Forage Fishes in the Western Gulf of Alaska: Variation in Productivity. Fisheries Science Center, 7600 Sand Point Way NE, Seattle, WA 98115

Transcription

1 Forage Fishes in the Western Gulf of Alaska: Variation in Productivity Matthew T. Wilson 1, Mike Mazur 2, Andre Buchheister 3, Janet Duffy-Anderson 1 1 National Oceanic and Atmospheric Administration, National Marine Fisheries Service, Alaska Fisheries Science Center, 7600 Sand Point Way NE, Seattle, WA Joint Institute for the Study of the Atmosphere and Ocean, Lisa Li Building, th Ave NE, University of Washington, Box , Seattle, WA Virginia Institute of Marine Science, P.O. Box 1346, Gloucester Point, Virginia PROJECT 0308 FINAL REPORT to NORTH PACIFIC RESEARCH BOARD 1007 West 3 rd Avenue, Suite 100 Anchorage, AK 99501

2 PREFACE In productive coastal ecosystems, zooplankton are consumed by small pelagic fishes that, in turn, are consumed by fishes, birds, and mammals. In this way, forage fishes facilitate the transfer of trophic energy up the food web to commercial fisheries and protected marine mammals. Because zooplankton are advected by ocean currents, hydrographic features can concentrate zooplankton and affect prey availability for forage fishes. This impacts forage fish production. In addition to spatial variability, seasonal cycles in zooplankton production affect forage fish production. In this Final Report, we present research resulting from a two-part, three-year project designed to examine geographic and seasonal effects on forage fish productivity in the western Gulf of Alaska (GOA). Part I focused on walleye pollock juveniles (top, left), capelin (top, center), and eulachon (top, right). Relevant prey taxa include euphausiids (middle, left and inset), and copepods (middle, right and inset) whose geographic distributions corresponded with hydrography. The main hydrographic feature is the bifurcation in the Alaska Coastal Current (ACC) as it flows into our study area (bottom, inset white box ) (current vectors adapted from Reed and Schumacher 1986). Part II provides a seasonal perspective by focusing on juvenile pollock before, during, and after the north temperate winter in a large part of the GOA (bottom, inset red box). i

3 ABSTRACT In 2000, the Ecosystems and Fisheries-Oceanography Coordinated Investigations Program (Eco- FOCI) in NOAA 1 initiated a two-part project to examine the effects of coastal hydrography and seasonality on forage fish productivity as mediated through bottom-up processes in the western Gulf of Alaska (GOA). These topics are relevant to understanding how large-scale, climaterelated perturbations force ecosystem dynamics, which takes place against a complicating backdrop of geographic and seasonal variation. Project Part I focused on Alaska Coastal Current (ACC)-related prey enrichment in the Shelikof sea valley relative to the adjacent slope and southwestern shelf. Target fishes were juvenile walleye pollock Theragra chalcogramma, capelin Mallotus villosus, and eulachon Thalichthys pacificus. A 48-station grid was occupied day and night during September 2000, 2001, and 2003 to measure physical conditions and obtain samples of zooplankton and fish. Influence of the buoyancy-driven ACC on local hydrography was clearly evident in all three years. Related to this, euphausiids and large copepods exhibited interannually consistent high abundance in and near the Shelikof sea valley as compared to the adjacent southwestern shelf. Euphausiids and large copepods were important, energy-rich, prey of the target fish species. The fish response to localized prey enrichment was affected by their body size. In terms of fish length and abundance, small fishes (age-0 pollock and capelin) were larger in prey-rich areas, medium-size fish (eulachon) increased in size and abundance in preyrich areas, and large fish (age-1+ pollock juveniles) were more abundant in prey-rich areas. Steller sea lion populations and other upper trophic-level foragers benefit from processes that result in high concentrations of forage fishes. In terms of diet and length-specific body weight (body condition), juvenile walleye pollock were more responsive than eulachon and capelin. The highest quality growth habitat was associated with the Shelikof sea valley for age-0 pollock but not for capelin. Age-0 pollock are dietary generalists relative to capelin. Relative to geographic variation in prey, thermal effects on fish physiology had a negligible effect on fish growth potentials. Estimated growth potential from bioenergetics models indicated that the location of high growth-potential habitat varied interannually with ACC velocity, which relates to freshwater influx and wind. Thus, climate-related forcing on hydrographic dynamics in the GOA most affected forage fish growth potentials by altering their potential prey field. 1 National Oceanic and Atmospheric Administration ii

4 Project Part II focused on changes in juvenile walleye pollock diet, feeding intensity, and body condition before, during, and after the north temperate winter. Individuals of the 2000 year class were collected during trawl surveys that sampled in three regions of the western GOA, Most fish had relatively little prey biomass in their stomachs during winter when euphausiid consumption was low. Correspondingly, body condition was relatively low during winter. As for many temperate fishes, winter was associated with decreased feeding, a shift in diet, and reduced body condition. There was, however, regional variation in stomach fullness and body condition indicating that pollock nurseries near Kodiak Island provide environmental conditions that temper wintertime adversity. Regional differences in wintertime nursery habitat might therefore affect juvenile walleye pollock survival and subsequent recruitment of individuals to spawning populations. Similar geographic modulation of temporal variation in nursery quality may occur over longer time scales. Many products have resulted from this research. Our findings have been reported in 14 scientific presentations (oral and poster) at national and international meetings. Five manuscripts have been published or are in press; five additional manuscripts are nearly ready for peer-review. Finally, we have demonstrated specific mechanisms that relate to understanding climateecosystem dynamics and the supply of prey to upper level trophic animals. Key words: juvenile walleye pollock, capelin, eulachon, distribution, size, diet, condition, bioenergetics, geographic variation, seasonal variation Citation: Wilson MT, Mazur M, Buchheister A, Duffy-Anderson JT (2006) Forage fishes in the western Gulf of Alaska: Variation in productivity. North Pacific Research Board Final Report 308, 130 p iii

5 TABLE OF CONTENTS STUDY CHRONOLOGY INTRODUCTION. 2 OBJECTIVES... 4 MANUSCRIPTS: PART I Chapter 1. Geographic distribution of pelagic forage fishes in relation to potential prey abundance and hydrography. 6 Chapter 2. Geographic variation in body weight, feeding intensity, and diet of pelagic forage fishes in relation to hydrography Chapter 3. Bioenergetics model-based estimates of the growth of two forage fish species Section 1. Geographic variation in growth of age-0 walleye pollock. 52 Section 2. Geographic variation in growth of capelin.. 72 Discussion and Conclusions 81 MANUSCRIPTS: PART II Chapter 1. Over-winter effect on the diet of juvenile walleye pollock 85 Chapter 2. Over-winter effect on body condition and energy density of juvenile walleye pollock.. 97 Discussion and Conclusions. 115 ADDITIONAL STUDIES 1. Ichthyoplankton community structure in relation to hydrography Comparative feeding ecology of age-0 walleye pollock and capelin Length conversions and preservation corrections for three forage fishes 119 FUTURE RESEARCH DIRECTION 120 PUBLICATIONS. 124 OUTREACH 125 ACKNOWLEDGMENTS 126 LITERATURE CITED iv

6 STUDY CHRONOLOGY In 2000, the Ecosystems and Fisheries-Oceanography Coordinated Investigations Program (eco- FOCI) in the National Oceanic and Atmospheric Administration (NOAA) initiated a project to investigate potential effects of coastal hydrography on forage fish productivity in the western Gulf of Alaska (GOA). The project involved Principal Investigators (PI) in the eco-foci Program and collaborators from academic institutions (see NPRB Project R0308 proposal entitled, Forage fishes in the western Gulf of Alaska: Variation in productivity ). From , the Project was partially supported by the NOAA/Steller Sea Lion Research Initiative (Grant 02FF-04 and Grant ). From , the Project was partially supported by the North Pacific Research Board (NPRB) Grant R0308. This Final Report, therefore, necessarily reflects work supported by many agencies but emphasizes research that particularly benefited from NPRB support. It includes updated information that was previously submitted as NPRB Semiannual Progress Reports on the 15 th day of July and January

7 INTRODUCTION Development of good resource management principles and predicted impact of climate change on ecosystems depend on understanding the mechanisms that link environmental variation to biotic population dynamics (Levin 1992). Thorough understanding of these mechanisms can be greatly facilitated by field studies of meso-scale spatial and temporal processes (Adams 2002). The funds provided by the North Pacific Research Board enabled us to conduct our proposed work involving two meso-scale field studies in the western Gulf of Alaska (GOA) to examine factors affecting forage fish productivity. Forage fishes are small, schooling fishes that are prevalent in the diet of other species (Springer and Speckman 1997). In coastal regions of the western GOA, forage fishes such as juvenile pollock Theragra chalcogramma, capelin Mallotus villosus, and eulachon Thalichthys pacificus provide a vital link in the flow of food energy through the ecosystem. These fishes eat zooplankton and, in turn, are eaten by sea birds, mammals, and piscivorous fishes. In fact, Anderson and Piatt (1999) suggest that a recent climate-related shift in the composition of the forage fish community may have contributed to the Steller sea lion population decline. A key forage fish species in the GOA is juvenile walleye pollock, though other co-occurring forage fishes also fulfill important trophic links in the North Pacific ecosystem. In the western GOA, the area between Shelikof Strait and Unimak Pass is an important pollock nursery that hosts an abundance of capelin and eulachon (Brodeur and Wilson 1996, Wilson 1996). There is some evidence, however, that the nursery area is fully exploited by these and other species (Bailey et al. 1996, Merati and Brodeur 1996, Wilson 2000) potentially making it sensitive to subtle influxes or depletions in zooplankton prey. Thus, constraints on the supply of zooplankton to this area may determine local forage fish productivity. One source of zooplankton exists at the northeast, upstream end of the area due to the southwestward-flowing Alaska Coastal Current (ACC). The ACC is the dominant coastal current in the Gulf, and it exhibits considerable seasonal (Stabeno et al. 2004) and interannual (Ingraham et al. 1998) variability. Southwest of Kodiak Island, most of the ACC water follows the Shelikof sea valley offshore, but there is some flow over the western margin of the sea valley (Sutwik Island Semidi Bank vicinity). Thus, if this water is rich in zooplankton, it should enrich the 2

8 western margin relative to more downstream areas of the nursery. Furthermore, if forage fish in the area are prey limited, their productivity should positively respond to zooplankton enrichment. Food-related constraint on productivity is likely to be most evident during autumn due to a seasonal decline in zooplankton abundance (Cooney 1986), and an influx of age-0 pollock. The magnitude of the pollock influx depends on larval survival (Bailey et al. 1996b) and transport (Hinckley et al. 1991, Wilson 2000), but the abundance of young pollock has the potential to exhaust local prey resources (Duffy-Anderson et al. 2002). In addition to constraining productivity, prey shortages will also intensify competition among predators. A second part of our proposed work addresses the hypothesis that winter conditions adversely affect juvenile pollock by eliciting reduced feeding, and lower body condition and decreased whole body energy content. Part of this work involves examining juvenile pollock otoliths for evidence that microstructural features may be useful for examining size-selective over-winter survival. For juvenile fish, slow growth translates into increased vulnerability to predation and consequent reduction in the chance of survival to harvest size (Sogard 1997). 3

9 OBJECTIVES Project objectives were organized into two parts. In each part, study objectives were primarily focused around a single hypothesis. For Part I, the hypothesis pertained to ACC-related prey enrichment in the vicinity of the Shelikof sea valley and the response of potentially prey-limited forage fishes: H-1: The forage-fish prey field is enriched over the western margin of the Shelikof sea valley; this enrichment is evident in forage fish productivity. This hypothesis was addressed in three studies (chapters) using data collected during September 2000, 2001, and 2003: 1. Geographic distribution of pelagic forage fishes in relation to potential prey abundance and hydrography, 2. Geographic variation in body weight, feeding intensity, and diet of pelagic forage fishes in relation to hydrography, 3. Bioenergetics model-based estimates of the growth of two forage fish species i. Geographic variation in growth of age-0 walleye pollock, ii. Geographic variation in growth of capelin. For Part II, the focal hypothesis pertained to winter-related effects on juvenile walleye pollock: H-1: Over-winter effects on juvenile walleye pollock result in reduced feeding, lower body condition and whole body energy content, and formation of the first-winter annulus on otoliths. This hypothesis was addressed in two studies (chapters), and is being addressed in an on-going study: 1. Over-winter effect on body condition and energy density of juvenile walleye pollock, 2. Over-winter effects on the diet of juvenile walleye pollock, 3. Utility of the otolith first-winter annulus as a tool in pollock recruitment studies (ongoing). 4

10 In addition to these two parts, we capitalized on collections made during Part I to leverage our contribution to forage fish research. The role of transport and competition were identified in the proposal as important factors in forage fish population dynamics. We therefore contributed information on the spatial structure of larval fishes relative to hydrography (Lanksbury et al. 2005), and on the comparative feeding ecology of age-0 walleye pollock and capelin (Wilson et al. in press). We were also able to provide length conversions and preservation corrections for three forage fish species so that methodological differences among field studies can be adjusted to yield a common size metric (Buchheister and Wilson 2005). 5

11 MANUSCRIPTS: PART I Chapter 1 Geographic distribution of pelagic forage fishes in relation to potential prey abundance and hydrography [NPRB Final Report -- Please do not cite or use information in this chapter without permission of the author. This material is being prepared for publication] Introduction: In many marine ecosystems, trophic energy is funneled through pelagic food webs by small midwater fishes (Cury et al. 2000). These fishes are usually planktivorous, highly abundant, and susceptible to predation; hence, they are often grouped under the common forage fish banner (Springer and Speckman 1997). High productivity of forage fishes is usually associated with plankton-rich upwelling areas (Mann and Lazier 1996); however, pelagic forage fishes abound in coastal waters (i.e. inshore of the continental slope) of the Gulf of Alaska (Anderson and Piatt 1999, Wilson et al. 1996, Abookire and Piatt 2005), which is considered a downwelling system (Stabeno et al. 2004). The Ecosystems and Fisheries-Oceanography Coordinated Investigations Program (eco- FOCI) in NOAA initiated a study in the western Gulf of Alaska (GOA) during 2000 to examine possible effects of coastal hydrography on the productivity of forage fishes as mediated through a bottom-up trophic link. The focal species in this study were juvenile walleye pollock (Theragra chalcogramma) and two osmerids: capelin Mallotus villosus and eulachon Thaleichthys pacificus. These are midwater zooplanktivores whose climate-related fluctuations in abundance have been associated with alterations in the trophic structure of the GOA ecosystem (Anderson and Piatt 1997). The underlying bottom-up controls on these fishes, however, are not well understood. Walleye pollock is a relatively long-lived, economically important gadid whose juveniles are commonly consumed by other fishes, birds, and mammals (Sinclair and Zeppelin 2002, Sanger and Hatch 1992). Capelin and eulachon are relatively small fishes valued as abundant, energyrich forage species (Payne et al. 1999, Iverson et al. 2002, Anthony et al. 2000, Yang et al. 2005). In turn, these fishes primarily consume euphausiids and copepods (Brodeur and Wilson 1996, M. Wilson unpublished data) with occasional high occurrences of larvaceans and thecosomate pteropods, at least in capelin and age-0 walleye pollock diets (Wilson et al. in 6

12 press). Unlike capelin and eulachon, which are protected from commercial harvest, there has been considerable research on juvenile walleye pollock. Merati and Brodeur (1996) report regional variation in the diet of age-0 juveniles in the western Gulf that generally corresponds with variation in their growth rates (Bailey et al. 1996) and body size (Wilson 2000). Indications are that a regional delineation in habitat quality exists at about 156 o W longitude, which is where the Shelikof sea valley crosses the continental shelf. This bathymetric feature has significant potential relevance to forage fish productivity due to its effect on hydrography, which in turn might affect zooplankton availability, but this potential has not been investigated. The shelf area between Shelikof Strait and the Shumagin Islands (Fig. 1) was strategically chosen for study based on prior oceanographic and biological information. Much of this information is from research related to the large, commercially important, aggregation of walleye pollock Theragra chalcogramma that spawn in Shelikof Strait during March (Kendall et al. 1996). Larvae from this spawning are advected southwestward by the Alaska Coastal Current (ACC), which is the main circulation feature in the Gulf (Stabeno et al. 2004). The ACC bifurcates near Sutwik Island with one branch continuing southwestward along the Alaska Peninsula while the other follows the Shelikof sea valley offshore (Schumacher and Reed 1986). The nearshore branch is indicated by relatively low-salinity (< 31.5 psu) while the offshore branch is indicated by relatively rapid flow. The offshore flow could export zooplankton and diminish the forage fish prey field over the shelf. Napp et al. (1996) have shown that, at least during spring, the ACC is rich in zooplankton. On the other hand, the offshore flow stimulates estuarine-like complementary influxes of oceanic water in the sea valley (Stabeno et al. 2004). This could result in localized prey enrichment especially during the period of maximum flow, which occurs in autumn (Stabeno et al. 2004). Such enrichment might be particularly important during the autumnal decline in zooplankton abundance (Coyle and Pinchuk 2003) when the demand for zooplankton might increase due to influxes of young-of-the-year (age-0) walleye pollock (Brodeur and Wilson 1996). The chance of detecting a prey-related hydrographic effect on the productivity of forage fishes might therefore be especially good during autumn. Thus, the objective of this study was to determine if specific geographic differences in forage fish distribution correspond to the distribution of major prey groups. The specific geographic effect of interest pertains to the ACC and was defined by empirical observations of salinity and model- 7

