Hippoglossus hippoglossus) fed a diet that promoted growth and survival (Hamre et al., 2002). Energy densities of larvae were similar to those of juveniles and suggests no change in energy allocation strategy between these two stages. This is in contrast to Walleye Pollock which undergo significant changes in both lipid content and energy density between lengths of 20 and 60 mm (Siddon et al., 2012), which is when Walleye Pollock undergo and complete transformation to the 201

juvenile stage (Brown et al., 2001). The only evidence of a shift in lipid content among both species of Atheresthes sampled here is associated with ontogeny in the larval stages.

Conclusion

This study has provided a method to genetically identify larval and early juvenile Arrowtooth Flounder and Kamchatka Flounder both in the field and in the laboratory, and has provided morphological and pigmentation characters to visually identify these two species at small (6.0–12.0 mm SL) and large sizes (≥ 18.0 mm SL). Future sampling effort during June–August when Arrowtooth Flounder and Kanchatka Flounder larvae 12.1–17.9 mm SL are in the water column could aid in completing the visual identification of these two species. Once all Arrowtooth Flounder and Kamchatka Flounder between 6.0 mm SL to the juvenile stage can be identified, all historical samples of Atheresthes collected in 1991– 2010 (19 years) by the AFSC from the eastern Bering Sea can be re-identified to the species level. This collection contains approximately 5,065 individuals that, when identified to the species level, will greatly increase our knowledge of distribution and abundance for these two species. Additionally, increased sampling effort in the deeper water off the shelf break during summer and fall would help define the distribution of the larger sizes of these two species and determine if there is a difference in timing of development for larvae in the summer.

Acknowledgements

The authors would like to thank all the scientists that collected Atheresthes spp., larvae over the years and the officers and crews of the following ships: NOAA ship Miller Freeman, NOAA ship Oscar Dyson, R/V Thomas G. Thompson, USCGC Healy, and the R/V Knorr. We would also like to thank Dan Cooper for his help with genetic identification of specimens while on the May 2010 cruise and for reviewing a draft of this manuscript, Melanie Paquin for her help with genetic identification and sequencing in the laboratory, Ashlee Overdick for her illustrations of both species, and Debbie Blood for also reviewing a draft of this manuscript. This is BEST-BSIERP Bering Sea Project publication number XX. This research is contribution EcoFOCI-XXX to NOAA’s Fisheries-Oceanography Coordinated Investigations. The findings and conclusions in this manuscript are those of the authors, and do not necessarily represent the views of the National Marine Fisheries Service.

Literature Cited

Aydin, K., Gaichas, S., Ortiz, I., Kinzey, D., and Friday, N. 2007. A comparison of the Bering Sea, Gulf of Alaska, and Aleutian Islands large marine ecosystems through food web modeling. U.S. Dept. Commer., NOAA Tech. Memo. NMFS-AFSC-178, 298 p.

202

Blood, D.M, Matarese, A.C., and Busby, M.S. 2007. Spawning, egg development, and early life history dynamics of Arrowtooth Flounder (Atheresthes stomias) in the Gulf of Alaska. NOAA Professional Paper NMFS 7, 28 p.

Brown, A.L., Busby, M.S., and Mier, K.L. 2001. Walleye Pollock Theragra chalcogramma during transformation from the larval to the juvenile stage: otolith and osteological development. Mar. Biol. 139: 845-851

Dawnay, N., Ogden, R., McEwing, R., Carvalho, G.R., and Thrope, R.S. 2007. Validation of the barcoding gene COI for use in forensic genetic species identification. Forensic Sci. Int. 173:1-6.

Hebert, P.D.N., Cywinska, A., Ball, S.L., and deWaard, J.R. 2002. Biological identifications through DNA barcodes. Proc. R. Soc. Lond. B. 270: 313-321.

Hamre, K., Opstad, I., Espe, M., Solbakken, J., Hemre, G.I., and Pittman, K. 2002. Nutrient composition and metamorphosis success of Atlantic halibut (Hippoglossus hippoglossus, L.) larvae fed natural zooplankton or Artemia. Aquac. Nutr. 8: 139-148

Ianelli, J.N., Barbeaux, S., Honkalehto, T., Kotwicki, S., Aydin, K., and Williamson, N. 2009. Assessment of the Walleye Pollock stock in the Eastern Bering Sea. In: Stock assessment and fishery evaluation report for the groundfish resources of the Bering Sea/Aleutian Islands regions. North Pacific Fishery Management Council, 605 W. 4th Ave., Suite 306, Anchorage, AK 99501.

Kendall, A.W., Jr., Ahlstrom, E.H., and Moser, H.G. 1984. Early life history stages and their characters. In: Ontogeny and systematic of fishes. (H.G. Moser, W.J. Richards, D.M. Cohen, M.P. Fahay, A.W. Kendall Jr., and S.L. Richardson, eds). P. 11-22. Spec. Publ. 1 Am. Soc. Ichthyol. Herpetol. Allen Press, Lawrence, KS.

Livingston, P.A., and Jurado-Molina, J. 2000. A multispecies virtual population analysis of the eastern Bering Sea. ICES J. Mar. Sci. 57: 294-299.

Matarese, A.C., Blood, D.M., Picquelle, S.J., and Benson, J.L. 2003. Atlas of abundance and distribution patterns of ichthyoplankton from the northeast Pacific Ocean and Bering Sea ecosystems based on research conducted by the Alaska Fisheries Science Center (1972-1996). NOAA Prof. Paper NMFS 1, 281 p.

Matarese, A.C., Kendall, A.W. Jr., Blood, D.M., and Vinter, B.M. 1989. Laboratory guide to early life history stages of northeast Pacific fishes. U.S Dep. Commer. NOAA Tech. Rep. NMFS 80, 652 p.

203

Mecklenburg, C.W., Mecklenburg, T.A., and Thorsteinson, L.K. 2002. Fishes of Alaska. American Fisheries Society, Bethesda, MD, 1116 p.

Methot, R.D. 1986. Frame trawl for sampling juvenile fish. Calif. Coop. Oceanic Fish. Invest. Rep. 27: 267-278

Ranck, C.L., Utter, F.M., Milner, G. B., and Smith, G. B. 1986. Genetic confirmation of specific distinction of Arrowtooth Flounder, Atheresthes stomias, and Kamchatka flounder, A. evermanni. Fish. Bull. 84: 222-226.

Rocha, L.A., Craig, M.T., and Bowen, B.W. 2007. Phylogeny and the conservation of coral reef fishes. Coral Reefs. 26: 501-512.

Siddon, E.C., Heintz, R.A., and Mueter, F.J. 2013. Conceptual model of energy allocation in Walleye Pollock (Theragra chalcogramma) from age-0 to age-1 in the southeastern Bering Sea. Deep-Sea Res. II. 94: 140-149.

Spies, I.B., Gaichas, S., Stevenson, D.E., Orr, J.W., and Canino, M.F. 2006. DNA-based identification of Alaska skates (Amblyraja, Bathyraja and Raja: Rajidae) using cytochrome c oxidase subunit I (coI) variation. J. Fish. Biol. 69(B): 283-292.

Spies, I.B., Wilderbuer, T.K., Nichol, D.G., and Aydin, K. 2011. Assessment of the Arrowtooth Flounder stock in the Bering Sea/Aleutian Islands. North Pacific Management Council, p. 727-744.

Tucker. G.H. 1951. Relation of fishes and other organisms to the scattering of underwater sound. J. Mar. Res. 10(2): 215–238.

Van Handel, E. 1985. Rapid determination of total lipids in mosquitoes. J. Am. Mosq. Control Assoc. 1 (3): 302-304.

Vollenweider, J.J., Heintz, R., Schaufler, L., and Bradshaw, R. 2011. Seasonal cycles in whole-body proximate composition and energy content of forage fish vary with water depth. Mar. Biol. 158, 413-27.

Wiebe, P.H., Burt, K.H., Boyd, S.H., and Morton, A.W. 1976. A multiple opening/closing net and environmental sensing system for sampling zooplankton. J. Mar. Res. 34: 313–326.

Wilderbuer, T.K., Nichol, D.G., and Lauth, R. 2011. Assessment of the Kamchatka flounder stock in the Bering Sea and Aleutian Islands. North Pacific Management Council, p. 795-801.

204

Yang, M.S. 1988. Morphological differences between two congeneric species of pleuronectid flatfishes: Arrowtooth Flounder, Atheresthes stomias, and Kamchatka flounder, A. evermanni. Fish. Bull. 86: 608-611.

Yang, M.S., and Livingston, P.A. 1986. Food habits and diet overlap of two congeneric species, Atheresthes stomias and Atheresthes evermanni, in the eastern Bering Sea. Fish. Bull. 82: 615- 623.

Zimmerman, M. and Goddard, P. 1996. Biology and distribution of arrowtooth, Athersthes stomias, and Kamchatka, A, evermanni, flounders in Alaskan waters. Fish. Bull. 94: 358-370.

205

Table 6.1. Year and sampling date of cruises on which larval and juvenile Atheresthes specimens were collected. Number of specimens used listed.

Year Dates Genetics Morphometrics Distribution Bioenergetics

11 April–30 April 1 1994 14 July–6 Sept 5 16 Sept–29 Sept 1 1995 20 May–29 May 1 17 July–2 Aug 7 1996 21 July–9 Aug 11 4 Sept–16 Sept 6 29 June–14 June 17 1997 18 July–2 Aug 4 8 Sept–18 Sept 6 1998 6 Sept–18 Sept 3 12 July–26 July 2 1999 20 July–1 Aug 8 2 Sept–19 Sept 17 2000 21 July–3 Aug 29 2001 14 July–29 July 15 2002 28 July–11 Aug 1 2003 19 July–26 July 3 2004 28 July–4 Aug 10 22 May–3 June 3 2005 15 July–21 July 28 8 May–19 May 44 2006 21 May–1 June 3 21 June–28 June 2 7 May–18 May 92 2007 25 July–1 Aug 2 18 Feb–26 Feb 16 1708 2008 19 Feb–28 Feb 6 12 May–21 May 26

206

24 May–30 May 4 3 July–17 July 7 7 25 Feb–4 March 7 278 24 April–4 May 7 7 May–20 May 46 39 2009 2 June–17 June 1 14 June–10 July 75 10 77 2 Sept–30 Sept 9 20 6 May–18 May 62 6 June–26 June 1 2010 17 June–4 July 27 90 27 30 June–17 July 6 16 Aug–26 Sept 59 TOTAL 415 39 2292 190

207

Table 6.2. Lengths of specimens used to measure condition. Measurements are mean ± standard error (number sampled).

Year Stage Arrowtooth Flounder Kamchatka Flounder 14.4 ± 1.0 18.6 2008 Larvae (6) (1)

12.4 ± 0.3 14.5 ± 0.3 Larvae (57) (20) 2009 37.2 ± 0.8 Juveniles n/a (9)

11.9 ± 0.8 15.5 ± 0.6 Larvae (11) (16) 2010 48.7 ± 0.5 49.2 ± 0.6 Juveniles (39) (20)

208

Figure 6.1. Illustrations of eastern Bering Sea a) Kamchatka Flounder (Atheresthes evermanni) and b) Arrowtooth Flounder (A. stomias). Arrows point to key identifying characters between the two species. Illustrations by Ashlee Overdick.

209

Figure 6.2. Distribution of abundance (catch/10m2) for formalin preserved larval and juvenile Atheresthes identified to species using visual identification techniques. a) Arrowtooth Flounder (ATF: Atheresthes stomias). b) Kamchatka Flounder (KF: A. evermanni). In both maps, circles depict larvae (<25 mm SL), triangles depict juveniles (≥25 mm SL), X depicts calculation of mean distribution generated by ArcGIS software of larvae based on abundance, and diamond depicts calculated mean distribution of juveniles based on abundance. The size of circles and triangles is proportional to abundance.

210

Figure 6.3. Length-dry mass relationships for Arrowtooth Flounder (Atheresthes stomias) and Kamchatka Flounder (Atheresthes evermanni) of larvae (a) and juveniles (b). Lines depict the least-squares relationships for all species combined.

0.012 Species Arrowtooth Flounder Kamchatka Flounder

) 0.008

(

r 0.004 D

0.000 9 12 15 18 21 SL (mm)

Species 0.32 Arrowtooth Flounder

Kamchatka Flounder )

g 0.24

(

r 0.16 D

B 0.08

30 36 42 48 54 SL211 (mm)

Figure 6.4. Measures of condition for Arrowtooth Flounder (Atheresthes stomias) and Kamchatka Flounder (A. evermanni) collected in June/July (larvae) and September (juveniles) 2008–2010. a) % lipid (dry mass) and b) energy density (kJ/g). Sample sizes are indicated above each bar.

212

Chapter 7: State of knowledge review and synthesis of the first year of life of Walleye Pollock (Gadus chalcogrammus) in the eastern Bering Sea with comments on implications for recruitment

Duffy-Anderson, J.T.1*, Barbeaux, S.1, Farley, E.2, Heintz, R.2, Horne, J.3, Parker-Stetter, S.3,Petrik, C.4,5, Siddon, E.C.2, Smart, T.I.3,6

1Alaska Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration, 7600 Sand Point Way NE, Seattle, WA 98115, USA

2Auke Bay Laboratories, Alaska Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration, 17109 Point Lena Loop Road, Juneau, AK 99801, USA

3School of Aquatic and Fishery Sciences, University of Washington, Seattle, WA 98195- 5020, USA

4 University of Alaska Fairbanks, School of Fisheries and Ocean Sciences, 17101 Point Lena Loop Road, Juneau, AK 99801 USA

5Present affiliation: University of California Santa Cruz, Institute of Marine Sciences, 110 Shaffer Rd., Santa Cruz, CA 95060, USA

6Present affiliation: Marine Resources Research Institute, South Carolina Department of Natural Resources, Charleston, South Carolina 29422, USA

Citation: Duffy-Anderson, J.T., Barbeaux, S., Farley, E., Heintz, R., Horne, J., Parker-Stetter, S.,Petrik, C., Siddon, E.C., Smart, T.I. In press. State of knowledge review and synthesis of the first year of life of Walleye Pollock (Gadus chalcogrammus) in the eastern Bering Sea with comments on implications for recruitment. Deep-Sea Research II.

213

Introduction

Walleye pollock (Gadus chalcogrammus, previously Theragra chalcogramma, Carr and Marshall 2008; Page et al., 2013) is a sub-arctic gadid found from Japan to the Chukchi Sea to central California. Walleye pollock (hereafter, pollock) occur over continental shelves, 50 – 300 m depth, migrate in large schools, and form seasonal spawning aggregations February – June. A major fishery for the species exists in the Pacific subarctic, where catches have ranged from approximately 0.7 – 1.7 million metric tons in the Eastern Bering Sea (EBS) and Gulf of Alaska (GOA) since 1984 (Hiatt et al., 2010). First-sale value for these catches is in excess of $1 billion US per year in the last decade. Fishery products include roe in the winter and fillets throughout the spring and summer. Not only are pollock of significant commercial interest, they are also a central component of the food web in the eastern Bering Sea, serving as prey for fish, marine mammals, and seabirds (Livingston, 1993; Napp et al., 2000; Wespestad et al., 2000; Sinclair et al., 2008). Since pollock are both economically and ecologically valuable, there is significant interest in understanding their recruitment, determining how and when recruitment is set, discerning the associated effects on demographics, and establishing the influences on the overall community structure, with the ultimate goal of developing appropriate management and conservation strategies for the species. Recruitment strength in pollock is particularly sensitive to events that occur during the early life phases because high abundances of small-sized offspring are more vulnerable to mortality than older, more established life stages (Houde, 1987; 1989).

In an effort to better understand recruitment in pollock it is appropriate to take a careful look at events occurring during the first year of life. This paper will review the current understanding of pollock ecology during the first year (spawning to age-1 juveniles), focusing on populations occurring in the EBS. The EBS was chosen as the system of study as it supports a major pollock fishery and it is experiencing a shift in atmospheric and oceanographic conditions (Stabeno et al., 2012), which may influence ecosystem dynamics and function, including pollock recruitment. The review focuses on pollock populations in the EBS, but makes reference to pollock in other large marine ecosystems in an effort to highlight differences or expand conceptual thought. This review will also identify areas of research that are needed to improve our understanding of EBS pollock recruitment, and provide comments on past and current EBS pollock recruitment paradigms that focus on the first year of life.

We define several early life history terminologies to describe young pollock ontogeny that are variously used throughout the pollock literature. Here, egg describes the life stage that occurs from spawning to hatching, also described as the embryonic phase of development (Miller and Kendall, 2009). The yolk-sac stage begins immediately after hatching and yolk is present for internal nourishment of the animal. Pollock larvae rely on maternally-provisioned yolk for several days to weeks and yolk exhaustion occurs when larvae are 5 – 7 mm standard length (SL). Early-stage larvae describe larvae in the 214

developmental period between exhaustion of the yolk sac and flexion of the notochord, a stage also referred to as preflexion in the literature (Dunn and Matarese, 1987). Late-stage larvae (also referred to as postflexion) are those stages between notochord flexion (10 – 17 mm SL) and the development of the full, adult complement of fin rays. The term age-0 juveniles (also, young-of-the-year) refers to individuals with a full complement of fin rays (transformation typically occurs 30 – 40 mm SL; Brown et al., 2001) that are less than 1 year old, and age-1 juveniles refers to those individuals that have experienced their first birthdays.

Collections of pollock during the first year of life in the EBS date from 1979, providing a rich historical data source from which observations are derived. Ichthyoplankton (eggs and larvae) have been collected using obliquely towed 60-cm bongo nets, 1-m2 Tucker nets, and 1-m2 depth-discrete Multiple Opening and Closing Net and Environmental Sensing Systems (MOCNESS, 333 or 505-µm mesh) by the National Oceanic and Atmospheric Administration (Alaska Fisheries Science Center) as well as by numerous academic institutions (University of Alaska, University of Washington, Oregon State University, among others). Age-0 juveniles have been collected with large rope trawls and with small- mesh trawls, (both of which typically use a set of trawls doors to open the net) and Methot nets (a fixed- frame mounted net) towed in the midwater and at the surface. Age-1 pollock have been collected from bottom trawl surveys and from midwater trawl surveys that are conducted in conjunction with acoustic surveys. Generally, spring surveys in the EBS collect pollock eggs and early-stage larvae, summer surveys collect late-stage larvae and transforming juveniles, and autumn surveys collect age-0 juveniles. There are no overwinter collections and age-1 individuals have been collected in spring and summer the year after spawning (Table 1). For the purpose of this review, focus will be on ecology of egg – age-0 pollock; age-1 pollock will only be addressed as they pertain to emergent overwintering age-0s.

Review

Oceanographic characteristics of the eastern Bering Sea shelf

The EBS is bordered to the east by the Alaska mainland, to the south by the Alaska Peninsula and eastern Aleutian Islands, to the west by the Aleutian Basin, and to the north by Bering Strait (Fig. 1). The continental shelf is very broad (~500 kilometers). Northwest flow is driven by the Alaska Coastal Current and flow through Aleutian passes (Napp et al., 2000). East-west flow is driven by the Aleutian North Slope Current. Together these two currents drive flow northward over the shelf. The EBS shelf can be divided into three bathymetric domains during spring and summer, each with its own characteristic hydrography (Coachman, 1986). The inner shelf (or coastal domain, < 50 m depth) is weakly stratified and influenced by freshwater run-off; the middle shelf (50 –100 m depth) is strongly stratified; and the 215

outer shelf (100 – 200 m depth) is an area of intermittent upwelling in the spring and summer, high productivity, and stratification (Hunt et al., 2002). Ocean thermal conditions are often categorized as warmer- or colder-than-average (hereafter “warm year” and “cold year,” respectively), based on sea-ice extent, timing of sea-ice retreat, water temperatures, and the extent of a pool of cold bottom water (Cold Pool, < 2 ºC) over the middle shelf during the summer (Stabeno et al., 2012). Currents also vary between cold and warm years; during cold years, currents are predominantly westward, while in warm years current direction is variable and seasonal, with northward flow during winter and weak flow in other seasons (Danielson et al., 2012; Stabeno et al., 2012).