13 based estimates of current velocity. Fish size was included in the analysis as a relevant factor in fish-prey relationships. Methods: The study area encompassed the Sutwik Island vicinity and included the continental shelf and upper continental slope from the Shelikof sea valley to the Shumagin Islands. A 48- station grid was occupied during September 2000, 2001, and 2003 (Fig. 1). Station locations were somewhat unevenly spaced to allow for navigational hazards, and to ensure that major bathymetric regions (<50 m, m, >200 m water depth) were sampled. Not all stations were sampled each year due to logistic constraints; however, most of those that were sampled each year were occupied once during the day (ca. 07:00 22:00 Alaska Daylight Time) and again at night (Table 1). Sampling was conducted aboard the NOAA Ship Miller Freeman and progressed from southwest to northeast. Station occupation involved measuring temperature and salinity throughout the water column and obtaining samples of the ambient zooplankton and fishes. Physical environment. Temperature, salinity, and depth measurements were obtained using a calibrated Sea-Bird Electronics 2 (SBE) 19 SeaCat profiler. The profiler was mounted on the tow wire just above the Tucker trawl and provided real-time depth information. Measurements were made during the upcast at 40 m depth (n=217) or, if shallower, at the maximum gear depth (n=11). Each year, several profiles obtained with the SeaCat were compared to profiles obtained with a calibrated SBE 911 CTD and validated the accuracy of the SeaCat data. Bottom depth was measured acoustically using a Simrad EK-500 echosounder (38 khz). Soundings were adjusted to account for the 9-m depth of the transducer. Current vectors within the study area were obtained from the rigid-lid semi-spectral primitive equation model (SPEM) (Stabeno and Hermann 1996). Daily vectors of current at 40-m water depth were generated at a spatial resolution of 4-km 2 for September 2000, 2001, and Rectilinear current vectors were converted to polar coordinates. Velocities were then averaged to give mean daily velocity for each year at the original 4-km 2 resolution, which were geographically contoured using the inverse distance weighted algorithm in ArcMap (ver. 8.2) (ESRI, Inc.). Velocity was also averaged at 90 o -bearing intervals (e.g. N, 315 o -44 o ; E, 45 o o ; S, 135 o -224 o ; W, 225 o -314 o ) to characterize geographic sub-areas with regard to model- 2 Use of trade names does not imply endorsement by the National Oceanic and Atmospheric Administration. 8

14 based flow estimates. Velocity estimates were not available for shallow (<25 m) parts of the study area due to model limitations. Zooplankton.--Samples were collected with a 1-m Tucker trawl (0.333-mm mesh) fished to 10 m off bottom but not deeper than 200 m. Ship speed was adjusted to maintain a wire angle of ca. 45 o. The Tucker trawl was equipped with two opening-closing collecting nets. A calibrated General Oceanics flowmeter was suspended in the mouth center of each net. The first sample was from at-depth to the bottom of the thermocline. The second sample was from the bottom of the thermocline to the surface. A net retrieval rate of 20 m min -1 was maintained for both sampling intervals. Only one net was used at shallow (ca. <50 m) locations where no thermocline was present. Samples collected in all successful tows (n=230) were preserved in 10% formalin and sorted at the Poland Plankton Sorting Center (Szeczin, Poland) according to established protocol. Standardized abundance estimates (ind m -2 ) of zooplankton were calculated by dividing the number caught by the volume filtered. When two samples were collected during a tow, the standardized abundances from each were summed. The plankton abundance data were grouped into five broad taxonomic and size categories commonly consumed by these fishes (Chapter 2). The four taxonomic groups were copepods, euphausiids, thecosomate pteropods, and larvaceans. Copepods were further separated into two size categories following Wilson et al. (in press): <2 mm prosome length (PL), or >2 mm PL. Thus, abundance estimates of each of five groups of potential forage fish prey were examined. Forage fishes.--each fish sample was collected using a Stauffer (a.k.a. anchovy trawl) midwater trawl. The trawl was fished with 1.5 x 2.1 m steel-v otter doors (566 kg each), and the codend was equipped with a liner (3-mm mesh). It was deployed at about 50 m min -1, allowed to settle at 200 m or 20 m (headrope depth) above bottom, whichever was shallower, and then retrieved at about 10 m/min. During retrieval, ship speed over ground was ca kts. Net depth was monitored with a Furuno net sounder. Animals collected in all successful tows (n=235) were sorted, enumerated, and weighed at sea to obtain abundance and weight estimates. Lengths of up to 100 individuals of each focal species were measured. Larval osmerids were rarely measured; these were generally <40 mm SL and exhibited larval capelin characteristics (Matarese et al. 1989). Most fish were measured to the nearest 1 mm standard length (SL). Age- 9

15 1+ pollock (>120 mm SL) were measured to the nearest 1 cm fork length (FL) due to fleshy growth over the hypural plate. Fork length measurements were converted to SL using established equations (Buchheister and Wilson 2005). Trawl catches were standardized to ind. m -2 based on volume filtered and maximum headrope depth. Volume filtered was estimated from trawl mouth area and the distance between the start of trawl door deployment and completion of trawl door recovery. The sampling duration at each m- interval of depth was assumed constant but trawl mouth area varied. The trawl mouth was assumed to be elliptical with area determined by height and width. Trawl mouth height and width was measured at sea using Scanmar net mensuration equipment. Based on these measurements, trawl mouth area (m 2 ) was related to wire out (m) (R 2 =0.98, n=697, unpublished data): Eqn 1) m 2 = (1-e (-0.011(m )) ) Data analysis.--three-factor analysis of variance (ANOVA, The SAS System for Windows release 8.02) tests were used to detect a sub-area effect on plankton abundance, and on fish abundance and size. Year and diel effects were included in the full model to allow for possible interaction with the sub-area main effect: Eqn 2) Y ijklm = + X ijklm + Year j + Diel k + Sub-area l + Year j *Diel k + Year j *Sub-area l + Diel k *Sub-area l + Year j *Diel k *Sub-area l + Station m {Sub-area l } + Sample i {Station m } + e i where Y ijklm is the dependent variable of the i th sample of plankton or fish from the j th year, k th diel period, l th sub-area, m th station, and e i is the random error component. The covariate, X, was SL. Year, diel, and sub-area were fixed effects. Sample was nested within station, and station was nested within sub-area; both nested terms were random effects. Following Milliken and Johnson (1996), the model was reduced by sequentially dropping non-significant (p > 0.05) terms beginning with the 3-way interaction, then 2-way interactions, and finally any main effects not included in any retained interaction term(s). Only one term was dropped between successive model runs. Residual and normal probability plots indicated that 4 th -root transformation of abundance estimates was necessary and satisfactorily normalized the distribution of residuals. 10

16 In the ANOVA tests of fish size, mean length per sample was used instead of the individual length measurements. Sample was therefore dropped from the model and the Sample i {Station m } term was omitted. This was necessary due to computer memory problems related to the large number of age-0 walleye pollock length measurements. For the other fish groups, conclusions based on tests of mean length versus individual measurements were similar. Mean length was weighted by the standardized catch (no m -2 ). For age-0 walleye pollock, a mean daily growth rate of 0.89 mm d -1, which is the longterm average for individuals in the Shumagin area (Bailey et al. 1996), and an instantaneous daily mortality of (Hollowed et al. 1996) was used to adjust mean size and abundance to the median survey date (12 September). However, these adjustments had little effect on the results and similar estimates were not available for the other fish; therefore, only results of the analysis of unadjusted data are presented. No samples were collected over the slope at night during This was problematic because it created an empty cell in the ANOVA test design. Although not ideal, the cell was filled using day/night abundance ratio estimators applied to the 3 daytime samples collected over the slope during The ratio estimator for each taxa was based on 111 day-night sample pairs. Following Cochran (1977), the estimator was calculated as the sum of abundances for stations sampled at day versus those sampled at night. Results: Sampling occurred during 1-19, 2-18, and 4-17 September in 2000, 2001, and 2003, respectively. A total of 235 and 230 samples of fish and zooplankton, respectively, were collected over these dates (Table 1, Fig. 1). All but 13 of the samples were paired by station into day-night pairs. Most of the 13 unpaired samples were collected during the day, which at this time of year lasts longer than night. Eight samples were from the sub-area near Chirikof Island, which was omitted from subsequent analysis due to the low number of samples. Few animals were collected in this sub-area. Physical environment.--salinity and model-based estimates of current velocity at 40-m depth could reasonably be used to divide the study area into five hydrographically distinct sub-areas. Salinity at 40-m depth ranged from 29.9 to 32.7 psu. Each year, relatively low values were observed over the inner shelf and adjacent to the Semidi Islands (Table 2, Fig. 2). Higher salinity was observed over the middle shelf and was assumed to indicate decreased ACC influence. Thus, 11

17 geographic stratification based on salinity yielded two shelf sub-areas: inshore and middle/outershelf. Velocity estimates from SPEM-model output indicated that water currents were generally fastest northeast of Sutwik and Semidi islands. Thus, geographic stratification based on velocity estimates yielded two sub-areas: southwest, and northeast. Superimposing the salinity- and velocity-based stratification schemes yielded four shelf sub-areas. The relatively high salinity and velocity associated with the slope presumably reflect Alaska Stream influence. The slope was therefore kept distinct as a fifth sub-area. The predominant current direction in these sub-areas, as inferred from the SPEM model, was toward the southwest (Fig. 3). Mean velocity estimates over the shelf were usually low (<100 mm s -1 ) relative to those estimated over the slope. During 2001, however, mean velocity estimates toward the south and west in the NEin Sub-area greatly exceeded 100 mm s -1. Water depth and thermal differences among sub-areas were also apparent (Table 2). Bottom depth increased from inshore to offshore with stations in the northeastern sub-areas about 60 m deeper than southwestern sampling locations. Mean water temperature at 40 m decreased from inshore to offshore. Because water salinity, current, temperature, and depth were all somewhat confounded, it was difficult to ascribe geographic variation in zooplankton and fish to a single physical attribute. Zooplankton.--The five select groups of zooplankton accounted for 90% of the abundance of all zooplankton collected (Table 3). Small copepods far surpassed the abundance of other groups due to large numbers of Acartia spp. and Pseudocalanus spp.. Large copepods were dominated by Calanus marshallae and Metridia spp.. About 60% of the euphausiids collected were unidentified furciliae; among the juveniles and adults identified, Thysannoessa inermis was the dominant species. Larvaceans and thecosomate pteropods were not identified to species but Oikopleura spp. and Limacina helicina are the only member taxa, respectively, found in the Gulf of Alaska (Cooney 1987). Estimates of abundance of small-size animals (small copepods, larvaceans, and pteropods) were not available for nine samples collected during A consistent geographic pattern of abundance was evident only for large copepods and for euphausiids. Statistically, the sub-area effect was significant and did not vary with year or diel period (Table 4). Mean abundance for both groups was relatively high in the northeastern (NEin, NEmid) and Slope sub-areas (Fig. 4). Post-hoc pairwise comparison tests indicated that mean 12

18 abundance of large copepods did not differ among these three sub-areas but it did differ between the northeastern and southwestern sub-areas (p < 0.05). Tests of euphausiid abundance resulted in the same conclusion; in addition, differences involving the Slope and the southwestern subareas (SWin and SWmid) were significant (p < 0.05). Significant year-sub-area interaction was detected for the other zooplankton groups indicating that abundance was not similarly distributed geographically among years (Table 4, Fig. 4). Geographic patterns of euphausiid abundance were largely due to furciliae (Fig. 5, top). Of the juveniles and adults that could be identified, most were T. inermis and T. spinifera. These two species, along with T. raschii, were most abundant in the NEin and NEmid sub-areas (Fig. 5, bottom). Four species, E. pacifica, T. oculatum, T. inspinata, and T. longipes, were most abundant in the Slope Sub-area. Thus, although euphausiid abundance was dominated by furciliae, species-specific differences in geographic distribution (shelf vs. slope) were pronounced. Geographic patterns in the abundance of large copepods were largely due to high abundance of Metridia spp. in the NEin and NEmid sub-areas, and E. bungii over the slope (Fig. 6, top). C. marshallae was the only species evenly distributed over all sub-areas. Similar to E. bungii, C. pacificus and Neocalanus spp. were relatively abundant over the slope (Fig. 6, bottom). Thus, geographic patterns of euphausiid and large copepod taxa indicate some distinction exists between the shelf and slope with regard to prey species potentially available to forage fishes. Forage fishes.--the forage fishes targeted in this study accounted for 26% and 17% of the abundance (no m -2 ) and biomass (mg m -2 ), respectively, of all animals collected in the Stauffer trawl (Table 5). Walleye pollock were collected in greater abundance and accounted for a larger portion of total biomass than capelin or eulachon. The frequency of occurrence of walleye pollock in catches was surpassed only by scyphozoans. Two persistent patterns were evident among the three target fish species. First, most were collected over the shelf with few occurrences over the slope indicating that they were largely confined to coastal (i.e. continental shelf) water. Second, the geographic patterns evident in fish abundance and mean size were similar between day and night although all fish were collected in greater abundance at night. Consistency in distribution was indicated statistically by no significant interaction between the diel and sub-area effects (p>0.05). The subsequent 13

19 presentation of results, therefore, focuses on differences in fish abundance and SL means detected among shelf sub-areas, which usually varied with year. Walleye pollock ranged in size from 38 to 677 mm SL with a multimodal size composition (Fig. 7). A distinct break in size at mm SL separates young-of-the-year (age-0) and older (age-1+) individuals (Brodeur and Wilson 1996); thus, age-0 and age-1+ walleye pollock were analyzed separately. For age-0 walleye pollock, geographic patterns of abundance and size varied with year. The year-sub-area interaction for abundance was significant (Table 6). During 2000, abundance was high in the SWmid Sub-area relative to all other sub-areas (p<0.05) (Fig. 8). In 2001, age-0 pollock abundance was high in the NEin Sub-area relative to the NEmid Sub-area (p=0.0022). In 2003, high catches in the NEin Sub-area were significantly different from all other sub-areas (p<0.05). Similarly, the year-sub-area interaction for mean standard length was significant (Table 7). Mean standard length of age-0 pollock generally declined from northeast to southwest and from inshore to offshore (Fig. 8). In 2000 and 2001, nearly all pairwise differences among subareas were significant (p<0.05). In 2003, fish from the NEin Sub-area were large relative to those captured in all other sub-areas (p<0.05). Thus, age-0 walleye pollock tended to be abundant and large in the NEin Sub-area, where the ACC bifurcates, but this varied with year. Geographic patterns of age-1+ walleye pollock abundance also varied by year (Tables 6). During 2000 and 2001, these fish were abundant in the NEin and NEmid sub-areas relative to the SWin and SWmid sub-areas (p<0.05) (Fig. 8). During 2003, few age-1+ pollock were collected although their abundance in the NEin Sub-area was high relative to the SWin (p=0.0387) and Slope (p=0.0297) sub-areas. An interesting distribution pattern, obscured by the stratification scheme, was that most age-1+ pollock collected in the SWmid Sub-area were caught at stations close to or in the Shumagin sea valley. With regard to mean length, the year-sub-area interaction was marginally not significant (Table 7), but post-hoc pairwise comparisons indicated a difference between the relatively large mean sizes in the SWin Sub-area relative to all other subareas (Fig. 8). The increase in mean length from 2000 to 2003, evident in Figure 8, is noteworthy because it probably reflects the numerical dominance and growth of individuals of the 1999 year class. Thus, age-1+ pollock were relatively abundant in the Shelikof sea valley vicinity but this varied with year. 14

20 Capelin size composition was unimodal with a size range of mm SL (Fig. 7). The distribution of capelin abundance did not significantly vary with year (Table 6). Capelin were most abundant in inshore sub-areas (Fig. 8), but only a marginally significant difference was detected between the SWin and NEmid sub-areas (p=0.0494). Mean size tended to increase from inshore to offshore with some evidence of a northeast-southwest difference (Fig. 5). The yearsub-area interaction was significant (Table 7). During 2000, mean capelin size in the SWin Subarea was low relative to all other sub-areas (p<0.05). During 2003, mean size was highest in the NEmid Sub-area, which was different from the NEin and SWin sub-areas (p<0.05). Thus, the most-evident geographic pattern in capelin abundance and size was among inshore and offshore sub-areas. Eulachon size composition was multi-modal with fish mm SL probably representing several year classes (Fig. 7). On average, eulachon tended to be more abundant in the northeastern sub-areas (Fig. 8), although significant variation among years was detected (Table 6). During all years, eulachon abundance in the NEin Sub-area was high relative to the SWin and SWmid sub-areas (p<0.05). During 2003, eulachon abundance in the NEmid Sub-area was low relative to the NEin Sub-area (p=0473). Eulachon mean size did not differ among sub-areas although the lack of significance was marginal (Table 7) due to relatively small mean size of eulachon in the SWin Sub-area (Fig. 5). Interestingly, this pattern was similar to capelin and opposite that of age-1+ pollock. Thus, eulachon abundance and, to a lesser extent, size tended to be greatest in the Shelikof sea valley vicinity. The rarity of these fishes over the slope, despite an abundance of large prey, raised questions about possible adverse environmental conditions. Current velocity as a potential constraint on fish distribution became apparent when it was related to fish length (Fig. 9). Based on mean length, age-0 pollock, capelin, and eulachon were in sub-areas where the expected mean current velocity at 40-m depth usually corresponded to swim speeds <1 body length s -1 (bl s -1 ) (Fig. 9). Over the slope, expected mean velocity corresponded to swim speeds >1 bl s -1. Age-1+ pollock, in contrast, were larger so that expected current velocities rarely exceeded 0.5 bl s -1. Other potentially important environmental constraints include salinity and bottom depth, which were also relatively high over the slope. The combined abundance of large copepods and euphausiids was consistently high (>1000 no m -2 ) in northeastern sub-areas (NEin and NEmid) but this did not always translate into high 15

21 abundance or large size of fish (Fig. 10). For walleye pollock, age-0 individuals tended to be larger and age-1+ fish more abundant in prey-rich areas; however, least-squares linear relationships were not significant (p>0.05). On the other hand, capelin and eulachon mean size linearly increased with large prey abundance (p<0.05). Eulachon was the only species for which the relationship between large prey and fish abundance was significant (p=0.027). Thus, the relatively swift, south-flowing branch of the ACC was associated with a high abundance of large copepods and euphausiids that, in turn, was positively associated with forage fish abundance and size. 16