Spawning Ecology

Pollock exhibit fidelity to general spawning regions and spawning occurs within predictable time intervals; however, small-scale spatio-temporal variability exists within broad-scale spawning regions and known spawning periods that is believed to be dependent on oceanographic conditions. Pollock are iterative spawners, releasing up to 10 cohorts of eggs each year per individual female (Brodeur et al., 1996). Maturity occurs at 3 – 4 years of age and female fecundity is in the millions (Hinckley, 1987). In the EBS, several active pollock spawning grounds have been identified (Fig. 1). Over the shelf, pollock spawn north of the Alaska Peninsula in the vicinity of Unimak Island in March – April, and proximal to the Pribilof Islands April – August (Jung et al., 2006; Bacheler et al., 2010). Over the slope, pollock spawn in the vicinity of Bogoslof Island February – April, although spawning stock biomass in this area has declined since the early 1990s (Bacheler et al., 2010). Farther north, a potential fourth spawning aggregation has been identified near Zhemchug Canyon, although the phenology of this aggregation is unknown (S. Barbeaux, unpublished data). Likewise, a very large, historical spawning aggregation over the Aleutian Basin, “The Donut Hole” aggregation, was depleted in the 1970s and 1980s, and there remains only a marginal, remnant population in the basin that does not contribute significantly to the shelf stock (Bailey, 2013). Spawning stock biomass on the EBS shelf increased substantially in the 1980s due to an influx of large year classes in the late 1970s and early 1980s (Ianelli et al., 2013). Large recruitment events continued through the 1980s and 1990s resulting in high spawning stock biomass during this period. It is believed that strong recruitment and a resultant increase in spawning biomass was due to a regime shift within the EBS in 1977 favoring juvenile pollock survival (Bailey, 2000; Hunt et al., 2002). Although a period of poor recruitment and continued fishing from 2001 to 2005 resulted in a decline in pollock spawning biomass through 2008, recent data suggest that favorable recruitment events in 2006 and 2008 and lower fishing quotas have led to an increasing spawning biomass trend since 2009 (Ianelli et al., 2013).

Previous work using egg and yolk-sac larval data has suggested a latitudinal cline in pollock spawning phenology (Hinckley 1987; Bacheler et al. 2010). This phenomenon can also be investigated 216

using pre-spawning data. The commercial roe fishery separates pollock roe into grades of different value; and hydrated roe can be applied as a proxy for spawning activity. The tonnage of roe recovered by week and grade was available from the EBS pollock commercial fishery (2001 – 2006) and we combined these data with weekly centroids of vessel fishing locations. A generalized additive model with the proportion of hydrated roe as the dependent variable and latitude and week by year as independent variables was fit using 2-dimensional cubic splines (Wood, 2006). The model explained 62% of the deviance. In general, hydration, taken as a proxy for spawning, occurred earlier in the south and later in the north, with 50% hydrated roe occurring 1 – 5 weeks earlier at 54°N than at 58°N latitude (Fig. 2). This was true for all but two years; in 2004 and 2006 there was little difference in the timing of hydration from north to south. Interannual differences in timing were also apparent with 50% hydrated roe occurring at 56°N as early as week 10 (mid-March) (2004 – 2006) and as late as week 14 (mid-April) (2001 – 2002). Results confirm the latitudinal cline in pollock spawning phenology but also demonstrate interannual plasticity in timing of spawning, the latter likely related to ocean thermal differences among years examined. Climate- mediated shifts in spawning phenology can influence density-independent mortality of pre-recruit life stages through variations in the timing of larval fish production relative to zooplankton production (match-mismatch hypothesis, Cushing, 1981) or variable dispersal of larvae relative to seasonally- established oceanographic features. Density-dependent mortality of older stages as a downstream result of phenological shifts can be manifested through temperature-mediated effects on growth, species interactions and predation, or juvenile condition.

Early studies suggested the existence of several stocks of pollock within the Bering Sea (Mulligan et al., 1989; Mulligan et al., 1992; Bailey et al. 1999), though more recent work on genetic differentiation of pollock populations demonstrated only a weak cline in population structuring from Puget Sound, Washington, USA, to Funka Bay, Japan (Canino et al., 2005). Nevertheless, Canino et al. (2005) suggested the existence of self-recruiting populations at moderate geographic scales, though whether these populations might contribute significantly to the overall population (Smith et al., 1990) remains unresolved. Other unknowns regarding pollock spawning ecology in the EBS include surprisingly little spatio-temporal information for how spawning aggregations are formed, or from which spawning locales resultant larvae are derived. Moreover, there has been work in other parts of the world to show that gadid fecundities are influenced by environmental factors (Marshall, 2009; Kjesbu et al., 1998), but comparable studies are lacking for pollock in the EBS. Increased research into the spawning ecology of adults will help to inform studies on the survival and recruitment of their offspring.

Fertilized Eggs

Eggs appear in the water column in the EBS as early as February in the Bogoslof Island area (Bacheler et al., 2010). Based on a 17-year time series of ichthyoplankton tows, the peak in the egg 217

production cycle for the EBS population occurs in April and May, with most eggs produced by the spawning aggregation north of the Alaska Peninsula and Unimak Island (Smart et al., 2012a). Egg mortality has been estimated on one occasion, in 1977, and was found to be approximately 0.6 d-1 in April and 0.3 d-1 in May (Jung et al. (2006). Eggs collected in the EBS tend to be larger than eggs collected from the Gulf of Alaska, and developmental rates at low temperatures are accelerated in the EBS relative to the Gulf (Blood, 2002) suggesting the EBS population is adapted to colder temperatures associated with winter ice cover. Normal development occurs at temperatures ranging from 0.4 to 3.8 ºC. Below 0 ºC development is abnormal and mortality is high.

The timing of peak abundance of pollock eggs occurs earlier in warm years compared to cold years (Smart et al., 2012b; Table 2), which likely is attributable to either earlier spawning (see Section 2.2) or temperature-dependent effects on hatching and development (Blood, 2002). In the EBS, eggs are found throughout the water column, occurring to depths as deep as 300 m (Smart et al., 2013), but also present in the neustonic layer (Waldron and Vinter, 1978). Sampling to document the deepest extent of pollock eggs in the Aleutian Basin has not occurred, but generally eggs collected over the basin are deeper than those collected over the continental shelf (Kendall, 2001). Over the continental shelf pollock eggs are found ≤ 30 m over all domains, but centers of distribution are < 30 m over the shelf and ≥ 100 m over the slope (Smart et al., 2013).

Larvae

Distribution

Pollock early- and late-stage larvae are found in the vicinity of spawning areas as well as hundreds of kilometers away from spawning grounds. Over the EBS shelf, larvae are concentrated near the Pribilof Islands, probably due to local anti-cyclonic circulation patterns (Stabeno et al., 2008) that act to retain larvae, and north and east of the Alaska Peninsula following transport by the Alaska Coastal Current (Stabeno et al., 1999) from spawning grounds located in the vicinity of Unimak Island. Very few late-stage larvae have been collected in proximity to the spawning grounds at Bogoslof Island, despite the prevalence of yolk-sac and early-stage larvae there (Smart et al., 2012a). This could be the result of low sampling intensity in this area in late spring when older larvae would be present in the water column, or it could indicate high mortality or transport away from this island along the Aleutian Island chain (Fig. 1). Large aggregations of larvae north of the Pribilof Islands have not been documented, but recent evidence of the occurrence of spawning-condition adult pollock over the outer shelf at latitudes north of 58ºN (S. Barbeaux, personal communication) suggests that pollock larvae may be concentrated there in late spring. Increased sampling effort over the outer shelf north of the Pribilof Islands is warranted.

218

Pollock larvae smaller than ~ 15 mm SL can be collected efficiently with several types of ichthyoplankton nets (primarily bongo or MOCNESS) with mesh sizes as small as 333 µm. Fish longer than ~ 50 mm SL (age-0 juveniles) are increasingly adept at avoiding capture by plankton nets but can still be collected by midwater trawls. A problematic area of research on larval pollock occurs at the juncture in lengths between fish collected efficiently with ichthyoplankton nets, and those sizes that are collected in small mesh trawls (age-0s, described below). Postflexion larvae 15 – 30 mm SL are relatively rare in collections by any presently-employed gear types, and constitute a portion of the size population that is significantly understudied. As such, there is little information on distribution, vertical distribution, diet, or growth of larvae within this size range.

During cold years in the EBS, centers of distribution of pollock eggs and larvae are located farther west (outer domain) within the general spawning region than in warm years (middle domain, Fig. 3; Smart et al. 2012b). In cold years, the Cold Pool persists over the southern middle shelf through the summer months (Maeda, 1977; Reed, 1995), but in warm years, the Cold Pool is restricted to the northern shelf. Pollock adults avoid water < 2 ºC (Wyllie-Echeverria and Wooster, 1998) and egg development and hatching is diminished at temperatures ≤ 0 ºC (Blood, 2002). An extensive Cold Pool likely reduces the available shelf habitat for spawning and for successful embryo development, and can also impact feeding success, as recent work has shown that zooplankton prey composition is shifted over the middle shelf in warm years compared to cold (Stabeno et al., 2012; Eisner et al. in press).

Pollock larvae in the EBS are found in the upper 100 m of the water column and depths of occurrence become shallower with ontogeny (Smart et al., 2013). Yolk-sac larvae are found throughout the upper 100 m of the water column, and non-feeding stages show no evidence of diel vertical migration. Late-stage larvae do exhibit diel vertical behavior, with most migration of feeding larval stages occurring above the pycnocline (~ 30 m in the EBS). Larvae are generally deeper during the day (10 – 40 m) than at night (0 – 20 m; Fig. 4), and movements are postulated to be dependent on prey levels, light levels, and predator presence. For comparison, feeding larvae collected form the Gulf of Alaska tended to be deepest at midday and shallowest at dusk, indicating a crepuscular pattern of movement (Kendall et al., 1987; Kendall et al., 1994). Depth occurrences of larvae have implications for variability in transport trajectories, which can ultimately influence the distribution of larvae. Pollock larvae spawned in the vicinity of Unimak Island are subject to a bifurcated flow, either northward along the 100 and 200 m isobaths of the EBS, or eastward along the 50 m isobath. Entrainment below the mixed layer depth (~25 – 30 m) increases the likelihood of northward transport (5 – 8 cm s-1) to the middle and outer shelves, while entrainment above the mixed layer increases the chance of eastward drift (2 – 4 cm s-1) and delivery to the middle and inner shelves (Stabeno et al., 1999). Near-surface entrainment increases exposure to wind- forced circulation, which has the potential to transport larvae far from nursery areas. Indeed, Danielson et

219

al. (2011) and Stabeno et al. (2012) have both demonstrated that variability in circulation over the EBS shelf is influenced by changes in wind direction.

Feeding Ecology and Growth

Pollock first-feeding larvae (preflexion stage, 4.5 – 6.5 SL) over the outer shelf primarily consume copepod eggs and nauplii (Hillgruber et al., 1995). As larvae grow and develop improved swimming capability and larger mouth gapes, the diets diversify to include copepodites, particularly those of Pseudocalanus spp. (Strasburger et al., 2014), barnacle cyprids, and euphausiid calyptopis. In 1995 (a cold year), pollock larvae were found at high densities in mid-May, but prey concentrations were below a hypothesized critical density of 20 prey l-1 derived for Shelikof Strait, Gulf of Alaska (Napp et al., 2000). However, critical prey abundances may be lower for Bering Sea larvae due to the cooler temperatures and lower metabolic rates of larvae. Theilacker and Porter (1995) did not find evidence of starvation at low prey densities in field-caught larvae in the Bering Sea. More recent work using a flow cytometric assay to identify starving larvae in the field suggests that pollock larvae collected over the EBS slope and basin were healthy, though a risk of food limitation (6% of individuals) was noted among larvae collected from over the adjacent shelf (Porter and Bailey, 2011). Finally, in a recent analysis of the Match/Mismatch Hypothesis (Cushing, 1981) for pollock larvae and their Pseudocalanus spp. prey, no relationship was found between an index of spatial mismatch and subsequent recruitment to age-1 (De Forest et al., in revision). Taken collectively, there is presently little evidence for mass starvation of pollock larvae in the EBS, though it is cautioned that weak or starving individuals are ready predation targets, making a true assessment of the food-limited fraction of population difficult to obtain. Larvae collected from the field are those that are healthy, and best able to avoid predator encounter and capture; mortality has already removed the others. In order to better assess the role of food limitation among larval pollock, laboratory studies of critical densities of prey needed over a range of larval sizes and temperatures, with models to describe consequent metabolism and growth, are required.

Energy allocation strategies in larval and juvenile fish reflect competing physiological demands of somatic growth versus lipid storage (Post and Parkinson, 2001) and are a response to differing survival constraints. By maximizing growth and transitioning through the larval period rapidly, larvae minimize exposure to size-dependent predation during this stage. However, overwinter survival is hypothesized to be higher in fish that are both larger and have increased lipid reserves, indicating that energy allocation during the juvenile stage will favor lipid storage while also increasing fish size (Beamish and Mahnken, 2001; Heintz and Vollenweider, 2010). Differing energy allocation strategies for larval (< 25 mm SL) and juvenile (25 mm < X < 100 mm SL) pollock indicate that distinct ontogenetic stages face different survival constraints. Larval fish favor energy allocation to somatic growth and development, presumably

220

in order to escape size-dependent predation, while juvenile fish begin to allocate energy to lipid storage in late summer (Siddon et al., 2013a; Fig. 5).

Historical laboratory efforts to estimate growth suggest rates of 0.1 – 0.7 mm d-1 (Davis and Olla, 1992; Porter and Bailey 2007; Hurst et al. 2013), while field estimates derived from otoliths suggest nearly 0.4 mm d-1 (Walline, 1985). Dell ‘Arciprete (1992) developed an age-length key from pollock otoliths collected in the EBS in 1988 and used this to estimate growth rates for 4 – 8 mm larvae at 0.21 mm d-1. There is some evidence for temperature-mediated growth, a phenomenon that has significant import to studies of atmospheric and oceanographic thermal variability on pollock. For example, Jung et al. (2006) used age-length keys to show that larvae collected in 1976 and 1977 (cold years) grew more slowly (0.12 mm d-1) than larvae collected in 1979 (warm year, 0.23 mm d-1), but suggested that differences were at least partly due to variations in timing of sampling, fish size, and length-specific growth. Smart et al. (2012b) estimated larval growth rates by assuming that most larvae collected in consecutive cruises belonged to the same or dominant cohort. They found that growth rates were dependent on water temperature, and showed that growth rates in cold years tended to be approximately half those of warm years (0.1 mm d-1and 0.2 mm d-1, respectively). Few studies have examined the consequences of short-term thermal stochasticity or longer, more gradual increases in ocean temperature on growth rate, and work to incorporate those consequences into a predictive framework is lacking.

Predation

Predation is likely a significant source of pollock pre-recruit mortality in the eastern Bering Sea; larvae are consumed by vertebrates and invertebrates alike. Most research on invertebrate predation of young pollock has been done either in the laboratory or in the Gulf of Alaska, where it has been shown that euphausiids are consumers of yolk-sac larvae (Bailey et al. 1993; Brodeur and Merati, 1993). Gelatinous macrozooplankton are predators of fish larvae in other marine systems, and their numbers appear to be increasing in the EBS, in particular the large jellyfish, Chrysaora melanaster (Brodeur et al., 2002; Brodeur et al. 2008), but we suggest they are not likely to be major predators of pollock eggs and larvae since the timing of production of jellyfish and larvae is mis-matched, with jellyfish occurring at small sizes in early spring when larvae are most vulnerable. Vertebrates certainly consume pollock early life stages. Spawning adult pollock in the EBS are cannibals of pollock eggs (Schabetsberger et al., 1999) and adult pollock, Pacific cod (Gadus macrocephalus) and arrowtooth flounder (Atheresthes stomias) all consume pre-recruit pollock (Livingston and Juardo-Molina, 2000; Bailey 2000). The minimum size of larval pollock in the diets of piscivorous fishes typically ranges from 10 – 20 mm SL, depending on species of predator. Predation and food limitation (section 2.4.2, above) are likely synergistic factors that act remove pollock larvae from the population, with predators selectively preying on undernourished,

221

weaker, and slower swimming or poorly-reactive individuals. Rates of predation on weaker individuals are not well documented but are likely quite high.

Age-0 Juveniles

Distribution

In their first summer, larvae transition to age-0 juveniles prior to reaching 40 mm SL (Brown et al., 2001). Juvenile pollock have been observed across the southeastern Bering Sea (Winter and Swartzman, 2006; Moss et al., 2009; Hollowed et al., 2012), with the area around the Pribilof Islands identified as an essential nursery habitat (Walters et al., 1988; Traynor and Smith, 1996; Brodeur et al., 1997; Hunt et al., 2002). In that region, age-0s are concentrated in inshore rather than offshore areas, despite better potential feeding conditions in offshore areas (Ciannelli, 2002). Patchiness in the distribution of age-0s is common with aggregations ranging in size from meters to kilometers (Benoit- Bird et al., 2013a). During late summer and early fall Parker-Stetter et al. (2013) found that in 2006 – 2010, age-0 pollock were distributed across the EBS, primarily within the middle and outer shelf regions. In 2006 – 2008, years that were classified as “average” or “cold” based on sea surface temperature (Stabeno et al., 2012), high numbers of age-0 pollock were associated with the pycnocline in the upper 30 m of the water column throughout the EBS (Fig. 6, Fig. 10 in Parker-Stetter et al., 2013) as previously identified in other studies (Bakkala et al.,1985; Swartzman et al., 2002; Moss et al., 2009; Hollowed et al., 2012). In 2009 and 2010, years that were colder than 2006 – 2008, higher numbers of age-0 pollock were observed in deep water below the pycnocline (Fig. 6, Fig. 10 in Parker-Stetter et al., 2013). The presence of age-0 pollock in midwater during summer has previously been observed in the EBS (Miyake et al., 1996; Tang et al., 1996) and in the Gulf of Alaska (Brodeur and Wilson, 1996) and has been attributed to survey timing coinciding with the ontogenetic transition from pelagic to demersal habits, and not due to water column attributes or differences in fish length (Brodeur and Wilson, 1996; Parker-Stetter et al. in review 1).

Two recent studies (Hollowed et al., 2012; Parker-Stetter et al., in review 2) have evaluated the horizontal distribution of age-0 pollock relative to physical, biological, and climate factors. Bottom depth and temperature were important predictors of both near-surface age-0 pollock presence and density in Hollowed et al. (2012). In contrast, Parker-Stetter et al. (in review 2) found that depth, quantity of zooplankton prey, and spring winds were important predictors of near-surface age-0 pollock presence and density. Previous studies have suggested that other factors such as prey abundance and/or predator overlap (Swartzman et al., 1999; Ciannelli et al., 2002a; Winter and Swartzman, 2006; Winter et al., 2007), habitat energetics (Ciannelli, 2002b), bottom depth (Brodeur et al., 1999), and temperature (Brodeur et al., 1999) are also important. 222

Age-0 pollock exhibit diel vertical migration, moving from the pyconcline or deeper into the near-surface waters at night. Swartzman et al. (2002) concluded that migration patterns were dependent on stratification, depth, zooplankton prey, and age-0 pollock size. Schabetsberger et al. (2000) suggested that vertical movements by age-0s were motivated by feeding potential and demonstrated that age-0 pollock and their prey had similar vertical distribution patterns. Bailey (1989) noted that shifts in vertical distribution of age-0s had significant implications for inter-cohort cannibalism, with age-1 and older fish feeding on age-0s occurring at similar depth strata.

Feeding Ecology and Growth

Diet of early age-0 juveniles (30 – 50 mm TL) includes small adult copepods (e.g. Pseudocalanus spp., Oithona spp., and Acartia spp.), euphausiids, amphipods, and chaetognaths (Brodeur, 1998; Brodeur et al., 2000; Strasburger et al., 2014). Feeding and body condition vary spatially, around frontal regions and across habitats (Brodeur et al., 2000; Brodeur et al., 2002; Ciannelli et al., 2002a; Ciannelli et al. 2002b; Swartzman et al. 2002; Schabetsberger et al. 2003), and interannually, which appears related to thermal conditions and associated changes in prey base (Heintz et al., 2013). Age-0 juveniles (30 – 100 mm SL) collected by surface trawls in 2003 – 2006, which were warm years, fed mainly on small copepods and euphausiids. In 2007 – 2009, which were cold years, age-0 juveniles within the same size range fed mainly on euphausiids and large copepods (Coyle et al., 2011). This shift is most likely the result of variation in prey availability rather than feeding ability as juveniles collected in surface trawls in 2004 and 2005 (warm years) were longer with larger mouth gapes than those collected in surface trawls in 2006 and 2007 (an average and a cold year respectively; Moss et al., 2009). A diet shift occurs after approximately 50 mm TL, and larger-sized age-0 juveniles feed on larger prey including euphausiids and large copepods (Calanus spp.) (Brodeur, 1998). The switch in diet coincides with the increased swimming speeds that allow predators to pursue larger, more evasive prey, as well as an increase in gape size that permits consumption of larger, higher-quality prey items. Moss et al. (2009) suggested that juveniles 100 – 130 mm TL are piscivorous and cannibalistic during warmer-than-average conditions, though it remains unclear if any larger pollock in their study might have been small-sized age-1s. Cannibalism has been documented among age-1 pollock in the EBS (Bailey, 1989; Duffy-Anderson et al., 2003).

Prey quality and availability relative to juvenile age-0 pollock energetic condition may also influence recruitment as a function of environmental conditions. Climate effects on prey field composition (Coyle et al., 2011) have been shown to lead to the consumption of high lipid diets in cold years (Heintz et al., 2013; Fig. 7). Estimates of total energy in late summer integrate the effects of climate, diet, growth, and prey abundance and quality at the end of the first growing season. The effects

223

of climate on diet quality and condition of age-0 pollock are consistent with observations of a generally inverse relationship between late summer sea surface temperatures and survival (Mueter et al., 2011).

Prey resources for age-0 juvenile pollock decline in autumn and reduced foraging opportunities coupled with near-freezing temperatures during winter heighten the importance of body size and physiological condition prior to winter onset. In high latitude systems, winter is a period of low light, cold temperatures, and reduced prey availability, and is therefore a significant source of mortality and a determinant of recruitment success of marine fishes (Hurst, 2007a). Availability of lipid-rich prey during this period is important as age-0 pollock face severe energy deficits during winter (e.g., Sogard and Olla, 2000). The window of time during which pollock can provision themselves with lipid occurs between the completion of metamorphosis in early August and the onset of oceanographic winter (Wilson et al., 2011; Siddon et al., 2013a). Those individuals that can consume high lipid prey during this critical period are expected to survive winter better than individuals consuming leaner prey. The energetic status of age-0 pollock in late summer is recognized as a predictor of age-1 abundance during the following summer in the EBS (Heintz et al., 2013). Therefore, late summer (July – September) represents a critical period for energy storage in age-0 pollock, and subsequent energy levels provide an early metric for the prediction of overwinter survival and recruitment success to age-1. Predation

Age-0 juvenile pollock are preyed upon by a wide variety of predators. Age-0s <100 mm SL are eaten by older pollock, yellowfin sole Limanda aspera, arrowtooth flounder, and Pacific cod (Livingston, 1993). Most consumption by adult pollock and arrowtooth flounder on age-0 pollock occurs late in the day (Lang et al., 2000). Some authors have contended that cannibalism is the most important source of mortality of age-0 juvenile pollock in the EBS (Livingston and Jurado-Molina, 2000). Hunsicker et al. (2013) found that arrowtooth flounder population abundance and temperature synergistically affect the spatial overlap of predator (arrowtooth flounder) and prey (pollock) suggesting that the magnitude of the primary source of predation of young pollock, arrowtooth flounder, may shift if projections of a warmer EBS become reality. Predation (and cannibalism) by fishes has been hypothesized to be the most significant contributor of age-0 juvenile mortality (Bailey, 1989; 2000; Livingston and Jurado-Molina, 2000) and an important regulator of overall recruitment, but recent work suggests a significant role for feeding, body condition, and overwinter survival.

Marine birds and mammals constitute a significant, although less pronounced, source of predation on juveniles (Livingston, 1993). Benoit-Bird et al. (2013b) showed that predation on age-0 juvenile pollock by seabirds and northern fur seals (Callorhinus ursinus) was dependent not on overall prey biomass, but on small-scale prey density.

224

Models

Biophysical Models

Several generations of biophysical models have been used to explain factors affecting distribution and abundance of young pollock over the Bering Sea shelf. Spatially explicit, coupled biological-physical models simulate dispersal of the early life stages and the biological processes that occur during this period (Table 3). One of the first biophysical models of the early life stages of pollock in the Bering Sea was used to test the effect of advection and sub-grid scale diffusion on the distribution of larvae (Walsh et al., 1981). The two-dimensional (2-D) physical model tracked distribution of two life history stages, nonfeeding and feeding larvae, in the vicinity of Unimak Island. The results of model simulations supported the idea that spawning date affected the drift trajectory and mortality of individuals.

Later, 2-D float-tracking models developed in the 1990s described near-surface wind-driven transport of pollock larvae in the EBS, but they contained no biological component (Wespestad et al., 1997; Wespestad et al. 2000) and pollock larvae do not typically occur at the surface. These models were used to examine how surface advection during the early life stages related to recruitment success by characterizing annual circulation as good or bad, with the effects of circulation on recruitment feeding into a stock assessment model for pollock. This model was further used to analyze how wind-driven transport affected the distribution of juvenile pollock in relation to the distribution of adults (Wespestad et al., 2000). Strong year classes, which often occurred in warm years, coincided with transport of early life stages to the north and onshore, which separated juveniles from adults. Average year classes corresponded with transport to the NW along the shelf, whereas weaker year classes resulted from little transport, which left juveniles on the outer shelf near adults.

A more complex biophysical model has recently been used to examine stage-specific, 3-D transport trajectories of pollock larvae over the EBS (Petrik et al., in press). The physical model used a version of the 3-D Regional Ocean Modeling System (ROMS) for the North East Pacific (NEP; Curchitser et al. 2010; Danielson et al., 2011) that included the EBS domain. Daily averages of velocity, diffusivity, and temperature were generated with ROMS-NEP6 and used to force the particle-tracking, biological model, TRACMASS (Döös, 1995), which simulated the development, growth and vertical behavior of pollock eggs and larvae. Results showed that shifts in spawning location, rather than changes in current patterns or temporal shifts in spawning time, contributed most to spatial shifts of larvae and juveniles observed over the EBS shelf. It was concluded that field-observed differences in distributions of eggs and larvae over the continental shelf between warm and cold years most likely occur because adults modify their use of spawning habitat in response to changes in the presence and extent of sea ice. Direct observations from the field are necessary to verify this hypothesis. 225

How has understanding of pollock early life ecology been shaped by biophysical modeling studies? Until only recently, physical models were primarily 2-D and incorporated little or no pollock- specific biology. Larvae were considered passive particles since there was limited ability to account for species-specific variations in physiological and behavioral characteristics within the model framework. While early models certainly had utility as predictive tools for evaluating horizontal dispersal relative to mean flow, and they had the advantage of modest computational requirements, their value for enhancing understanding of pollock biology was significantly hampered by the inability to account for species- specific attributes that have the potential to influence, or even change, the dispersal outcome. The legacy of biophysical modeling in the study of pollock biology has been vastly improved since the development of spatially-explicit, 3-D circulation models coupled to Lagrangian particle tracking and behavioral models, and they have the potential for further advancement with the more recent development of full life cycle models than span generations (reviewed by Lett et al., 2010). As they exist now, biophysical models of pelagic fishes, inclusive of pollock, provide powerful tools with which to evaluate long-term shifts in growth, distribution, and recruitment. Use of biophysical models of pollock in the Bering Sea in recruitment prediction scenarios is forthcoming (C. Petrik and F. Mueter, personal communication), but remain regional and single species.

Trophic Models

Trophic models have been developed that link age-0 pollock juveniles with other components of the EBS ecosystem in an effort to determine energy flows, relative importance, and consequences on pollock survivorship and subsequent recruitment. All too often however, these models minimize large portions of the first year period, typically egg and larval, favoring to model those life history phases as dispersed propagules rather than treating them as part of the larger trophic framework. This approach assumes that primary sources of larval mortality are density-independent, an assumption that is increasingly turning out to be biased. While evidence for mass starvation of pollock larvae in the Bering Sea is equivocal (section 2.5.3, above), the role of predation upon larvae weakened by poor nutrition should not be overlooked. For decades predation has been acknowledged to be a major source of mortality among marine fish larvae (Hunter et al. 1981; Bailey and Houde 1989; Leggett and Deblois 1994; Houde, 2002; Gallego et al. 2012), and predation likely plays a critical role in structuring age-0 pollock cohort strength. Knowledge of the links and energy pathways that put larvae at risk of predation, including prey availability, foraging capability, feeding, temperature, biochemical condition, swimming competency, and sensory development provides critical mechanistic understanding of how young pollock interact with the environment, which factors are strongly limiting, and over what developmental stages recruitment controls are likely to act. Given the importance of the first year of life to recruitment success, we encourage development of trophic models that are specifically focused on the first year to better

226

understand mechanisms regulating early life survival and connections between vulnerable young fish and the broader ecosystem.

Despite the limitations described above, trophic models have been successful in identifying energy flows from the age-0 juvenile phase forward. Earlier food web models that included young pollock in the EBS took the form of mass balance food web approaches, such that flow of juvenile age-0 biomass entering and leaving the ecosystem was tracked and accounted for, allowing biomass loss or gain by the system to be realized (Aydin et al., 2007). Mass balance models achieved success in identifying and quantifying key sources of juvenile pollock mortality in the EBS (cannibalism), determining the sensitivity of pollock to fluctuations in prey availability, and identifying key energy flows to upper trophic levels that rely on juvenile pollock production.

Physiological trophic modeling approaches have been applied in the EBS to test the hypothesis that variability in pollock growth and survival is structured, in part, by climate-driven, bottom-up control of zooplankton composition. To this end, the broadly applied Wisconsin bioenergetics modeling approach (Kitchell et al., 1977; Ney, 1990) has been adapted for pollock (Ciannelli et al., 1998; Mazur et al., 2007; Holsman and Aydin, in review). The spatially-explicit model estimates temperature- and weight-specific consumption and assesses changes in predicted growth under varying prey or climate conditions. Model results suggest variability in age-0 pollock growth rates relative to hydrographic features which may influence prey availability and food supply (Ciannelli et al., 2002), and pronounced, thermally-induced changes in the relative foraging rates and daily ration of arrowtooth flounder and Pacific cod (Holsman and Aydin, in review).

A mechanistic, 3-D individual-based trophic model (IBM) has also been developed and applied to age-0 juvenile pollock to predict growth over geographic area. The IBM, based on that of Kristiansen et al. (2009), predicts growth using a mechanistic prey selection component that simulates the feeding behavior of age-0 juvenile pollock on zooplankton. The simulated feeding ecology depends on age-0 development (e.g., swimming speed, gape width, eye sensitivity) and vertical behavior, prey densities and size, as well as light and physical oceanographic conditions. This approach has been used to examine mechanisms underlying observed differences in juvenile pollock energy content between warm and cold years in the EBS (Siddon et al., 2013b). Model results revealed that poor growth conditions resulted from spatial mismatches between juvenile pollock and spatial hot spot areas that are conducive to growth, and were correlated to interannual variations in pollock recruitment.

Most recently, efforts have focused on fully integrated model frameworks, in which output from one model feeds another in a vertical suite of model hierarchies. Efforts such as the Bering Sea Integrated Ecosystem Research Program (http://bsierp.nprb.org/) have provided a platform for the development of 227

linked biophysical- energetics models that offer the opportunity to evaluate spatial predictions of fisheries based on Intergovernmental Panel on Climate Change scenario inputs (2000–2050). At its base, hindcast (1970 – 2005, for model validation) or forecast climate scenarios force a 3-D, 10-km resolution physical oceanographic model (ROMS) that contains an embedded Nutrient-Phytoplankton-Zooplankton-Detritus (NPZD) component. Upper trophic level dynamics are modeled through the age-length based bioenergetics model FEAST (Forage-Euphausiid Abundance in Space and Time), which adopts a layered landscape approach to modeling foraging, growth, movement, and survival of 15 distinct fish groups (I. Ortiz, personal communication). FEAST provides two-way feedback between the NPZD model and the fish groups so that it can capture bottom-up and top-down influences on ecosystem dynamics.

Finally, multi-species stock-assessment models (MSMs) quantify the indirect effects of fisheries harvest on populations and evaluate management trade-offs for fisheries that target several species. MSMs include population dynamics models for each stock and link these through annual predation mortality rates (Jesus-Molina and Livingston, 2004). Holsman et al. (in review) recently used this approach to demonstrate that interactions between fishing pressure and Pacific cod and arrowtooth flounder predation on age-0 juvenile pollock strongly influence adult pollock population biomass. These authors have clearly shown that trophic interactions and harvest rates affect biomass estimates, so the ultimate goal would be to incorporate this information into stock assessment.

Model Caveats

It should be cautioned that the output of any model is only as sound as the input data used in its parameterization and the observational data used in model validation. Therefore, while model efforts have certainly helped researchers understand factors influencing pollock connectivity, trophic interactions, phenology, distribution, survival, and recruitment in the EBS, their value remains contingent upon the supporting laboratory and fieldwork that are indispensable for model success. In the Bering Sea, key observational data are lacking in several critical areas, which compromises the value of hypothesis testing using a model context. One crucial consideration is the need for stage-specific information, given that different developmental stages have varied physiological requirements, behaviors, and ecological constraints that potentially influence model outcomes. Basic information by developmental stage that is wanting includes robust temperature- and size-dependent growth rates, information on prey preferences, consumption rates as a function of prey type, prey biomass, and temperature, assimilation rates, resting metabolism as a function of size and temperature, and maximum growth rates under unlimited food conditions. In order to be fully operative, studies must be conducted across the range of temperatures experienced during the larval period and with prey sources that represent the natural mix of zooplankton encountered in the field.

228

Of course, the argument could be made that rich data sources exist for congeneric species, particularly Atlantic cod, Gadus morhua, which could be used as proxy data in model development, or that field work from other ecosystems, namely from studies of pollock from the Gulf of Alaska, are ready substitutes for the data-poor Bering Sea. This tactic should be approached with caution, as species- specific or regional differences in physiology, development, ecology, and behavior all have the potential to influence model-predicted outcome. An exercise conducted for the purposes of this review highlights these constraints. Growth rates (mm d-1) across temperatures were derived from the literature for gadids from the Bering Sea, Gulf of Alaska, Funka Bay, Gulf of Maine, and Norwegian Sea and plotted relative to one another. Derivations were based on either original empirical models fit to laboratory estimates or otolith derived estimates of age/length, or growth rates back-calculated from field-observed lengths (or weights converted to length using published algorithms). Results show significant differences in temperature-dependent growth among species, within regions, and across ecosystems (Figure 8A). Moreover, comparative age-length plots (Figure 8B) show that growth trajectories vary greatly among species examined, with several species reaching asymptotic size thresholds at temperatures that are well within the optional thermal range for others.

Other examples of subtle but meaningful differences include regional differences in egg buoyancy (Kendall, 2001), which can impact drift rate and trajectory in depth-specific currents, durations of larval development and the pelagic larval period (Auditore et al., 1994), which affect time in the plankton and dispersal potential, vertical movements of juveniles in response to environmental cues (Brodeur and Wilson, 1996; Smart et al., 2013), which influence growth and survival, and predation risk (as reviewed in Ottersen et al., 2014) as it has the potential to control predicted recruitment biomass (Hunsicker et al., 2013). Use of data derived from other sources may be necessary at times, such as when model predictions are the only option for making fisheries management decisions despite incomplete understanding. In these cases, careful consideration and thorough vetting a must when results are to be applied in a predictive framework.

Recruitment paradigms

The term recruitment generally refers to the age at which a fish species becomes vulnerable to the fishery; in the case of pollock this is typically age 3 – 4 (Ianelli et al., 2013). To ecologists however, recruitment is a general term that describes survival from one phase in the life cycle to another. For the purposes of this work, we use recruitment to refer to the survival of an annual cohort to the second year of life (age-1). Here we review the recruitment hypotheses that center on the cohort success of pollock during the first year of life, focusing on hypotheses that affect transport, variability in nutritional status, and vulnerability to predation, to better understand the role of fish early life history in population fluctuation of pollock in the EBS. 229

Transport

The Transport-Cannibalism Hypothesis predicts that strong pollock year classes are produced in years that support high rates of juvenile transport away from the adult habitat (Wespestad et al., 2000; Yamamura, 2005; Sakurai, 2007), reducing the distributional overlap of age-0 pollock with cannibalistic adults. This hypothesis assumes that vertical distributions of predators and prey are similar as well, which has been shown to be the case among cannibalistic age 1+ pollock in the EBS (Duffy-Anderson et al., 2003). Cannibalism of young pollock in the EBS is indeed high; Aydin et al. (2007) estimated that cannibalism accounted for nearly 40% of the total mortality of juvenile pollock in the 1990s. The Transport-Cannibalism hypothesis assumes that 1) juvenile distribution patterns are determined by passive drift of eggs and larvae in the upper water column, and 2) warmer-than-average ocean conditions are associated with transport to the middle and inner shelves where the presence of adults is reduced. Recent evidence suggests eastward distributional shifts in pollock eggs, larvae, and early juveniles in warm years (Smart et al., 2012b), but concomitant eastward shifts in the distribution of adults (Kotwicki et al., 2005) may, in fact, maintain the overlap between adults and juveniles. Generally however, we find that the Transport-Cannibalism Hypothesis retains its applicability to the present. The EBS is broad, flat, and relatively homogeneous, providing little landscape refuge for vulnerable age-0 pollock to shelter from large, cannibalistic adults. Empirical and theoretical evidence (Danielson et al., 2012; Stabeno et al., 2012; Wilderbuer et al., 2013) document the differential oceanographic currents that act to disperse larvae, and studies have shown that overlap indices of adult and juvenile pollock can explain up to 50% of recruitment variability (Mueter et al., 2006). Environmental conditions can modify the strength of predator-prey overlap as has been shown between predatory arrowtooth flounder and age-1 pollock (Hunsicker et al., 2013), but the interplay of spatial co-occurrence and top-down predation clearly continues to measurably modify juvenile pollock survivorship and subsequent recruitment.

The Recruitment Routes Hypothesis was developed to understand recruitment variability in the Oyashio ecosystem off of Japan (Suzaki et al., 2003), but may have application to the Bering Sea. This hypothesis explores the link between physical oceanographic shifts and changes in distribution and abundance of pollock, suggesting that differential transport of eggs from spawning grounds underlies shifts in recruitment success. These authors proposed that when the southward flow of the Oyashio Current was strong and nearshore, successful transport to the Doto nursery resulted in average or good recruitment due to the presence of feeding conditions that favor growth potential. When the Oyashio Current was weakened and offshore, poor year classes were observed. The Recruitment Routes idea has been applied to the EBS to describe factors that contribute to the fluctuation of flatfish populations (Wilderbuer et al., 2002), and is currently under investigation for pollock in the EBS (Petrik et al., in prep.).