22 Tables: Table 1. Number of midwater trawl (A) and Tucker trawl (B) hauls conducted to collect forage fishes and their zooplankton prey in the western Gulf of Alaska during September 2000, 2001, and Samples were collected at day and night in six geographic sub-areas. A) B) Geographic sub-area All subarea Year Diel NEin NEmid SWin SWmid Slope omit 2000 day night day night day night * 1 31 All times Geographic sub-area All subarea Year Diel NEin NEmid SWin SWmid Slope omit 2000 day night day night day night * 1 28 All times *) species-specific ratio estimators used to generate nighttime catch estimates from the three daytime catches. These are not included in the sample tally, but were included as independent observations in all statistical tests. 17

23 Table 2. Means and standard errors (SE) of water temperature and salinity at 40 m, and bottom depth (m) at n locations in six geographic sub-areas sampled in the western Gulf of Alaska during September 2000, 2001, and Temp. ( o C) at 40 m Salinity (psu) at 40 m Bottom depth (m) Year Sub-area n Mean SE Mean SE n Mean SE 2000 NEin NEmid SWin SWmid Slope omit all areas NEin NEmid SWin SWmid Slope omit all areas NEin NEmid SWin SWmid Slope omit all areas

24 Table 3. Taxonomic composition of zooplankton collected in the upper 200 m depth with a 1-m 2 Tucker trawl (333-μm mesh) (n=230) in the western Gulf of Alaska during September 2000, 2001, and Copepodite stages 1-6 are indicated as C1-CVI. Frequency Abundance Taxonomic group Occurrence (%) (no m -2 ) Copepoda, sml Acartia spp., CVI Pseudocalanus spp., CI-CVI Oithona spp., CV-CVI Calanus marshallae, CII Metridia pacifica/lucens, CIV Metriididae, C1-CIII Eucalanus bungii, CI Neocalanus cristatus, CII Metridia spp., CIV Neocalanus spp., CII Unidentified copepoda < 2mm Copepoda, lg Calanus marshallae, CIII-CVI Metridia pacifica/lucens, CV-CVI Eucalanus bungii, CIII-CVI Calanus pacificus, CIV-CVI Neocalanus cristatus, CIII-CVI Neocalanus spp., CIII-CVI Metridia spp., CV-CVI Unidentified copepoda > 2mm Chaetognatha Euphausiacea Thysaneossa inermis Thysanoessa spinifera Euphausia pacifica Thysanoessa raschii Thysanoessa inspinata Thysanoessa longipes Tessarabrachion oculatum euphausiid furciliae Cnidaria Teleost larvae Thecosomata Misc. small zooplankton 1, Siphonophora Natantia Reptantia Larvacea Cirripedia Amphipoda Cladocera Ostracoda Mysidacea Ctenophora All groups combined ) n=221 samples, abundance estimates were unavailable for nine samples from ) adult and juvenile 3) small chaetognaths, amphipods, euphausiid calyptopis, and mysids. 19

25 Table 4. F-statistics and p-values from analysis of variance tests of the effect of year, diel, and sub-area on the abundance of each zooplankton group. Dashes indicate terms that were omitted from the reduced models due to non-significance (p>0.05). The 3-way interaction term, not shown, was not significant in any model (p > 0.05). F statistic (p value) zooplankton group year diel sub-area year-diel year-subarea diel-sub-area Euphausiacea (<0.0001) (<0.0001) (0.0002) Copepoda, lg (0.0072) (0.0064) (0.0031) Copepoda, sml (0.1661) -- (0.4660) -- (0.0002) -- Larvacea (<0.0001) -- (0.0479) -- (<0.0001) -- Thecosomata (<0.0001) -- (0.0001) -- (<0.0001) -- 20

26 Table 5. Taxonomic composition of midwater trawl catches collected in the upper 200-m depth with a Stauffer trawl (3-mm mesh) (n=235) in the western Gulf of Alaska during September 2000, 2001, and Frequency Abundance Biomass Taxonomic group Occurrence (%) (no m -2 ) (mg m -2 ) Scyphozoa Theragra chalcogramma, age Mallotus villosus, post-larval Theragra chalcogramma, age Pleuronectidae Thaleicthys pacificus Stichaeidae Teleostei larvae Salmonidae Scorpaenidae Ammodytidae Gadidae Zaproridae Myctophidae Trichodontidae Bathymasteridae Clupeidae Cottidae Cyclopteridae Liparididae Anoplopomatidae Agonidae Bathylagidae Hexagrammidae Gasterosteidae Macrouridae Zoarcidae Rajidae Lamnidae Squalidae All groups combined

27 Table 6. F-statistics and p-values from analysis of variance tests of the effect of year, diel, and sub-area on the abundance of forage fish groups. Dashes indicate terms that were not included in the reduced models. The 3-way interaction term, not shown, was not significant in any model (p > 0.05). F statistic (p value) forage fish group year diel sub-area year-diel year-subarea diel-sub-area walleye pollock, age-0 (0.0022) (<0.0001) (<0.0001) -- (<.0001) -- walleye pollock, age-1+ (<0.0001) (0.0012) (0.0002) -- (0.0002) -- capelin (<0.0001) (0.0150) eulachon (0.0081) (0.0041) (0.0004) -- (0.0300) -- 22

28 Table 7. F-statistics and p-values from analysis of variance tests of the effect of year and subarea on mean standard length of each group of fish. F statistic (p value) diel year-subarea Fish/prey group year sub-area walleye pollock, age-0 (<0.001) (<0.001) (<0.001) walleye pollock, age-1+ (<0.001) (0.0128) (0.0598) capelin (0.9264) (0.0029) (0.0637) (0.0006) eulachon (0.0970) 23

29 Figures: Figure 1. Sites sampled with a midwater trawl (A) and Tucker trawl (B) to collect forage fishes and their zooplankton prey, respectively, in the western Gulf of Alaska during September 2000, 2001, and Most sites were occupied twice per year (day and night) but these are not differentiated. Thick lines indicate six geographic sub-area, which are labeled in the lower panel. Inset depicts general circulation in the Gulf as indicated by Reed and Schumacher (1986). 24

30 Figure 2. Salinity (A) and SPEM-model estimates of current velocity (B) at 40-m water depth during September 2000, 2001, and Salinity was measured at each site concurrent with zooplankton sampling. Current velocity is the daily average for the month of September. 25

31 Figure 3. Mean daily current velocity (mm s -1 + SE) by direction of flow relative to true north (E, deg.; S, deg.; W, deg.; N, deg.) as estimated by the SPEM model (Stabeno and Hermann 1996) for September 2000, 2001, and The dashed reference line indicates 100 mm s Velocity (mm s E S W N 2001 E S W N NEin NEmid SWin SWmid Slope E S W N Direction 26

32 Figure 4. Back-transformed least-squares mean abundance (+ SE) of five zooplankton groups collected in six geographic sub-areas in the western Gulf of Alaska during September 2000, 2001, and Symbol distinguishes year where the sub-area effect significantly (p<0.05) varied by year. 27

33 Figure 5. Taxonomic composition of euphausiid abundance (top, includes taxa >1% of total abundance) and distribution (bottom, includes furciliae and all identifiable juvenile and adult stages) among five geographic sub-areas in the western Gulf of Alaska during September 2000, 2001, and 2003 (n=222). 28

34 Figure 6. Taxonomic composition of large copepod abundance (top) and the distribution of identifiable taxa (bottom) among five geographic sub-areas in the western Gulf of Alaska during September 2000, 2001, and 2003 (n=222). 29

35 Figure 7. Size composition of walleye pollock, capelin, and eulachon collected in the western Gulf of Alaska during September 2000, 2001, and Vertical bar at 130 mm SL delineates age-0 and age-1+ walleye pollock. 30

36 Figure 8. Back-transformed least-squares mean abundance (+ SE) and abundance-weighted least-squares means of standard length (+ SE) for four fish groups collected in six geographic sub-areas in the western Gulf of Alaska during September 2000, 2001, and Year-specific symbols indicate significant ( p<0.05) interannual variation among sub-areas. 31

37 Figure 9. Standard length of age-0 and age-1+ walleye pollock, capelin, and eulachon versus SPEM-model estimates of current velocity at 40-m during September 2000, 2001, and 2003 in the western Gulf of Alaska. Points indicate averages by year and sub-area (NEin, NEmid, SWin, SWmid, and Slope). Dashed lines depict isolines of fish swim speed (0.5, 1, and 2 body lengths s -1 ). Vertical line at velocity = 160 mm s -1 delineates shelf (<160 mm s -1 ) and slope (>160 mm s - 1 ) sub-areas. 32

38 Figure 10. Mean abundance and standard length of age-0 and age-1+ walleye pollock, capelin, and eulachon versus large zooplankton (large copepods and euphausiids) abundance during September 2000, 2001, and 2003 in the western Gulf of Alaska. Points (n=12) indicate averages by year and shelf sub-area (NEin, NEmid, SWin, and SWmid). Solid lines represent significant (p < 0.05) least-squares linear relationships. 33

39 Chapter 2 Geographic variation in body weight, feeding intensity, and diet of pelagic forage fishes in relation to hydrography [NPRB Final Report -- Please do not cite or use information in this chapter without permission of the author. This material is being prepared for publication] Introduction: Geographic variation in zooplankton abundance presumably affects the productivity of midwater forage fishes. Many of these fishes consume zooplankton. They are often abundant, small, and prevalent in the diets of upper trophic-level piscivores. Hence, forage fish productivity affects ecosystem structure. Within coastal areas, bathymetry and hydrography are relevant to zooplankton abundance but the link between the physical environment and forage fish productivity is not well defined. The Ecosystems and Fisheries-Oceanography Coordinated Investigations Program (eco- FOCI) in NOAA initiated a multi-year project in the western Gulf of Alaska (GOA) during 2000 to examine the link between a large hydrographic feature and forage fish productivity. The study area over the shelf between Shelikof Strait and the Shumagin Islands (see Fig. 1 in Chapter 1) was strategically chosen based on prior oceanographic and biological information. Much of this information relates to research conducted on the early life history stages of walleye pollock Theragra chalcogramma, which is a commercially important gadid that spawns in Shelikof Strait during March (Kendall et al. 1996). The Alaska Coastal Current (ACC), which is the main circulation feature in the Gulf (Stabeno et al. 2004), transports larval walleye pollock southwestward. It bifurcates near Sutwik Island with one branch continuing southwestward along the Alaska Peninsula while the other follows the Shelikof sea valley offshore (Schumacher and Reed 1986). Merati and Brodeur (1996) found that young-of-the-year (age-0) walleye pollock southwest of the sea valley primarily consumed larvaceans while those to the northeast fed mostly on euphausiids. This generally corresponded with variation in growth rates (Bailey et al. 1996) and body size (Wilson 2000) implying that exportation of prey from the shelf in the offshore limb of the ACC led to impoverished feeding conditions over the adjacent southwestern shelf. In the first part of this project (Chapter 1), high abundance of large copepods (>2 mm prosome length, PL) and euphausiids was associated with the sea valley vicinity as compared to the adjacent southwestern shelf. These zooplankton are important prey of juvenile walleye 34

40 pollock (Brodeur and Wilson 1996), eulachon Thaleichthys pacificus (Sturdevant 1999), and capelin Mallotus villosus (Wilson et al. in press). These three groups constituted the major forage fishes collected within the study area. Juvenile walleye pollock and eulachon exhibited geographic patterns of distribution similar to these prey groups, but capelin did not. Thus, the second study was designed with the following three objectives in mind. First, compare body condition and diet among species to provide insight on how these fishes might differentially respond to enrichment of specific prey types. Second, determine if forage fish diets reflect the relative abundance of euphausiids and large copepods in the Shelikof sea valley vicinity. Third, determine if the relative abundance of these prey in the zooplankton translates into high stomach fullness and body weight. The general analytical approach used to address the first objective was to compare syntopic collections of fishes, the second and third objectives were addressed by examining species-specific differences among relevant geographic sub-areas. Methods: Sampling occurred day and night during September 2000, 2001, and 2003 over a 48- station grid.the grid was stratified into five geographic sub-areas. This study used only samples from the four shelf sub-areas (for field sampling details and sub-area definition, see Chapter 1). Juvenile walleye pollock, capelin, and eulachon were randomly selected from small-mesh trawl catches. Age-0 and older walleye pollock were preserved separately. Age-0 juveniles ( mm SL) were distinctly smaller than older individuals mm SL) (Brodeur and Wilson 1996). These older walleye pollock are hereafter referred to as age-1+. Fish were frozen (- 80 o C then transferred to -20 o C) within one hour of landing for subsequent examination in the laboratory. In the laboratory, fish were thawed in seawater. Individuals were selected by size from each sample so that the different sizes available were reasonably well represented. Up to 20 age-0 walleye pollock were selected. Due to logistic constraints on the number of stomachs that could be examined, only 10 individuals were selected from each of the remaining groups: capelin, eulachon and age-1+ walleye pollock. Each fish was examined to determine body length, weight, gut content weight, and the number, weight, and state of digestion of all prey. Standard length and whole wet body weight of each thawed fish was used to estimate species-specific length-weight relationships. Each fish was measured to the nearest 1 mm SL, blotted dry, and weighed to the nearest 1 mg. Stomach content weights (see below) were 35

41 subtracted to give somatic body weight. Finally, lengths and weights were converted to freshfish equivalents (Buchheister and Wilson 2005) assuming that the effect of freezing on capelin standard length was equivalent to its effect on fork length. Stomachs were excised between the esophagus and pylorus and stored individually in a sodium borate-buffered 10% formalin solution. None of the stomachs were flaccid so regurgitation was not believed to be a problem. Stomach contents were dissected from the preserved samples, blotted dry and weighed to the nearest 0.01 mg. Mean stomach content weight, or fullness, was the back-transformed mean of 4 th -root transformed weights. Stomach contents were identified to broad taxonomic categories and quantified following Brodeur et al. (2000). Each taxonomic category was comprised of subcategories to indicate digestion and fragment size: 1 = <50% intact and well-digested, 2 = >50% intact but welldigested, 3 = largely intact but digested, 4 = intact and fresh. Most prey showed signs of digestion (i.e. 1-3) so net feeding was probably negligible. Taxonomic composition was quantified by prey count and prey weight. Numerical composition of species diets was based on percentages of total prey count. This study emphasizes numerical abundance of prey because the previously observed enrichment of zooplankton in the sea valley was based on abundance not biomass. Prey size, however, was indicated by mean weight of largely intact (i.e. fragment categories 3 and 4) individuals in each taxonomic category. Following Cochran (1977), mean weight was the ratio of the sum of all largely intact individuals divided by the sum of their weights. From these examinations, four dependent variables were used in subsequent statistical examinations: somatic body weight, stomach fullness, number of euphausiids, and number of large copepods. Inter-species comparison. Walleye pollock, capelin, and eulachon were compared using data from 1745 fish representing 47 trawl hauls. This subset of hauls was selected because body size (length and weight), stomach fullness, euphausiid counts, and large copepod counts were available for all three species in each haul. Prior to statistical analysis, standard length and somatic weight data were ln-transformed to create normally distributed variables that were linearly related with homogeneous variance. Stomach fullness and prey counts were 4 th -root transformed, then averaged across hauls by 10-mm length intervals to create normally distributed populations of means. Means of <5 fish were omitted from analysis. Differences among species were statistically examined using analysis of covariance (ANCOVA) as 36

42 implemented by the General Linear Model module of Systat for Windows (ver. 11). Species was included as a main effect, and SL was the independent variable. The regression slope was not allowed to vary among species for stomach fullness and prey count data due to the narrow range of capelin lengths relative to variation within the dependent variables. Geographic variation. Variables were analyzed with analysis of variance (ANOVA) tests designed to detect significant differences among geographic sub-areas, which delineate main hydrographic features within the study area (Chapter 1). One test was conducted on each data set (body weight, stomach fullness, euphausiid count and large copepod count) from each fish species (age-0 and age-1+ walleye pollock were treated separately, see Results). Thus, 16 separate tests were conducted. The ANCOVA model was based on a fully crossed, blocked, and nested design that included year and diel effects: Eqn 1) Y ijklm = + X ijklm + Year j + Diel k + Sub-area l + Year j *Diel k + Year j *Sub-area l + Diel k *Sub-area l + Year j *Diel k *Sub-area l + Station m {Sub-area l } + Sample i {Station m } + e i where Y ijklm is the dependent variable of the i th sample of fish from the j th year, k th diel period, l th sub-area, m th grid station, and e i is the random error component. The covariate, X, is SL. Year, diel, and sub-area were fixed effects. Sample was nested within station, and station was nested within sub-area; both nested terms were random effects. Observations on individual fish were considered replicates within sample. Elimination of terms from the model, or model reduction, was accomplished following Milliken and Johnson (1996) with alpha=0.05. Non-significant (p > 0.05) terms were sequentially eliminated beginning with the 3-way interaction, then 2-way interactions, and finally any main effects not included in any retained interaction term(s). Only one term was dropped between successive model runs. All tests were conducted using SAS System for Windows (ver. 8.02), Procedure Mixed. Sample coverage by year, diel period, and sub-area was complete or nearly complete for age- 0 walleye pollock and eulachon, respectively. There were more empty statistical cells for capelin and age-1+ pollock, which required elimination of effect levels to maintain a fully crossed ANOVA design. For capelin, the NEmid level of the sub-area effect (n=8 samples, 30 fish) was not included; for age-1+ walleye pollock, the SWin sub-area (n=3 samples, 28 fish) 37