230

Nutrition and condition

The Production-Competition Hypothesis predicts that strong year classes in the EBS are produced in years with strong storm activity in the summer (Bond and Overland, 2005) based on the observation that four out of five high recruitment years from 1977 to 2000 included warm and windy summers. Storms that occur after the establishment of summer stratification promote mixing of nutrients from below the thermocline into the euphotic zone where nutrients enhance a prolonged period of primary productivity (Sambrotto et al., 1986; Whitledge et al., 1986). Prolonged productivity then may be converted to high availability of preferred prey for larval pollock, reducing competition. Indeed, increased abundances of late larvae have been associated with moderate to vigorous summer wind mixing, and high abundances of early juveniles were associated with high summer wind mixing from 1988 – 2008 (Smart et al., 2012a). Increased abundances of feeding larvae and early juveniles were associated with high concentrations of copepods in these years as well. However, recent warm and windy years (2003, 2004, 2005) have not produced strong year classes (Coyle et al., 2011) and not all warm and windy years prior to 2000 produced strong year classes (Bond and Overland, 2005). The Production-Competition Hypothesis is probably only partially valid in that bottom-up forcing plays a role in pollock larval survivorship, but agents of mortality acting between the larval stage and recruitment to age-1 can modify or even nullify effects. As an example, Gann et al. (in revision) found that a weak year class of pollock was derived from 2007, despite lower trophic level conditions that initially appeared favorable for recruitment, due to weak summer wind patterns that led to poor missing, high stratification, and low nutrient replenishment.

The Oscillating Control Hypothesis (OCH; Hunt et al., 2002) predicts that the EBS ecosystem alternates between primarily bottom-up control in cold phases and primarily top-down control in warm phases. Warm years are predicted to be favorable to pollock recruitment. This hypothesis is based on the impact of interannual variation in the timing of sea-ice retreat on spring and summer temperature conditions and the extent of the Cold Pool, similar to the Production-Competition Hypothesis. The OCH diverges from the Production-Competition Hypothesis in that zooplankton production is not tied to the onset or persistence of the spring bloom, but rather zooplankton production is directly tied to the water temperature in which zooplankters are grazing on the spring bloom. During cold periods with late ice retreat, there is an ice-associated spring bloom while during warm periods there is an open-water spring bloom, so prey availability for zooplankton is not necessarily limiting in either scenario. In cold periods, however, zooplankton grazing rates, subsequent growth rates, and the onset of reproduction are diminished by temperature (Walsh and McRoy 1986; Huntley and Lopez 1992), thus reducing availability of zooplankton prey to larval fish. Biomasses of copepods that serve as the primary prey items for pollock larvae (Calanus marshallae, Pseudocalanus spp., and Acartia spp.) were reduced in the cool springs of

231

1980 and 1981 relative to the warm springs of 1997 and 1998 (Stockwell et al., 2001; Napp et al., 2002). The hypothesis predicts that several successive warm years producing strong year classes of fish contribute to a build-up on the pollock population in the EBS. However, limitations of the hypothesis were realized after the failure of several successive warm-year year classes (2001 – 2005), and the success of several pollock year classes produced from a series of cold years (2006 – 2009). A recent modification (Hunt et al., 2011) to the original OCH suggests a more dynamic interplay of temperature, sea ice, copepod production, and young pollock than originally conceived, and predicts that cold years are favorable to pollock recruitment. In the revised view, early ice retreat and warm waters still enhance primary, and in turn secondary, production contributing to large numbers of age-0 pollock in spring and summer. However, a lack of large, lipid-rich copepod species during warm years is a factor in poor body condition of age-0 pollock in autumn, contributing to high overwinter mortality and poor survival to age- 1.

The Critical Size Hypothesis is similar to ideas outlined in the OCH but also recognizes the need for juvenile fish to store energy prior to winter (Beamish and Mahnken, 2001; Heintz et al., 2013). The OCH states that recruitment of a cohort depends on both the size and diet of the fish during the period in which they provision themselves prior to their first winter. The time at which larval transformation is complete is a direct function of temperature (Smart et al., 2012b), so under warm spring and summer conditions age-0 pollock should complete larval development early, increasing the length of the provisioning period. However, fish must also encounter abundant and energy-rich food in order to meet the metabolic demands of growth while maximizing energy storage (Siddon et al., 2013a). Actively finding food is particularly important for smaller fish with reduced energy reserves compared to a larger fish with ample stored energy and reduced incentive to search for prey. Failure to encounter food of sufficient quality will produce fish entering winter with lower levels of stored energy. Therefore, the conditions that maximize survival beyond the first summer of life are those that lead to early larval transformation, production of abundant energy-rich prey, spatial overlap between pollock and prey, and maximal energy storage in the form of lipids.

While the body of evidence supporting the link between pre-winter condition and survivorship is building (Sogard and Olla, 2000; Heintz and Vollenweider, 2010), critical questions remain unanswered that complicate the two hypotheses outlined above. Laboratory studies rearing larval pollock at low temperatures (<2 ºC) indicate that pollock not only survive extreme thermal conditions, but are capable of growth during the winter (T. Hurst, personal communication). This observation is supported by results from spatially-explicit bioenergetics models of young-of-the-year pollock in the Bering Sea suggesting that age-0 juveniles grow at temperatures as low as 2 ºC (Ciannelli et al., 1998; Siddon et al., 2013a). If at least some growth can occur over the winter, it may be possible for overwintering fish to compensate for

232

poor initial condition. Metabolic adjustments may be possible, and certainly behavioral thermoregulation is documented for overwintering pollock (Sogard and Olla, 1996; Hurst, 2007b), leaving open the possibility that age-0s could offset some of the expense of small initial body size or poor condition. Additionally, in the absence of good in situ measurements, it is difficult to know how much spatial plasticity of food resources exists overwinter, which can potentially offer low, but growth-sustainable, levels of localized prey resources to small age-0s. Of course, though small fish have reduced metabolic requirements they also have lower overall energy reserves, so the potential to deplete energy stores faster than larger pollock remains. Nevertheless, small, winter-emergent juvenile pollock are collected in spring, indicating that survival of small fish does occur overwinter. Finally, the influence of winter severity on survivorship has not been investigated, but mild winter temperatures could preclude winter starvation if sufficient food resources persist. In light of these unresolved issues, the risk of overwinter mortality due to small body size or poor condition needs further investigation.

The Gas Tank Hypothesis (Sigler et al., in review) builds from the Critical Size Hypothesis to recognize that timing of production across trophic levels is critical to pre-winter conditioning success in pollock during the first year of life. This theory maintains that timing of the spring phytoplankton bloom is central to phytoplankton-zooplankton-pollock energy transfer, but is careful to recognize that dynamics in summer and autumn, particularly temperature-metabolic requirements and predation, remain important to pre-recruit production. Moreover, the hypothesis maintains that production is also dependent upon spatial co-occurrences of phytoplankton-zooplankton-pre-recruit pollock, and that successive warmer- than-average or colder-than-average conditions over the EBS shelf influence overall trophic control of recruitment success to age-1.

Predation

Conceptual models of pollock recruitment as related to predation impacts on age-0s have been best developed in the Gulf of Alaska. The Shifting Control Hypothesis (Bailey, 2000) suggests that control of pollock recruitment in the Gulf shifted from bottom-up control of abiotic parameters acting on feeding and survival of pollock larvae in spring, to top-down control by large piscivorous predators, most notably arrowtooth flounder and Pacific cod feeding on age-0 pollock in summer and autumn. Increases in biomass of these and other piscivores, coupled with the Gulf of Alaska’s narrow continental shelf and constricted landscape geography that acts to bring predators and pollock into close proximity, makes this a feasible hypothesis in the Gulf system but less likely to exert population level control in the EBS. Biomass of predator species is increasing in the EBS, but the broad continental shelf allows for greater spatial separation of predators and prey, making top-down predation control by piscivores less problematic unless factors act to bring predators and prey into close proximity (ex: Cold Pool presence/absence, Transport Cannibalism Hypothesis). Multi-species modeling efforts in the EBS suggest 233

that top-down recruitment control is possible, but caution that it is temporally variable and dependent upon predator population levels (Livingston and Jurado-Molina, 2000).

A corollary to the Shifting Control Hypothesis is the Climate-Biology Hypothesis (Ciannelli et al., 2005). In this refinement of the original Shifting Control idea, the authors proposed that the dominant mechanisms of juvenile pollock mortality (predation) in the Gulf of Alaska could change over contrasting climate regimes, suggesting a dynamic interplay of climate and predation pressure. They demonstrated that during ecosystem phases characterized by elevated sea surface temperatures and high predation on juvenile pollock, pollock recruitment variability and abundance were below average. In contrast, periods characterized by low temperatures and relaxed predation were associated with high pollock abundance and high population variability. The recognition of the interplay of physical and biological forcing factors on pollock recruitment is not unique, but has important implications for the management of the species in the Gulf of Alaska and elsewhere.

Data Gaps

Even after nearly 40 years of active research, basic information on pollock early life dynamics in the EBS remains unresolved. Researchers have disentangled certain aspects of the complex ecology of young pollock, though changes in climate, oceanography, trophic structure, and food web dynamics act synchronously to interrupt, modify, or break down known relationships, prompting a continuing need to re-evaluate and redefine conceptual frameworks. The following sections identify gaps in knowledge of pollock early life ecology, and reflect on the variables that are appropriate for continued monitoring as they may either be uniquely influenced by changing ecosystem dynamics or exert disproportionate control on pollock recruitment.

Spawning location

The vast geographic area over which pollock spawning grounds are located, coupled with the large spatial discontinuities in their locations, make comprehensive sampling of putative pollock spawning locations in the Bering Sea extremely difficult. Moreover, logistical constraints of sampling during late winter and early spring when storms are frequent and sea ice is present complicate complete pre-spawning and spawning sampling. Recent work (De Robertis and Cokelet, 2012) suggests that pollock shift their distribution away from areas of cold water and extensive ice cover, indicating spawning under the ice is probably minimal. There is indirect evidence of spawning north of 60 ºN latitude over the outer shelf in the vicinity of Zhemchug Canyon (see section 2.2), but direct evidence is lacking. Critical questions linger: Does spawning area expand and contract interannually with variable thermal conditions? Does spawning habitat become fragmented by the presence of the Cold Pool? What happens to spawning activity when the preferred spawning area is covered by ice? Pollock in the BS exhibit broad-scale 234

spawning site fidelity, but do they move into deeper water over the basin where temperatures are warmer and sea ice is lacking, thus spawning in locations where currents are less conducive to on-shelf transport of offspring? Continued monitoring of spawning populations, complemented by long-term observations of egg and early larval distributions, is critical to elucidating environmental tolerances of adult pollock during spawning, and potential effects on recruitment.

Climate-induced variability in timing of spawning has been suggested in several studies but questions remain. Predominant currents in the EBS are seasonally altered due to the early spring establishment of the Inner Front along the Alaska Coastal Current (Stabeno et al., 1999; Kachel et al. 2002), which has major implications for transport and distribution of larvae and ultimately for spatial distribution of age-0 juveniles. Pollock adults that spawn in the vicinity of Unimak Island produce larvae that are entrained in one of two major flows, the cross shelf flow of the Alaska Coastal Current that delivers propagules over the southeastern middle shelf, and the northward flowing current along the 100 and 200 m isobaths (Fig. 1) delivering larvae over the outer shelf. Set up of the Inner Front typically occurs in late spring (Kachel et al., 2002), a month or more after the presumed spawning time for pollock in that vicinity (March). Shifts in spawning time relative to establishment of the Inner Front have the potential to affect the proportion of offspring that are delivered to the middle and outer shelves.

Some progress can be made in addressing the questions posed above with regard to spawning ecology by examining long-term variations in maturity scales and by examining shifts in distributions of eggs. Collections of ovaries from annual pre-spawning acoustic surveys have suggested differences in timing of gamete development with thermal regime (Smart et al, 2012a) and differences in spawning time between warm and cold periods in the EBS have been inferred through examination of timing of peak abundances of pollock eggs (Smart et al., 2012b). These approaches suggest that spawning can be delayed by up to a month during cold years. In addition, multi-year surveys of egg distribution can provide indirect evidence for spatial shifts in spawning areas due to climate and oceanographic shifts, as has been shown for Arcto-Norwegian cod (Sundby and Nakken, 2008).

Finally, until now we have not had the technology to localize the natal area for any particular juvenile pollock collected at sea. This information would be extremely valuable in clarifying stock origin, resolving the degree of mixing among Bering Sea source areas (Bogoslof, Unimak, Pribilof, Zhemchug), and answering source-sink connectivity questions. Analysis of otolith elemental composition could help to resolve these issues since otoliths incorporate trace metals into their growth rings that can be related to environmental conditions at the time of the growth period. Otoliths have been used to resolve natal origin, connectivity, and nursery habitat for anadromous species (Thorrold et al., 1998; Kennedy et al., 2002), reef fishes (Ruttenberg et al., 2008), and flatfishes (Cuveliers et al., 2010), but the effort for pelagic fishes has been complicated by the temporal scale of residency relative to dispersal potential 235

(Gillanders, 2011). However, recent work using gadoids shows some promise. Thorisson et al. (2010) were able to trace natal areas of age-0 Atlantic cod in a year with limited oceanic mixing, DiMaria et al. (2011) determined that elemental signatures in larval Pacific cod (Gadus macrocephalus) could reflect thermal as well as chemical variations in rearing water, and a pilot study using pollock collected from the Gulf of Alaska indicated that broad-scale geographic differences in residency may be resolved among age-0s (Fitzgerald et al., 2004). These approaches, if successfully applied in the EBS, would be extremely valuable to begin addressing questions such as, “Which source areas contribute significantly to the overall population biomass?”; “To what degree does pollock input from the Gulf of Alaska affect EBS population structure?;” “What are the connectivity links, rates, and routes between larval, juvenile, and adult fish?;” and “How much local recruitment is there?”. The EBS, with its broad, homogenous seascape and well- mixed water column during the pollock spring-spawning period present challenges, but continued efforts are encouraged.

Ageing

Ambiguities in ability to accurately assign daily ages to larvae and age-0 juveniles from otoliths have hampered early life studies of walleye pollock in the EBS. Determining ages of larval and age-0 pollock in the EBS is of significant interest as several studies indicated the existence of multiple cohorts (Walline, 1985; Nishimura et al., 1996). However, difficulties determining growth increments for fishes at low temperatures are well documented (Campana and Neilson, 1985), and it appears that pollock larvae may not consistently accumulate daily otolith rings in the EBS (K. Bailey, personal communication), a phenomenon that has been demonstrated for other larval fish species as well (Geffen, 1982; McGurk, 1987; Collins et al., 1989). More research into improved or new approaches to assigning daily ages to larval and age-0 pollock in the EBS is an area of much-needed effort.

Another area in which ageing information is needed is to determine whether 100 –150 mm TL juvenile pollock collected in late-summer are large age-0 fish that have not experienced a first winter or small age-1 fish that have undergone a slow-growth period. Typically, inferences of age are made based on fish size and length-frequency histograms, but size alone is not a reliable indicator of age. Moreover, length-frequency histograms are confounded in the EBS where offspring are produced from multiple spawning events over a wide geographic area, and length-based separations into year classes are often imprecise. Examination of otoliths for the presence of a winter hyaline zone (winter check) is possible, but unusual growth zones over the otolith occasionally confound that interpretation (T. Helser, personal communication). Historically-collected pollock (70 – 150 mm TL) from the EBS are available at the Alaska Fisheries Science Center and can be examined for the presence of overwinter signatures, as determined by relative ratios of stable oxygen isotopes. Studies of Atlantic cod demonstrate that peaks in δ18O concentrations in the otolith indicate the animal experienced winter conditions (Gao, 2002), and a 236

similar approach for juvenile pollock would help to resolve age discrepancies. Such information is extremely useful as it can not only help to resolve year classes in the EBS, but it can also be used to determine whether there are spatial or temporal shifts in the size range of age-0s with varying environmental conditions (prey availability, temperature, etc).

Maternal history

It is not presently known what the influence of maternal history is on resultant offspring. Significant work on other species has shown varying oceanographic conditions experienced by the female affect the traits of their offspring, including size at hatching, larval growth, behavior, and survival (Solemdal, 1997; Berkeley et al., 2004). Abiotic influences such as temperature and photoperiod as well as biotic factors such as maternal age, crowding, prey availability, and habitat quality exert measurable phenotypic effects on progeny. Once thought to be comparatively inconsequential to overall population structure, phenotypic effects have been shown to have the potential for demographic influences propagated by variations in offspring performance, including how young experience food limitation, predator encounters, thermal stress, and oceanographic shifts. Moreover, recent work to examine whether pollock reproductive biology affects stock productivity among Gulf of Alaska pollock has shown that weight-specific relative fecundity and maternal weight-at-age influence stock status, which in turn have the potential to influence estimation of fisheries harvest points (Spencer and Dorn, 2013). Accordingly, more effort to quantify the effects of maternal history on population differences in progeny is recommended.

Condition and overwinter success

Work to examine links between environmental conditions and recruitment has focused on the role of juvenile condition and overwintering success. Field work on EBS age-0 pollock provides evidence of a link between survivorship and condition of YOY pollock in near-surface waters. However, it is known that large aggregations of juvenile pollock school below the pycnocline (30 m), and these deeper fish may have different energy content relative to counterparts near surface. It has been hypothesized that poor condition pollock occur in near surface waters during daylight hours because they do not possess the energy reserves to vertically migrate. Vertically-migrating age-0 pollock may have higher energy contents since they remain in close proximity to vertically migrating zooplankton prey. A pilot study which included directed, paired sampling of pollock from the surface and midwater regions suggests no at-station differences in energy content from fish collected above and below the pycnocline (Parker Stetter et al., 2013) but additional work is needed to fully resolve whether there are differences in energy content of age-0s collected across vertical strata.

237

Some laboratory work has been done to examine compensatory growth among food-limited age-0 pollock at optimal temperatures (Sogard and Olla, 2002), but work specifically examining the potential for accelerated growth under simulated late autumn/winter conditions is needed. If prey quality and quantity regulates size and body condition prior to first winter onset then the capacity to compensate for inadequate provisioning during summer seems likely. Spawning in the Bering Sea is seasonally protracted and occurs over several spawning areas, yielding multiple groups of larvae spawned months apart that co- mingle in the water column. Can late-spawned, smaller juveniles catch up to early-spawned larger-sized individuals under low-thermal, winter-simulated conditions if prey resources are available? Or are all the smaller, late-spawned cohorts fated to disproportionate overwinter mortality? If not, and overwinter compensatory growth is possible, are there associated costs?

If size- or condition-selective overwinter mortality occurs in juvenile pollock, a significant body of research is needed to determine demographic effects on the resultant population. For example, are early-spawned progeny that have a protracted opportunity for growth prior to winter onset selected for, potentially scaling the population for early reproduction? If late-spawned cohorts are subject to high rates of overwinter mortality, is there a competitive growth release from population-level density-dependent regulation among survivors? If factors act to delay spawning, or limit habitat areas over which early spawning can take place (e.g., sea ice), what are the effects on resultant overwintering juveniles? Are there tendencies toward expedited or disproportionate energy allocation prior to or during the overwintering period? Are demographic effects modulated by interactions between growth and/or predation? Are effects propagated as the population ages? These and other questions remain unanswered, but warrant attention and study.

Multi-species interactions

Pollock research has benefited from significant single-species attention, but more could be done to investigate interactions between pollock and other species. As an example, it is known that young pollock and Pacific cod spatially and temporally co-occur during the first year of life, but larval pollock and Pacific cod show dissimilarities in diet (Strasburger et al., 2014), while age-0 diets are similar (Farley et al., in press). Do initial diet dissimilarities and resource partitioning help to mitigate the effects of ontogenetic convergence later on? Are diet convergences at the age-0 stage constrained or modulated by habitat separations between demersal Pacific cod and pelagic pollock? Are these relationships maintained during oscillating thermal phases of the EBS ecosystem? Further, studies from the Gulf of Alaska have demonstrated that capelin have a similar prey base as age-0 pollock (Logerwell et al., 2010), and studies in the Bering Sea that examine the relationship between pollock and other midwater zooplanktivores (capelin, eulachon, Pacific herring, jellyfish) are warranted. Multispecies studies that examine these and

238

other interactions, both at the local and broad-scale level, are necessary to better understand community response to climate variation and effects on recruitment potential.

Conclusions and Recommendations

Factors affecting production, growth, and survivorship of pollock during the first year of life underlie population fluctuations in the species, so it remains critical to continue to monitor and study pollock during this period. Because the Bering Sea represents the northern end of their latitudinal range, environmental constraints have the potential to exert significant control over intrinsic rates of pollock population increase. Understanding these processes is critical given that climate conditions in the EBS are changing, with documented increases in air and ocean temperatures, changes in frequency and intensity of storms, declines in sea ice thickness and extent, and affects to ocean stratification, to name only a few. In order to effectively understand the complex ecological processes governing EBS pollock population variability, we must seek mechanistic understanding of the factors influencing processes occurring over the first year of life. To do so we recommend, in order of importance,:

1) Development and implementation of a seasonal monitoring program to examine pollock ecology and dynamics over the entire first year of life, through the winter to the ensuing spring (age-1), with focus on process-oriented research that can resolve mechanistic linkages and pathways of recruitment control; 2) Studies directed to examining predation dynamics and trophic consequences of predation, with particular emphasis on the relationship between food availability and food quality, larval and juvenile body condition, and vulnerability to predators; 3) Studies that focus on larval and age-0 juvenile diet and condition, as well as studies that examine physiological and behavioral responses to prey shifts; 4) Investment in additional laboratory studies to resolve stage-specific baseline information such as temperature-dependent growth, compensatory growth, prey preferences, consumption rates, assimilation rates, and rates of metabolism that can be used to parameterize models; 5) Development of spatially-explicit, mechanistic biophysical, bioenergetics, and trophic models that specifically focus on the first year of life to resolve energy flows, connectivity, trophic constraints, and sources of mortality that may be indistinct using present approaches; 6) Development of stage-specific predictive models (statistical, trophic, and biophysical) to assess implications of variable early life mortality to future recruitment; 7) Work to resolve critical production areas (spawning areas, feeding grounds, nursery habitats) of pre-recruit pollock; 8) Evaluation of potential effects of competition throughout the first year of life from species such as Pacific cod, capelin, Pacific herring, and jellyfish; and 239

9) Development of novel approaches to resolving daily ages of larval and age-0 juvenile pollock in the Bering Sea. The importance of factors influencing the growth and survival of early life stages on recruitment potential has been recognized since the earliest days of Hjort (1914), and certainly recruitment variability during early life remains a leading factor in determining year class strength (Ottersen et al., 2014). For walleye pollock occurring in the Bering Sea, a myriad of influences, environmental, trophic, and intrinsic, act in concert over a suite of developmental phases throughout the critical first year to influence recruitment variability and regulate recruitment strength. Future research should move toward mechanistic understanding of how survival is modulated throughout the first year of life and strive for inclusion in predictive methodologies that can be used in successful management of the species.

Acknowledgements

Special thanks to Patrick Ressler, Franz Mueter, Kevin Bailey, and Thomas Hurst for discussion and to Ann Matarese, Jeff Napp, Mike Sigler, and three anonymous reviewers for comments. Debbie Blood assisted with editing an earlier version of this manuscript. This research was supported, in part, with funds from the Ecosystems and Fisheries Oceanography Coordinated Investigation’s North Pacific Climate Regimes and Ecosystems Program, the North Pacific Research Board’s Bering Sea Integrated Ecosystem Program, and the National Science Foundation’s Bering Ecosystem Study Program. This paper is contribution EcoFOCI-0780 to NOAA's Fisheries-Oceanography Coordinated Investigations Program, BEST-BSIERP Project publication number XX, and NPRB publication number XXX. The findings and conclusions in the paper are those of the authors and do not necessarily represent the views of the National Marine Fisheries Service. Reference to trade names does not imply endorsement by the National Marine Fisheries Service, NOAA.

Literature Cited

Auditore, P.J., Lough, R.G., and Broughton, E.A., 1994. A review of the comparative development of Atlantic cod (Gadus morhua L.) and haddock (Melanogrammus aeglefinus L.) based on an illustrated series of larvae and juveniles from Georges Bank. NAFO Sci. Coun. Studies 20: 7-18.

Aydin, K., Gaichas, S., Ortiz, I., Kinzey, D., Friday, N., 2007. A comparison of the Bering Sea, Gulf of Alaska, and Aleutian Islands large marine ecosystems though food web modeling. NOAA Technical Memorandum NMFS-AFSC-178. 298 pp.

240

Bailey, K.M., 1989. Interaction between vertical distribution of juvenile walleye pollock Theragra chalcogramma in the eastern Bering Sea, and cannibalism. Mar. Ecol. Prog. Ser. 53, 205–213.

Bailey, K.M., 2000. Shifting control of recruitment of walleye pollock Theragra chalcogramma after a major climatic and ecosystem change. Mar. Ecol. Prog. Ser. 198, 215–224.

Bailey, K.M., 2013. Billion Dollar Fish: The Untold Story of Alaska Pollock. University of Chicago Press. 280 pp.

Bailey, K.M., Brodeur, R. D., Merati, N. Yoklavich, M., 1993. Predation on walleye pollock (Theragra chalcogramma) eggs and yolk-sac larvae by pelagic crustacean invertebrates in the western Gulf of Alaska. Fish. Oceanogr. 2, 30-39.

Bailey, K.M., Houde, E.D. 1989. Predation on eggs and larvae of marine fishes and the problem of recruitment. Adv. Mar. Biol. 25: 1-83.

Bailey, K.M., Quinn, T.J., Bentzen, P. and Grant, W.S., 1999. Population structure and dynamics of walleye pollock, Theragra chalcogramma. In: Advances in Marine Biology (A.J. Southward, P. Tyler, and C.M. Young, eds.), Vol.37, Academic Press, pp. 179-255 (September 1999).

Bakkala, R.G., Traynor, J.J., Teshima, K., Shimada, A.M., Yamaguchi, H., 1985. Results of cooperative U.S.-Japan groundfish investigations in the eastern Bering Sea during June-November 1982. NOAA Technical Memorandum, NMFS F/NWC-87, U.S. Department of Commerce, 448 pp.

Beamish, R.J., Mahnken, C., 2001. A critical size and period hypothesis to explain natural regulation of salmon abundance and the linkage to climate and climate change. Prog. Oceanogr. 49, 423–437.

Benoit-Bird, K.J., McIntosh, N.E., Heppell, S.A., 2013a. Nested scales of spatial heterogeneity in juvenile walleye pollock Theragra chalcogramma in the southeastern Bering Sea. Mar. Ecol. Prog. Ser. 484, 219–238.

Benoit-Bird, K.J., Battaile, B.C., Heppell, S.A., Hoover, B., Irons, D., Jones, N., Kuletz, K.J., Nordstrom, C.A., Paredes, R., Suryan, R.M., Waluk, C.M., Trites, A.W., 2013b. Prey patch patterns predict habitat use by top marine predators with diverse foraging strategies. PLoS ONE. 8, 1–12.

Berkeley, S.A., Chapman, C., Sogard, S.M., 2004. Maternal age as a determinant of larval growth and survival in a marine fish, Sebastes melanops. Ecology. 85, 1258–1264.

Blood, D.M., 2002. Low-temperature incubation of walleye pollock (Theragra chalcogramma) eggs from the southeastern Bering Sea shelf and Shelikof Strait, Gulf of Alaska. Deep-Sea Res. Part II. 49, 6095–6108.

Bond, N.A., Overland, J.E., 2005. The importance of episodic weather events to the ecosystem of the Bering Sea shelf. Fish. Oceanogr. 14, 97–111.

241

Brodeur, R.D., 1998. Prey selection by age-0 walleye pollock, Theragra chalcogramma, in nearshore waters of the Gulf of Alaska. Environ. Biol. Fish. 51, 175–186.

Brodeur, R.D., Decker, M.B., Ciannelli, L., Purcell, J.E., Bond, N.A., Stabeno, P.J., Acuna, E., Hunt, G.L., Jr., 2008. Rise and fall of jellyfish in the eastern Bering Sea in relation to climate regime shifts. Prog. Oceanogr. 77, 103-111.

Brodeur, R.D. and Merati, N., 1993. Predation on walleye pollock (Theragra chalcogramma) eggs in the western Gulf of Alaska: the role of vertebrate and invertebrate predators. Mar. Biol. 117, 483- 493.

Brodeur, R.D., Sugisaki, H., Hunt, G.L., Jr., 2002. Increases in jellyfish biomass in the Bering Sea: implications for the ecosystem. Mar. Ecol. Prog. Ser. 233, 89-103.

Brodeur, R.D., Wilson, M.T., 1996. A review of the distribution, ecology and population dynamics of age-0 walleye pollock in the Gulf of Alaska. Fish. Oceanogr. 5, 148-166.

Brodeur, R.D., Picquelle, S.J., Blood, D.M., Merati, N., 1996. Walleye pollock egg distribution and mortality in the western Gulf of Alaska. Fish. Oceanogr. 5, 91–111.

Brodeur, R.D., Wilson, M.T., Ciannelli, L., 2000. Spatial and temporal variability in feeding and condition of age-0 walleye pollock (Theragra chalcogramma) in frontal regions of the Bering Sea. ICES J. Mar. Sci. 57: 256-264.

Brodeur, R.D., Wilson, M.T., Ciannelli, L., Doyle, M., Napp, J.M., 2002 Interannual and regional variability in distribution and ecology of juvenile pollock and their prey in frontal structures of the Bering Sea. Deep Sea Res. II. 49, 6051-6067.

Brodeur, R.D., Wilson, M.W., Napp, J.M., Stabeno, P.J., Salo, S., 1997. Distribution of juvenile pollock relative to frontal structure near the Pribilof Islands, Bering Sea. In Proceedings of the International Symposium on the Role of Forage Fishes in Marine Ecosystems, pp. 573–589. Alaska Sea Grant Pub. AK-SG-97-01.

Brodeur, R.D., Wilson, M.T., Walters, G.E., Melnikov, I.V., 1999. Forage fishes in the Bering Sea: distribution, species associations, and biomass trends. In: Loughlin, T.R, Ohtani, K. (Eds.) Dynamics of the Bering Sea, Alaska Sea Grant, Fairbanks, pp. 509-536.

Brown, A.L., M.S. Busby, K.L. Mier., 2001. Walleye pollock Theragra chalcogramma during transformation from the larval to juvenile stage: otolith and osteological development. Mar. Biol. 139, 845-851.

Campana, S.E., Hurley, P.C.F., 1989. An age- and temperature-mediated growth model for cod (Gadus morhua) and haddock (Melanogrammus aeglefinus) larvae in the Gulf of Maine. Can. J. Fish. Aquat. Sci. 46, 603-613.

Campana, S.E., Neilson, J.D., 1985. Microstructure of fish otoliths. Can. J. Fish. Aquat. Sci. 42, 1014– 1032.

242

Canino, M.F., O’Reilly, P.T., Hauser, L., Bentzen, P., 2005. Genetic differentiation in walleye pollock (Theragra chalcogramma) in response to selection at the pantophysin (PanI) locus. Can. J. Fish. Aquat. Sci. 62, 2519–2529.

Carr, S. M., Marshall, H.D.. 2008. Phylogeographic analysis of complete mtDNA genomes from Walleye Pollock (Gadus chalcogrammus Pallas, 1811) shows an ancient origin of genetic biodiversity. Mitochondrial DNA 19:490-496.

Ciannelli, L., 2002. Effects of spatial variability associated with a frontal structure on predictions of age-0 walleye pollock (Theragra chalcogramma) growth around the Pribilof Islands, Bering Sea. Estuar. Coast. Shelf Sci. 55, 151–165.

Ciannelli, L., Brodeur, R.D., Swartzman, G.L., Salo, S., 2002a. Physical and biological factors influencing the spatial distribution of age-0 walleye Pollock (Theragra chalcogramma) around the Pribilof Islands, Bering Sea. Deep Sea. Res. II 49, 6109-6126.

Ciannelli, L., Paul, A.J., Brodeur, R.D., 2002b. Regional, interannual and size-related variation of age 0 year walleye pollock whole body energy content around the Pribilof Islands, Bering Sea. J. Fish. Biol. 60, 1267-1279.

Ciannelli, L., Brodeur, R.D., Buckley, T.W., 1998. Development and application of a bioenergetics model for juvenile walleye pollock. J. Fish Biol. 52, 879–898.

Coachman, L.K., 1986. Circulation, water masses, and fluxes on the southeastern Bering Sea shelf. Cont. Shelf Res. 5, 23–108.

Collins, M.R., Schmidt, D.J., Waltz, C.W., Pickney, J.L., 1989. Age and growth of king mackerel, Scomberomorus cavalla, from the Atlantic coast of the United States. Fish. Bull. 87, 49–61.

Curchitser, E.N., Hedstrom, K., Danielson, S., Weingartner, T.J., 2010. Modeling of circulation in the North Aleutian Basin. OCS Study BOEMRE 2010-028, Bureau of Ocean Management, U.S. Department of the Interior, Washington D.C. http://alaska.boemre.gov/reports/2010rpts/2010_028.pdf

Cushing, D.H., 1981. Fisheries biology: a study in population dynamics. The University of Wisconsin Press, Madison, Wisconsin.

Cuveliers, E.L., Geffen, A.J., Guelinckx, J., Raeymaekers, A.M., Skadal, F., Volkaert, A.M., Maes, G.E., 2010. Microchemical variation in juvenile Solea solea otoliths as a powerful tool for studying connectivity in the North Sea. Mar. Ecol. Prog. Ser. 401, 211–220.

243

Danielson, S., Curchitser, E., Hedstrom, K., Weingartner, T., Stabeno, P., 2011. On ocean and sea ice modes of variability in the Bering Sea. J. Geophys. Res. 116, C12034, doi:10.1029/2011JC007389.

Danielson, S., Hedstrom, K., Aagaard, K., Weingartner, T., Curchitser, E., 2012. Wind-induced reorganization of the Bering shelf circulation. Geophys. Res. Lett. 39, L08601, doi:10.1029/2012GL051231.

Davis, M.W., Olla, B.L., 1992. Comparison of growth, behavior and lipid concentrations of walleys pollock Theragra chalcogramma larvae fed lipid-enriched, lipid deficient and field-collected prey. Mar. Ecol. Prog. Ser. 90, 23–30.

De Forest, L.G., Napp, J.M., Mier, K.L., Ferm, N., In review. Internannual variability in distribution of zooplankton communities and potential prey for larval walleye pollock in a major fish spawning area of the eastern Bering Sea. Deep Sea Res. II.

Dell’Arciprete, O.P. 1992. Growth, mortality, and transport of walleye pollock larvae (Theragra chalcogramma) in the eastern Bering Sea. MS thesis, Univ. Wash., 105 pp.

De Robertis, A. and Cokelet, E.D. 2012. Distribution of fish and macrozooplankton in ice-covered and open water areas of the eastern Bering Sea. Deep-Sea Research II. 65-70: 217-229.

DiMaria, R.A., Miller, J.A., Hurst, T.P., 2010. Temperature and growth effects on otolith elemental chemistry of larval Pacific cod, Gadus macrocephalus. Environ. Biol. Fish. 89, 453–462.

Döös, K., 1995. Inter-ocean exchange of water masses. J. Geophys. Res. 100, 13499–13514.

Duffy-Anderson, J.T., Ciannelli, L., Honkalehto, T., Bailey, K.M., Sogard, S.M., Springer, A., Buckley, T., 2003. Distribution of age-1 and age-2 walleye pollock in the Gulf of Alaska and Eastern Bering Sea: sources of variation and implications for higher trophic levels. Pp. 381–394. In: The Big Fish Bang: Proceedings from the 26th Annual Larval Fish Conference. Browman, H and A. Skiftesvik (Eds.).

Dunn, J.R., Matarese, A.C., 1987. A review of the early life history of Northeast Pacific gadoid fishes. Fish. Res. 5, 163–184.

Eisner, L.B., Napp, J.M., Mier, K.L., Pinchuk, A.I., Andrews, A.G., In press. Climate-mediated changes in zooplankton community structure for the eastern Bering Sea. Deep-Sea Research II.

Farley, E.V., Heintz, R.A., Andrews, A.G., Hurst, T.P., In press. Size, diet, and condition of age-0 Pacific cod (Gadus macrocephalus) during warm and cold climate states in the eastern Bering Sea. Deep- Sea Research II.

Fitzgerald, J.L., Thorrold, S.R., Bailey, K.M., Brown, A.L., Severin, K.P., 2004. Elemental signatures of larval walleye pollock (Theragra chalcogramma) from the northeast Pacific Ocean. Fish. Bull. 102, 604-616.

Folkvord, A., 2005. Comparison of size-at-age of larval Atlantic cod (Gadus morhua) from different populations based on size- and temperature-dependent growth models. Can. J. Fish. Aquat. Sci. 62, 1037-1052. 244

Gallego, A., North, E.W, and Houde, E.D., 2012. Understanding and quantifying mortality in pelagic, early life stages of marine organisms — Old challenges and new perspectives. J. Mar. Syst. 93, 1-3.

Gann, J.C., Eisner, L.B., Mordy, C., Porter, S., Watson, J.T., Cieciel, K., Yasumiishi, E.M., and Farley, E.V. In revision. Potential links between silicate and age-0 walleye pollock (Gadus chalcogrammus) in the southeastern Bering Sea during late summer/early fall. Deep Sea Res. II.

Gao, Y.W., 2002. Regime shift signatures from stable oxygen isotopic records of otoliths of Atlantic cod (Gadus morhua). Isot. Environ. Health Stud. 38, 251–263.

Geffen, A. J., 1982. Otolith ring deposition in relation to growth rate in herring (Clupea herangus) and turbot (Scopthalmus maximus) larvae. Mar. Biol. 71, 317–326.

Gillanders, B., 2011. Temporal and spatial variability in elemental composition of otoliths: implications for determining stock identity and connectivity of populations. Can. J. Fish. Aquat. Sci. 59, 669– 679.

Heintz, R.A., Siddon, E.C., Farley, E.V. Jr., Napp, J.M., 2013. Correlation between recruitment and fall condition of age-0 walleye pollock (Theragra chalcogramma) from the eastern Bering Sea under varying climate conditions. Deep-Sea Res. Part II. 94, 150–156.

Heintz, R.A., Vollenweider, J.J., 2010. Influence of size on the sources of energy consumed by overwintering walleye pollock (Theragra chalcogramma). J. Exp. Mar. Biol. Ecol. 393, 43–50.

Hjort, J. 1914. Fluctuations in the great fisheries of northern Europe viewed in the light of biological research. Rapports et Procès-Verbeaux des Reunions Conseil International pour l’Exploration de la Mer. 20, 1-228.

Hiatt, T., Dalton, M., Felthoven, R., Fissel, B., Garber-Yonts, B., Haynie, A., Kasperski, S., Lew, D., Package, C., Sepez, J., Seung, C., 2010. Economic status of the groundfish fisheries off Alaska, 2009. In: Stock assessment and fishery evaluation report for the groundfish fisheries of the Gulf of Alaska and Bering Sea/Aleutian Islands area, 254 p. [Available from North Pacific Fishery Management Council, 605 W. 4th Ave., Anchorage, AK 99510.]

Hillgruber, N., Haldorson, L.J., Paul, A.J., 1995. Feeding selectivity of larval walleye pollock Theragra chalcogramma in the oceanic domain of the Bering Sea. Mar. Ecol. Prog. Ser. 120, 1–10.

Hinckley, S., 1987. The reproductive biology of walleye pollock, Theragra chalcogramma, in the Bering Sea, with reference to spawning stock structure. Fish. Bull. 85, 481–498.

Hollowed A.B., Barbeaux, S.J., Cokelet, E.D., Kotwicki, S., Ressler, P.H., Spital, C., Wilson, C.D., 2012. Effects of climate variations on pelagic ocean habitats and their role in structuring forage fish distributions in the Bering Sea. Deep-Sea Res. Part II. 65–70, 230–250.

Holsman, K.K., Aydin, K., In review. Comparative methods for evaluating climate change impacts on the foraging ecology of Alaskan groundfish. Mar. Ecol. Prog. Ser.

245

Holsman, K.K., Ianelli, J., Aydin, K., Punt, A.E., In review. Comparative biological reference points estimated from temperature-specific multispecies and single species stock assessment models. Deep-Sea Res. Part II.