43 and the 2003 year (n=5 samples, 30 fish) levels were not included. Within each data set, observations were weighted by fish standardized abundance estimates (Chapter 1). For each species and sample, length-specific standardized abundance estimates (fish m -2 cm -1 ) were apportioned among individuals examined in the laboratory that were of the same cm-size interval. Data transformations were applied prior to statistical analysis. Relatively few, highmagnitude stomach fullness and prey counts resulted in right-skewed distributions. For stomach fullness, this was largely remedied using a power transformation (4 th -root). For large copepod and euphausiid counts, it was necessary to average the data by sample and length interval prior to applying the 4 th -root transformation. Weighted means of prey counts were calculated by 10- mm interval (e.g , 55-64, and so on) for capelin and age-0 walleye pollock, and by 25-mm interval for eulachon and older walleye pollock ( , , and so on), which spanned a relatively large length range. Weights were the apportioned, standardized fish abundances described in the preceding paragraph. Length-interval midpoints (e.g. 40, 50, 125, 150) were used as the covariate in ANOVA tests of count data. For somatic body weight, the dependent variable and the covariate were log e -transformed to normalize the residual error and linearize the relationship between body weight and length. Visual examination of the residuals and normal probability plots confirmed that the transformed variables conformed to the underlying assumptions of normality and homogeneous variance. Results: Body size and stomach fullness were collected from 4263 fish representing 376 samples (Table 1). The largest range in body size was exhibited by walleye pollock (Fig. 1). Age-0 pollock were mm SL, and age-1+ fish were mm SL. Capelin and eulachon were mm SL and mm SL, respectively. Stomach fullness increased with size for all species (Fig. 2). Maximum mean stomach content weight of pollock was 0.9 g fish -1 for mm SL fish. Maximum mean stomach content weights of capelin and eulachon were similar, 0.1 g fish -1 for mm and mm SL fish, respectively. The target prey items, euphausiids and large copepods, occurred in 3036 (71%) of the stomachs. Remaining stomachs were either empty (n=406) or contained alternate or unidentifiable prey (n=821). Euphausiids occurred in >43% of the stomachs from each species group whereas large copepods occurred in >49% of walleye pollock stomachs but were 38

44 relatively rare in capelin (26%) and eulachon (4%) stomachs. Numerically, these prey comprised <50% of most capelin and age-0 walleye pollock diets but >50% of eulachon and age-1+ walleye pollock diets (Fig. 2). The numerical importance of euphausiids in particular increased with fish length (Fig 2). Euphausiids were large relative to large copepods and most other prey types (Fig. 3). Size-related dietary transitions were especially pronounced for capelin and walleye pollock (Fig 2 and 3). Because age-0 and age-1+ walleye pollock were also distinguishable by body length, separate statistical tests were used to examine geographic variation among each of these age/size groups. Inter-species comparison. Species-specific differences were evident for all variables examined. For somatic body weight, interpretation of the data was complicated by significant variation in the slope of the length-weight relationship (ANCOVA, F= , p<0.001); however, among fish <120 mm SL, walleye pollock tended to be heavier than capelin and eulachon (Fig. 4). Stomach fullness and euphausiid counts both increased with fish length (F= , p<0.001; F=95.068, p<0.001, respectively) whereas large copepod counts decreased with length (F=4.866, p=0.035) (Fig. 4). Stomach fullness, number of euphausiids, and number of large copepods were all highest for walleye pollock, intermediate for capelin, and lowest for eulachon (Fig. 4); all pairwise differences were significant (p<0.05). Geographic variation. For each fish group, stomach fullness did not vary significantly (p>0.05) with geographic sub-area. Thus, the following ANOVA results pertain only to the body weight and prey-count variables. Results are organized by fish group. For age-0 walleye pollock, significant sub-area effects were detected on somatic body weight and prey counts (Table 2). Somatic body weights in the NEmid (p=0.0079) and SWin (p=0.0062) sub-areas were significantly different from those in the SWmid Sub-area, which were relatively low (Fig. 5). The numbers of euphausiids and large copepods recovered from age-0 pollock stomachs were also affected by sub-area; however, this varied by year (Table 2). During 2000 and 2003, the NEmid Sub-area, which encompasses the Shelikof sea valley, was often associated with high numbers of euphausiids and large copepods (Fig. 5). In 2000, prey counts associated with this Sub-area were high relative to the SWmid (p<0.05) Sub-area. In addition, comparatively few large copepods were recovered from the SWin (p=0.0088) fish. Similarly, during 2003, high counts of large copepods in fish from the NEmid Sub-area differed from counts in fish from the NEin (p=0.0409) Sub-area, which also differed from the SWin 39

45 (p=0.0403) Sub-area. In contrast, during 2001, large copepods were relatively abundant in fish from the SWin Sub-area as compared to those from the NEin (p=0.0473) Sub-area. Thus, with a possible exception in 2001, age-0 walleye pollock somatic body weights and consumption of euphausiids and large copepods were generally consistent with an enriched prey field in the Shelikof sea valley vicinity. For age-1+ walleye pollock, sub-area was significantly associated with somatic body weight and prey counts. The sub-area effect on body weight varied by year (Table 3). During 2000, significant differences were detected between the relatively low weight of fish in the SWmid Sub-area and those in the NEin (p=0.0285) and NEmid (p=0.0089) sub-areas (Fig. 5). During 2001, fish in the NEin Sub-area were heavier, and significantly different from, those in the NEmid (p=0.0160) and SWmid (p=0.0268) sub-areas. With regard to prey counts (Table 3), fish collected from the NEmid Sub-area had consumed more euphausiids than had those from either the NEin (p=0.0137) or SWmid (p=0.0055) sub-areas (Fig. 5). The count of large copepods in age-1+ walleye pollock varied between day and night samples (Table 3). Among nighttime collections, large copepod counts in fish from the NEmid Sub-area were high compared to the NEin (p=0.0008) and SWmid (p=0.0043) sub-areas (Fig. 5). Thus, somatic body weight and prey count data suggest that age-1+ walleye pollock also benefited from the relatively high abundance of euphausiids and large copepods in the Shelikof sea valley vicinity. For capelin, sub-area was significantly associated with the number of euphausiids recovered from stomachs. This effect varied with diel period (Table 4). Among nighttime collections, euphausiid counts in capelin from the NEin (p<0.0001) and SWmid (p<0.0001) sub-areas were high relative to the SWin Sub-area (Fig. 5). No sub-area effect on body weight or large copepod counts was detected. Thus, evidence that capelin benefit from enriched prey abundance in the Shelikof sea valley vicinity was weak. For eulachon, sub-area was significantly associated with somatic body weight and prey counts (Table 5). Body weights of eulachon in the SWin Sub-area were low and significantly different from all other sub-areas: NEin, p=0.0047; NEmid, p=0.0012; SWmid, (p=0.0021) (Fig. 5). The sub-area effect on prey counts varied between day and night samples. Among daytime samples, the number of euphausiids consumed by eulachon in the NEin Sub-area was high relative to fish in the SWmid Sub-area (p=0.0424) (Fig. 5). No sub-area differences were detected among samples collected at night. Robustness of the test on large copepods was 40

46 weakened by few non-zero counts. Thus, there was weak evidence of increased euphausiid consumption by eulachon in proximity to the Shelikof sea valley. 41

47 Tables: Table 1. Number of individuals examined to determine the diets of juvenile walleye pollock (age-0, age-1+), capelin, and eulachon collected at day and night in four geographic sub-areas of the western Gulf of Alaska during September 2000, 2001, and Dashes indicate no samples available. year and diel Species Sub-area day night day night day night walleye pollock, age-0 walleye pollock, age-1+ All combined NEin NEmid SWin SWmid all subareas NEin NEmid SWin SWmid all subareas capelin NEin NEmid SWin SWmid all subareas NEin eulachon NEmid SWin SWmid all subareas All combined

48 Table 2. ANOVA test results of year, diel, and sub-area effects on somatic body weight and the number of euphausiids and large copepods in the stomachs of age-0 walleye pollock in the western Gulf of Alaska, 2000, 2001, and Standard length, SL, was included as a covariate. dependent effect / degrees of freedom variable covariate numerator denominator F p somatic body weight Year <.0001 Sub-area SL <.0001 no. euphausiids Year Sub-area Year*Sub-area SL <.0001 no. large copepods Year Diel Sub-area Year*Diel Year*Sub-area SL

49 Table 3. ANOVA test results of year, diel, and sub-area effects on somatic body weight and the number of euphausiids and large copepods in the stomachs of age-1+ walleye pollock in the western Gulf of Alaska, 2000, 2001, and Standard length, SL, was included as a covariate. dependent effect / degrees of freedom variable covariate numerator denominator F p somatic body weight Year Sub-area Year-Sub-area SL <.0001 no. euphausiids Year <.0001 Sub-area SL no. large copepods Year Diel Sub-area Diel*Sub-area

50 Table 4. ANOVA test results of year, diel, and sub-area effects on somatic body weight and the number of euphausiids and large copepods in the stomachs of capelin in the western Gulf of Alaska, 2000, 2001, and Standard length, SL, was included as a covariate. dependent effect / degrees of freedom variable covariate numerator denominator F p somatic body weight Year <.0001 Diel SL <.0001 no. euphausiids Diel Sub-area Diel*Sub-area SL <.0001 no. large copepods SL

51 Table 5. ANOVA test results of year, diel, and sub-area effects on somatic body weight and the number of euphausiids and large copepods in the stomachs of eulachon in the western Gulf of Alaska, 2000, 2001, and Standard length, SL, was included as a covariate. dependent effect / degrees of freedom variable covariate numerator denominator F p somatic body weight Year Sub-area SL <.0001 no. euphausiids Year Diel Sub-area Year*Diel Diel*Sub-area no. large copepods Year Diel Sub-area Year*Diel Diel*Sub-area SL <

52 Figures: Figure 1. Fresh somatic body weight versus fresh standard length for capelin, eulachon, and juvenile walleye pollock collected in the western Gulf of Alaska during September 2000, 2001, and

53 Figure 2. Length-specific taxonomic composition of the diets of walleye pollock, capelin, and eulachon as determined by counts of individual animals in predator stomachs (unweighted by CPUE). Fish were collected during the day and night in the western Gulf of Alaska during September 2000, 2001, and Panels are aligned horizontally according to fresh fish length. 48

54 Figure 3. Weight (g) of largely intact individual prey consumed by capelin, eulachon, and juvenile walleye pollock collected at day and night in the western Gulf of Alaska during September 2000, 2001, and PL = prosome length Individual prey wt (g) walleye pollock, age-0 walleye pollock, age-1+ capelin eulachon Larvaceans Calanoid copepods, <2mm PL Pteropods Calanoid copepods, >2mm PL Amphipods Crabs Chaetognath Prey Mysids Euphausiids Shrimps Fish 49

55 Figure 4. Length-specific comparison of syntopically collected walleye pollock, capelin, and eulachon with regard to body size, stomach content weight, and prey counts. Prey counts indicate the number of euphausiids and large copepods that were recovered from fish stomachs. Lines represent least-squares regression equations. Symbols represent means based on >5 fish within each 10-mm length interval. 50

56 Figure 5. Means (+ SE) of fish body weight and prey counts from stomach content analysis of juvenile walleye pollock (age-0, and age-1+), capelin, and eulachon that were collected in four geographic sub-areas sampled day and night in the western Gulf of Alaska, September 2000, 2001, and Target prey items were euphausiids and large copepods. Symbols used in each panel portray ANOVA test results relevant to the geographic sub-area effect. For example, geographic variation in large copepod counts in age-0 pollock stomachs varied by year; hence, symbols distinguish year-sub-area statistical cells. Grey symbols indicate combined data. 51

57 Chapter 3 Bioenergetics model-based estimates of the growth of two forage fish species Section 1. Geographic variation in growth of age-0 walleye pollock Theragra chalcogramma [NPRB Final Report -- Please do not cite or use information in this section without permission of the author. Much of this section has been submitted for publication: Mazur et al. accepted pending revision] Introduction: Growth is an important component of fish life histories and the rate and timing of growth, especially in larval and juvenile phases, can have implications to the survival and eventual recruitment of individuals into adult populations (Sogard 1997). Growth, through its control of body size, can alter predator-prey dynamics. Increases in body size for larval and juvenile fishes are often associated with a reduction in vulnerability to piscivores (Sogard 1997) and a concurrent increase in the size range of available prey (Chapter 2). If body size and growth rate are important characteristics of the survival and recruitment of immature fish, then repeated measures of the quality, quantity, and spatial extent of available growth habitat could contribute to understanding inter-annual variation in recruitment. For walleye pollock Theragra chalcogramma in the western Gulf of Alaska (GOA), recruitment control of a cohort appears to have shifted from the larval to the juvenile life stage (Bailey 2000). Broad scale patterns in recent pollock recruitment seem to be largely driven by finer scale biological processes (Bailey et al. 2005) such as juvenile pollock and piscivore spatial overlap and interactions (Ciannelli et al. 2004a). Therefore, accounting for the spatial and inter-annual variation in juvenile pollock growth and body size across the fall rearing habitat could provide insight to pollock recruitment variability. Previous investigation of the spatial distribution of juvenile pollock growing conditions during the fall in the western GOA suggested that prey quality had a stronger influence on pollock growth than the available range of temperatures observed in the rearing environment (Mazur et al. in press). Further, the proportion of energy rich euphausiids in the diet of juvenile pollock explained most of the observed variation in 52

58 growth rates among geographic locations. Winter survival of juvenile pollock may be greatly influenced by the availability of euphausiid prey in the fall. Estimates of age-0 pollock cohort consumption of euphausiids during the fall of 1990 indicated that only about 50% of the euphausiid biomass was consumed over a 34-day simulation period within the rearing habitat (Ciannelli et al. 1998). Localized prey depletion, however, was not addressed. Though juvenile pollock consumption demand, in general, appears to not exceed the supply of euphausiids within the entire rearing area, the potential for localized shortages of euphausiids could lead to growth suppression. Bioenergetics models are powerful tools for isolating and exploring the nonlinear effects of temperature, prey quality, prey quantity, and predator size on the growth of fish (Madenjian et al. 2004). A bioenergetics model for juvenile walleye pollock (hereafter referred to as the pollock model) was developed by Ciannelli et al. (1998) and has been extensively used to investigate the foraging impact of pollock on their prey (e.g. Sturdevant et al. 2001, Duffy-Anderson et al. 2002, Ciannelli et al. 2004b). Attempts to corroborate the pollock model using field data and alternate models have produced mixed results, with both comparable (Ciannelli et al. 1998) and dissimilar (Ciannelli et al. 2004b) consumption estimates among models. However, both of these corroboration attempts consisted of single point estimates of consumption from a restricted size range of fish. In this study we estimated how the growth of age-0 walleye pollock varied spatially within an important nursery area (Hinckley et al. 1991, Wilson 2000) in the western Gulf of Alaska during September of 2000, 2001, and We used a bioenergetics model to isolate environmental influences on growth to investigate which components exerted the most influence on the observed variability in juvenile pollock growth across this area. Because our approach relies on the performance of the pollock model, we also wanted to corroborate the model in the field using independent estimates of consumption across a broad size range of juvenile fish. Measurements of the energy content of prey groups found in the juvenile pollock diet were also used to investigate the contribution of each prey group to the size-specific growth of juvenile pollock. We then used the bioenergetics model to evaluate if higher growth rates were associated with the downstream portion of the Shelikof sea valley due to prey field enhancement with an influx of Alaska Coastal 53

59 Current water. Further, we investigated whether site-specific growing conditions were relatively consistent from year to year or if the growing conditions varied spatially within the rearing area inter-annually. Because euphausiids are an important prey resource for juvenile pollock in the fall we examined the potential for localized euphausiid depletion to occur at a finer spatial scale as a result of juvenile pollock consumption. Methods: Study samples were collected September 2000, 2001, 2003 in the western GOA from southern Shelikof Strait to the northeastern Shumagin Islands (Fig. 1). This area is thought to be an important rearing area for juvenile walleye pollock that recruit to the Shelikof spawning population (Wilson et al., 1996). The rearing area is primarily contained over the shelf of the Alaska Peninsula with water depth generally less than 200 m. The Alaska Costal Current (ACC) is an important feature in this area. The ACC follows the Shelikof sea valley, joining the Alaska Stream in the southeastern edge of the sampling grid (Schumacher and Reed 1986). Some water from the ACC flows across the shelf and into the pollock rearing area (Schumacher and Reed 1986). Variable fresh water input contributes to a large range of temperatures ( ºC) and salinities (30 33 PSU) (Lanksbury et al. 2005). Sample collection.--day and night paired tows for forage fish, plankton, and oceanographic conditions were conducted at 48 stations along a grid across the rearing area from the NOAA Ship Miller Freeman. Juvenile walleye pollock were captured in the water column using an Anchovy trawl equipped with 568-kg, steel-v doors, and a 3- mm mesh codend liner (Chapter 1). The trawl was fished from 200 m depth to the surface or 10 m off the bottom, if the bottom was shallower than 200 m, with a warp retrieval of 10 m min -1 and ship speed of knots. Plankton were sampled using a Tucker trawl with a 1 m 2 net opening and two 333 μm mesh sequential closing nets. The first net was opened at 200 m or 10 m off of the bottom and fished at a 45º-wire angle during retrieval (20 m min -1 ) until the bottom of the thermocline was reached. The second net was then fished from the bottom of the thermocline to the surface. Measures of net depth, temperature, and salinity were collected using a Sea-Bird SeaCat profiler attached to the lead wire of the Tucker trawl. 54