Houde, E.D., 1987. Fish early life dynamics and recruitment variability. Am. Fish. Soc. Symp. 2, 17-29.

Houde, E.D., 1989. Comparative growth, mortality, and energetic of marine fish larvae: temperature and implied latitudinal effects. Fish. Bull. 87, 471-495.

Houde, E.D. 2002. Predation. In: Fishery Science: The Unique Contributions of Early Life Stages. (Eds. Fuiman, L.A., Verner, R.G.). Blackwell Science, Oxford. Pp. 64-87.

Hunsicker, M.E., Ciannelli, L., Bailey, K.M., Zador, S., Stige, L. C., 2013. Climate and demography dictate the strength of predator-prey overlap in a subarctic marine ecosystem. PLoS ONE. 8(6), e66025. Doi:10.137/journal.pone.066025

Hunt, G.L., Coyle, K.O., Eisner, L.B., Farley, E., Heintz, R., Mueter, F., Napp, J.M., Overland, J.E., Ressler, P.H., Salo, S., Stabeno, P.J., 2011. Climate impacts on eastern Bering Sea food webs: A synthesis of new data and an assessment of the Oscillating Control Hypothesis. ICES J. Mar. Sci. 68, 1230–1243.

Hunter, J.R. 1981. Feeding ecology and predation of marine fish larvae. In: Marine Fish Larvae: Morphology, Ecology, and Relation to Fisheries. (Ed. Lasker, R.) University of Washington Press, Seattle. Pp 33-77.

Huntley, M.E., Lopez, M.D.G., 1992. Temperature-dependent production of marine copepods: a global analysis. Amer. Nat. 140, 201–242.

Hurst, T.P., 2007a. Causes and consequences of winter mortality in fishes. J. Fish Biol. 71, 315–345.

Hurst, T.P., 2007b. Thermal effects on behavior of juvenile walleye pollock (Theragra chalcogramma): implications for energetic and food webs models. Can. J. Fish. Aquat. Sci. 64, 449–457.

Hurst, T.P., Fernandez, E.R., Mathis, J.T., 2013. Effects of ocean acidification on hatch size and larval growth of walleye pollock (Theragra chalcogramma). ICES J. Mar. Sci. 70, 812–822.

Hurst, T.P., Laurel, B.J., Ciannelli, L., 2010. Ontogenetic patterns and temperature-dependent growth rates in early life stages of Pacific cod (Gadus macrocephalus). Fish. Bull. 108, 382-392.

Ianelli, J.N., Honkalehto, T., Barbeaux, S., Kotwicki, S., Aydin, K., Williamson, N., 2013. Assessment of the walleye pollock stock in the eastern Bering Sea. In: Stock Assessment and Fishery Evaluation Report for the Groundfish Resources of the Bering Sea/Aleutian Islands Regions. North Pacific Fishery Management Council, Anchorage, AK. [Available from North Pacific Fishery Management Council, 605 W. 4th Ave., Anchorage, AK 99510.]

246

Juadro-Molina, J. and Livingston, P., 2004. Sensitivity analysis of the multispecies virtual population analysis model parameterized for a system of trophically-linked species from the eastern Bering Sea. Ciencias Marinas. 30 (2) 1-12.

Jung, K.M., Kang, S., Kim, S., Kendall, A.W., Jr., 2006. Ecological characteristics of walleye pollock eggs and larvae in the southeastern Bering Sea during the late 1970s. J. Oceanogr. 62, 859–871.

Kachel, N., Hunt, G.L., Jr., Salo, S., Schumacher, J.D. Stabeno, P.J., Whitledge, T.E., 2002. Characteristics and variability of the inner front of the southeastern Bering Sea. Deep- Sea Res. Part II. 49, 5889–5909.

Kendall, A.W., Jr., 2001. Specific gravity and vertical distribution of walleye pollock (Theragra chalcogramma) eggs. NOAA AFSC Processed Report 2001–01. 88 pp.

Kendall, A.W., Jr., Clarke, M.E., Yoklavich, M.M. Boehlert, G.W., 1987. Distribution, feeding, and growth of larval walleye pollock, Theragra chalcogramma, from Shelikof Strait, Gulf of Alaska. Fish. Bull. US. 85, 499-521.

Kendall, A.W., Incze, L., Ortner, P.B., Cummings, S.R., Brown, P.K., 1994. The vertical distribution of eggs and larvae of walleye pollock, Theragra chalcogramma, in Shelikof Strait, Gulf of Alaska. Fish. Bull. US. 540-554.

Kennedy, B.P., Klaue, A., Blum, J.D., Folt, C.L., Nislow, K.H., 2002. Reconstructing the lives of fish using Sr isotopes in otoliths. Can. J. Fish. Aquat. Sci. 59, 925–929.

Kitchell, J.F., Stewart, D.J., Weininger, D., 1977. Applications of a bioenergetics model to perch (Perca flavescens) and walleye (Stizostedion vitreum). J. Fish. Res. Board Can. 34, 1922–1935.

Kjesbu, O.S., Witthames, P.R., Solemdal, P., Greer Walker, M., 1998. Temporal variations in the fecundity of Arcto-Norwegian cod (Gadus morhua) in response to natural changes in food and temperature. J. Sea Res. 40, 303–321.

Kotwicki, S., Buckley, T.W., Honkalehto, T., Walters, G., 2005. Variation in the distribution of walleye pollock (Theragra chalcogramma) with temperature and implications for seasonal migration. Fish. Bull. 103, 574–587.

Kristiansen, T., Lough, R.G., Werner, F.E., Broughton, E.A., Buckley, L.J., 2009. Individual-based modeling of feeding ecology and prey selection of larval cod on Georges Bank. Mar. Ecol. Prog. Ser. 376, 227–243.

Lang, G.M., Brodeur, R.D., Napp, J.M., Schabetsberger, R., 2000. Variation in groundfish predation on juvenile walleye pollock relative to hydrographic structure near the Pribilof Islands, Alaska. ICES J. Mar. Sci. 57, 265–271.

Leggett, W., Deblois, E., 1994. Recruitment in marine fishes: Is it regulated by starvation and predation in the egg and larval stages? Neth. J. Sea Res. 32, 119-134.

Lett C., Rose K. A., Megrey B.A., 2010. Biophysical models. In: Climate change and small pelagic fish (Eds. Checkley D., Alheit J., Oozeki Y., Roy C.), pp. 88–111. Cambridge, UK: Cambridge University Press. 247

Livingston, P.A., 1993. Importance of predation by groundfish, marine mammals, and birds on walleye pollock Theragra chalcogramma and Pacific herring Clupea pallasi in the eastern Bering Sea. Mar. Ecol. Prog. Ser. 102, 205–215.

Livingston, P.A., Jurado-Molina, J., 2000. A multispecies virtual population analysis of the eastern Bering Sea. ICES J. Mar. Sci. 57, 294–299.

Maeda, T., 1977. Reproduction mechanism of the Alaska pollock in the subarctic waters of the North Pacific Ocean and the adjacent waters. Bull. Jap. Soc. Fish. Oceanogr.. 31, 32–39.

Marshall, C.T., 2009. Implementing information on stock reproductive potential in fisheries management: the motivation, challenges and opportunities. In: T. Jakobsen, M. J. Fogarty, B.A. Megrey, and E.S. Moksness, (eds.). Fish Reproductive Biology, pp. 395–420, Wiley-Blackwell ,Oxford, UK;

Mazur M.M., Wilson, M.T., Dougherty, A.B., Buchheiste,r A., Beauchamp, D.A., 2007. Temperature and prey quality effects on growth of juvenile walleye pollock Theragra chalcogramma (Pallas): a spatially explicit bioenergetics approach. J. Fish Biol. 70, 816–836.

McGurk, M.D., 1987. Age and growth of Pacific herring larvae based on length-frequency data and otolith ring number. Environ. Biol. Fish. 20, 33–47.

Miller, B., Kendall, A.W., Jr., 2009. Early Life History of Marine Fishes. University of California Press. 376 pp.

Miyake, H., Yoshida, H., Ueda, Y., 1996. Distribution and abundance of age-0 juvenile walleye pollock, Theragra chalcogramma, along the Pacific coast of southeastern Hokkaido, Japan. in: Brodeur, R., Livingston, P.A., Loughlin, T.R., Hollowed, A.B. (Eds.), NOAA Technical Report, NMFS 126, U.S. Department of Commerce, pp. 3-10.

Moss, J.H., Farley, E.V., Feldmann, A.M., Ianelli, J.N., 2009. Spatial distribution, energetics status, and food habits of eastern Bering Sea age-0 walleye pollock. Trans. Am. Fish. Soc. 138, 497–505.

Mueter, F.J., Bond, N.A., Ianelli, J.N., Hollowed, A.B., 2011. Expected declines in recruitment of walleye pollock (Theragra chalcogramma) in the eastern Bering Sea under future climate change. ICES J. Mar. Sci. 68, 1284–1296.

Mueter, F.J., Ladd, C., Palmer, M.C., Norcross, B.L., 2006. Bottom-up and top-down controls of walleye pollock (Theragra chalcogramma) on the eastern Bering Sea shelf. Prog. Oceanogr. 68, 152–183.

Mulligan, T.J., Bailey, K.M , Hinckley, S., 1989. The occurrence of larval and juvenile walleye pollock in the eastern Bering Sea with implications for stock structure, pp. 471–489. In: Proc. Int. Symp. Biol. Mgmt. Walleye pollock, Alaska Sea Grant Report, 89-1.

Mulligan, T.J., Chapman, R.W., Brown, B.L., 1992. Mitochondrial DNA analysis of walleye pollock Theragra chalcogramma from the eastern Bering Sea and Shelikof Strait, Gulf of Alaska. Can. J. Fish. Aquatic Sci. 49, 319-326.

Napp, J.M., Kendall, A.W., Jr., Shumacher, J.D., 2000. A synthesis of biological and physical processes affecting the feeding environment of larval walleye pollock (Theragra chalcogramma) in the eastern Bering Sea. Fish. Oceanogr. 9, 147–162. 248

Napp, J.M., Baier, C.T., Brodeur, R.D., Coyle, K.O., Shiga, N., Mier, K., 2002. Interannual and decadal variability in zooplankton communities of the southeast Bering Sea shelf. Deep- Sea Res. Part II. 49, 5991–6008.

Ney, J.J., 1990. Trophic economics in fisheries: assessment of demand-supply relationships between predators and prey. Rev. Aquat. Sci.. 2, 55–81.

Nishimura, A., Mito, K-I, Yanagimoto, T., 1996. Hatch date and growth estimation of juvenile walleye pollock, Theragra chalcogramma, collected in the Bering Sea in 1989 and 1990, pp. 81–87. In: Ecology of Juvenile walleye pollock, Theragra chalcogramma. Papers from the workshop “The Importance of Prerecruit walleye pollock to the Bering Sea and North Pacific Ecosystems.” Seattle, Washington, 28–30 October 1993. R.D. Brodeur, P.A. Livingston, T.R. Loughlin, A.B. Hollowed (eds.) NOAA Technical Report NMFS 126, 227 p.

Nishimura, A., Yamada, J., 1984. Age and growth of larval and juvenile walleye pollock, Theragra chalcogramma (Pallas), as determined by otolith daily growth increments. J. Exp. Mar. Biol. Ecol. 82, 191-205.

Otterlei, E., Nyhammer, G., Folkvord, A. Stefansson, S.O., 1999. Temperature- and size-dependent growth of larval and early juvenile Atlantic cod (Gadus morhua): a comparative study of Norwegian coastal cod and northeast cod. Can. J. Fish. Aquat. Sci. 56, 2099-2111.

Ottersen, G., Bogstad, B., Yaragina, N.A., Stige, L.C., Vikebø, F.B., Dalpadado, P., 2014. A review of early life dynamics of Barents Sea cod (Gadus morhua). ICES Journal of Marine Science. doi: 10.1093/icesjms/fsu037.

Page, L.M., Espinosa-Perez, H., Findley, L.T., Gilbert, C.R., Lea, R.N., Mandrak, N.E., Mayden, R.L., Nelson, J.S., 2013. Common and Scientific names of Fishes from the United States, Canada, and Mexico, 7th edition. American Fisheries Society, Special Publication 34. 243 pp.

Parker-Stetter, S.L., Horne, J.K., Farley, E.V., Barbee, D.H., Andrews, A.G. III, Eisner, L.B., Nomura, J.M., 2013. Summer distribution of forage fish in the eastern Bering Sea. Deep- Sea Res. Part II. 94,211–230.

Parker-Stetter, S., Horne, J.K., Urmy, S., Heintz, R., Eisner, L. Farley. E.V., In review 1. Vertical distribution of age-0 pollock during late-summer: biophysical coupling or ontogeny? Fisheries Oceanography.

Parker-Stetter, S., Urmy, S., Horne, J.K., Eisner, L. Farley, E.V., In review 2. Factors affecting summer distributions of Bering Sea forage fish species: assessing competing hypotheses. Deep Sea Research II.

Petrik, C.M., Duffy-Anderson, J.T., Mueter, F., Hedstrom, K., Curchitser, E., In press. Biophysical transport model suggests that climate variability determines distribution of walleye pollock early life stages in the eastern Bering Sea through effects on spawning. Prog. Oceanogr. DOI: 10.1016/j.pocean.2014.06.004

249

Petrik, C.M., Duffy-Anderson, J.T., Curchitser, E., Mueter, F., Hedstrom, K., In preparation Oceanographic transport anomalies influence spawning-nursery habitat connectivity and recruitment of walleye pollock in the eastern Bering Sea. ICES Journal of Marine Science.

Porter, S.M., Bailey, K.M., 2007. Optimization of feeding and growth conditions for walleye pollock Theragra chalcogramma (Pallas) larvae reared in the laboratory. AFSC Processed Rep. 2007-06, 20 p. Alaska Fish. Sci. Cent., NOAA, Natl. Mar. Fish. Serv., 7600 Sand Point Way NE, Seattle WA 98115.

Porter, S.M., Bailey, K.M., 2011. Assessing the condition of walleye pollock (Theragra chalcogramma (Pallas) larvae using muscle-based flow cytometric cell cycle analysis. J. Exp. Mar. Biol. Ecol. 399, 101–109.

Post, J.R., Parkinson, E.A., 2001. Energy allocation strategy in young fish: allometry and survival. Ecology 82,1040-1051.

Reed, R.K., 1995. Water properties over the Bering Sea shelf: climatology and variations. NOAA Technical Report ERL 452-PMEL 42, 15 pp.

Ruttenberg, B.I., Hamilton, S.L., Warner, R.R., 2008. Spatial and temporal variation on the natal otolith chemistry of a Hawaiian reef fish: prospects for measuring population connectivity. Can. J. Fish. Aquat. Sci. 65, 1181–1192.

Sakurai, Y., 2007. An overview of the Oyashio ecosystem. Deep-Sea Res. Part II . 54, 2526–2542.

Sambrotto, R.N., Niebauer, H.J., Goering, J.J., Iverson, R.L., 1986. Relationships among vertical mixing, nitrate uptake and phytoplankton growth during the spring bloom in the southeast Bering Sea middle shelf. Cont. Shelf Res. 5, 161–198.

Schabetsberger, R., Brodeur, R.D., Ciannelli, L., Napp, J.M., Swartzman, G.L., 2000. Diel vertical migration and interaction of zooplankton and juvenile walleye pollock (Theragra chalcogramma) at a frontal region near the Pribilof Islands, Bering Sea. ICES J. Mar. Sci. 57, 1283–1295.

Schabetsberger, R., Brodeur, R.D., Honnkalehto, T., and Mier, K., 1999. Sex-specific egg cannibalism in spawning walleye pollock: the role of reproductive behavior. Env. Biol. Fishes. 54,175-190.

Schabetsberger, R., Sztatecsny, M., Drozdowski, G. Brodeur, R.D., Swartzman, G. Wilson, M.T., Winter, A.G., Napp, J.M. 2003. Size-dependent, spatial, and temporal variability of juvenile walleye pollock (Theragra chalcogramma) feeding at a structural front in the southeast Bering Sea. Mar. Ecol. 54, 175-190.

Siddon, E.C., Heintz, R.A., Mueter, F.J., 2013a. Conceptual model of energy allocation in walleye pollock (Theragra chalcogramma) from age-0 to age-1 in the southeastern Bering Sea. Deep-Sea Res. Part II. 94,140–149.

Siddon, E.C., Kristiansen, T., Mueter, F.J., Holsman, K., Heintz, R.A., Farley, E.V., 2013b. Spatial match-mismatch between juvenile fish and prey explains recruitment variability across climate regimes in the eastern Bering Sea. PLoS ONE. 8, e84526.

250

Sigler, M.F., Heintz, R.A., Hunt, G.L., Jr., Lomas, M.W., Napp, J.M., Stabeno, P.J., In review. A mid- trophic view of subarctic productivity: lipid storage, location matters, and historical context. Deep-Sea Research II.

Sinclair, E.H., Vlietstra, L.S., Johnson, D.S., Zeppelin, T.K., Byrd, G.V., Springer, A.M., Ream, R.R., Hunt, G.L., 2008. Patterns in prey use among fur seals and seabirds in the Pribilof Islands. Deep- Sea Res. Part II. 55, 1897–1918.

Smart, T., Duffy-Anderson, J.T., Horne, J., Farley, E., Wilson, C., Napp, J., 2012a. Influence of environment on walleye pollock eggs, larvae, and juveniles in the Southeastern Bering Sea. Deep- Sea Res. Part II. 65–70, 196–207.

Smart, T., Duffy-Anderson, J.T., Horne, J., 2012b. Alternating climate states influence walleye pollock life stages in the southeastern Bering Sea. Mar. Ecol. Prog. Ser. 455, 257–267.

Smart, T.I., Siddon, E.C., Duffy-Anderson, J.T., 2013. Vertical distributions of the early life stages of a common gadid in the eastern Bering Sea (Theragra chalcogramma, walleye pollock). Deep-Sea Res. Part II. 94, 201–210.

Smith, P.J., Jamieson, A., Birley, A.J., 1990. Electrophoretic studies and stock concept in marine teleosts. J. Council Internat’l Explor. Sea, 47, 231–245.

Sogard, S.M., Olla, B.L., 1996. Diel patterns of behavior in juvenile walleye pollock, Theragra chalcogramma. Environ. Biol. Fish. 47, 379-386.

Sogard, S.M., Olla, B.L., 2000. Endurance of simulated winter conditions by age-0 walleye pollock: effects of body size, water temperature and energy stores. J. Fish Biol. 56,1–21.

Sogard, S.M., Olla, B.L., 2002. Contrasts in capacity and underlying mechanisms for compensatory growth in two pelagic marine fishes. Mar. Ecol. Prog. Ser. 243, 165-177.

Solemdal, P., 1997. Maternal effects – a link between past and future. J. Sea Res. 37, 213–227.

Spencer, P.D., Dorn, M.W., 2013. Incorporation of weight-specific relative fecundity and maternal effects in larvae survival into stock assessments. Fish. Res. 138, 159–167.

Stabeno, P.J., Kachel, N., Mordy, C., Righi, D., Salo, S., 2008. An examination of the physical variability around the Pribilof Islands in 2004. Deep-Sea Res. Part II 55, 1701–1716.

Stabeno, P.J., Kachel, N.B., Moore, S.E., Napp, J.M., Sigler, M., Yamaguchi, A., Zerbini, A.N., 2012. Comparison of warm and cold years on the southeastern Bering Sea shelf and some implications for the ecosystem. Deep-Sea Res. Part II. 65–70, 31–45.

Stabeno.P.J., Schumacher, J.D., Ohtani, K., 1999. The physical oceanography of the Bering Sea, pp. 1– 28. In: Dynamics of the Bering Sea (Loughlin and Ohtani, Eds). University of Alaska Sea Grant, Fairbanks, AK, 825 pp.