60 Juvenile walleye pollock captured in the Anchovy trawl were weighed and enumerated on deck with up to 50 individuals randomly selected for preservation. Age-0 pollock were differentiated from older pollock using the previously identified body length threshold of <130-mm fork length (L F ) (Brodeur and Wilson 1996). The sub-samples of pollock were either flash frozen at 80 ºC or preserved in 10% buffered formalin for gut content analysis in the NOAA AFSC in Seattle WA (Buchheister and Wilson 2005). Zooplankton collected in the Tucker trawl was preserved in 5% buffered formalin and sent to the Poland Plankton Sorting and Identification Center (Szezcin, Poland) for identification and enumeration. Prey energy density estimates.--for estimates of prey energy densities, zooplankton and ichthyoplankton samples were collected from the study area in September, 2003 from the NOAA ship Miller Freeman. Samples were collected using the Tucker trawl using the same procedures outlined above. Prey groups commonly found in the diets of juvenile walleye pollock from 2000 were sorted from the samples into broad taxonomic groupings, including gammariid and hyperiid amphipods, chaetognaths, copepods, crab megalopae, euphausiids, fish larvae, larvaceans, and pteropods. Sorted samples were frozen ( 80 o C) in vials with seawater and thawed approximately 5 months later in the laboratory. Additional sorting of samples using dissecting microscopes was used to verify that other prey groups had not contaminated the samples. Due to the prevalence of two major species of euphausiids in the diets of juvenile pollock, euphausiid samples were separated by species, resulting in four groups: Thysanoessa inermis (Hansen), T. spinifera (Holmes), T. raschii (Hansen), and Euphausia pacifica (Hansen). Euphausiids were further differentiated by size, either greater than or less than 15-mm total length. Copepods found in the diets of the fish from this study were not identified to species but rather distinguished by size, therefore copepod samples were also divided by size groups using 2-mm, 850-μm, and 48-μm sieves. The resultant copepod groups were roughly mm (small), mm (medium), and mm (large) in prosome length. Most fish larvae were Mallotus villosus (Müller). All samples were gently blotted, weighed to the nearest 0 01mg, and dried at 65 o C until reaching a constant mass. Once dry, each prey group was ground to a homogenous mixture using a mortar and pestle, and energy densities were measured using a Parr 1425 Semi-micro bomb calorimeter. When 55

61 enough dry material was available, two replicates of a sample weighing between 0 04 and 0 16 g dry were combusted. Differences between the two replicates were minimal (CV 1%). Prey energy density values were converted to kj gram -1 wet mass for use in the bioenergetics model using the ratio of wet to dry weight for each of the samples. Diet analysis.--preserved juvenile pollock were thawed in seawater or removed from formalin and rinsed in seawater for analysis in the lab. Individual fish were blotted dry, measured to the nearest mm standard length (L S ), and weighed (0 001 g). Individual fish lengths and masses were corrected for shrinkage due to preservation (Buchheister and Wilson 2005) prior to use in any analysis. Stomachs were dissected, and prey items removed. Prey items were identified to the lowest possible taxonomic level under a dissecting microscope, sorted, counted, blotted, and weighed (wet mass). Otoliths were removed from up to 20 thawed fish selected to represent the available size range, cleaned and stored in 95% ethanol for later analysis. Diet data was analyzed by grouping prey into broad taxonomic groups (e.g. Ciannelli et al. 2004b) and calculating the mean proportional diet content, by weight, of each prey group for size-specific fish diets. Because body size strongly influences fish consumption and to parallel the temporal and spatial extent of gastric evacuation estimates, pollock diet estimates from 2000 were grouped into 5-mm size bins and pooled from across the sampling grid to ensure adequate sample sizes and a frequent dielsampling interval for all size classes for the comparison between bioenergetics and gastric evacuation models. To evaluate if higher growth rates were associated with the downstream portion of the Shelikof sea valley diets for juvenile pollock used to estimate station-specific diets and growth estimates were grouped into three size classes (35-59, 60-85, and > 85 mm L S ) to ensure adequate sample sizes at each independent station for each year. Otolith growth.--otoliths for age and growth analysis used in the model comparison were selected from the size range of available age-0 pollock captured during 2000 ( mm L S ) and were stratified across the sample grid (Fig. 1). Otoliths from 139 juvenile pollock were processed following the methods of Brown and Bailey (1992). Measures of the ventral radius and outer growth increments for the last 5 days were 56

62 identified and recorded. Lengths were back-calculated using an equation relating the ventral radius of the otolith to fish length (r 2 = 0 88): Eqn 1) L S = O VR where L S is the standard fish length and O VR is the ventral radius of the otolith. Separate length-mass regressions were derived for the NE, SE, NW, and SW sections of the sample grid and were used to convert back-calculated length estimates into mass: Eqn 2) W = a L S b where W is the estimated wet mass, L S the preservative-corrected standard length, with area-specific intercepts a and slopes b (Table I). Daily growth, in mass, was estimated by subtracting the back-calculated mass 5-day before capture from the estimated capture mass and dividing by 5 days, producing an individual daily average for the 5-day period prior to capture. Individual estimates were then averaged to produce daily growth estimates for each sample station and 5-mm size bin. Bioenergetics and gastric evacuation models.--the Wisconsin bioenergetics model, version 3 0 (Hanson et al. 1997) was used to estimate both the daily growth of juvenile walleye pollock and the size-specific consumption for model corroboration. This model, parameterized for juvenile walleye pollock (Ciannelli et al. 1998), uses a balanced energy budget where energy available for growth equals the total energy consumed minus the energy lost to waste, activity, and respiration (Hanson et al. 1997). The bioenergetics model operates on a daily time step and incorporates allometric weight-dependent and temperature-dependent functions for maximum consumption and metabolism (Hanson et al. 1997). Ciannelli et al. (1998) conducted a sensitivity analysis of the model; however, our results do not integrate parameter estimate uncertainty. To corroborate the pollock model in the field we compared estimates from September 2000 of consumption from the pollock model with independent estimates from an in situ gastric evacuation model, similar to the methods used by (Arrhenius and Hansson 1994) for young-of-the-year (YOY) herring. Because consumption is dependent on fish size, 57

63 estimates from both models were generated for 5-mm size classes of pollock from across the sampling grid. Size-specific bioenergetics model inputs were averaged from across the grid for thermal experience (8 6º C), proportional diet content (Table 2), growth (Table 3), and prey energy content (Table 4) to parallel the temporal and spatial extent of gastric evacuation estimates. Juvenile pollock whole body energy contents (WBEC) were measured previously (Buchheister and Wilson in press) and the mean WBEC for each 10-mm size class was applied as the predator energy density within the pollock model (Table 3). The prey energy density for the euphausiid prey group was modeled using one energy value estimated by multiplying the proportion of each identified euphausiid species in the diet of juvenile pollock by the species-specific wet WBEC and summing. The diets of pollock contained 900 euphausiid, 488 were unidentifiable and another 171 were only broadly identifiable as being Thysanoessa. Of the 241 remaining identifiable euphausiid, 2% (5) were T. rashi, 12% (29) E. pacifica, 42% (102) T. spinifera, and 44% (105) T. inermis. The size structure of consumed copepods was only roughly estimated from the diets of juvenile pollock, so we modeled the energy content of small copepods by taking the mean of the mm (small) and mm (medium) energy estimates. Copepods larger than 2 mm were modeled using the energy estimate obtained from the mm (large) size group. For the gastric evacuation estimates used to estimate sub-area ration sizes fish diets were pooled across all stations contained within the six sub-areas previously defined in Chapter 1 to ensure adequate sample sizes for all size classes, years, and a frequent dielsampling interval necessary for gastric evacuation estimates. Differences in the stomach fullness of the annual size-specific diet and diel variation were negligible across each year within each sub-area. Therefore, gastric evacuation estimates of consumption were modeled using a Bajkov equation (Ney 1990) modified by Eggers (1979) to fit an exponential evacuation rate without applying a correction for differences in diet at the beginning and end of model estimates (Boisclair and Marchand 1993, Richter et al. 2004): Eqn 3) C D = 24 S median R 58

64 Daily consumption C D (g day -1 ) was estimated for each size class from S median, the median amount of food in the stomach over all diel sample periods and R, the gastric evacuation rate assuming a simple exponential evacuation rate (Eggers 1979). We applied the evacuation rate of 0 25 day -1 estimated by Merati and Brodeur (1996) using the MAXIMS program (Jarre-Teichmann et al. 1993) generated from samples collected during September of 1990 from the same juvenile walleye pollock nursery area. To reduce the influence of several very large individual stomach contents and to maintain comparability with previous studies conducted on juvenile pollock, we used median stomach content rather than the geometric mean (Cochran and Adelman 1982, Merati and Brodeur 1996). To compare model (gastric evacuation vs. bioenergetics) estimates of daily consumption during 2000, we used a concordance correlation analysis (Zar, 1996) in conjunction with a partitioned mean squared error (MSE) analysis (Rice and Cochran 1984, Yang and Arritt 2002). The concordance correlation analysis is most commonly used to assess the reproducibility of measurements made using two independent techniques. The partitioned-mse analysis technique was used to identify the degree and source of error associated with the deviation between predicted values. Eqn 4) MSE = 1 n ( Pi Ai ) n i= 1 2 = 2 ( P A) + ( A ) 2 S P S + 2(1 r)s S P A Where P i and A i are the series of predicted and actual values with mean values P and A, standard deviations S P and S A, and correlation coefficient r. The MSE analysis partitions the sources of error between the mean component (M C ), slope component (S C ) and the random error component (R C ) based on the proportional contribution of each to the observed deviations in the relationship between model estimates. Eqn 5) 1 = M C + S C + R C = ( P A) MSE 2 + ( S S MSE 2(1 r S PS + MSE 2 P A ) ) A Spatial growth predictions.--daily growth for juvenile walleye pollock was estimated using the pollock model at every sample station in the western Gulf of Alaska where fish of that size were captured and diet information was available. Bioenergetics model inputs of proportional diet content, thermal experience, and ration size were estimated for each sample station in the grid. Mean ration sizes (p-values) for each sub-area were 59

65 calculated for each year as the proportional difference between the gastric evacuation estimate of consumption and the bioenergetically estimated maximum consumption level. Ration sizes were fixed within each sub-area and year to estimate growth across the sample grid. By fixing the ration size, any unknown errors associated with the bioenergetics model in estimating growth induced by altering the ration size would be fixed at a consistent level within each sub-area (Bajer et al. 2004). A multiple linear regression was used to investigate the relative contribution of temperature, prey quantity, and prey quality to the estimated growth of juvenile pollock. Euphausiid depletion.--site-specific estimates of the age-0 pollock consumptive demand for euphausiid prey in relation to the amount of available standing stock biomass was estimated using consumption estimates from the bioenergetics model runs. The consumption to biomass of euphausiids at each site was estimated for three size classes of age-0 pollock (35-59, 60-85, and > 85 mm SL) using the mean density of each size class in midwater trawl catches at each site (Chapter 1) combined with their size-specific bioenergetics consumption estimate. Standing stock biomass of euphausiids was estimated using only the nocturnal tows of a 1-m Tucker trawl (0.333-mm mesh; Chapter 1) when euphausiid capture efficiency was thought to be highest. Results: Model corroboration.--size-specific consumption estimates for juvenile walleye pollock using the bioenergetics model and the gastric evacuation rate model during 2000 were correlated (r c = 0 945, transformed 95% confidence limits L 1 = 0 834, L 2 = 0 982; Fig. 2) and highly significant (r 2 = 0 98, P < 0 001). Both estimates of daily consumption (g day -1 ) consistently fell below the theoretical level of maximum consumption (C-max) for pollock (Fig. 3) with no apparent trend associated with body size. The partitioned MSE analysis indicated that the slight differences between consumption estimates were primarily associated with the random R C (0 49) and mean M C (0 45) components, while the slope component S C (0 06) had little influence. The M C error identified the systematic difference between model estimates where the pollock model consistently estimated a larger amount of consumption than the gastric evacuation model. The lack of error associated with the S C illustrated that the difference between estimates was consistent 60

66 across the size range of pollock in this analysis. Similarly, the strong correlation of the consumption estimates indicates that the systematic error was small and would not produce significantly different estimates from those expected from a null model (Rice and Cochran, 1984). Pollock growth.--growth estimates generated from otoliths for all size classes of juvenile pollock for September 2000 in the sample grid ranged between 0 03 to 0 19 g day -1 with a mean of 0 06 g day -1. The corresponding specific growth rates ranged from 0 01 to 0 06 g g -1 day -1 with a mean of 0 03 g g -1 day -1. To investigate if higher growth rates were associated with the downstream portion of the Shelikof sea valley due to prey field enhancement growth was estimated across the sample grid for 70 mm L S pollock at every sample station where fish of that size were captured and diet information was available (Fig. 3). Estimates of growth for the selected sizes of juvenile pollock ranged between 0 04 and 0 10 g day -1 during 2000, 0 05 and 0 14 g day -1 during 2001, and 0 04 and 0 15 g day -1 during 2003 (Fig. 4). The spatial plots of growth for all three years and sub-areas of pollock (Fig. 3) indicated that the best areas for growth varied among the years with the highest growth potential occurring in sub-areas containing the Shelikof sea valley during 2000 and 2003, whereas higher growth potential during 2001 occurred in areas farther removed from the sea valley and the influence of the ACC (Fig. 3). Estimates of growth for all sub-areas were highly variable between the three years, but the NEmid sub-area that contains the sea valley had the lowest amount of inter-annual variability associated with it (Fig. 5). Sub-areas of higher growth for 70 mm L S juvenile pollock contained significantly higher proportions of euphausiids in captured fish diets than areas of lower growth (r 2 = 0 38, P = 0 032). The multiple regression analysis indicated that the amount of consumption estimated by gastric evacuation in each sub-area explained 83% (P < 0 001) of the variation observed in estimates of growth and the proportion of prey with a WBEC > 4000 J/wet g in pollock diets explained the remaining 17% ( P = 0 005). In contrast, differences in thermal exposure among sample stations had an insignificant influence on estimates of growth for 70 mm L S pollock (P > 0 36). Prey energy density.--prey WBEC collected during September of 2003 in the western GOA varied by prey group and body size (Table 4). Larvaceans and chaetognaths had 61

67 the lowest specific WBEC while large euphausiids, gammariids, and large copepods contained the highest WBEC. Energy densities may have been slightly underestimated for larvaceans because we were only able to measure the energy contained in their body and not their house. The amount of water in prey tissue was more correlated with wet energy content (Pearsons r = 0 85, P < 0 001), than were estimates of calories in a gram of dry tissue (r = 0 60, P = 0 007). E. pacifica contained the lowest energy content observed among the euphausiid prey group (WBEC = kj wet g -1 ), whereas large T. inermis contained 1 7 times higher energy content (WBEC = kj wet g -1 ). Similarly, the specific energy content of large bodied (> 15 mm L T ), euphausiids of the species T. inermis and T. spinifera were 1 3 and 1 5 times greater, respectively, than smaller (< 15 mm L T ) individuals of the same species. Euphausiid depletion.--localized juvenile pollock consumption of euphausiid prey on a daily basis exceeded 5% of the standing stock biomass at only 6% of the 71 sample sites (Fig. 6). The majority of sites (83%) experienced a daily reduction of euphausiid standing stock of 1% or less. Two stations estimated to have the highest amount of euphausiid depletion (47-48%) were both located near shore in shallow water (5E, 2000 and 9E, 2003) where overlap between predator and prey should have been highest and or sampling bias was most likely to occur. Similarly, in 2000 two sites (2B and 5B) had greater than 10% daily reductions in euphausiids primarily due to low euphausiid standing stock estimates near the slope zone along the B sampling line (Fig. 1). At the spatial scale of the entire sampled rearing area age-0 pollock consumption of euphausiids had C/B ratios of for 2000, for 2001 and for

68 Tables: Table 1. The sample stations contained in each habitat area and area-specific intercepts (a) and slopes (b) of length-weight equations for juvenile walleye pollock used to estimate daily growth (g) during 2000 for bioenergetics model runs. Habitat area Sample stations a b r 2 N P SW 1-4B,1-4C 6 7E < NW 1-4D, 1-4E 5 1E < SE 5-8B, 5-10C 7 7E < NE 5-10D, 5-10E 4 4E < Table 2. Proportional diet composition of prey for size specific groups of juvenile walleye pollock from all stations sampled in the grid during Proportional diet content was determined from the wet masses of prey items identified in stomach samples. Size class Amphipods Copepods Copepods Chaetognath Crab Fish Mm L S n & Shrimp < 2 mm > 2 mm & Larvaceans larvae Euphausiids larvae Pteropods

69 Table 3. Mean size-specific estimates of daily growth rate in wet mass and the corresponding 95% CI. Growth was estimated using sectioned otoliths from juvenile walleye pollock captured across all stations in 2000 and was used in bioenergetics model runs to predict prey consumption during 2000 for model corroboration. The whole body energy content (WBEC) of juvenile walleye pollock was altered in bioenergetics model runs to correspond with observed WBEC measured in the western GOA (Buchheister and Wilson in press). Size class WBEC Mm L S n Wet mass g day -1 95% CI J g wet -1 mass

70 Table 4. The whole body energy content (WBEC), wet mass, dry mass, % dry weight, and bioenergetics modeled energy values of juvenile walleye pollock prey groups collected in September of 2003 from the western GOA sampling grid. WBEC was estimated for size- and species-specific differences when feasible. Prey Group Species Size (mm L T ) Wet Mass (g) Dry Mass (g) % Dry Mass KJ dry g -1 KJ wet g -1 Modeled J wet g -1 Chaetognaths Crab larvae Cancer spp Fish larvae M.Villosus Gammariids Hyperiids Pteropods Larvaceans 8 * Copepods Calanoid spp ** Calanoid spp ** Calanoid spp ** Euphausiids E. pacifica > T. inermis > T. inermis < T. raschi T. spinifera > T. spinifera < * Mean lengths for the prey group ** Indicates prosome length for copepods 65

71 Figures: Figure1. Study area and sampling stations (+) in the western Gulf of Alaska. Station labels correspond to sample stations in Table 1. 66

72 Figure 2. Comparison of gastric evacuation rate model estimates of consumption (g wet mass day -1 ) and bioenergetics model estimates of consumption (g wet mass day -1 ) for size-specific groups of walleye pollock in the western Gulf of Alaska, September The dashed line represents the 1:1 line used for comparison. Bioenergetics model estimates of consumption (g/day) Gastric evacuation model estimate of consumption (g/day) 67

73 Figure 3. Bioenergetics model estimates of daily growth for 70 mm standard length walleye pollock during 2000, 2001, and 2003 (top, middle, and bottom panels respectively) in the western Gulf of Alaska. Depth contours in gray are 100 m and 200 m are in black. Heavy black lines delineate sub-areas boundaries. 68

74 Figure 4. Frequency distribution of daily growth for 70 mm standard length walleye pollock during 2000, 2001, and 2003 estimated using the bioenergetics model Frequency Growth (g/day) 69

75 Figure 5. Inter-annual and sub-area variation in estimates of growth for 70 mm standard length walleye pollock during 2000 ( ), 2001 ( ), and 2003 ( ) estimated using the bioenergetics model. Mean growth (g/day) NE mid NE in SW in SW mid 70

76 Figure 6. Frequency distribution of site-specific age-0 pollock population consumption demand compared to the available standing stock biomass of euphausiids. Distribution consists of 71 individually sampled sites during 2000, 2001, and Consumption at each site was estimated using the bioenergetics model and site-specific populations of age-0 pollock determined using midwater trawls. Euphausiid biomass was estimated from nocturnal tows of a 1-m Tucker trawl (0.333-mm mesh). The two highest C/B ratio stations (0.47 and 0.48) were near shore whereas the 0.11 and 0.22 C/B ratio stations were located off shore near the outer edge of the shelf Euphausiids Frequency Consumption / Biomass Ratio 71

77 Section 2. Geographic variation in growth of capelin Mallotus villosus [NPRB Final Report -- Please do not cite or use information in this section without permission of the author. This material is being prepared for publication] Introduction: Capelin Mallotus villosus, are an abundant planktivore with a circumpolar distribution in arctic and semi-arctic ocean ecosystems of the northern hemisphere and as such functions as a link between primary and secondary production and upper trophic level consumers in boreal food chains (Gjøsaeter 1998, Rose 2005). Capelin are thought to have originated in the North Pacific and migrated to the North Atlantic and Barents Sea in response to climate change during interglacial periods (Rose 2005). In the Barents Sea, Norwegian and Russian biologists have been studying capelin for over 100 years (Gjøsaeter et al. 2002). However, in the North Pacific little information is available regarding the ecology of capelin, primarily due to the lack of a commercial fishery (Brown 2002). The distribution, abundance and feeding ecology of capelin appears to be strongly linked to climatic fluctuations in temperature (Fiksen et al. 1995, Anderson and Piatt 1999, Orlova et al. 2002, Huse et al. 2004, Orlova et al. 2005, Rose 2005). In the northern Gulf of Alaska (GOA) capelin were very abundant in a near shore small-mesh trawl surveys during the 1950 s through the 1970 s, but began to decline in the 1980 s and collapsed during the 1990 s presumably in response to a climate regime shift in 1977 (Anderson and Piatt 1999). The direct mechanisms responsible for the observed decline in capelin are unknown, but it is believed to be associated with a temperature-induced shift in the composition and timing of zooplankton abundance (Anderson and Piatt 1999) and a steady increase in the abundance of piscivorous fishes (Bailey 2000). These broader scale patterns in survival and recruitment of fishes result from the accumulation of finer scale processes that occur at spatial scales relevant to individuals. Therefore, by looking at how individual fish respond to alterations in the environment that influence their feeding, growth, and survival we may gain an understanding of what processes are influencing population level shifts in abundance and distribution. The growth rate of individual fish is strongly tied to the thermal exposure and the availability of prey experienced over time by individual fish (Kitchell et al. 1977). Similarly, the resulting growth and body condition of fish, in general, appears to influence which 72

78 individuals survive through winter (Sogard 1997, Pangle et al. 2004). If growth is important to fish survival and recruitment, then being able to link individual fish growth with changes in temperature and prey availability at the appropriate spatial scale is an important step in linking individual level processes with community level responses. In this study, spatially explicit estimates of daily growth and prey consumption were estimated for capelin in the western GOA. Growth estimates for capelin were generated using bioenergetics models borrowed from closely related species to investigate the influence of inter-annual differences in temperature and prey quality. To evaluate which existing bioenergetics model was the most appropriate surrogate for capelin, size-specific estimates of maximum consumption from the competing models were compared to empirical estimates of capelin gut fullness collected in September 2000, 2001, and Methods: Study samples were collected September 2000, 2001, 2003 in the western GOA from the southern end of the Shelikof Strait to the northeastern edge of the Shumagin Islands. The sampling area is primarily contained over the shelf of the Alaska Peninsula with water depth generally less than 200 m. The Alaska Coastal Current (ACC) is an important feature in this area. The ACC follows the Shelikof sea valley, joining the Alaska Stream in the southeastern edge of the sampling grid (Schumacher and Reed, 1986). Some water from the ACC flows across the shelf and into the sample area (Schumacher and Reed 1986). Variable fresh water input from snowmelt contributes to a large range of temperatures ( ºC) and salinities (30 33 PSU) (Lanksbury et al. 2005). Sample collection- Day and night paired tows for forage fish, plankton, and oceanographic conditions were conducted at 42 stations along a grid across the rearing area from the NOAA Ship Miller Freeman. Age-1 and greater capelin were captured in the water column using a Stauffer trawl equipped with 568-kg, steel-v doors, and a 3-mm mesh codend liner (Wilson et al. 1996). The trawl was fished from 200 m depth to the surface or 10 m off the bottom, if the bottom was shallower than 200 m, with a warp retrieval of 10 m min -1 and ship speed of knots. Effective net mouth for the Stauffer trawl was previously estimated at ca. 100 m 2 (Wilson et al. 1996). Plankton were sampled using a Tucker trawl with a 1 m 2 net opening and two 333 μm mesh 73

79 sequential closing nets. The first net was opened at 200 m or 10 m off of the bottom and fished at a 45º-wire angle during retrieval (20 m min -1 ) until the bottom of the thermocline was reached. The second net was then fished from the bottom of the thermocline to the surface. Measures of net depth, temperature, and salinity were collected using a Sea-Bird SeaCat profiler attached to the lead wire of the Tucker trawl. Capelin captured in the Stauffer trawl were weighed and enumerated on deck with up to 30 individuals randomly selected for preservation. The sub-samples of capelin were either flash frozen at 80 ºC or preserved in 10% buffered formalin for gut content analysis in the NOAA AFSC in Seattle WA (Buchheister and Wilson 2005). Zooplankton collected in the Tucker trawl was preserved in 5% buffered formalin and sent to the Poland Plankton Sorting and Identification Center (Szezcin Poland) for identification and enumeration. Diet analysis- Preserved capelin were thawed in seawater or removed from formalin and rinsed in seawater for analysis in the lab. Individual fish were blotted dry, measured to the nearest mm standard length (L S ), and weighed (0 001 g). Individual fish lengths and masses were corrected for shrinkage due to preservation (Buchheister and Wilson 2005) prior to use in any analysis. Stomachs were dissected, and prey items removed. Prey items were identified to the lowest possible taxonomic level under a dissecting microscope, sorted, counted, blotted, and weighed (wet mass). Diet data was analyzed by grouping prey into broad taxonomic groups (e.g. Ciannelli et al. 2004b) and calculating the mean proportional diet content, by weight, of each prey group for size-specific fish diets. To evaluate the spatial positioning of growth rates diets for capelin used to estimate station-specific growth estimates were grouped into a mm L S size class to ensure adequate sample sizes at each independent station for each year. Total prey weight (blotted wet g) was measured for each capelin to determine individual gut fullness for use in the gastric evacuation rate model estimate of consumption. Bioenergetics models- The Wisconsin bioenergetics model, version 3 0 (Hanson et al. 1997) was parameterized for herring Clupea harengus (Rudstam et al. 1994), european smelt Osmerus eperlanus (Karjalainen et al. 1997), and rainbow smelt Osmerus mordax (Lantry and Stewart 1993) to estimate the daily growth of age-1 + capelin. These models 74

80 use a balanced energy budget where energy available for growth equals the total energy consumed minus the energy lost to waste, activity, and respiration (Hanson et al. 1997). The bioenergetics model operates on a daily time step and incorporates allometric weight-dependent and temperature-dependent functions for maximum consumption and metabolism (Hanson et al. 1997). To compare the various model estimates to determine which model best represents an initial approximation for capelin bioenergetics we compared estimates from September 2000, 2001, and 2003 of specific gut fullness (g g -1 ) from field captured capelin with model derived estimates of size-specific maximum daily ration sizes. Sizespecific bioenergetics model inputs were averaged from across the grid for thermal experience (8 6º C), proportional diet content (available upon request), and prey energy content (Mazur et al. in press, this chapter) to parallel the temporal and spatial extent of gastric evacuation estimates. Capelin whole body energy content (WBEC) was borrowed from estimates of rainbow smelt (Lantry and Stewart 1993) and was applied as the predator energy density within each bioenergetics model. The prey energy density for the euphausiid prey group was modeled using one energy value estimated by multiplying the proportion of each identified euphausiid species in the diet of capelin by the speciesspecific wet WBEC (Mazur et al. in press, this chapter) and summing. The diets of capelin contained 5.7% T. raschii, 3.2% E. pacifica, 44.6% T. spinifera, and 45.8% T. inermis resulting in a euphausiid energy estimate of 6081 J g -1 wet weight. The size structure of consumed copepods was only roughly estimated from the diets of capelin, so we modeled the energy content of small copepods by taking the mean of the mm (small) and mm (medium) energy estimates. Copepods larger than 2 mm were modeled using the energy estimate obtained from the mm (large) size group. Gastric evacuation model- For the gastric evacuation estimates of ration sizes; capelin diets were pooled across all stations for each year to ensure adequate sample sizes for all size classes and a frequent diel-sampling interval necessary for gastric evacuation estimates. Gastric evacuation estimates of consumption were modeled using a Bajkov equation (Ney 1990) modified by Eggers (1979) to fit an exponential evacuation rate without applying a correction for differences in diet at the beginning and end of model estimates (Boisclair and Marchand 1993, Richter et al. 2004): 75

81 Eqn 1) C D = 24 S median R Daily consumption C D (g day -1 ) was estimated for capelin between 75 and 100 mm SL using S median, the median amount of food in the stomach over all diel sample periods and R, the gastric evacuation rate assuming a simple exponential evacuation rate (Eggers 1979). We applied the evacuation rate of 0 35 day -1 estimated by Wilson et al. (2006) using the MAXIMS program (Jarre-Teichmann et al. 1993) generated from samples collected during this study. To reduce the influence of several very large individual stomach contents we used median stomach content rather than the geometric mean (Cochran and Adelman 1982, Merati and Brodeur 1996). Spatial growth predictions- Daily growth for age-1+ capelin was estimated using the most appropriate model as a surrogate for capelin at every sample station in the western Gulf of Alaska where capelin between 75 and 100 mm size were captured and diet information was available. Bioenergetics model inputs of proportional diet content, thermal experience, and ration size were estimated for each sample station in the grid. Mean ration sizes (p-values) for each year were calculated as the proportional difference between the gastric evacuation estimate of consumption and the bioenergetically estimated maximum consumption level. Ration sizes were fixed for each year to estimate growth across the sample grid. By fixing the ration size, any unknown errors associated with the bioenergetics model in estimating growth induced by altering the ration size would be fixed at a consistent level for each year (Bajer et al. 2004). A multiple linear regression was used to investigate the relative contribution of temperature and prey quantity to the estimated growth of capelin. Results: Similarities in maximum consumption estimates and the peak observed gut fullness estimates for capelin indicate that the bioenergetics model for rainbow smelt Osmerus mordax was the most appropriate substitute for capelin (Fig. 1). The best-fit model was determined for capelin to be the model that most closely approached the sizespecific field estimates of gut fullness without crossing over the maximum consumption line. 76

82 Growth potentials for capelin using the O. mordax model ranged from to g d -1 (Fig. 2). Growth potentials were relatively low in 2000, and high in This was evident from shifts in growth potential frequency distributions and modes. Thus, capelin growth potentials varied with year. Geographic distributions indicate that capelin growth potentials were not consistently organized within the study area. However, during 2001, the highest growth potentials (>0.63 g d -1 ) did occur relatively close to the Alaska Peninsula (Figure 3). In contrast, during 2000 and 2003, high growth potentials occurred haphazardly throughout the area sampled. Thus, the high growth potentials in 2001 occurred in nearshore habitat not related to the Shelikof sea valley. 77

83 Figure 1. Size-specific gut fullness (g/g) observed in individual capelin captured during September 2000, 2001, and 2003 ( ) in the western Gulf of Alaska compared with the size-specific theoretical maximum daily consumptions estimated from bioenergetics models of related species. Maximum daily consumption estimates for the rainbow smelt model (. ) are a closer approximation to observed capelin gut fullness estimates than those from the european smelt model (---) or the most commonly used surrogate for capelin, Baltic herring model ( ). Maximum consumption (g/g/day) and Gut fullness (g/g) Capelin body weight (wet g) Individual capelin gut fullness herring model rainbow smelt model european smelt model 78

84 Figure 2. Frequency of growth potentials for 85 mm SL capelin estimated using the Osmerus mordax bioenergetics model during September of 2000, 2001, and 2003 in the western Gulf of Alaska. 0.3 September Frequency September 2001 September Growth (g/day) 79

85 Figure 3. Spatially explicit bioenergetic estimates of growth potential for 85 mm SL capelin using the Osmerus mordax model during September of 2000, 2001, and 2003 in the western Gulf of Alaska. Depth contours are 100 m in gray and 200 m in black. 80

86 Discussion and Conclusions Some components of the forage-fish prey field were enriched in the Shelikof sea valley vicinity. These were euphausiids and large copepods. Other zooplankton groups of potential importance as forage fish prey did not exhibit consistent patterns of distribution within the study area. However, none of the sampling locations were devoid of all zooplankton relevant to forage fish diets. Thus, the Shelikof sea valley vicinity was enriched but prey were available elsewhere. The mechanisms responsible for the enriched prey field are not clear. The abundance of large copepods and euphausiids over the slope suggests an influx from the slope to the sea valley. This could result from downstream transport in the ACC, or from at-depth influxes of slope water as part of the estuarine-like flow generated by the ACC (Stabeno et al. 2004). Based on distribution patterns of specific zooplankton, which suggest some constraint on mixing between shelf and slope populations, enrichment within the sea valley is probably not exclusively due to influxes from the slope. There may be physical process on the shelf that interact with behaviorally mediated movements of large copepods and euphausiids to facilitate concentration of select species in the sea valley (Greene et al. 1988). The combined abundance of large copepods and euphausiids was consistently high in the Shelikof sea valley vicinity but this did not translate into high abundance and large size for all forage fish species. It was, however, associated with increased mean size of small-size forage fishes (age-0 pollock and capelin), and increased abundance of large forage fish (age-1+ pollock juveniles). Eulachon, which were intermediate in size, exhibited both increased size and abundance. Thus, fish body size seems relevant to how these fishes differentially responded to the enrichment. The rarity of these fishes over the slope, despite an abundance of large copepods and euphausiids, raised questions about other factors that might affect forage fish habitat quality. Salinity and bottom depth were relatively high over the slope, but direct mechanistic effects seem unlikely due to the relatively small changes in salinity and because these fishes are not closely associated with the sea floor. Water velocity was another possible factor. Over the slope, expected mean velocity corresponded to swim 81

87 speeds >1 bl s -1. It is unclear how this compares to optimal swim speeds for these fishes, given the existing environmental conditions, but Ryer et al. (2002) report swim speeds <1 bl s -1 for age-0 walleye pollock and noted that swim speed increased with fish size. Juvenile walleye pollock better responded to the enriched prey field, in terms of body weight and diet, than did capelin and eulachon. Wilson et al. (in press) showed that juvenile walleye pollock had a broader diet than capelin. Eulachon seem more like capelin in that they appear to depend on a relatively narrow suite of prey types. When foraging in areas where euphausiids and large copepods are relatively low in abundance, walleye pollock may switch to alternate prey. In contrast, capelin and eulachon may continue to forage efficiently on preferred prey. It is, however, difficult to separate species-specific differences from ontogenetic effects. Nutritional demands may be more pronounced for the pollock since all of those examined were sub-adult. The spatial association of high quality growth habitat for juvenile pollock varied among the years with the highest growth occurring in areas associated with the Shelikof sea valley and the ACC during 2000 and Higher pollock growth potential during 2001 was farther removed from the sea valley and the direct influence of the ACC. High estimates of growth were strongly associated with the amount of euphausiids that were found in the diets of juvenile pollock and, to some degree, the amount of euphausiids observed in the environment from plankton tows of the entire water column. It is unclear why growth estimates from 2001 were farther removed from the Shelikof sea valley. However, water velocity estimates from the SPEM model were higher during 2001 and velocity may be influencing the distribution of both juvenile pollock and their euphausiid prey. Estimates of growth for all areas demonstrated a relatively high inter-annual variation among the three years, but the area encompassing the sea valley had the lowest amount of variability in growth. The reduced variability near the sea valley may result from the consistent high abundance of euphausiid and large copepod prey observed in the area. Areas of high growth for juvenile pollock contained significantly more euphausiids in captured fish diets. Similarly, the observed variation in growth estimates was primarily explained by the quantity of prey in the diets of pollock, with the quality of that prey explaining the remaining variation. Larvaceans and chaetognaths had the lowest specific 82