Stockwell, D.A., Whitledge, T.E., Zeeman, S.I., Coyle, K.O., Napp, J.M., Brodeur, R.D., Pinchuk, A.I., 2001. Anomalous conditions in the south-eastern Bering Sea, 1997: nutrients, phytoplankton, and zooplankton. Fish. Oceanogr. 10, 99–116. 251

Strasburger, W.W. Hillgruber, N. Pinchuk, A.I., Mueter, F.J., In press. Feeding ecology of age-0 walleye pollock (Gadus chalcogramma) and Pacific cod (Gadus macrocephalus) in the southeastern Bering Sea. Deep-Sea Research Part II.

Sundby, S., Nakken, O., 2008. Spatial shifts in spawning habitats of Arcto-Norwegian cod related to multidecadal climate oscillations and climate change. ICES J. Mar. Sci. 65, 953–962.

Suzaki, A., Sakurai, Y., Yamamoto, J., Hamatsu, T., Ito, S., Hattori, T., 2003. Influence of Oyashio Current on stock fluctuation of walleye pollock in the Tohoku region, northern Japan. Abstract of 12th PICES Annual Meeting, 145.

Swartzman, G., Napp, J.M., Brodeur, R.D., Winter, A., Ciannelli, L., 2002. Spatial patterns of pollock and zooplankton distribution in the Pribilof Islands, Alaska nursery area and their relationship to pollock recruitment. ICES J. Mar. Sci. 59, 1167-1186.

Swartzman, G., Brodeur, R., Napp, J., Hunt, G., Demer, D., Hewitt, R., 1999. Spatial proximity of age-0 walleye pollock (Theragra chalcogramma) to zooplankton near the Pribilof Islands, Bering Sea, Alaska. ICES J. Mar. Sci. 56, 545-560.

Swartzman, G., Napp, J., Brodeur, R., Winter, A., Ciannelli, L., 2002. Spatial patterns of pollock and zooplankton distribution in the Pribilof Islands, Alaska nursery area and their relationship to pollock recruitment. ICES J. Mar. Sci. 59, 1167-1186.

Tang, W., Jin, X., Li, F., Chen, J., Wang, W., Chen, Y., Zhao, X., Dai, F., 1996. Summer distribution and abundance of age-0 walleye pollock, Theragra chalcogramma, in the Aleutian Basin. in: Brodeur, R., Livingston, P.A., Loughlin, T.R., Hollowed, A.B. (Eds.), NOAA Technical Report, NMFS 126, U.S. Department of Commerce, pp. 35-46.

Theilacker, G.H., Porter, S.M., 1995. Condition of larval walleye pollock, Theragra chalcogramma, in the western Gulf of Alaska assessed with histological and shrinkage indices. Fish. Bull. 93, 333– 344.

Thorisson, K., Jónsdóttir, I., Marteinsdottir, G., Campana, S.E., 2011. The use of otolith chemistry to determine the juvenile source of spawning cod in Icelandic waters. ICES J. Mar. Sci. 68, 98–106.

Thorrold, S.R., Jones, C.M., Campana, S.E., McLaren, J.W., Lam, J.W.H., 1998. Trace element signatures in otolths record natal river of juveniles American shad (Alosa sapidissima). Limnol. Oceanogr. 43, 1826–1835.

Traynor, J.J., Smith, D., 1996. Summer distribution and abundance of age-0 walleye pollock, Theragra chalcogramma, in the Bering Sea. NOAA Technical Report, NMFS 126, U.S. Department of Commerce, pp. 57-59.

Waldron, K.D. and Vinter, B.M., 1978. Ichthyoplankton of the eastern Bering Sea. NMFS, NWAFSC, Processed Report. Final report (RU 380), environmental assessment of the Alaskan continental shelf, sponsored by United States Department of Interior Bureau of Land Management. 88 p.

Walline, P.D. 1985. Growth of larval walleye pollock related to domains within the SE Bering Sea. Mar. Ecol. Prog. Ser. 21, 197–203.

252

Walsh, J.J., McRoy, C.P., 1986. Ecosystem analysis in the southeastern Bering Sea. Cont. Shelf Res. 5, 259–288.

Walsh, J.J., Wirick, C.D., Dieterle, D.A., Tingle, A.G., 1981. Environmental constraints on larval fish survival in the Bering Sea. Rapp. P.-v. Réun. Cons. Int. Explor. Mer. 178, 24–27.

Walters, G.E., Teshima, K., Traynor, J.J., Bakkala, R.G., Sassano, J.A., Halliday, K.L., Karp, W.A., Mito, K., Williamson, N.J., Smith, D.M., 1988. Distribution, abundance, and biological characteristics of groundfish in the eastern Bering Sea based on results of the U.S.-Japan triennial bottom trawl and hydroacoustic surveys during May-September, 1985. NOAA Technical Memorandum, NMFS F/NWC-154, U.S. Department of Commerce, 400 pp.

Wespestad, V.G., Ianelli, J., Fritz, L.W., Honkalehto, T., Walters, G., 1997. Bering Sea-Aleutian

Islands walleye pollock assessment for 1997. In: Stock Assessment and Fishery Evaluation Report for the Groundfish Resources of the Bering Sea/Aleutian Islands Regions, pp. 1–73. Ed by Groundfish Plan Team for the Bering Sea-Aleutian Islands Region, North Pacific Fishery Management Council, Anchorage, AK, USA. 571 p.

Wespestad, V.G., Fritz, L.W., Ingraham, W.J., Megrey, B.A., 2000. On relationships between cannibalism, climate variability, physical transport, and recruitment success of Bering Sea walleye pollock (Theragra chalcogramma). ICES J. Mar. Sci.. 57, 268–274.

Whitledge, T.E., Reeburgh, W.S., Walsh, J.J., 1986. Seasonal inorganic nitrogen distribution and dynamics in the southeastern Bering Sea. Cont. Shelf Res. 5, 109–132.

Wilderbuer, T., Hollowed, A.B., Ingraham, W.J. Jr., Spencer, P.D., Conners, M.E. and Walters, G.E. 2002. Flatfish recruitment response to decadal climatic variability and ocean conditions in the eastern Bering Sea. Prog. Oceanogr. 55, 235–247.

Wilderbuer, T., Stockhausen, W., and Bond, N., 2013. Updated analysis of flatfish recruitment response to climate variability and ocean conditions in the eastern Bering Sea. Deep-Sea Res. Part II. 94, 157–164.

Wilson, M.T., Buchheister, A., Jump, C., 2011. Regional variation in the annual feeding cycle of juvenile walleye pollock (Theragra chalcogramma) in the western Gulf of Alaska. Fish. Bull. 109, 316- 326.

Winter, A.G., Swartzman, G.L., 2006. Interannual changes in distribution of age-0 walleye pollock near the Pribilof Islands, Alaska, with reference to the prediction of pollock year-class strength. ICES J. Mar. Sci. 63, 1118-1135.

Winter, A., Coyle, K.O., Swartzman, G., 2007. Variations in age-0 pollock distribution among eastern Bering Sea nursery areas: A comparative study through acoustic indices. Deep-Sea Res. II 54, 2869-2884.

Wood, S., 2006. Generalized Additive Models: An Introduction with R. Chapman and Hall - CRC. Boca Raton, FL.

253

Wyllie-Echeverria, T., Wooster, W.S., 1998. Year-to-year variations in Bering Sea ice cover and some consequences for fish distributions. Fish. Oceanogr. 7, 159–170.

Yamamura, O., 2005. Trophodynamic modeling of walleye pollock (Theragra chalcogramma) in the Doto area, northern Japan: model description and baseline simulations. Fish. Oceanogr. 13, 138– 154.

Yoklavich, M.M. Bailey, K.M., 1990. Hatching period, growth and survival of young walleye pollock, Theragra chalcogramma, as determined from otolith analysis. Mar. Ecol. Prog. Ser. 64, 13-23.

254

Table 7.1. Collections of the early life stages of Walleye Pollock in the eastern Bering Sea. Eco-FOCI = Ecosystems and Fisheries Oceanography Coordinated Investigations, EMA = Ecosystem Monitoring and Assessment, MACE = Midwater Assessment and Conservation Engineering, UAF = University of Alaska Fairbanks, UW = University of Washington.

Life Stage Program Years Season Biological Physical Sampling Gear Sampling Gear

Eggs, larvae Eco-FOCI 1988-present Spring 60 cm bongo, CTD, Seacat 1m2 MOCNESS, Tucker nets

Larvae, age-0 Eco-FOCI, 2008-2010 Spring, 1m2 CTD, Seacat juveniles UAF Summer MOCNESS

Larvae, age 0 Eco-FOCI, 1995-2001 Summer Methot trawl Seacat juveniles Hokkaido University

Age-0 EMA, UW 2001-present Summer CanTrawl, CTD, Seacat juveniles acoustics, midwater trawl

Pre-spawning MACE Spring Midwater adults trawl, acoustics

Age-1 MACE Winter & Midwater juveniles summer the trawl, acoustics following year

255

Table 7.2. Phenological shifts in temporal distributions of Walleye Pollock early life stages showing day of year when each life stage appears over the eastern Bering Sea shelf. NA: not available to seasonal sampling limitations. Adapted from Smart et al., 2012a.

WARM COLD

Life stage Start Peak End Start Peak End

Eggs NA 65 130 NA 105 150

Yolksac NA 100 170 100 145 180

Early stage larvae NA 125 185 100 145 195

Late stage larvae 100 150 210 125 180 240

Juveniles 120 200 NA 135 225 NA

256

Table 7.3. Hydrodynamic models used to examine Walleye Pollock early life ecology in the eastern Bering Sea.

Domain; Physical model and Model Biology Results/Outputs Dimensions forcing

Wind stress, bottom 2 stages friction, topography, (nonfeeding, Trajectories and Walsh et al. EBS; 2D Coriolis force, feeding); mortalities based on 1981 geostrophic pressure Growth; spawning date gradient Mortality

Wespestad et al. OSCURS; Sea level Relate year-class success 1997; Ianelli et pressures, mean to wind-driven advection al. 1998; geostrophic currents, 1 stage; No EBS; 2D and spatial overlap of Wespestad et al. satellite-tracked biology final distributions with 2000 drifter speed adults coefficients

4 stages (egg, Trajectories, distributions, yolksac, patchiness, and Petrik et al. in preflexion, EBS; 3D ROMS connectivity as influenced preparation a late); Growth; by environmental and Vertical spawning conditions behavior

257

Figure 7.1. The southeastern Bering Sea with schematic representation of the major flows over the Aleutian Basin and adjacent shelf. Walleye Pollock spawning areas (known and putative) indicated as shaded ellipses.

258

Figure 7.2. Proportion of harvested roe that is hydrated from the commercial fishery by latitude (LAT) and week of the year (Week) for 2001 through 2006. The surfaces were derived from a 2-dimensional cubic splines fit within a generalized additive model (GAM) for latitude and week by year as the independent variables. Points are the data used in the GAM for each year. Color ramp indicates density.

259

Figure 7.3. Centers of distribution of Walleye Pollock early life history stages over the eastern Bering Sea shelf under years of warm (light grey) and cold (black) oceanographic conditions. Bars indicate +/- one standard deviation. Eg = Egg, YL = Yolksac larvae, PL = Preflexion larvae, LL = late larvae, JV = Juvenile.

260

Figure 7.4. Vertical distribution of feeding stages of Walleye Pollock larvae (preflexion, late larvae, early juvenile) during day and night. Bars represent 1 standard deviation; n indicates the number of positive tows.

261

Figure 7.5. Generalized additive mixed model (GAMM) regression analysis showing the estimated effect of standard length (SL; mm) on energy density (kJ/g dry mass) for age-0 Walleye Pollock. Dashed lines denote 95% confidence interval. Energy densities are plotted as anomalies because actual values depend on location and year of sampling. Adapted from Siddon et al. (2013).

262

Figure 7.6. Relationship between catch-averaged weight, energy density and total energy of YOY Walleye Pollock with survival. Survival is estimated as the number of age-1 recruits per female spawner and was not available for 2011year class at time of publication. Adapted from Heintz et al. (2013).

R 2 = 0.496 0 1.5 2.0 2.5 3.0

Weight (g)

e n

w Warm year a 20

p Average year

- r

e Cold year

s t

i 10

- R 2 = 0.677 e

g 0

A 3.5 4.0 4.5 5.0 Energy density (kJ/g)

R 2 = 0.736

0 5.0 7.5 10.0 12.5 15.0 Total Energy (kJ/fish)

263

Figure 7.7. Horizontal distribution of late-juvenile pollock in 2008-2010. The top panel shows shallow (surface to pycnocline) and bottom panel shows deep (pycnocline to bottom) distributions. Deep data were not available for 2008. Bubble size is proportional to density, a thin black line shows 0 density, and a thin grey line indicates that data were not available for this section of transect. Adapted from Parker- Stetter et al. (2013).

264

Figure 8. A. Temperature-dependent growth in length (mm d-1) and B. Age-dependent growth-in-length at four temperatures derived for Bering Sea pollock (BSP; Petrik et al. in press), Gulf of Alaska Pacific cod (GOAPC; Hurst et al. 2010), Norweigian Coastal cod (NCC; Otterlei et al., 1999), Northeast arctic cod (Otterlei et al.; 1999), Norweigian cod (NC; Folkvord 2005), Norweigian arctic cod (NAC; Folkvord 2005), Shelikof Strait pollock (SSP K; Kendall et al., 1987), Shelikof Strait pollock (SSP Y&B; Yoklavich and Bailey 1990), Funka Bay pollock (FBP; Nishimura and Yamada, 1984), Gulf of Maine haddock (GOMH; Campana and Hurley, 1989), and Gulf of Maine cod (GOMC; Campana and Hurley, 1989).

265

266

Overall Conclusions

Project B53 was executed within a framework of hypotheses:

BSIERP Hypothesis 4a. Climate-ocean conditions impact advection of larvae and juveniles to favorable nursery habitat.

i. Shoreward wind-driven advection to favorable nursery habitat increases larval and juvenile walleye pollock survival.

Results from Project B53 demonstrate that shifts in ichthyoplankton community composition occur in response to environmental variability, and establish that delineations in species associations are based on hydrographic conditions in the eastern Bering Sea (Chapter 1). We showed that transitions occur over a relatively short time period that included both pronounced warm and cold conditions. We observed significant differences in assemblage structure, supporting the hypothesis that early life stages may be primary indicators of environmental change. Larval abundances were generally higher at the time of sampling in warm years with high abundances of Walleye Pollock contributing most to differences between warm and cold periods in Unimak Pass, the outer domain, and shelf areas. These observations can be coupled with the work that showed that for Walleye Pollock, centers of distribution of early life stage are shifted eastward during warm years (Chapter 3), which is related to coincident shifts in wind- induced circulation and spatial shifts in spawning distribution (Chapter 5). We found that growth rates of Walleye Pollock larvae are higher in warm years relative to cold, but collaborative work with Project B54 (Bioenergetics) suggests that nutritional condition, not size, is a better indicator of overwinter survival and recruitment.

BSIERP Hypothesis 4b. Climate-ocean conditions impact predator-prey spatial and temporal overlaps.

Onshore currents separate juvenile fish from outer domain piscivores by transporting larvae inshore, away from adults.

ii. Strength of frontal boundaries will weaken due to absence of the summer cold pool and low summer winds. A weakened inner front will open gateways to the inner domain for predators from the middle and outer domains.

We found that cross-shelf assemblage structure was primarily associated with a geographic and/or salinity gradient that distinguished slope and shelf communities (Chapter 1), but we did not resolve whether

267

strength of the inner front influenced advection to the inner shelf. We did observe however, that the advection of Alaska Coastal Current waters through Unimak Pass affects the distribution of larval fish on the EBS shelf. Species entrained in, or advected by, ACC waters within Unimak Pass and the Bering Sea shelf included Pacific Cod (Gadus macrocephalus) and Northern Rock Sole (Lepidopsetta polyxystra), with higher overall abundances of these species in warm years. Unfortunately, our sampling design could not resolve whether these larvae originated in the Gulf of Alaska or were entrained in ACC waters within Unimak Pass and nearby spawning grounds. The impact of Gulf of Alaska larvae on Bering Sea populations, and the degree to which the populations are connected, are important ecological (i.e., competition, predation) and fisheries management (number of sub-populations) questions. However, while some studies have demonstrated larval drift from the Gulf of Alaska to the Bering Sea, for most species the role of advection through Unimak Pass and the resulting connectivity between Gulf of Alaska and Bering Sea populations is still largely unknown.

BSIERP Hypothesis 4c. Climate-ocean conditions impact the strength of fronts between domains and the sizes of the domains.

iv. Strength of frontal boundaries will weaken due to absence of the summer cold pool and low summer winds. Out-migrations of anadromous species will shift away from shore due to the weakening of the inner front. v. Strength of the inner front will weaken, allowing expansion of the inner domain, which will increase the carrying capacity of the inner domain for juveniles.

A coupled ocean-sea ice model, conducted in hindcast mode for the period from 1994-2012, simulated winds, stratification, ice, currents, and other ocean conditions over nearly two decades, encompassing both warm and cold years (Chapter 5). The physical model was coupled to an IBM to simulate drift and transport of Walleye Pollock larvae over the same period. We found that differences in dispersal were due primarily to differences in spawning locations of adults, rather than to climate-mediated differences in circulation or to temporal shifts in spawning time. Our results suggest that spatio-temporal variations in frontal boundaries are not the primary motivator for observed spatial changes in distribution of young Walleye Pollock.

268

Physical conditions vary interannually over the Bering Sea shelf in response to atmospheric forcing, but our results suggest that decision-making parameters exhibited by biological organisms responding to abiotic variation may have a greater role in biotic aggregation (Chapters 3, 4, 5). We found that vertical distribution of feeding larvae exhibited diel vertical migration, with results suggesting that vertical distributions and diel migration potentially are driven by prey availability at sufficient light levels for preflexion larvae to feed and a trade-off between prey access and predation risk for postflexion larvae. Moreover, model results indicate that horizontal shifts in Walleye Pollock distribution between warm and cold years were most likely due to spatial shifts in the distribution of spawning adults rather than to oceanographic changes in frontal structure, stratification, or current direction and velocity alone. We hypothesize that the spawning area for Walleye Pollock expands and contracts within certain undescribed geographic boundaries of the broader spawning region, and is dependent upon sea ice presence, temperature, and population biomass.

The majority of information learned during Project B53 supports a better understanding of the early life history and ecology of Walleye Pollock. Unfortunately, despite a 20-year history of sampling in the eastern Bering Sea led by the National Oceanic and Atmospheric Administration, collections of Pacific Cod and Arrowtooth Flounder larvae are either very few (Pacific Cod) or not to species level (Arrowtooth Flounder). Instances of catch of Pacific Cod larvae in bongo tows collected from the Bering Sea are fewer than 300 (http://access.afsc.noaa.gov/ichthyo/index.php), and catches during the BSIERP program years were less than 20. Records of Arrowtooth Flounder are greater than for Pacific Cod, and catches were higher during BSIERP years as well, though historical difficulties with taxonomic resolution between this species and a congeneric, Kamchatka Flounder, Atheresthes evermanni, required that samples be identified only to the species level. We made great progress in reclassifying historical samples to the species level using a combination of genetic and morphometric approaches, but ultimately were not successful at species-specific resolution across all larval size classes. This limited what data could be put into ecological context and the conclusions that could be derived. Accordingly, data for Pacific Cod and Arrowtooth Flounder were deemed insufficiently robust to be utilized to the extent we originally envisioned. As such, we focused on understanding the ecology of Walleye Pollock early life, and provided information on Arrowtooth Flounder and Pacific Cod where possible.

269

BSIERP and Bering Sea Project Connections

We successfully collaborated with BSIERP projects B54 Seasonal Bioenergetics (Heintz et al.), B59 Surface Trawl Survey Acoustics (Horne, B52 Biophysical Moorings (Stabeno et al.), B90 Surface Trawl Survey (Farley et al.), and B60 Walleye Pollock and Pacific Cod Distribution (Ciannelli et al.).