88 WBEC while large euphausiids, gammariids, and large copepods contained the highest WBEC. Euphausiids are an important prey for pollock growth because they have relatively large body sizes and high WBEC. In contrast, differences in thermal exposure among sample stations had an insignificant direct influence on estimates of growth for juvenile pollock. Overall, age-0 walleye pollock foraging did not appear to be causing either broad scale depletions of euphausiid standing stocks nor depletions at the finer spatial scale of individual sampling stations. However, these results should not be misconstrued as suggesting that age-0 pollock are not causing localized depletions of euphausiids that have the potential to lead to future growth suppression. Whereas our estimates of euphausiid standing stock were fairly precise on a horizontal scale, we are unable to account for differences in the vertical distribution of euphausiids and how these differences are influencing the amount of euphausiid biomass that is available to foraging pollock. Because pollock are visual foragers they depend on adequate light conditions to forage effectively. It is possible that age-0 pollock consumption could have caused depletions in euphausiid biomass on the vertical scale (Aksnes et al. 2004) at times and depths that pollock foraging is maximized. Our analysis was not designed to detect vertical depletions of euphausiids nor account for how these depletions were influencing pollock foraging and growth. Similarly, depending on the functional response relationship between pollock and euphausiids, minor reductions in the abundance of available euphausiids could result in reciprocal reductions in pollock foraging and growth. We also did not account for the entire community consumptive demand for euphausiids made up of other foraging vertebrates (fishes, mammals, and birds) and invertebrates. Growth potentials for capelin using the O. mordax model were higher during 2001 than in 2000 or 2003 with the majority of the high growth in 2001 occurring near shore. Similarly, high growth potentials for capelin were less geographically organized during 2000 and 2003 than in It is unclear if the near shore habitat in general provided a higher growth habitat for capelin, or if the higher water velocities abnormally influenced the 2001 results. Previous research suggests that capelin abundance is higher near shore (Anderson and Piatt 1999), but it is not clear if this association is common or if a more 83

89 near shore distribution confers a growth advantage. Capelin daily growth estimates were roughly half those of pollock of a similar body size and lacked a consistent spatial trend, suggesting that they are less influenced by the spatial gradients in prey observed in the sampled area and are able to obtain sufficient amounts of prey to satisfy the level of growth compatible with their life history strategy. 84

90 MANUSCRIPTS: PART II Chapter 1 Over-winter effects on the diet of juvenile walleye pollock. [NPRB Final Report -- Please do not cite or use information in this chapter without permission of the author. This material is being prepared for publication] Introduction: Seasonal variation in diet is an important but neglected aspect of the biology of walleye pollock Theragra chalcogramma in the western Gulf of Alaska. In this region, walleye pollock support a large fishery, and are commonly consumed by upper trophic level animals. In fact, Bailey et al. (2000) suggest that predation of early juveniles effectively controls recruitment to the fishery. Bailey et al. (2000) suggest that young-of-the-year (age-0) juveniles become vulnerable to groundfishes in late summer. Vulnerability to predation often relates to body size (Sogard 1997). Brodeur and Wilson (1996) showed that the body size of GOA juvenile walleye pollock increased little during winter due, presumably, to relatively slow growth. Thus, the seasonal period of vulnerability may be extended by a low over-winter growth rate. This may partly reflect relatively low water temperature, but it might also relate to seasonal reduction in zooplankton abundance (Coyle and Pinchuk 2003). For the 2000 year class of walleye pollock, a decrease in body condition during their first winter suggested dietary stress; interestingly, the magnitude of decrease varied regionally (Buchheister et al. in press). Correspondence between body condition and diet has been demonstrated among juvenile pollock in bays and over the continental shelf adjacent to east Kodiak Island (Wilson et al. 2000). Juvenile walleye pollock feed mostly on zooplankton. In the Gulf of Alaska, during autumn, Merati and Brodeur (1996) found that diets of young-of-the-year (age-0) individuals consisted mostly of copepods and euphausiids with other crustaceans, pteropods, and gelatinous (e.g. larvaceans) prey of minor components. They noted, however, a regional shift in diet, which generally corresponds with regional differences in growth rate (Bailey et al. 1996) and body length (Wilson 2000). In addition, Brodeur and Wilson (1996) noted an increase in feeding on epibenthic prey during winter and early 85

91 spring. In the Bering Sea (Dwyer et al. 1987) and off northern Japan (Yamamura et al. 2002), euphausiids were an important staple of juvenile walleye pollock but the fish examined were probably older than age-1. To better understand the role of diet as a factor in the relatively low body condition, slow growth rate, and probable high rate of predation-related mortality of juvenile walleye pollock during their first winter, this study focused on two objectives. First, feeding intensity and diet were compared for evidence of a winter-related shift, and 2) seasonal patterns were compared among sub-areas for evidence of a modulating geographic effect on seasonality. This study was conducted in conjunction with an investigation of seasonal and regional variation in the body condition of juvenile walleye pollock (Buchheister et al. in press). Methods: Juvenile (age-0 and age-1) walleye pollock of the 2000 year class were sampled along the continental shelf of the western Gulf of Alaska between 151 o 161 o W longitude (Fig. 1). The study area was divided into three regions (Kodiak, Semidi, and Shumagin), similar to those used by Wilson (2000) and Brown and Bailey (1992). Sampling occurred during four time periods: August-September, 2000 (late summer, autumn - Aut00); January-March, 2001 (winter - Win01); June-July, 2001 (late spring, summer - Sum01); and August-September, 2001 (late summer, autumn - Aut01). Briefly, samples were obtained by trawling in midwater or on bottom during various groundfish assessment and research cruises conducted by the National Marine Fisheries Service (NMFS) and the University of Alaska (see Chapter 2 for more at-sea sampling details). Trawls were equipped with small mesh codend liners and were fished mainly during daylight hours. Members of the 2000 year class were distinguished from older individuals based on length. In Aut00, age-0 individuals were separated from older individuals based on a clear break in the size frequency distribution (Brodeur and Wilson 1996). After the year class became age-1 (Win01, Sum01 and Aut01), the upper size limit of the cohort was estimated using NMFS length at age data (M. Dorn, Alaska Fisheries Science Center, unpublished data). Up to 150 individuals from each trawl catch were randomly selected and frozen (-20 o C) for subsequent analysis in the laboratory. 86

92 The fish used in this study were selected from a larger set of available samples. Selection was based on sampling location and fish size. Two sampling locations (stations) were chosen within each season-region cell (Table 1, Fig. 1). Choice of location was somewhat arbitrary with preference given to samples used to determine body condition (Chapter 2), and to stations that were geographically separate. From each sample, up to 20 age-0 fish and 10 age-1 fish were selected to represent available sizes. Fish were thawed in seawater and measured prior to stomach excision. Standard length (SL) was measured to the nearest 1 mm. Individuals were blotted dry, then weighed to the nearest 1 mg. Stomachs were excised from the esophagus to the pylorus and preserved in 10% formalin. The contents of each stomach were sorted under a dissecting microscope and identified to the lowest possible taxonomic level. Each taxon was enumerated, blotted dry, and weighed to the nearest 0.01 mg. Due to the difficulty in identifying copepods to species when highly digested, this prey group was identified by order and by prosome length (<2.0 mm or >2.0 mm). In addition, euphausiids were separated into two categories: juveniles and adults (juv./adult), and calyptopis and furcilia (cal./fur.). Stomach fullness. For each fish, stomach content weight (SCW), or fullness, was expressed in terms of percent body weight (%BW): SCW Eqn 1) % BW = * 100 BW SCW Seasonal differences in %BW were examined by region using the nonparametric Kruskal-Wallis test. Significant test results were followed with pairwise tests of consecutive seasons using a Mann-Whitney test (Zar 1999). The same statistical procedure was used to test for regional differences by season. Diet composition. Taxonomic composition of the diets was expressed as a percentage of prey count or weight. Percentages were based on total prey number or weight in each season-region cell. To simplify the presentation of results, prey groups that accounted for <2% of the diet of all fish were combined into an other category. 87

93 This category varied slightly depending on whether percent composition was based on prey count or weight. Multivariate tests.--analysis of similarity (ANOSIM, Clarke and Warwick 2001a) was used to test for differences in the taxonomic composition of walleye pollock diets with regard to region and season. ANOSIM is a nonparametric, multivariate, permutation test where one first computes a station by station Bray-Curtis dissimilarity matrix from a species by station matrix, then uses a permutation method to determine if an R statistic, Eqn 2) R = ( r r ) B W, 0.5( n( n 1)) is significantly different from random, where r W is the average of rank similarities (1- Bray-Curtis coefficient) of all paired samples within regions or seasons, and r B is the average of rank similarities of all paired samples among regions or seasons. This number can vary from 1 to 1, but is usually between 0 and 1 with numbers near 0 indicating weak separation of groups and numbers close to 1 indicating strong separation. A nested design was used to account for fish being grouped by trawl haul. For each test, minor prey groups (<2% of total stomach content) were excluded to better resolve differences among major prey groups. There were seven major prey groups but the complement varied depending on composition by number or composition by weight. Day-night variation in collection time was ignored because this effect was not significantly associated with taxonomic composition of stomach contents by prey number (p = 0.998, R = ) or by prey weight (p = 0.999, R = ). We first tested for a significant region effect and then for a significant season effect because sample sizes did not support a fully crossed design. Significant effects were followed by a percent similarity (SIMPER) analysis (Clarke and Warwick 2001b) in order to determine which species were most responsible for dietary differences. SIMPER determines the percent contribution that each prey item makes to the overall dissimilarity of each paired difference by calculating the average dissimilarity between all pairs of samples. 88

94 Results: A total of 284 juvenile walleye pollock were examined. Standard lengths ranged from mm (Table 1). The available fish sizes increased with season and, among age-1 samples (Win01-Aut01), tended to be relatively large for individuals from the Kodiak Region. Stomach fullness (%BW) varied significantly by season. Within the Shumagin and Semidi regions, fish tended to have low stomach fullness during Win01 as compared to collections made during the preceding autumn or subsequent summer (Fig. 2). This pattern was supported statistically with significant differences between consecutive season pairs (Table 2); however, the Aut00-Win01 comparison was marginally not significant (p=0.082). Stomach fullness did not vary from Sum01 to Aut01. In contrast, stomach fullness of fish from the Kodiak Region declined steadily as the seasons progressed. Significant differences were detected between Aut00 and Win01, and between Sum01 and Aut01 (Table 2). During Aut00 and Win01, the Kodiak Region was associated with large median stomach fullness relative to the Semidi and Shumagin regions (Table 2). The opposite occurred in Aut01 when the lowest median stomach fullness was from fish in the Kodiak Region, but this was based on only seven fish. Juvenile walleye pollock primarily consumed seven prey types. Of these, juvenile and adult euphausiids, small and large calanoid copepods, and larvaceans accounted for approximately 75% by number or weight. Other abundant prey types (>2% of total counts) included larval euphausiids (calyptopis and furcilia stages), cumaceans, and crab larvae (Fig. 3). Other prey types >2% by weight included fish, chaetognaths, cumaceans, and amphipods (Fig. 4). Interestingly, the percentage of euphausiids was relatively low during Win01 and Sum01 in all regions. During these times, the alternate prey generally included large portions of copepods and larvaceans; in the Kodiak Region, cumaceans and fish were also relatively important. Regional differences in diet composition, however, were not statistically significant for prey counts (ANOSIM, p = 1.000, R = ) or weight (ANOSIM, p = 1.000, R = ). Statistical examination of the season effect, therefore, was based on data pooled for all regions. Season was significantly associated with juvenile pollock diet whether characterized by prey count (ANOSIM, p = 0.005, R = 0.302) or weight (ANOSIM, p = 0.001, R = 0.318). Most differences were between autumn and winter or summer (Table 3). The 89

95 SIMPER analysis indicated that the difference between Aut00 and Win01 in composition by weight was largely due to juvenile and adult euphausiids (37%) and large copepods (29%) (Table 4). These two prey categories accounted for a majority (>50%) of the between-season dissimilarity in all significant differences of composition by weight (Table 4). The euphausiid portion of the diet generally declined during the winter and summer while that due to large copepods increased (Fig. 4). Small copepods contributed 22-27% of the dissimilarity in composition by prey count between Aut00 and Sum01 or Aut01. Interestingly, the Aut00-Aut01 difference was not significant by weight due to the gravimetric importance of euphausiids in both time periods. Admittedly, seasonal changes in diet were confounded by predator length. However, the percentage of euphausiids in the diets of small fish in Aut00 and of large fish in Aut01 was large relative to intermediate-size fish collected during winter and summer. Thus, at least some of the seasonal variation in diet can not be explained by fish size. 90

96 Tables: Table1. Sampling information for walleye pollock collected in the Gulf of Alaska. Number of fish indicated by n, with number of stations in parentheses. Stations were sampled during day (D), night (N), or both (B). SE is standard error. Standard length (mm) Region Season n Date Diel Mean (SE) Range Shumagin Aut00 40 (2) 5 Sep 00 B 62 (1.3) Win01 20 (2) Feb 01 B 99 (2.9) Sum01 20 (2) 5 Jun 01 D 114 (2) Aut01 20 (2) 5-7 Sep 01 B 180 (6.2) Kodiak Aut00 40 (2) Aug 00 N 71 (0.9) Win01 20 (2) Feb 01 D 121 (3.1) Sum01 20 (2) 26 Jun - 8 Jul 01 D 165 (2.7) Aut01 7 (1) 3 Sep 01 N 201 (4.5) Semidi Aut00 37 (2) Sep 00 B 87 (1.7) Win01 20 (2) Mar 01 D 110 (3.1) Sum01 20 (2) Jun 01 D 115 (2.4) Aut01 20 (2) Sep 01 D 169 (5.1) Table 2. P-values for pairwise Mann-Whitney tests of stomach fullness. Bold numbers indicate a significant difference. Pairwise tests between regions in Sum01 were not conducted due to non-significant Kruskal Wallis results. Region Aut00 vs. Win01 Win01 vs. Sum01 Sum01 vs. Aut01 Kodiak Semidi Shumagin Season Kod. vs. Sem. Kod. vs. Shu. Sem. vs. Shu. Aut00 < Win01 < Sum Aut

97 Table 3. Post-hoc pairwise test results of Analysis of Similarity (ANOSIM) indicating significant (p < 0.05; bold) differences between season pairs with regard to taxonomic composition of stomach contents of juvenile pollock. Composition was characterized by prey number and by prey weight. Taxonomic composition Season pairs p-value R by number Aut00 vs. Win Aut00 vs. Sum Aut00 vs. Aut Win01 vs. Sum Win01 vs. Aut Sum01 vs. Aut by weight Aut00 vs. Win Aut00 vs. Sum Aut00 vs. Aut Win01 vs. Sum Win01 vs. Aut Sum01 vs. Aut Table 4. Percent similarity (SIMPER) results indicating the percent contribution of each major prey group to significant between-season dissimilarity (see Table 3) in juvenile walleye pollock diet. Diet was indicated by taxonomic composition of stomach contents by prey number and by prey weight. Dash (--) indicates prey group not > 2% of total stomach contents (see Methods). Pairwise comparison between seasons Composition by number Composition by weight Aut00 Win01 Sum01 Aut00 Aut00 Win01 vs vs vs vs vs vs Aut01 Aut01 Aut01 Win01 Sum01 Aut01 Aut00 vs Sum01 Sum01 vs Aut01 Prey group euphausiids, j+a copepods, >2mm larvaceans cumaceans copepods, < 2mm crabs euphausiids, c+f chaetognath fish amphipods

98 Figures: Figure 1. Sampling locations of juvenile walleye pollock in the Gulf of Alaska through four different time periods: Aut00 (Δ), Win01 (O), Sum01 ( ), and Aut01 (+). 93

99 Figure 2. Box plot of stomach fullness of juvenile walleye pollock in terms of percent body weight (%BW). Solid lines denote the 25 th, 50 th, and 75 th percentiles, and error bars indicate the 10 th and 90 th percentiles. Circles represent 5 th and 95 th percentiles. Mean values are included as dotted lines. 94

100 Figure 3. (A) Diet composition of juvenile walleye pollock by percent of total number. Top seven prey groups are shown. All other prey groups were combined into one category. Numbers of fish stomachs examined are noted above the bars. (B) Ranges and means of walleye pollock standard length by region and season. 95

101 Figure 4. Diet composition of juvenile walleye pollock by percent of total weight. Top seven prey groups are shown. All other prey groups were combined into one category. Numbers of fish stomachs examined are noted above the bars. 96

102 Chapter 2 Over-winter effect on body condition and energy density of juvenile walleye pollock. [NPRB Final Report -- Please do not cite or use information in this chapter without permission of the author. This chapter has been accepted for publication: Buchheister et al. in press] Introduction: Seasonal variation in productivity and water temperature largely determines the reproductive strategy and growth schedule of fishes in temperate marine environments (Pitcher and Hart 1982). In these waters, spawning usually occurs during spring so that young fish are maximally exposed to periods of abundant food and warmer temperatures conducive to rapid growth (Match/Mismatch hypothesis; Cushing 1990). With the onset of winter, fish growth slows, energy reserves may be depleted, and mortality rates of juveniles can increase (Sogard 1997, Hurst and Conover 1998, Gotceitas et al. 1999), likely due to cooler water temperatures and diminished prey fields (Shuter and Post 1990, Schultz and Conover 1999, Hurst and Conover 2001). Large fish often have higher rates of survival than smaller individuals, due to decreased vulnerability to predation and a capability of storing more energy that can be relied upon as growth and foraging conditions change in the winter (Sogard 1997). Large body size also results in reduced weight-specific metabolic rate, which lowers the rate of depletion of stored energy. But, size-selective mortality does not occur for all temperate water species (Sogard 1997), presumably due to complex interactions between water temperature, prey availability, and predation (Garvey et al. 1998). However, survival was positively related to body size for age-0 walleye pollock Theragra chalcogramma held without feeding in low temperature conditions in the laboratory (Sogard and Olla 2000). Thus, accumulation of sufficient energy reserves and attainment of a relatively large body size prior to winter may be an important factor in walleye pollock winter survival and recruitment. Also, the ability of these fish to cope with deleterious overwinter conditions in the wild may vary geographically due to spatial differences in the size, growth, and diet of age-0 individuals (Paul et al. 1998b, Wilson 2000, Wilson et al. 2005). Much research has focused on the factors influencing recruitment of walleye pollock into the adult population (Kendall et al. 1996), and the juvenile stage of development has been shown to be of importance (Bailey 2000). 97

103 However, few studies have investigated how winter affects juvenile condition and energetics, and in this regard, little is known about the relatively large population in the western Gulf of Alaska (GOA). Therefore, the objective of this study was to examine the seasonal and geographic variation in body length and condition that occurs within a cohort of juvenile walleye pollock before, during, and after its first winter in the western GOA. Two indices of body condition were used: 1) length-specific weight (Cone 1989, Patterson 1992) and 2) length-specific whole body energy content (Paul et al. 1998a, Foy and Paul 1999). Comparison of these two condition indices was an ancillary objective to explore the suitability of using at-sea measurements of fish length and weight as a proxy for whole body energy content, which is more difficult to obtain. Methods: Juvenile (age-0 and age-1) walleye pollock of the 2000 year class were sampled along the continental shelf of the western Gulf of Alaska between 151 o 161 o W longitude (Fig. 1). The study area was divided into three regions (Kodiak, Semidi, and Shumagin), similar to those used by Wilson (2000) and Brown and Bailey (1992). The 2000 year class was sampled during four different time periods: August-September, 2000 (late summer, autumn - Aut00); January-March, 2001 (winter - Win01); June-July, 2001 (late spring, summer - Sum01); and August-September, 2001 (late summer, autumn - Aut01) (Table 1). Samples were obtained during various groundfish assessment and research cruises conducted by the National Marine Fisheries Service (NMFS) and the University of Alaska (Table 1). Trawls were equipped with small mesh codend liners and were fished mainly during daylight hours. In Aut00, age-0 members of the 2000 year class were distinguished from older individuals based on a clear break in the size frequency distribution (Brodeur and Wilson 1996). After the year class became age-1 (Win01, Sum01 and Aut01), the upper size limit of the cohort was estimated using NMFS length at age data (M. Dorn, Alaska Fisheries Science Center, unpublished data). Up to 150 individuals from each trawl catch were randomly selected and frozen (-20 o C) for subsequent analysis in the laboratory. 98

104 Water temperature was measured at each sampling location using a calibrated SBE- 19, SBE-39 (Sea-Bird Electronics), or microbathythermograph (Richard Brancker Research Ltd.). Temperature profiles were averaged for all stations by season and region. Laboratory analysis. In the laboratory, frozen samples were thawed in seawater within 2-31 months of collection. An average of 13 fish (range: 1-49) were selected from each station to represent the available size range. Standard length (SL) was measured to the nearest millimeter. Whole body weight was measured to the nearest milligram after fish were blotted dry. All length and weight measurements were corrected for preservation effects using shrinkage conversion equations (Buchheister and Wilson 2005). No correction was applied to account for differences in the time samples spent frozen because Buchheister and Wilson (2005) demonstrated that preservation durations of months did not significantly affect walleye pollock length reduction or weight loss. Whole body energy content (WBEC) was determined for a total of 168 fish. These samples were obtained from a minimum of two stations within each season-region combination (Fig. 1). Fish were selected to represent the available size range. Individuals were measured, weighed, and then dried whole at 65 o C to a constant weight. Fish were homogenized, pelletized ( g dry weight), and the energy density (kj/g dry weight) of each fish was measured using a Parr 1425 Semi-micro bomb calorimeter (n = 134). The mean difference in energy density measurements between replicate pellets from 41 fish was low (1.7%) and values were highly concordant (r c = 0.95; Zar, 1999), therefore, subsequent energy determinations were based on one pellet per fish. Energy densities were converted to kj/g wet weight using the ratio of dry to wet weight for each individual. A freezer failure spoiled many of the WBEC samples from the Kodiak region collected during Sum01 and Aut01. We therefore include energy density data (n = 34) from a proximate analysis of age-1 walleye pollock collected from the corresponding region and seasons using the F/V Alaska Beauty (Table 1) (R. J. Foy, unpublished data). To determine proximate composition, frozen samples were ground into a homogenate, and lipid was extracted using the Supercritical Fluid Extraction method in a LECO FA- 100 Fat Analyzer. Protein content was determined with a LECO FP-2000 Nitrogen 99

105 Analyzer, whereas ash content was determined by combustion of homogenate in a muffle furnace at 510 o C until constant weight was attained. Proximate composition estimates were converted to energy density using conversions of kj/g and kj/g for lipid and protein, respectively (Brett 1995), which have been shown to yield similar results to values obtained by bomb calorimetry (Vollenweider 2004). Stomach contents were removed from fish processed for proximate composition. Analysis of body length and condition. Temporal and regional differences in standard length were examined by plotting mean standard length per station against date of sampling. Data from this study were compared graphically to a polynomial regression model of NMFS groundfish survey data ( ) published by Brodeur and Wilson (1996). Coefficients for their regression model were converted to the appropriate length type (SL) following Buchheister and Wilson (2005). Where appropriate, regional differences in mean length were tested by season using analysis of variance (ANOVA). Analysis of covariance (ANCOVA) was used to test for season and region differences in both length-weight and length-wbec relationships. For each ANCOVA, length was the covariate while season and region were main effects. For season and region effects to be tested, lack of significant differences between regression slopes was first confirmed. Then, differences in length-specific weight (or length-specific WBEC) were examined by testing for differences in y-intercepts. Any significant season or region effects in the ANCOVA models were followed by pairwise comparison tests between pairs of consecutive seasons or among the three regions. Critical P-values were Bonferronicorrected for the number of comparisons. For pairwise comparisons, relative differences between values were also calculated as percentages. There were a few methodological differences in the ANCOVAs of the length-weight and WBEC data. Natural log transformations were used to linearize the length-weight relationships and homogenize their variances, but no transformation was necessary for the WBEC data. The length-weight ANCOVA utilized a split-plot design to account for trawl haul (the sampling unit) being nested within season and region (Milliken and Johnson 1992), but due to small sample sizes, the ANCOVA for WBEC did not incorporate a nested design. Also, only stations with four or more fish were used in the length-weight analysis. 100

106 Correlation analysis of condition indices. Correlation analysis was conducted using residual WBEC and residual weight (WT) to see if individuals that were relatively heavy for their length were also relatively energy-rich. Residuals were calculated using linear least-squares regressions fit to all data pooled. WT residuals were based on natural logtransformed data. Only the 168 fish with both WBEC and length-weight data were used. Following Suthers et al. (1992), significance of the correlation was based on the Bartlett Chi-square test. Results: Water temperature at all sampling stations ranged from 4 to 12 o C. There was a pronounced seasonal signal in thermal structure that was similar among regions (Fig. 2). Stratification in the water column was most notable in the autumn (Aut00 and Aut01) with mean surface and bottom waters near 11 o C and 5 o C, respectively. During winter, the water column was uniform and well-mixed with a mean temperature of approximately 5 o C. Stratification was again evident in the summer with generally cooler temperatures relative to autumn. However, summer water temperatures in the Kodiak region were warmer and more similar to those in autumn, most likely due to later sampling dates. Body length.--a total of 2241 walleye pollock from the 2000 year class were measured. Fresh standard lengths ranged from 39 to 263 mm (Table 2). Seasonal progression of mean length corresponded well with NMFS data from (Fig. 3). The seasonal rate of increase in length was rapid in Aut00, slowed to practically no increase by Win01, and then increased through Sum01 and Aut01. Regional differences in fish length were also evident, particularly in the winter and summer (Fig. 3). In Win01, the size of fish from the Kodiak region was significantly different from fish from the other regions combined (ANOVA, P < 0.001), with Kodiak fish larger by an average of 18 mm. Likewise, fish from the Kodiak region had larger mean lengths in Sum01, although this is confounded by later sampling dates. However, an average growth rate of 1.6 mm/day would be required for fish from the Semidi and Shumagin regions to attain the same length as fish from Kodiak, but realistic growth rates are typically below 1 mm/day (Brodeur and Wilson 1996). Temporal differences in sampling and a possible bias in gear selectivity (Table 1) prevented use of ANOVA in 101

107 Aut00, although Kodiak fish appeared to have larger mean lengths-at-date (Fig. 3). Patterns in Aut01 could not be determined due to insufficient sampling. Body condition.--whole wet body weight was measured on all 2241 individuals that were lengthed. Fresh weights ranged from 0.5 to 173 g. Coefficients of determination were high for most regressions of length-weight data, except in the Kodiak region in Aut01 (Table 2). Slope heterogeneity among seasons (P = 0.031) prevented the simultaneous analysis of all seasons and regions in one all-inclusive ANCOVA test. Therefore, the season effect was tested for each region individually, and likewise, the region effect was tested for each season individually. In each of these resultant lengthweight ANCOVA models, slopes of regressions were not significantly different (Table 2, Table 3), except for the test of season in the Semidi region (P = 0.011) (Table 3). Consequently, tests for seasonal differences were conducted pairwise for the Semidi region, omitting the comparisons of nonconsecutive seasons which were not of interest. In all three regions, length-specific weight was significantly lower during winter than during other seasons (Table 3, Fig. 4). Length-specific weight decreased significantly from Aut00 to Win01 by as much as 17.3%, but subsequently rebounded by up to 19.8% in Sum01. Fish from the Semidi region also exhibited a significant increase from Sum01 to Aut01 by 11.4%, but this was not reflected in the other regions. Significant regional differences in length-specific weight were only detected during Aut00, with fish from the Kodiak region nearly 6% heavier than similar sized fish from the Semidi region (Table 4). Although not significant, Kodiak fish also had a tendency towards greater length-specific weight in Win01. No regional differences in weight at length were detected in Sum01 and Aut01. Whole body energy content of juvenile walleye pollock ranged from 3.00 to 6.37 kj/g (Fig. 5). In the ANCOVA for WBEC versus standard length, slopes were not significantly different among regions or seasons; thus, a common-slope model was used. Both the season and region effects were found to be significant (P = and P = 0.007, respectively). Although the effect of the covariate (SL) was marginally non-significant (P = 0.072), it was retained in the model due to its significant effect on WBEC over the entire data set (P < 0.001, WBEC = 0.009(SL) , r 2 = 0.372, n = 168). 102

108 Post-hoc multiple comparison tests of length-specific WBEC identified three significant differences between season pairs (Table 5; Fig. 5). First, length-specific WBEC decreased from Aut00 to Win01 in the Semidi region (0.584 kj/g; 14.0%). Second, length-specific energy density of Shumagin fish increased by kj/g (23.1%) from Win01 to Sum01. And lastly, WBEC increased significantly by kj/g (11.4%) from Sum01 to Aut01 in the Kodiak region. Thus, the seasonal effect was not consistent among regions but two of the three significant differences involved relatively low wintertime energy density estimates. The post-hoc tests also detected three significant regional differences in lengthspecific WBEC (Table 5; Fig. 5). In Aut00, fish from the Kodiak region exhibited higher length-specific energy densities than fish from the Shumagin region by kj/g (18.3%). During winter, length-specific WBEC for Kodiak fish was significantly greater than fish from the Semidi region (0.607 kj/g; 14.1%). During Sum01, length-specific energy densities were high in the Shumagin region, with average values over 0.8 kj/g (17.6%) greater than in the Semidi region. Although the regional effect was not consistent among seasons, two of the three significant differences involved relatively high energy density estimates for fish from the Kodiak region. Comparison between condition indices.--the correlation between residual WBEC and residual WT was significant (r = 0.464, P < 0.001). However, although energy density residuals tended to increase with weight residuals, the correlation was weak and WT residuals did not explain a large amount of the variance (Fig. 6). We did note, however, that percent dry weight (DW) was a reasonably good predictor of energy density (WBEC = 0.314(DW) 2.125, r 2 = 0.872, n = 168). 103

109 Tables: Table 1. Sampling information for collection of juvenile walleye pollock from the western Gulf of Alaska. Abbreviations used for fishing gears are: MMT Marinovich Midwater Trawl, AWT Aleutian Wing 30/26 Trawl, MSMT Modified Stauffer Midwater Trawl, HST High-opening Shrimp Trawl, POLYNE Poly-Nor eastern Bottom Trawl, and BBMT (Dan Trawl) Bering Billionaire Midwater Trawl. Codend Liner Season Vessel Name Regions Sampled Start Date End Date Gear Mesh Size (mm) Aut00 Miller Freeman Kodiak 12 Aug Aug 2000 MMT, AWT 32 Miller Freeman Kodiak, Semidi, Shumagin 3 Sep Sep 2000 MSMT, HST 3 Win01 Miller Freeman Kodiak, Semidi 31 Jan Feb 2001 POLYNE 32 Sea Storm Kodiak 11 Feb Feb 2001 POLYNE 32 Miller Freeman Shumagin 14 Feb Feb 2001 POLYNE, AWT 32 Ocean Harvestor Shumagin 8 Mar Mar 2001 POLYNE 32 Miller Freeman Kodiak, Semidi 13 Mar Mar 2001 POLYNE, AWT 32 Sum01 Vesteraalen Kodiak, Semidi, Shumagin 2 Jun Jul 2001 POLYNE 32 Morning Star Kodiak, Semidi, Shumagin 8 Jun Jul 2001 POLYNE 32 Alaska Beauty Kodiak 25 Jul Jul 2001 BBMT 22 Aut01 Miller Freeman Kodiak, Semidi, Shumagin 3 Sep Sep 2001 MSMT, HST 3 Alaska Beauty Kodiak 11 Sep Sep 2001 BBMT

110 Table 2. Least-squares linear regressions of natural log (ln)-transformed juvenile walleye pollock standard length (SL, mm) and natural log-transformed whole wet weight (WT, g) by region and season. The coefficient of determination (r 2 ), number of fish (with number of stations in parentheses), standard error of estimate of the regression (SEE) a, and range of standard lengths for each regression are also included. Region Season Equation r 2 N SEE a SL Range (mm) Kodiak Aut00 ln(wt) = ln(sl) (8) Win01 ln(wt) = ln(sl) (23) Sum01 ln(wt) = ln(sl) (14) Aut01 ln(wt) = ln(sl) (4) Semidi Aut00 ln(wt) = ln(sl) (48) Win01 ln(wt) = ln(sl) (4) Sum01 ln(wt) = ln(sl) (5) Aut01 ln(wt) = ln(sl) (26) Shumagin Aut00 ln(wt) = ln(sl) (16) Win01 ln(wt) = ln(sl) (15) Sum01 ln(wt) = ln(sl) (7) Aut01 ln(wt) = ln(sl) (7) a SEE are reported for accurate back-calculation of untransformed length-weight relationships (Sprugel 1983). 105

111 Table 3. Summary results of ANCOVA tests of pairwise seasonal differences in lengthweight relationships by region. Differences among slopes and Y-intercepts were tested at critical values of P = and P = 0.005, respectively (Bonferroni-corrected, α = 0.05). Significance is indicated by bold numbers. For each comparison, the season with the greater Y-intercept, or length-specific weight, is included. Percent changes (% Change) in length-specific weight from the earlier to the later season were calculated from backtransformed weights. H 0 : Slopes are equal H 0 : Y-intercepts are equal Greater Y-intercept Seasonal Comparison % Change Kodiak All Seasons P = P < Aut00 vs. Win01 P < Aut Win01 vs. Sum01 P < Sum Sum01 vs. Aut01 P = Semidi All Seasons P = Aut00 vs. Win01 P = P < Aut Win01 vs. Sum01 P = P = Sum Sum01 vs. Aut01 P = P = Aut Shumagin All Seasons P = P < Aut00 vs. Win01 P < Aut Win01 vs. Sum01 P < Sum Sum01 vs. Aut01 P = Aut01 106

112 Table 4. Summary results of ANCOVA tests of regional differences in length-weight relationships by season. Differences among slopes and Y-intercepts were tested at critical values of P = and P = 0.005, respectively (Bonferroni-corrected, α = 0.05). Bold numbers indicate significance. For each comparison, the region with the greater Y- intercept is included. Percent differences between the regional length-specific weights were calculated from back-transformed values. H 0 : Slopes are equal H 0 : Y-intercepts are equal Relationship (greater Y-int) Regional Comparison Aut00 All regions P = P = Kodiak vs. Semidi P = Kodiak 5.9 Kodiak vs. Shumagin P = Kodiak 4.0 Semidi vs. Shumagin P = Shumagin 1.8 % Difference Win01 All regions P = P = Kodiak vs. Semidi P = Kodiak 4.8 Kodiak vs. Shumagin P = Kodiak 5.7 Semidi vs. Shumagin P = Sum01 All regions P = P = Aut01 All regions P = P =

113 Table 5. P-values for post-hoc pairwise comparisons conducted after an ANCOVA of whole body energy content with standard length as the covariate. Differences between consecutive seasons were tested for each region at a critical value of P = (Bonferroni-corrected, α = 0.05) and differences between regions were tested for each season at a critical value of P = (Bonferroni-corrected, α = 0.05). Bold numbers indicate significant test results. Region Aut00 vs. Win01 Win01 vs. Sum01 Sum01 vs. Aut01 Kodiak a a Semidi Shumagin < Season Kod. vs. Sem. Kod. vs. Shu. Sem. vs. Shu. Aut < Win Sum a a < Aut a a a Most of the energy density measurements for Sum01 and Aut01 in Kodiak were calculated from proximate composition and not determined using bomb calorimetry. 108

114 Figures: Figure 1. Study area and location of regions (Kodiak, Semidi, Shumagin) where the 2000 year class of walleye pollock was sampled as age-0 and age-1 juveniles during autumn 2000 autumn Length and weight data were collected from juveniles sampled at all locations ( ) and whole body energy content was measured using juveniles from a subset of the stations (O). 109