Working within an integrated research project such as BSIERP resulted in a far greater understanding of ichthyoplankton dynamics within the larger framework of the Bering Sea ecosystem. The progression of BSIERP began with more focused research efforts (i.e., individuals working to meet project-specific goals), but as projects moved forward, cross-integration became a larger focus. At the onset of the project, PI meetings felt disparate as projects were just getting “off the ground”. In hindsight, however, this may be an inevitable part of the progression towards better integration. We felt we had sufficient time to meet project-specific objectives, while also collaborating with several other efforts to enhance our understanding of larval ecology across contrasting climate regimes in the EBS.

Seasonal Bioenergetics

On all BSIERP field surveys (2008, 2009, 2010) samples were collected and sent to Juneau, AK for use in energetic and condition analyses. Samples from age-0 Walleye Pollock and age-0 Arrowtooth Flounder were provided. Arrowtooth Flounder were conclusively identified at sea prior to shipment from DNA analyses. Limited samples of Pacific Cod were made available from historical AFSC collections, but samples were few. Very few Pacific Cod were collected during BSIERP field years. Analyses resulted in a paper published by a graduate student (Siddon et al. 2013). Collaborations with B53 also included a co-authored paper on the ecology and bioenergetics of Arrowtooth Flounder (De Forest et al., in press), and co-authorship on a review paper on pollock ecology during the first year of life in the Bering Sea (Duffy-Anderson et al., in press). Heintz and Duffy-Anderson also served as external graduate committee members for graduate student E. Siddon (PhD, UAF). Work from the Siddon dissertation applied to both projects. Other work included a collaborative paper on condition and recruitment of age-0 pollock (Heintz et al. 2013)

Surface Trawl Survey Acoustics

We collaborated with project B59 by jointly advising postdoctoral researcher (T. Smart). Dr. Smart worked with J. Horne and S. Parker-Stetter to examine factors affecting abundance of pollock in the EBS (Smart et al. 2012 a, b) and vertical distribution of pollock in the EBS (Smart et al. 2013). Collaboration also included co-authorship on a review paper on pollock ecology during the first year of life in the Bering Sea (Duffy-Anderson et al. in press).

Biophysical Moorings

270

Collaborations with this project included joint ship time to recover/deploy BSIERP moorings collect zooplankton samples around moorings along the 70 m isobath. Data from a portion of the joint sampling was used in Smart et al. 2011. Collaborations also included discussion of implications of results from Smart et al. 2011 and Smart et al. 2012a. Other collaborative papers include Stabeno et al. 2012, Stabeno et al. in press.

Surface Trawl Survey

All ichthyoplankton from surface trawl surveys (BASIS) were sorted and identified by the FOCI Program as matching contribution (Ann Matarese lead). All data were provided to BSIERP research. We also collaborated with this project via sample sharing (3 BSIERP target fish species, zooplankton analyses and information), and through the efforts of UAF graduate student E. Siddon and postdoctoral researcher T. Smart. Collaborations resulted in two co-authored publications (Smart et al. 2012a) and Duffy-Anderson et al. (in press). Other collaborations include Hunt et al. 2011 and Sigler et al. 2012.

Walleye Pollock and Pacific Cod Distribution

We collaborated with this project by providing historical data from AFSC ichthyoplankton surveys to postdoctoral researcher N. Bacheler for use in data analyses and interpretation. Data sharing resulted in three publications, Bacheler et al. 2009, Bachelor et al. 2010 (co-authored with J. Duffy-Anderson), and Bacheler et al. (2012). These publications helped elucidate spawning areas of Walleye Pollock in the Bering Sea, effects of water temperature on pollock spawning phenology, and landscape ecology and density-dependent effects of spawning biomass on egg density. We also collaborated on a paper to examine how variations in major currents (Bering Slope Current, cross-shelf currents) affect recruitment of Walleye Pollock, Pacific Cod, Arrowtooth Flounder (Vestfals et al. in press). In addition, L. Ciannelli trained postdoctoral associate T. Smart in spatial statistics for use in data analyses.

271

Management Implications

Predictions of continued warming trends in the Bering Sea heighten the need for mechanistic understanding of factors that contribute to spatial and temporal shifts in Walleye Pollock biomass and to variations in Walleye Pollock recruitment. It has long been recognized that population variability of fecund marine species like Walleye Pollock is linked to factors influencing growth and survivorship of early life history stages. Results from this project contribute to the growing body of knowledge of factors influencing Walleye Pollock distribution, growth, abundance, survivorship, and recruitment. Our work is useful to stock assessment and research managers and stakeholders as it identifies critical areas of young Walleye Pollock production, which appear to be spatially and temporally variable and dependent on climate conditions. This has important implications for management of the Walleye Pollock roe fishery, which targets pre-spawning adults for commercial harvest. In addition, these dynamic areas could be considered Essential Fish Habitat (EFH) for Walleye Pollock during spawning periods. Next, we demonstrate that oceanographic conditions, in particular, temperature and zooplankton prey availability, have pronounced effects not only on distribution of young fish but also on Walleye Pollock rates of growth and mortality in the months prior to onset of the first winter. There is increasing awareness that the quality of prey available to juvenile Walleye Pollock has an impact on recruitment success. Results from our study, combined with those from B54 Seasonal Bioenergetics (Heintz et al.) demonstrate mechanisms leading to differences in recruitment success under varying climate conditions. Such biological responses to broad-scale climate changes can be incorporated into the stock assessment to better predict recruitment success under future climate scenarios. Finally, our work also points to the likelihood of thermally-mediated spatial expansions and contractions of spawning areas, which is of significant interest to managers since contractions in spatial spawning extent can be indicative of overall declines in population biomass. Additionally, changes in spatial extent may result in changes in predator-prey dynamics; such information could be used in Integrated Ecosystem Assessments (IEA; www.noaa.gov/iea). We note that continued monitoring of Walleye Pollock populations in the eastern Bering Sea is central to honing predictive capability. Long-term programs to monitor and evaluate interannual spatial and temporal spawning variations, diet and nutritional condition of larvae and juveniles, and stage- specific shifts in centers of distribution is central to validating the above observations and providing stakeholders useful quantitative tools to successfully manage the fishery.

272

Publications (*primary and collaborative with other BSIERP projects)

*De Forest, L., Duffy-Anderson, J.T., Heintz, R., Matarese, A.M., Siddon, E., Smart, T. and Spies, I. In press. Ecology and taxonomy of the early life stages of arrowtooth (Atheresthes stomias) and Kamchatka (Atheresthes evermanni) flounder in the eastern Bering Sea. Deep Sea Research II: Topical Studies in Oceanography. http://dx.doi.org/10.1016/j.dsr2.2014.05.005

*Duffy-Anderson, J.T., Barbeaux, S., Farley, E., Heintz, R., Horner, J., Parker-Stetter, S., Petrik, C., Siddon, E.C., and Smart, T.I. In review. An ecological synthesis of the first year of life of Walleye Pollock (Gadus chalcogrammus) in the eastern Bering Sea. Deep Sea Research II: Topical Studies in Oceanography.

Heintz, R.A., Siddon, E.C., Farley Jr, E.V., Napp, J.M. 2013. Correlation between recruitment and fall condition of age-0 pollock (Theragra chalcogramma) from the eastern Bering Sea under varying climate conditions. Deep Sea Research II: Topical Studies in Oceanography. 94:150-156.

Hunt, G.L., Coyle, K.O., Eisner, L.B., Farley, E.V., Heintz, R., Mueter, F.J., Napp, J.M., Overland, J.E., Ressler, P.H., Salo, S., Stabeno, P.J. 2011. Climate impacts on eastern Bering Sea foodwebs: a synthesis of new data and an assessment of the Oscillating Control Hypothesis. ICES Journal of Marine Science. 68: 1230-1243.

*Petrik, C., Duffy-Anderson, J.T., Mueter, F.J., Hedstrom, K., and Curchitser, E. In press. Modeling the effect of climate variations on the transport and distribution of Walleye Pollock early life stages in the eastern Bering Sea. Progress in Oceanography. http://www.sciencedirect.com/science/article/pii/S0079661114001128

*Siddon, E.C., Duffy-Anderson, J.T., and Mueter, F. 2011. Community-level response of ichthyoplankton to environmental variability in the eastern Bering Sea. Marine Ecology Progress Series. 426: 225-239.

Sigler, M.F., Harvey, H.R., Ashjian, C.J., Lomas, M.W., Napp, J.M., Stabeno, P.J., Van Pelt, T. 2012. How does climate change affect the Bering Sea Ecosystem? EOS. 91: 457-468.

*Smart, T., Duffy-Anderson, J.T., Horne, J. 2012a. Alternating climate states influence Walleye Pollock life stages in the southeastern Bering Sea. Marine Ecology Progress Series. 455: 257-267.

*Smart, T., Duffy-Anderson, J.T., Horne, J., Farley, E., Wilson, C., and Napp, J. 2012b. Influence of environment on Walleye Pollock eggs, larvae, and juveniles in the Southeastern Bering Sea. Deep Sea Res. II: Topical Studies in Oceanography. 65-70: 196-207.

273

*Smart, T., Duffy-Anderson, J.T., and Siddon, E. 2013. Vertical distribution of early life stages of walleye Pollock and implications for transport and connectivity. Deep Sea Research II: Topical Studies in Oceanography. 94: 201-201.

Stabeno P.J., Farley E.V. Jr, Kachel N.B., Moore S., Mordy C.W., Napp J.M., Overland J.E., Pinchuk A.I., Sigler M.F. 2012. A comparison of the physics of the northern and southern shelves of the eastern Bering Sea and some implications for the ecosystem. Deep Sea Research II: Topical Studies in Oceanography. 65-70: 14-30.

Stabeno, P.J., Kachel, N.B., Moore, S., Napp, J.M., Sigler, M.F., Yamaguchi, A., Zerbini, A.N. In press. Comparison of warm and cold years on the southeastern Bering Sea shelf and some implications for the ecosystem. Deep Sea Research II: Topical Studies in Oceanography. 65-70: 31-45.

Stabeno P.J., Napp J.M., Mordy C., Whitledge T. 2010. Factors influencing physical structure and lower trophic levels of the eastern Bering Sea shelf in 2005: Sea ice, tides and winds. Progress in Oceanography. 85: 180-196.

Vestfals, C., Ciannelli, L., Duffy-Anderson, J.T., and Ladd, C. In press. Effects of seasonal and interannual variability of along-shelf and cross-shelf transport on groundfish recruitment in the eastern Bering Sea. Deep Sea Research II: Topical Studies in Oceanography.

274

Poster and Oral Presentations (*Indicates award winner)

Bacheler, N., Ciannelli, L., Duffy-Anderson, J.T. and Bailey, K. Distribution of Walleye Pollock, Atheresthes. spp., and Pacific Cod early life history stages in the Bering Sea. 2008. Poster, North Pacific Research Board Annual Meeting. Anchorage, AK. October.

De Forest, L., Duffy-Anderson, J., Heintz, R., Matarese, A., Siddon, E., Smart, T., and Spies, I. 2011. Current knowledge of the early life history of arrowtooth flounder (Atheresthes stomias) in the eastern Bering Sea: with comments on Kamchatka flounder (A. evermanni). Gilbert Ichthyological Society, Seabeck, WA. October.

De Forest, L., Smart, T., Duffy-Anderson, J., Matarese, A., Heintz, R., Siddon, E., Spies, I., and Cooper, D. 2011. Current knowledge of the early life history of arrowtooth flounder (Atheresthes stomias) in the eastern Bering Sea: with comments on Kamchatka flounder (A. evermanni). Poster, International Flatfish Symposium, Ijmuiden, Netherlands. November.

De Forest, L., Duffy-Anderson, J.T., Heintz, R., Matarese, A., Siddon, E., Smart, T., and Spies, I. 2013. Current knowledge on the ecology and taxonomy of the early life stages of arrowtooth (Atheresthes stomias) and Kamchatka flounder (A. evermanni) in the eastern Bering Sea. (Poster) Alaska Marine Science Symposium Anchorage, AK. January.

De Forest, L., Duffy-Anderson, J., Heintz, R., Matarese, A., Siddon, E., Smart, T., and Spies, I. 2013. Ecology and taxonomy of the early life stages of arrowtooth flounder (Atheresthes stomias) and Kamchatka flounder (A. evermanni) in the eastern Bering Sea. Larval Fish Conference, Miami, FL. June.

Duffy-Anderson, J.T. 2011. Ichthyoplankton dynamics and seasonal bioenergetics in the southeast Bering Sea. NPRB BSIERP PI Meeting, Anchorage, AK. March.

Duffy-Anderson, J.T., Hillgruber, N., Napp, J., Matarese, A., Eisner, L., and Heintz, R. 2009. Ichthyoplankton dynamics and seasonal bioenergetics in the eastern Bering Sea. North Pacific Research Board Annual Meeting. Girdwood, AK. October.

Duffy-Anderson, J.T., Siddon, E., Smart, T., and Hillgruber, N. 2009. Vertical distribution of larval Walleye Pollock, Pacific Cod, and Arrowtooth Flounder in the eastern Bering Sea. North Pacific Research Board Annual Meeting. Girdwood, AK. November.

Duffy-Anderson, J.T., Smart, T.I., Siddon, E., Mueter, F., Matarese, A., Eisner, L., Napp, J., and Heintz, R. 2011. Ichthyoplankton and seasonal bioenergetics. BSIERP PI Meeting, Anchorage, AK. January. 275

Matarese, A., De Forest, L., Duffy-Anderson, J., Smart, T., and Spies, I. 2013. Identification and distribution of the early life stages of arrowtooth (Atheresthes stomias) and Kamchatka flounder (A. evermanni) in the eastern Bering Sea. (Poster) Larval Fish Conference, Miami, FL. June.

Napp, J.M., Eisner, L.B., Farley, E.W., Stabeno, P.J., Hunt, Jr., G.L. 2010. Recent changes in zooplankton abundance and biomass in the eastern Bering Sea, Alaska Marine Science Symposium, Anchorage, AK, January.

Napp, J.M., Sigler, M.F., and Stabeno, P.J. Understanding ecosystem processes in the Bering Sea. 2010. NOAA Headquarters Lunch Seminar, Silver Spring, MD, March.

Hillgruber, N., Duffy-Anderson, J.T., Eisner, L., Heintz, R., Matarese, A., Napp, J., and Siddon, E. 2008. Distribution, transport, and condition of early life stages of Walleye Pollock, Pacific Cod, and Arrowtooth Flounder in the eastern Bering Sea under the auspice of changing climatic conditions. Contributed poster. Larval Fish Conference. Kiel, Germany. August.

Petrik, C.M., Duffy-Anderson, J.T., Mueter, F., Hedstrom, K., Danielson, S., Curchitser, E., and Barbeaux, S. 2013. Modeling climate effects on the dispersal and distribution of pollock early life stages in the eastern Bering Sea. ICES Annual Science Conference. Reykjavik, Iceland. September.

*Petrik, C.M., Duffy-Anderson, J.T., Mueter, F., Hedstrom, K., Danielson, S., Curchitser, E., and Barbeaux, S. 2013. Modeling climate effects on the dispersal and distribution of pollock early life stages in the eastern Bering Sea. PICES Annual Science Conference. Nanaimo, British Columbia. October. *Best Paper, Physical Oceanography section

*Siddon, E., Duffy-Anderson, J.T., and Mueter, F. 2010. Community-level response of ichthyoplankton to environmental variability in the eastern Bering Sea. Western Groundfish Conference. April. Juneau, AK.*Best Student Paper Award

Smart, T.I. and Duffy-Anderson, J.T. The Use of Generalized Additive Models to Investigate Alternating Environmental States and the Early Life Stages of Walleye Pollock in the Eastern Bering Sea. 2010. Quantitative Seminar Series, School of Aquatic and Fishery Sciences, UW, January.

Smart, T.I. and Duffy-Anderson, J.T. Alternating Environmental States and the Early Life Stages of Walleye Pollock in the Eastern Bering Sea. 2010. EcoFOCI Seminar Series, NOAA/AFSC, December.

276

Smart, T.I., Siddon, E., Duffy-Anderson, J.T., and Hillgruber, N. 2009. Vertical distribution of larval Walleye Pollock (Theragra chalcogramma), Pacific Cod (Gadus macrocephalus), and Atheresthes spp. in the eastern Bering Sea (poster). BEST-BSIERP Principal Investigators Meeting, October.

Smart, T., Duffy-Anderson, J.T., and Horne, J. 2011. Alternating climate states influence Walleye Pollock early life stages in the southeastern Bering Sea. PICES/ESASS Symposium. May.

Smart, T., Duffy-Anderson, J.T., Horne, J., and Farley, E. 2011. Alternating climate states influence Walleye Pollock early life stages in the southeastern Bering Sea. Alaska Marine Science Symposium. January.

Smart, T.I., Duffy-Anderson, J.T., Horne, J., Farley, E., Stabeno, P. 2010. Alternating climate states and the effects on vital rates of Walleye Pollock early life stages in the Bering Sea. Oral presentation, Larval Biology Symposium, Wellington, NZ. August.

Smart, T., Siddon, E.C., Duffy-Anderson, J.T., and Hillgruber, N. 2010. Vertical distribution of two common larval gadids, Walleye Pollock (Theragra chalcogramma) and Pacific Cod (Gadus macrocephalus), in the eastern Bering Sea and their relationships with climate variables. Ocean Sciences Meeting. Portland, OR. February.

*Siddon, E., Duffy-Anderson, J.T., Mueter, F., and Hillgruber, N. 2010. Interannual variability in cross- shelf assemblage structure of ichthyoplankton in the eastern Bering Sea. Alaska Marine Science Symposium, January. Anchorage, AK. *Best Student Paper Award

Siddon, E.C. Mueter, F.J., Heintz, R., and Duffy-Anderson, J.T. 2012. Shifts in community structure reflect species-specific responses to climate variability in sub-arctic seas. PICES Early Career Scientist Conference. Mallorca, Spain. April.

Smart, T., Siddon, E., Duffy-Anderson, J.T., and Hillgruber, N. 2009. Vertical distribution of larval Walleye Pollock, Pacific Cod, and Arrowtooth Flounder in the eastern Bering Sea. North Pacific Research Board Annual Meeting. Girdwood, AK. October.

277

Outreach

Exhibit/Display:

Tracey Smart developed a hands-on activity with the Pacific Science Center in Seattle, WA, to demonstrate the collection and identification of marine larvae in the Bering Sea. This activity was presented during Polar Science Weekend (March 3-6, 2011) and was spotlighted during several Research Weekends at the Pacific Science Center. Duplicates of this activity were provided to Nora Deans at NPRB.

Marine Larvae of the Bering Sea. Presented at the Pacific Science Center.

Website: The zooplankton/ichthyoplankton sampling effort aboard the WHOI ship Knorr BEST/BSIERP cruise was supported by the generous help provided by the PolarTrec teacher-at-sea, Mr. Mark McKay. Mr. McKay reported enthusiastically on the MOCNESS sampling effort in his weblog (http://www.polartrec.com/bering-ecosystem-study-09).

Presentations in Schools: May 2009: Janet Duffy-Anderson made presentation to a Seattle Public School (North Beach Elementary) on work in the sub-arctic and on NOAA’s involvement in BEST-BSIERP activities April 2010: Janet Duffy-Anderson prepared a hands-on activity in Seattle Public School (North Beach Elementary) on measuring wind speed and examining effects of water currents. 278

March 2012: Janet Duffy-Anderson presented a marine plankton display as part of Science Day in Seattle Public Schools. The presentation included several data examples from this project.

Community Meeting:

Franz Mueter presented a public seminar in Nome, Alaska, on May 29, 2012, on the spatial dynamics of fishes in the Bering Sea with a special focus on the northern Bering Sea. The presentation included examples of the distribution of early life stages from this project.

279

Acknowledgements

Funding for this project was provided by, the North Pacific Research Board’s Bering Sea Integrated Ecosystem Research Program, the National Science Foundation’s Bering Ecosystem Study program, NOAA Fisheries Alaska Fisheries Science Center’s Ecosystem and Fisheries Oceanography Coordinated Investigations program, the University of Alaska, the University of Washington’s School of Aquatic and Fishery Sciences, and Oregon State University’s College of Earth, Ocean, and Atmospheric Sciences.

280